zig/lib/std/zig/AstGen.zig

13897 lines
544 KiB
Zig

//! Ingests an AST and produces ZIR code.
const AstGen = @This();
const std = @import("std");
const Ast = std.zig.Ast;
const mem = std.mem;
const Allocator = std.mem.Allocator;
const assert = std.debug.assert;
const ArrayListUnmanaged = std.ArrayListUnmanaged;
const StringIndexAdapter = std.hash_map.StringIndexAdapter;
const StringIndexContext = std.hash_map.StringIndexContext;
const isPrimitive = std.zig.primitives.isPrimitive;
const Zir = std.zig.Zir;
const BuiltinFn = std.zig.BuiltinFn;
const AstRlAnnotate = std.zig.AstRlAnnotate;
gpa: Allocator,
tree: *const Ast,
/// The set of nodes which, given the choice, must expose a result pointer to
/// sub-expressions. See `AstRlAnnotate` for details.
nodes_need_rl: *const AstRlAnnotate.RlNeededSet,
instructions: std.MultiArrayList(Zir.Inst) = .{},
extra: ArrayListUnmanaged(u32) = .{},
string_bytes: ArrayListUnmanaged(u8) = .{},
/// Tracks the current byte offset within the source file.
/// Used to populate line deltas in the ZIR. AstGen maintains
/// this "cursor" throughout the entire AST lowering process in order
/// to avoid starting over the line/column scan for every declaration, which
/// would be O(N^2).
source_offset: u32 = 0,
/// Tracks the corresponding line of `source_offset`.
/// This value is absolute.
source_line: u32 = 0,
/// Tracks the corresponding column of `source_offset`.
/// This value is absolute.
source_column: u32 = 0,
/// Used for temporary allocations; freed after AstGen is complete.
/// The resulting ZIR code has no references to anything in this arena.
arena: Allocator,
string_table: std.HashMapUnmanaged(u32, void, StringIndexContext, std.hash_map.default_max_load_percentage) = .{},
compile_errors: ArrayListUnmanaged(Zir.Inst.CompileErrors.Item) = .{},
/// The topmost block of the current function.
fn_block: ?*GenZir = null,
fn_var_args: bool = false,
/// Whether we are somewhere within a function. If `true`, any container decls may be
/// generic and thus must be tunneled through closure.
within_fn: bool = false,
/// The return type of the current function. This may be a trivial `Ref`, or
/// otherwise it refers to a `ret_type` instruction.
fn_ret_ty: Zir.Inst.Ref = .none,
/// Maps string table indexes to the first `@import` ZIR instruction
/// that uses this string as the operand.
imports: std.AutoArrayHashMapUnmanaged(Zir.NullTerminatedString, Ast.TokenIndex) = .{},
/// Used for temporary storage when building payloads.
scratch: std.ArrayListUnmanaged(u32) = .{},
/// Whenever a `ref` instruction is needed, it is created and saved in this
/// table instead of being immediately appended to the current block body.
/// Then, when the instruction is being added to the parent block (typically from
/// setBlockBody), if it has a ref_table entry, then the ref instruction is added
/// there. This makes sure two properties are upheld:
/// 1. All pointers to the same locals return the same address. This is required
/// to be compliant with the language specification.
/// 2. `ref` instructions will dominate their uses. This is a required property
/// of ZIR.
/// The key is the ref operand; the value is the ref instruction.
ref_table: std.AutoHashMapUnmanaged(Zir.Inst.Index, Zir.Inst.Index) = .{},
const InnerError = error{ OutOfMemory, AnalysisFail };
fn addExtra(astgen: *AstGen, extra: anytype) Allocator.Error!u32 {
const fields = std.meta.fields(@TypeOf(extra));
try astgen.extra.ensureUnusedCapacity(astgen.gpa, fields.len);
return addExtraAssumeCapacity(astgen, extra);
}
fn addExtraAssumeCapacity(astgen: *AstGen, extra: anytype) u32 {
const fields = std.meta.fields(@TypeOf(extra));
const extra_index: u32 = @intCast(astgen.extra.items.len);
astgen.extra.items.len += fields.len;
setExtra(astgen, extra_index, extra);
return extra_index;
}
fn setExtra(astgen: *AstGen, index: usize, extra: anytype) void {
const fields = std.meta.fields(@TypeOf(extra));
var i = index;
inline for (fields) |field| {
astgen.extra.items[i] = switch (field.type) {
u32 => @field(extra, field.name),
Zir.Inst.Ref,
Zir.Inst.Index,
Zir.Inst.Declaration.Name,
Zir.NullTerminatedString,
=> @intFromEnum(@field(extra, field.name)),
i32,
Zir.Inst.Call.Flags,
Zir.Inst.BuiltinCall.Flags,
Zir.Inst.SwitchBlock.Bits,
Zir.Inst.SwitchBlockErrUnion.Bits,
Zir.Inst.FuncFancy.Bits,
Zir.Inst.Declaration.Flags,
=> @bitCast(@field(extra, field.name)),
else => @compileError("bad field type"),
};
i += 1;
}
}
fn reserveExtra(astgen: *AstGen, size: usize) Allocator.Error!u32 {
const extra_index: u32 = @intCast(astgen.extra.items.len);
try astgen.extra.resize(astgen.gpa, extra_index + size);
return extra_index;
}
fn appendRefs(astgen: *AstGen, refs: []const Zir.Inst.Ref) !void {
return astgen.extra.appendSlice(astgen.gpa, @ptrCast(refs));
}
fn appendRefsAssumeCapacity(astgen: *AstGen, refs: []const Zir.Inst.Ref) void {
astgen.extra.appendSliceAssumeCapacity(@ptrCast(refs));
}
pub fn generate(gpa: Allocator, tree: Ast) Allocator.Error!Zir {
var arena = std.heap.ArenaAllocator.init(gpa);
defer arena.deinit();
var nodes_need_rl = try AstRlAnnotate.annotate(gpa, arena.allocator(), tree);
defer nodes_need_rl.deinit(gpa);
var astgen: AstGen = .{
.gpa = gpa,
.arena = arena.allocator(),
.tree = &tree,
.nodes_need_rl = &nodes_need_rl,
};
defer astgen.deinit(gpa);
// String table index 0 is reserved for `NullTerminatedString.empty`.
try astgen.string_bytes.append(gpa, 0);
// We expect at least as many ZIR instructions and extra data items
// as AST nodes.
try astgen.instructions.ensureTotalCapacity(gpa, tree.nodes.len);
// First few indexes of extra are reserved and set at the end.
const reserved_count = @typeInfo(Zir.ExtraIndex).Enum.fields.len;
try astgen.extra.ensureTotalCapacity(gpa, tree.nodes.len + reserved_count);
astgen.extra.items.len += reserved_count;
var top_scope: Scope.Top = .{};
var gz_instructions: std.ArrayListUnmanaged(Zir.Inst.Index) = .{};
var gen_scope: GenZir = .{
.is_comptime = true,
.parent = &top_scope.base,
.anon_name_strategy = .parent,
.decl_node_index = 0,
.decl_line = 0,
.astgen = &astgen,
.instructions = &gz_instructions,
.instructions_top = 0,
};
defer gz_instructions.deinit(gpa);
// The AST -> ZIR lowering process assumes an AST that does not have any
// parse errors.
if (tree.errors.len == 0) {
if (AstGen.structDeclInner(
&gen_scope,
&gen_scope.base,
0,
tree.containerDeclRoot(),
.auto,
0,
)) |struct_decl_ref| {
assert(struct_decl_ref.toIndex().? == .main_struct_inst);
} else |err| switch (err) {
error.OutOfMemory => return error.OutOfMemory,
error.AnalysisFail => {}, // Handled via compile_errors below.
}
} else {
try lowerAstErrors(&astgen);
}
const err_index = @intFromEnum(Zir.ExtraIndex.compile_errors);
if (astgen.compile_errors.items.len == 0) {
astgen.extra.items[err_index] = 0;
} else {
try astgen.extra.ensureUnusedCapacity(gpa, 1 + astgen.compile_errors.items.len *
@typeInfo(Zir.Inst.CompileErrors.Item).Struct.fields.len);
astgen.extra.items[err_index] = astgen.addExtraAssumeCapacity(Zir.Inst.CompileErrors{
.items_len = @intCast(astgen.compile_errors.items.len),
});
for (astgen.compile_errors.items) |item| {
_ = astgen.addExtraAssumeCapacity(item);
}
}
const imports_index = @intFromEnum(Zir.ExtraIndex.imports);
if (astgen.imports.count() == 0) {
astgen.extra.items[imports_index] = 0;
} else {
try astgen.extra.ensureUnusedCapacity(gpa, @typeInfo(Zir.Inst.Imports).Struct.fields.len +
astgen.imports.count() * @typeInfo(Zir.Inst.Imports.Item).Struct.fields.len);
astgen.extra.items[imports_index] = astgen.addExtraAssumeCapacity(Zir.Inst.Imports{
.imports_len = @intCast(astgen.imports.count()),
});
var it = astgen.imports.iterator();
while (it.next()) |entry| {
_ = astgen.addExtraAssumeCapacity(Zir.Inst.Imports.Item{
.name = entry.key_ptr.*,
.token = entry.value_ptr.*,
});
}
}
return Zir{
.instructions = astgen.instructions.toOwnedSlice(),
.string_bytes = try astgen.string_bytes.toOwnedSlice(gpa),
.extra = try astgen.extra.toOwnedSlice(gpa),
};
}
fn deinit(astgen: *AstGen, gpa: Allocator) void {
astgen.instructions.deinit(gpa);
astgen.extra.deinit(gpa);
astgen.string_table.deinit(gpa);
astgen.string_bytes.deinit(gpa);
astgen.compile_errors.deinit(gpa);
astgen.imports.deinit(gpa);
astgen.scratch.deinit(gpa);
astgen.ref_table.deinit(gpa);
}
const ResultInfo = struct {
/// The semantics requested for the result location
rl: Loc,
/// The "operator" consuming the result location
ctx: Context = .none,
/// Turns a `coerced_ty` back into a `ty`. Should be called at branch points
/// such as if and switch expressions.
fn br(ri: ResultInfo) ResultInfo {
return switch (ri.rl) {
.coerced_ty => |ty| .{
.rl = .{ .ty = ty },
.ctx = ri.ctx,
},
else => ri,
};
}
fn zirTag(ri: ResultInfo) Zir.Inst.Tag {
switch (ri.rl) {
.ty => return switch (ri.ctx) {
.shift_op => .as_shift_operand,
else => .as_node,
},
else => unreachable,
}
}
const Loc = union(enum) {
/// The expression is the right-hand side of assignment to `_`. Only the side-effects of the
/// expression should be generated. The result instruction from the expression must
/// be ignored.
discard,
/// The expression has an inferred type, and it will be evaluated as an rvalue.
none,
/// The expression will be coerced into this type, but it will be evaluated as an rvalue.
ty: Zir.Inst.Ref,
/// Same as `ty` but it is guaranteed that Sema will additionally perform the coercion,
/// so no `as` instruction needs to be emitted.
coerced_ty: Zir.Inst.Ref,
/// The expression must generate a pointer rather than a value. For example, the left hand side
/// of an assignment uses this kind of result location.
ref,
/// The expression must generate a pointer rather than a value, and the pointer will be coerced
/// by other code to this type, which is guaranteed by earlier instructions to be a pointer type.
ref_coerced_ty: Zir.Inst.Ref,
/// The expression must store its result into this typed pointer. The result instruction
/// from the expression must be ignored.
ptr: PtrResultLoc,
/// The expression must store its result into this allocation, which has an inferred type.
/// The result instruction from the expression must be ignored.
/// Always an instruction with tag `alloc_inferred`.
inferred_ptr: Zir.Inst.Ref,
/// The expression has a sequence of pointers to store its results into due to a destructure
/// operation. Each of these pointers may or may not have an inferred type.
destructure: struct {
/// The AST node of the destructure operation itself.
src_node: Ast.Node.Index,
/// The pointers to store results into.
components: []const DestructureComponent,
},
const DestructureComponent = union(enum) {
typed_ptr: PtrResultLoc,
inferred_ptr: Zir.Inst.Ref,
discard,
};
const PtrResultLoc = struct {
inst: Zir.Inst.Ref,
src_node: ?Ast.Node.Index = null,
};
/// Find the result type for a cast builtin given the result location.
/// If the location does not have a known result type, emits an error on
/// the given node.
fn resultType(rl: Loc, gz: *GenZir, node: Ast.Node.Index) !?Zir.Inst.Ref {
return switch (rl) {
.discard, .none, .ref, .inferred_ptr, .destructure => null,
.ty, .coerced_ty => |ty_ref| ty_ref,
.ref_coerced_ty => |ptr_ty| try gz.addUnNode(.elem_type, ptr_ty, node),
.ptr => |ptr| {
const ptr_ty = try gz.addUnNode(.typeof, ptr.inst, node);
return try gz.addUnNode(.elem_type, ptr_ty, node);
},
};
}
fn resultTypeForCast(rl: Loc, gz: *GenZir, node: Ast.Node.Index, builtin_name: []const u8) !Zir.Inst.Ref {
const astgen = gz.astgen;
if (try rl.resultType(gz, node)) |ty| return ty;
switch (rl) {
.destructure => |destructure| return astgen.failNodeNotes(node, "{s} must have a known result type", .{builtin_name}, &.{
try astgen.errNoteNode(destructure.src_node, "destructure expressions do not provide a single result type", .{}),
try astgen.errNoteNode(node, "use @as to provide explicit result type", .{}),
}),
else => return astgen.failNodeNotes(node, "{s} must have a known result type", .{builtin_name}, &.{
try astgen.errNoteNode(node, "use @as to provide explicit result type", .{}),
}),
}
}
};
const Context = enum {
/// The expression is the operand to a return expression.
@"return",
/// The expression is the input to an error-handling operator (if-else, try, or catch).
error_handling_expr,
/// The expression is the right-hand side of a shift operation.
shift_op,
/// The expression is an argument in a function call.
fn_arg,
/// The expression is the right-hand side of an initializer for a `const` variable
const_init,
/// The expression is the right-hand side of an assignment expression.
assignment,
/// No specific operator in particular.
none,
};
};
const coerced_align_ri: ResultInfo = .{ .rl = .{ .coerced_ty = .u29_type } };
const coerced_addrspace_ri: ResultInfo = .{ .rl = .{ .coerced_ty = .address_space_type } };
const coerced_linksection_ri: ResultInfo = .{ .rl = .{ .coerced_ty = .slice_const_u8_type } };
const coerced_type_ri: ResultInfo = .{ .rl = .{ .coerced_ty = .type_type } };
const coerced_bool_ri: ResultInfo = .{ .rl = .{ .coerced_ty = .bool_type } };
fn typeExpr(gz: *GenZir, scope: *Scope, type_node: Ast.Node.Index) InnerError!Zir.Inst.Ref {
return comptimeExpr(gz, scope, coerced_type_ri, type_node);
}
fn reachableTypeExpr(
gz: *GenZir,
scope: *Scope,
type_node: Ast.Node.Index,
reachable_node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
return reachableExprComptime(gz, scope, coerced_type_ri, type_node, reachable_node, true);
}
/// Same as `expr` but fails with a compile error if the result type is `noreturn`.
fn reachableExpr(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
reachable_node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
return reachableExprComptime(gz, scope, ri, node, reachable_node, false);
}
fn reachableExprComptime(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
reachable_node: Ast.Node.Index,
force_comptime: bool,
) InnerError!Zir.Inst.Ref {
const result_inst = if (force_comptime)
try comptimeExpr(gz, scope, ri, node)
else
try expr(gz, scope, ri, node);
if (gz.refIsNoReturn(result_inst)) {
try gz.astgen.appendErrorNodeNotes(reachable_node, "unreachable code", .{}, &[_]u32{
try gz.astgen.errNoteNode(node, "control flow is diverted here", .{}),
});
}
return result_inst;
}
fn lvalExpr(gz: *GenZir, scope: *Scope, node: Ast.Node.Index) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const node_tags = tree.nodes.items(.tag);
const main_tokens = tree.nodes.items(.main_token);
switch (node_tags[node]) {
.root => unreachable,
.@"usingnamespace" => unreachable,
.test_decl => unreachable,
.global_var_decl => unreachable,
.local_var_decl => unreachable,
.simple_var_decl => unreachable,
.aligned_var_decl => unreachable,
.switch_case => unreachable,
.switch_case_inline => unreachable,
.switch_case_one => unreachable,
.switch_case_inline_one => unreachable,
.container_field_init => unreachable,
.container_field_align => unreachable,
.container_field => unreachable,
.asm_output => unreachable,
.asm_input => unreachable,
.assign,
.assign_destructure,
.assign_bit_and,
.assign_bit_or,
.assign_shl,
.assign_shl_sat,
.assign_shr,
.assign_bit_xor,
.assign_div,
.assign_sub,
.assign_sub_wrap,
.assign_sub_sat,
.assign_mod,
.assign_add,
.assign_add_wrap,
.assign_add_sat,
.assign_mul,
.assign_mul_wrap,
.assign_mul_sat,
.add,
.add_wrap,
.add_sat,
.sub,
.sub_wrap,
.sub_sat,
.mul,
.mul_wrap,
.mul_sat,
.div,
.mod,
.bit_and,
.bit_or,
.shl,
.shl_sat,
.shr,
.bit_xor,
.bang_equal,
.equal_equal,
.greater_than,
.greater_or_equal,
.less_than,
.less_or_equal,
.array_cat,
.array_mult,
.bool_and,
.bool_or,
.@"asm",
.asm_simple,
.string_literal,
.number_literal,
.call,
.call_comma,
.async_call,
.async_call_comma,
.call_one,
.call_one_comma,
.async_call_one,
.async_call_one_comma,
.unreachable_literal,
.@"return",
.@"if",
.if_simple,
.@"while",
.while_simple,
.while_cont,
.bool_not,
.address_of,
.optional_type,
.block,
.block_semicolon,
.block_two,
.block_two_semicolon,
.@"break",
.ptr_type_aligned,
.ptr_type_sentinel,
.ptr_type,
.ptr_type_bit_range,
.array_type,
.array_type_sentinel,
.enum_literal,
.multiline_string_literal,
.char_literal,
.@"defer",
.@"errdefer",
.@"catch",
.error_union,
.merge_error_sets,
.switch_range,
.for_range,
.@"await",
.bit_not,
.negation,
.negation_wrap,
.@"resume",
.@"try",
.slice,
.slice_open,
.slice_sentinel,
.array_init_one,
.array_init_one_comma,
.array_init_dot_two,
.array_init_dot_two_comma,
.array_init_dot,
.array_init_dot_comma,
.array_init,
.array_init_comma,
.struct_init_one,
.struct_init_one_comma,
.struct_init_dot_two,
.struct_init_dot_two_comma,
.struct_init_dot,
.struct_init_dot_comma,
.struct_init,
.struct_init_comma,
.@"switch",
.switch_comma,
.@"for",
.for_simple,
.@"suspend",
.@"continue",
.fn_proto_simple,
.fn_proto_multi,
.fn_proto_one,
.fn_proto,
.fn_decl,
.anyframe_type,
.anyframe_literal,
.error_set_decl,
.container_decl,
.container_decl_trailing,
.container_decl_two,
.container_decl_two_trailing,
.container_decl_arg,
.container_decl_arg_trailing,
.tagged_union,
.tagged_union_trailing,
.tagged_union_two,
.tagged_union_two_trailing,
.tagged_union_enum_tag,
.tagged_union_enum_tag_trailing,
.@"comptime",
.@"nosuspend",
.error_value,
=> return astgen.failNode(node, "invalid left-hand side to assignment", .{}),
.builtin_call,
.builtin_call_comma,
.builtin_call_two,
.builtin_call_two_comma,
=> {
const builtin_token = main_tokens[node];
const builtin_name = tree.tokenSlice(builtin_token);
// If the builtin is an invalid name, we don't cause an error here; instead
// let it pass, and the error will be "invalid builtin function" later.
if (BuiltinFn.list.get(builtin_name)) |info| {
if (!info.allows_lvalue) {
return astgen.failNode(node, "invalid left-hand side to assignment", .{});
}
}
},
// These can be assigned to.
.unwrap_optional,
.deref,
.field_access,
.array_access,
.identifier,
.grouped_expression,
.@"orelse",
=> {},
}
return expr(gz, scope, .{ .rl = .ref }, node);
}
/// Turn Zig AST into untyped ZIR instructions.
/// When `rl` is discard, ptr, inferred_ptr, or inferred_ptr, the
/// result instruction can be used to inspect whether it is isNoReturn() but that is it,
/// it must otherwise not be used.
fn expr(gz: *GenZir, scope: *Scope, ri: ResultInfo, node: Ast.Node.Index) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const main_tokens = tree.nodes.items(.main_token);
const token_tags = tree.tokens.items(.tag);
const node_datas = tree.nodes.items(.data);
const node_tags = tree.nodes.items(.tag);
const prev_anon_name_strategy = gz.anon_name_strategy;
defer gz.anon_name_strategy = prev_anon_name_strategy;
if (!nodeUsesAnonNameStrategy(tree, node)) {
gz.anon_name_strategy = .anon;
}
switch (node_tags[node]) {
.root => unreachable, // Top-level declaration.
.@"usingnamespace" => unreachable, // Top-level declaration.
.test_decl => unreachable, // Top-level declaration.
.container_field_init => unreachable, // Top-level declaration.
.container_field_align => unreachable, // Top-level declaration.
.container_field => unreachable, // Top-level declaration.
.fn_decl => unreachable, // Top-level declaration.
.global_var_decl => unreachable, // Handled in `blockExpr`.
.local_var_decl => unreachable, // Handled in `blockExpr`.
.simple_var_decl => unreachable, // Handled in `blockExpr`.
.aligned_var_decl => unreachable, // Handled in `blockExpr`.
.@"defer" => unreachable, // Handled in `blockExpr`.
.@"errdefer" => unreachable, // Handled in `blockExpr`.
.switch_case => unreachable, // Handled in `switchExpr`.
.switch_case_inline => unreachable, // Handled in `switchExpr`.
.switch_case_one => unreachable, // Handled in `switchExpr`.
.switch_case_inline_one => unreachable, // Handled in `switchExpr`.
.switch_range => unreachable, // Handled in `switchExpr`.
.asm_output => unreachable, // Handled in `asmExpr`.
.asm_input => unreachable, // Handled in `asmExpr`.
.for_range => unreachable, // Handled in `forExpr`.
.assign => {
try assign(gz, scope, node);
return rvalue(gz, ri, .void_value, node);
},
.assign_destructure => {
// Note that this variant does not declare any new var/const: that
// variant is handled by `blockExprStmts`.
try assignDestructure(gz, scope, node);
return rvalue(gz, ri, .void_value, node);
},
.assign_shl => {
try assignShift(gz, scope, node, .shl);
return rvalue(gz, ri, .void_value, node);
},
.assign_shl_sat => {
try assignShiftSat(gz, scope, node);
return rvalue(gz, ri, .void_value, node);
},
.assign_shr => {
try assignShift(gz, scope, node, .shr);
return rvalue(gz, ri, .void_value, node);
},
.assign_bit_and => {
try assignOp(gz, scope, node, .bit_and);
return rvalue(gz, ri, .void_value, node);
},
.assign_bit_or => {
try assignOp(gz, scope, node, .bit_or);
return rvalue(gz, ri, .void_value, node);
},
.assign_bit_xor => {
try assignOp(gz, scope, node, .xor);
return rvalue(gz, ri, .void_value, node);
},
.assign_div => {
try assignOp(gz, scope, node, .div);
return rvalue(gz, ri, .void_value, node);
},
.assign_sub => {
try assignOp(gz, scope, node, .sub);
return rvalue(gz, ri, .void_value, node);
},
.assign_sub_wrap => {
try assignOp(gz, scope, node, .subwrap);
return rvalue(gz, ri, .void_value, node);
},
.assign_sub_sat => {
try assignOp(gz, scope, node, .sub_sat);
return rvalue(gz, ri, .void_value, node);
},
.assign_mod => {
try assignOp(gz, scope, node, .mod_rem);
return rvalue(gz, ri, .void_value, node);
},
.assign_add => {
try assignOp(gz, scope, node, .add);
return rvalue(gz, ri, .void_value, node);
},
.assign_add_wrap => {
try assignOp(gz, scope, node, .addwrap);
return rvalue(gz, ri, .void_value, node);
},
.assign_add_sat => {
try assignOp(gz, scope, node, .add_sat);
return rvalue(gz, ri, .void_value, node);
},
.assign_mul => {
try assignOp(gz, scope, node, .mul);
return rvalue(gz, ri, .void_value, node);
},
.assign_mul_wrap => {
try assignOp(gz, scope, node, .mulwrap);
return rvalue(gz, ri, .void_value, node);
},
.assign_mul_sat => {
try assignOp(gz, scope, node, .mul_sat);
return rvalue(gz, ri, .void_value, node);
},
// zig fmt: off
.shl => return shiftOp(gz, scope, ri, node, node_datas[node].lhs, node_datas[node].rhs, .shl),
.shr => return shiftOp(gz, scope, ri, node, node_datas[node].lhs, node_datas[node].rhs, .shr),
.add => return simpleBinOp(gz, scope, ri, node, .add),
.add_wrap => return simpleBinOp(gz, scope, ri, node, .addwrap),
.add_sat => return simpleBinOp(gz, scope, ri, node, .add_sat),
.sub => return simpleBinOp(gz, scope, ri, node, .sub),
.sub_wrap => return simpleBinOp(gz, scope, ri, node, .subwrap),
.sub_sat => return simpleBinOp(gz, scope, ri, node, .sub_sat),
.mul => return simpleBinOp(gz, scope, ri, node, .mul),
.mul_wrap => return simpleBinOp(gz, scope, ri, node, .mulwrap),
.mul_sat => return simpleBinOp(gz, scope, ri, node, .mul_sat),
.div => return simpleBinOp(gz, scope, ri, node, .div),
.mod => return simpleBinOp(gz, scope, ri, node, .mod_rem),
.shl_sat => return simpleBinOp(gz, scope, ri, node, .shl_sat),
.bit_and => return simpleBinOp(gz, scope, ri, node, .bit_and),
.bit_or => return simpleBinOp(gz, scope, ri, node, .bit_or),
.bit_xor => return simpleBinOp(gz, scope, ri, node, .xor),
.bang_equal => return simpleBinOp(gz, scope, ri, node, .cmp_neq),
.equal_equal => return simpleBinOp(gz, scope, ri, node, .cmp_eq),
.greater_than => return simpleBinOp(gz, scope, ri, node, .cmp_gt),
.greater_or_equal => return simpleBinOp(gz, scope, ri, node, .cmp_gte),
.less_than => return simpleBinOp(gz, scope, ri, node, .cmp_lt),
.less_or_equal => return simpleBinOp(gz, scope, ri, node, .cmp_lte),
.array_cat => return simpleBinOp(gz, scope, ri, node, .array_cat),
.array_mult => {
// This syntax form does not currently use the result type in the language specification.
// However, the result type can be used to emit more optimal code for large multiplications by
// having Sema perform a coercion before the multiplication operation.
const result = try gz.addPlNode(.array_mul, node, Zir.Inst.ArrayMul{
.res_ty = if (try ri.rl.resultType(gz, node)) |t| t else .none,
.lhs = try expr(gz, scope, .{ .rl = .none }, node_datas[node].lhs),
.rhs = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, node_datas[node].rhs),
});
return rvalue(gz, ri, result, node);
},
.error_union => return simpleBinOp(gz, scope, ri, node, .error_union_type),
.merge_error_sets => return simpleBinOp(gz, scope, ri, node, .merge_error_sets),
.bool_and => return boolBinOp(gz, scope, ri, node, .bool_br_and),
.bool_or => return boolBinOp(gz, scope, ri, node, .bool_br_or),
.bool_not => return simpleUnOp(gz, scope, ri, node, coerced_bool_ri, node_datas[node].lhs, .bool_not),
.bit_not => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, node_datas[node].lhs, .bit_not),
.negation => return negation(gz, scope, ri, node),
.negation_wrap => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, node_datas[node].lhs, .negate_wrap),
.identifier => return identifier(gz, scope, ri, node),
.asm_simple,
.@"asm",
=> return asmExpr(gz, scope, ri, node, tree.fullAsm(node).?),
.string_literal => return stringLiteral(gz, ri, node),
.multiline_string_literal => return multilineStringLiteral(gz, ri, node),
.number_literal => return numberLiteral(gz, ri, node, node, .positive),
// zig fmt: on
.builtin_call_two, .builtin_call_two_comma => {
if (node_datas[node].lhs == 0) {
const params = [_]Ast.Node.Index{};
return builtinCall(gz, scope, ri, node, &params);
} else if (node_datas[node].rhs == 0) {
const params = [_]Ast.Node.Index{node_datas[node].lhs};
return builtinCall(gz, scope, ri, node, &params);
} else {
const params = [_]Ast.Node.Index{ node_datas[node].lhs, node_datas[node].rhs };
return builtinCall(gz, scope, ri, node, &params);
}
},
.builtin_call, .builtin_call_comma => {
const params = tree.extra_data[node_datas[node].lhs..node_datas[node].rhs];
return builtinCall(gz, scope, ri, node, params);
},
.call_one,
.call_one_comma,
.async_call_one,
.async_call_one_comma,
.call,
.call_comma,
.async_call,
.async_call_comma,
=> {
var buf: [1]Ast.Node.Index = undefined;
return callExpr(gz, scope, ri, node, tree.fullCall(&buf, node).?);
},
.unreachable_literal => {
try emitDbgNode(gz, node);
_ = try gz.addAsIndex(.{
.tag = .@"unreachable",
.data = .{ .@"unreachable" = .{
.src_node = gz.nodeIndexToRelative(node),
} },
});
return Zir.Inst.Ref.unreachable_value;
},
.@"return" => return ret(gz, scope, node),
.field_access => return fieldAccess(gz, scope, ri, node),
.if_simple,
.@"if",
=> {
const if_full = tree.fullIf(node).?;
no_switch_on_err: {
const error_token = if_full.error_token orelse break :no_switch_on_err;
switch (node_tags[if_full.ast.else_expr]) {
.@"switch", .switch_comma => {},
else => break :no_switch_on_err,
}
const switch_operand = node_datas[if_full.ast.else_expr].lhs;
if (node_tags[switch_operand] != .identifier) break :no_switch_on_err;
if (!mem.eql(u8, tree.tokenSlice(error_token), tree.tokenSlice(main_tokens[switch_operand]))) break :no_switch_on_err;
return switchExprErrUnion(gz, scope, ri.br(), node, .@"if");
}
return ifExpr(gz, scope, ri.br(), node, if_full);
},
.while_simple,
.while_cont,
.@"while",
=> return whileExpr(gz, scope, ri.br(), node, tree.fullWhile(node).?, false),
.for_simple, .@"for" => return forExpr(gz, scope, ri.br(), node, tree.fullFor(node).?, false),
.slice_open => {
const lhs = try expr(gz, scope, .{ .rl = .ref }, node_datas[node].lhs);
const cursor = maybeAdvanceSourceCursorToMainToken(gz, node);
const start = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, node_datas[node].rhs);
try emitDbgStmt(gz, cursor);
const result = try gz.addPlNode(.slice_start, node, Zir.Inst.SliceStart{
.lhs = lhs,
.start = start,
});
return rvalue(gz, ri, result, node);
},
.slice => {
const extra = tree.extraData(node_datas[node].rhs, Ast.Node.Slice);
const lhs_node = node_datas[node].lhs;
const lhs_tag = node_tags[lhs_node];
const lhs_is_slice_sentinel = lhs_tag == .slice_sentinel;
const lhs_is_open_slice = lhs_tag == .slice_open or
(lhs_is_slice_sentinel and tree.extraData(node_datas[lhs_node].rhs, Ast.Node.SliceSentinel).end == 0);
if (lhs_is_open_slice and nodeIsTriviallyZero(tree, extra.start)) {
const lhs = try expr(gz, scope, .{ .rl = .ref }, node_datas[lhs_node].lhs);
const start = if (lhs_is_slice_sentinel) start: {
const lhs_extra = tree.extraData(node_datas[lhs_node].rhs, Ast.Node.SliceSentinel);
break :start try expr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, lhs_extra.start);
} else try expr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, node_datas[lhs_node].rhs);
const cursor = maybeAdvanceSourceCursorToMainToken(gz, node);
const len = if (extra.end != 0) try expr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, extra.end) else .none;
try emitDbgStmt(gz, cursor);
const result = try gz.addPlNode(.slice_length, node, Zir.Inst.SliceLength{
.lhs = lhs,
.start = start,
.len = len,
.start_src_node_offset = gz.nodeIndexToRelative(lhs_node),
.sentinel = .none,
});
return rvalue(gz, ri, result, node);
}
const lhs = try expr(gz, scope, .{ .rl = .ref }, node_datas[node].lhs);
const cursor = maybeAdvanceSourceCursorToMainToken(gz, node);
const start = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, extra.start);
const end = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, extra.end);
try emitDbgStmt(gz, cursor);
const result = try gz.addPlNode(.slice_end, node, Zir.Inst.SliceEnd{
.lhs = lhs,
.start = start,
.end = end,
});
return rvalue(gz, ri, result, node);
},
.slice_sentinel => {
const extra = tree.extraData(node_datas[node].rhs, Ast.Node.SliceSentinel);
const lhs_node = node_datas[node].lhs;
const lhs_tag = node_tags[lhs_node];
const lhs_is_slice_sentinel = lhs_tag == .slice_sentinel;
const lhs_is_open_slice = lhs_tag == .slice_open or
(lhs_is_slice_sentinel and tree.extraData(node_datas[lhs_node].rhs, Ast.Node.SliceSentinel).end == 0);
if (lhs_is_open_slice and nodeIsTriviallyZero(tree, extra.start)) {
const lhs = try expr(gz, scope, .{ .rl = .ref }, node_datas[lhs_node].lhs);
const start = if (lhs_is_slice_sentinel) start: {
const lhs_extra = tree.extraData(node_datas[lhs_node].rhs, Ast.Node.SliceSentinel);
break :start try expr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, lhs_extra.start);
} else try expr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, node_datas[lhs_node].rhs);
const cursor = maybeAdvanceSourceCursorToMainToken(gz, node);
const len = if (extra.end != 0) try expr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, extra.end) else .none;
const sentinel = try expr(gz, scope, .{ .rl = .none }, extra.sentinel);
try emitDbgStmt(gz, cursor);
const result = try gz.addPlNode(.slice_length, node, Zir.Inst.SliceLength{
.lhs = lhs,
.start = start,
.len = len,
.start_src_node_offset = gz.nodeIndexToRelative(lhs_node),
.sentinel = sentinel,
});
return rvalue(gz, ri, result, node);
}
const lhs = try expr(gz, scope, .{ .rl = .ref }, node_datas[node].lhs);
const cursor = maybeAdvanceSourceCursorToMainToken(gz, node);
const start = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, extra.start);
const end = if (extra.end != 0) try expr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, extra.end) else .none;
const sentinel = try expr(gz, scope, .{ .rl = .none }, extra.sentinel);
try emitDbgStmt(gz, cursor);
const result = try gz.addPlNode(.slice_sentinel, node, Zir.Inst.SliceSentinel{
.lhs = lhs,
.start = start,
.end = end,
.sentinel = sentinel,
});
return rvalue(gz, ri, result, node);
},
.deref => {
const lhs = try expr(gz, scope, .{ .rl = .none }, node_datas[node].lhs);
_ = try gz.addUnNode(.validate_deref, lhs, node);
switch (ri.rl) {
.ref, .ref_coerced_ty => return lhs,
else => {
const result = try gz.addUnNode(.load, lhs, node);
return rvalue(gz, ri, result, node);
},
}
},
.address_of => {
const operand_rl: ResultInfo.Loc = if (try ri.rl.resultType(gz, node)) |res_ty_inst| rl: {
_ = try gz.addUnTok(.validate_ref_ty, res_ty_inst, tree.firstToken(node));
break :rl .{ .ref_coerced_ty = res_ty_inst };
} else .ref;
const result = try expr(gz, scope, .{ .rl = operand_rl }, node_datas[node].lhs);
return rvalue(gz, ri, result, node);
},
.optional_type => {
const operand = try typeExpr(gz, scope, node_datas[node].lhs);
const result = try gz.addUnNode(.optional_type, operand, node);
return rvalue(gz, ri, result, node);
},
.unwrap_optional => switch (ri.rl) {
.ref, .ref_coerced_ty => {
const lhs = try expr(gz, scope, .{ .rl = .ref }, node_datas[node].lhs);
const cursor = maybeAdvanceSourceCursorToMainToken(gz, node);
try emitDbgStmt(gz, cursor);
return gz.addUnNode(.optional_payload_safe_ptr, lhs, node);
},
else => {
const lhs = try expr(gz, scope, .{ .rl = .none }, node_datas[node].lhs);
const cursor = maybeAdvanceSourceCursorToMainToken(gz, node);
try emitDbgStmt(gz, cursor);
return rvalue(gz, ri, try gz.addUnNode(.optional_payload_safe, lhs, node), node);
},
},
.block_two, .block_two_semicolon => {
const statements = [2]Ast.Node.Index{ node_datas[node].lhs, node_datas[node].rhs };
if (node_datas[node].lhs == 0) {
return blockExpr(gz, scope, ri, node, statements[0..0]);
} else if (node_datas[node].rhs == 0) {
return blockExpr(gz, scope, ri, node, statements[0..1]);
} else {
return blockExpr(gz, scope, ri, node, statements[0..2]);
}
},
.block, .block_semicolon => {
const statements = tree.extra_data[node_datas[node].lhs..node_datas[node].rhs];
return blockExpr(gz, scope, ri, node, statements);
},
.enum_literal => return simpleStrTok(gz, ri, main_tokens[node], node, .enum_literal),
.error_value => return simpleStrTok(gz, ri, node_datas[node].rhs, node, .error_value),
// TODO restore this when implementing https://github.com/ziglang/zig/issues/6025
// .anyframe_literal => return rvalue(gz, ri, .anyframe_type, node),
.anyframe_literal => {
const result = try gz.addUnNode(.anyframe_type, .void_type, node);
return rvalue(gz, ri, result, node);
},
.anyframe_type => {
const return_type = try typeExpr(gz, scope, node_datas[node].rhs);
const result = try gz.addUnNode(.anyframe_type, return_type, node);
return rvalue(gz, ri, result, node);
},
.@"catch" => {
const catch_token = main_tokens[node];
const payload_token: ?Ast.TokenIndex = if (token_tags[catch_token + 1] == .pipe)
catch_token + 2
else
null;
no_switch_on_err: {
const capture_token = payload_token orelse break :no_switch_on_err;
switch (node_tags[node_datas[node].rhs]) {
.@"switch", .switch_comma => {},
else => break :no_switch_on_err,
}
const switch_operand = node_datas[node_datas[node].rhs].lhs;
if (node_tags[switch_operand] != .identifier) break :no_switch_on_err;
if (!mem.eql(u8, tree.tokenSlice(capture_token), tree.tokenSlice(main_tokens[switch_operand]))) break :no_switch_on_err;
return switchExprErrUnion(gz, scope, ri.br(), node, .@"catch");
}
switch (ri.rl) {
.ref, .ref_coerced_ty => return orelseCatchExpr(
gz,
scope,
ri,
node,
node_datas[node].lhs,
.is_non_err_ptr,
.err_union_payload_unsafe_ptr,
.err_union_code_ptr,
node_datas[node].rhs,
payload_token,
),
else => return orelseCatchExpr(
gz,
scope,
ri,
node,
node_datas[node].lhs,
.is_non_err,
.err_union_payload_unsafe,
.err_union_code,
node_datas[node].rhs,
payload_token,
),
}
},
.@"orelse" => switch (ri.rl) {
.ref, .ref_coerced_ty => return orelseCatchExpr(
gz,
scope,
ri,
node,
node_datas[node].lhs,
.is_non_null_ptr,
.optional_payload_unsafe_ptr,
undefined,
node_datas[node].rhs,
null,
),
else => return orelseCatchExpr(
gz,
scope,
ri,
node,
node_datas[node].lhs,
.is_non_null,
.optional_payload_unsafe,
undefined,
node_datas[node].rhs,
null,
),
},
.ptr_type_aligned,
.ptr_type_sentinel,
.ptr_type,
.ptr_type_bit_range,
=> return ptrType(gz, scope, ri, node, tree.fullPtrType(node).?),
.container_decl,
.container_decl_trailing,
.container_decl_arg,
.container_decl_arg_trailing,
.container_decl_two,
.container_decl_two_trailing,
.tagged_union,
.tagged_union_trailing,
.tagged_union_enum_tag,
.tagged_union_enum_tag_trailing,
.tagged_union_two,
.tagged_union_two_trailing,
=> {
var buf: [2]Ast.Node.Index = undefined;
return containerDecl(gz, scope, ri, node, tree.fullContainerDecl(&buf, node).?);
},
.@"break" => return breakExpr(gz, scope, node),
.@"continue" => return continueExpr(gz, scope, node),
.grouped_expression => return expr(gz, scope, ri, node_datas[node].lhs),
.array_type => return arrayType(gz, scope, ri, node),
.array_type_sentinel => return arrayTypeSentinel(gz, scope, ri, node),
.char_literal => return charLiteral(gz, ri, node),
.error_set_decl => return errorSetDecl(gz, ri, node),
.array_access => return arrayAccess(gz, scope, ri, node),
.@"comptime" => return comptimeExprAst(gz, scope, ri, node),
.@"switch", .switch_comma => return switchExpr(gz, scope, ri.br(), node),
.@"nosuspend" => return nosuspendExpr(gz, scope, ri, node),
.@"suspend" => return suspendExpr(gz, scope, node),
.@"await" => return awaitExpr(gz, scope, ri, node),
.@"resume" => return resumeExpr(gz, scope, ri, node),
.@"try" => return tryExpr(gz, scope, ri, node, node_datas[node].lhs),
.array_init_one,
.array_init_one_comma,
.array_init_dot_two,
.array_init_dot_two_comma,
.array_init_dot,
.array_init_dot_comma,
.array_init,
.array_init_comma,
=> {
var buf: [2]Ast.Node.Index = undefined;
return arrayInitExpr(gz, scope, ri, node, tree.fullArrayInit(&buf, node).?);
},
.struct_init_one,
.struct_init_one_comma,
.struct_init_dot_two,
.struct_init_dot_two_comma,
.struct_init_dot,
.struct_init_dot_comma,
.struct_init,
.struct_init_comma,
=> {
var buf: [2]Ast.Node.Index = undefined;
return structInitExpr(gz, scope, ri, node, tree.fullStructInit(&buf, node).?);
},
.fn_proto_simple,
.fn_proto_multi,
.fn_proto_one,
.fn_proto,
=> {
var buf: [1]Ast.Node.Index = undefined;
return fnProtoExpr(gz, scope, ri, node, tree.fullFnProto(&buf, node).?);
},
}
}
fn nosuspendExpr(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const body_node = node_datas[node].lhs;
assert(body_node != 0);
if (gz.nosuspend_node != 0) {
try astgen.appendErrorNodeNotes(node, "redundant nosuspend block", .{}, &[_]u32{
try astgen.errNoteNode(gz.nosuspend_node, "other nosuspend block here", .{}),
});
}
gz.nosuspend_node = node;
defer gz.nosuspend_node = 0;
return expr(gz, scope, ri, body_node);
}
fn suspendExpr(
gz: *GenZir,
scope: *Scope,
node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const gpa = astgen.gpa;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const body_node = node_datas[node].lhs;
if (gz.nosuspend_node != 0) {
return astgen.failNodeNotes(node, "suspend inside nosuspend block", .{}, &[_]u32{
try astgen.errNoteNode(gz.nosuspend_node, "nosuspend block here", .{}),
});
}
if (gz.suspend_node != 0) {
return astgen.failNodeNotes(node, "cannot suspend inside suspend block", .{}, &[_]u32{
try astgen.errNoteNode(gz.suspend_node, "other suspend block here", .{}),
});
}
assert(body_node != 0);
const suspend_inst = try gz.makeBlockInst(.suspend_block, node);
try gz.instructions.append(gpa, suspend_inst);
var suspend_scope = gz.makeSubBlock(scope);
suspend_scope.suspend_node = node;
defer suspend_scope.unstack();
const body_result = try fullBodyExpr(&suspend_scope, &suspend_scope.base, .{ .rl = .none }, body_node);
if (!gz.refIsNoReturn(body_result)) {
_ = try suspend_scope.addBreak(.break_inline, suspend_inst, .void_value);
}
try suspend_scope.setBlockBody(suspend_inst);
return suspend_inst.toRef();
}
fn awaitExpr(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const rhs_node = node_datas[node].lhs;
if (gz.suspend_node != 0) {
return astgen.failNodeNotes(node, "cannot await inside suspend block", .{}, &[_]u32{
try astgen.errNoteNode(gz.suspend_node, "suspend block here", .{}),
});
}
const operand = try expr(gz, scope, .{ .rl = .ref }, rhs_node);
const result = if (gz.nosuspend_node != 0)
try gz.addExtendedPayload(.await_nosuspend, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = operand,
})
else
try gz.addUnNode(.@"await", operand, node);
return rvalue(gz, ri, result, node);
}
fn resumeExpr(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const rhs_node = node_datas[node].lhs;
const operand = try expr(gz, scope, .{ .rl = .ref }, rhs_node);
const result = try gz.addUnNode(.@"resume", operand, node);
return rvalue(gz, ri, result, node);
}
fn fnProtoExpr(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
fn_proto: Ast.full.FnProto,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const token_tags = tree.tokens.items(.tag);
if (fn_proto.name_token) |some| {
return astgen.failTok(some, "function type cannot have a name", .{});
}
const is_extern = blk: {
const maybe_extern_token = fn_proto.extern_export_inline_token orelse break :blk false;
break :blk token_tags[maybe_extern_token] == .keyword_extern;
};
assert(!is_extern);
var block_scope = gz.makeSubBlock(scope);
defer block_scope.unstack();
const block_inst = try gz.makeBlockInst(.block_inline, node);
var noalias_bits: u32 = 0;
const is_var_args = is_var_args: {
var param_type_i: usize = 0;
var it = fn_proto.iterate(tree);
while (it.next()) |param| : (param_type_i += 1) {
const is_comptime = if (param.comptime_noalias) |token| switch (token_tags[token]) {
.keyword_noalias => is_comptime: {
noalias_bits |= @as(u32, 1) << (std.math.cast(u5, param_type_i) orelse
return astgen.failTok(token, "this compiler implementation only supports 'noalias' on the first 32 parameters", .{}));
break :is_comptime false;
},
.keyword_comptime => true,
else => false,
} else false;
const is_anytype = if (param.anytype_ellipsis3) |token| blk: {
switch (token_tags[token]) {
.keyword_anytype => break :blk true,
.ellipsis3 => break :is_var_args true,
else => unreachable,
}
} else false;
const param_name = if (param.name_token) |name_token| blk: {
if (mem.eql(u8, "_", tree.tokenSlice(name_token)))
break :blk .empty;
break :blk try astgen.identAsString(name_token);
} else .empty;
if (is_anytype) {
const name_token = param.name_token orelse param.anytype_ellipsis3.?;
const tag: Zir.Inst.Tag = if (is_comptime)
.param_anytype_comptime
else
.param_anytype;
_ = try block_scope.addStrTok(tag, param_name, name_token);
} else {
const param_type_node = param.type_expr;
assert(param_type_node != 0);
var param_gz = block_scope.makeSubBlock(scope);
defer param_gz.unstack();
const param_type = try fullBodyExpr(&param_gz, scope, coerced_type_ri, param_type_node);
const param_inst_expected: Zir.Inst.Index = @enumFromInt(astgen.instructions.len + 1);
_ = try param_gz.addBreakWithSrcNode(.break_inline, param_inst_expected, param_type, param_type_node);
const main_tokens = tree.nodes.items(.main_token);
const name_token = param.name_token orelse main_tokens[param_type_node];
const tag: Zir.Inst.Tag = if (is_comptime) .param_comptime else .param;
const param_inst = try block_scope.addParam(&param_gz, tag, name_token, param_name, param.first_doc_comment);
assert(param_inst_expected == param_inst);
}
}
break :is_var_args false;
};
if (fn_proto.ast.align_expr != 0) {
return astgen.failNode(fn_proto.ast.align_expr, "function type cannot have an alignment", .{});
}
if (fn_proto.ast.addrspace_expr != 0) {
return astgen.failNode(fn_proto.ast.addrspace_expr, "function type cannot have an addrspace", .{});
}
if (fn_proto.ast.section_expr != 0) {
return astgen.failNode(fn_proto.ast.section_expr, "function type cannot have a linksection", .{});
}
const cc: Zir.Inst.Ref = if (fn_proto.ast.callconv_expr != 0)
try expr(
&block_scope,
scope,
.{ .rl = .{ .coerced_ty = .calling_convention_type } },
fn_proto.ast.callconv_expr,
)
else
Zir.Inst.Ref.none;
const maybe_bang = tree.firstToken(fn_proto.ast.return_type) - 1;
const is_inferred_error = token_tags[maybe_bang] == .bang;
if (is_inferred_error) {
return astgen.failTok(maybe_bang, "function type cannot have an inferred error set", .{});
}
const ret_ty = try expr(&block_scope, scope, coerced_type_ri, fn_proto.ast.return_type);
const result = try block_scope.addFunc(.{
.src_node = fn_proto.ast.proto_node,
.cc_ref = cc,
.cc_gz = null,
.align_ref = .none,
.align_gz = null,
.ret_ref = ret_ty,
.ret_gz = null,
.section_ref = .none,
.section_gz = null,
.addrspace_ref = .none,
.addrspace_gz = null,
.param_block = block_inst,
.body_gz = null,
.lib_name = .empty,
.is_var_args = is_var_args,
.is_inferred_error = false,
.is_test = false,
.is_extern = false,
.is_noinline = false,
.noalias_bits = noalias_bits,
});
_ = try block_scope.addBreak(.break_inline, block_inst, result);
try block_scope.setBlockBody(block_inst);
try gz.instructions.append(astgen.gpa, block_inst);
return rvalue(gz, ri, block_inst.toRef(), fn_proto.ast.proto_node);
}
fn arrayInitExpr(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
array_init: Ast.full.ArrayInit,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const node_tags = tree.nodes.items(.tag);
const main_tokens = tree.nodes.items(.main_token);
assert(array_init.ast.elements.len != 0); // Otherwise it would be struct init.
const array_ty: Zir.Inst.Ref, const elem_ty: Zir.Inst.Ref = inst: {
if (array_init.ast.type_expr == 0) break :inst .{ .none, .none };
infer: {
const array_type: Ast.full.ArrayType = tree.fullArrayType(array_init.ast.type_expr) orelse break :infer;
// This intentionally does not support `@"_"` syntax.
if (node_tags[array_type.ast.elem_count] == .identifier and
mem.eql(u8, tree.tokenSlice(main_tokens[array_type.ast.elem_count]), "_"))
{
const len_inst = try gz.addInt(array_init.ast.elements.len);
const elem_type = try typeExpr(gz, scope, array_type.ast.elem_type);
if (array_type.ast.sentinel == 0) {
const array_type_inst = try gz.addPlNode(.array_type, array_init.ast.type_expr, Zir.Inst.Bin{
.lhs = len_inst,
.rhs = elem_type,
});
break :inst .{ array_type_inst, elem_type };
} else {
const sentinel = try comptimeExpr(gz, scope, .{ .rl = .{ .ty = elem_type } }, array_type.ast.sentinel);
const array_type_inst = try gz.addPlNode(
.array_type_sentinel,
array_init.ast.type_expr,
Zir.Inst.ArrayTypeSentinel{
.len = len_inst,
.elem_type = elem_type,
.sentinel = sentinel,
},
);
break :inst .{ array_type_inst, elem_type };
}
}
}
const array_type_inst = try typeExpr(gz, scope, array_init.ast.type_expr);
_ = try gz.addPlNode(.validate_array_init_ty, node, Zir.Inst.ArrayInit{
.ty = array_type_inst,
.init_count = @intCast(array_init.ast.elements.len),
});
break :inst .{ array_type_inst, .none };
};
if (array_ty != .none) {
// Typed inits do not use RLS for language simplicity.
switch (ri.rl) {
.discard => {
if (elem_ty != .none) {
const elem_ri: ResultInfo = .{ .rl = .{ .ty = elem_ty } };
for (array_init.ast.elements) |elem_init| {
_ = try expr(gz, scope, elem_ri, elem_init);
}
} else {
for (array_init.ast.elements, 0..) |elem_init, i| {
const this_elem_ty = try gz.add(.{
.tag = .array_init_elem_type,
.data = .{ .bin = .{
.lhs = array_ty,
.rhs = @enumFromInt(i),
} },
});
_ = try expr(gz, scope, .{ .rl = .{ .ty = this_elem_ty } }, elem_init);
}
}
return .void_value;
},
.ref => return arrayInitExprTyped(gz, scope, node, array_init.ast.elements, array_ty, elem_ty, true),
else => {
const array_inst = try arrayInitExprTyped(gz, scope, node, array_init.ast.elements, array_ty, elem_ty, false);
return rvalue(gz, ri, array_inst, node);
},
}
}
switch (ri.rl) {
.none => return arrayInitExprAnon(gz, scope, node, array_init.ast.elements),
.discard => {
for (array_init.ast.elements) |elem_init| {
_ = try expr(gz, scope, .{ .rl = .discard }, elem_init);
}
return Zir.Inst.Ref.void_value;
},
.ref => {
const result = try arrayInitExprAnon(gz, scope, node, array_init.ast.elements);
return gz.addUnTok(.ref, result, tree.firstToken(node));
},
.ref_coerced_ty => |ptr_ty_inst| {
const dest_arr_ty_inst = try gz.addPlNode(.validate_array_init_ref_ty, node, Zir.Inst.ArrayInitRefTy{
.ptr_ty = ptr_ty_inst,
.elem_count = @intCast(array_init.ast.elements.len),
});
return arrayInitExprTyped(gz, scope, node, array_init.ast.elements, dest_arr_ty_inst, .none, true);
},
.ty, .coerced_ty => |result_ty_inst| {
_ = try gz.addPlNode(.validate_array_init_result_ty, node, Zir.Inst.ArrayInit{
.ty = result_ty_inst,
.init_count = @intCast(array_init.ast.elements.len),
});
return arrayInitExprTyped(gz, scope, node, array_init.ast.elements, result_ty_inst, .none, false);
},
.ptr => |ptr| {
try arrayInitExprPtr(gz, scope, node, array_init.ast.elements, ptr.inst);
return .void_value;
},
.inferred_ptr => {
// We can't get elem pointers of an untyped inferred alloc, so must perform a
// standard anonymous initialization followed by an rvalue store.
// See corresponding logic in structInitExpr.
const result = try arrayInitExprAnon(gz, scope, node, array_init.ast.elements);
return rvalue(gz, ri, result, node);
},
.destructure => |destructure| {
// Untyped init - destructure directly into result pointers
if (array_init.ast.elements.len != destructure.components.len) {
return astgen.failNodeNotes(node, "expected {} elements for destructure, found {}", .{
destructure.components.len,
array_init.ast.elements.len,
}, &.{
try astgen.errNoteNode(destructure.src_node, "result destructured here", .{}),
});
}
for (array_init.ast.elements, destructure.components) |elem_init, ds_comp| {
const elem_ri: ResultInfo = .{ .rl = switch (ds_comp) {
.typed_ptr => |ptr_rl| .{ .ptr = ptr_rl },
.inferred_ptr => |ptr_inst| .{ .inferred_ptr = ptr_inst },
.discard => .discard,
} };
_ = try expr(gz, scope, elem_ri, elem_init);
}
return .void_value;
},
}
}
/// An array initialization expression using an `array_init_anon` instruction.
fn arrayInitExprAnon(
gz: *GenZir,
scope: *Scope,
node: Ast.Node.Index,
elements: []const Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const payload_index = try addExtra(astgen, Zir.Inst.MultiOp{
.operands_len = @intCast(elements.len),
});
var extra_index = try reserveExtra(astgen, elements.len);
for (elements) |elem_init| {
const elem_ref = try expr(gz, scope, .{ .rl = .none }, elem_init);
astgen.extra.items[extra_index] = @intFromEnum(elem_ref);
extra_index += 1;
}
return try gz.addPlNodePayloadIndex(.array_init_anon, node, payload_index);
}
/// An array initialization expression using an `array_init` or `array_init_ref` instruction.
fn arrayInitExprTyped(
gz: *GenZir,
scope: *Scope,
node: Ast.Node.Index,
elements: []const Ast.Node.Index,
ty_inst: Zir.Inst.Ref,
maybe_elem_ty_inst: Zir.Inst.Ref,
is_ref: bool,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const len = elements.len + 1; // +1 for type
const payload_index = try addExtra(astgen, Zir.Inst.MultiOp{
.operands_len = @intCast(len),
});
var extra_index = try reserveExtra(astgen, len);
astgen.extra.items[extra_index] = @intFromEnum(ty_inst);
extra_index += 1;
if (maybe_elem_ty_inst != .none) {
const elem_ri: ResultInfo = .{ .rl = .{ .coerced_ty = maybe_elem_ty_inst } };
for (elements) |elem_init| {
const elem_inst = try expr(gz, scope, elem_ri, elem_init);
astgen.extra.items[extra_index] = @intFromEnum(elem_inst);
extra_index += 1;
}
} else {
for (elements, 0..) |elem_init, i| {
const ri: ResultInfo = .{ .rl = .{ .coerced_ty = try gz.add(.{
.tag = .array_init_elem_type,
.data = .{ .bin = .{
.lhs = ty_inst,
.rhs = @enumFromInt(i),
} },
}) } };
const elem_inst = try expr(gz, scope, ri, elem_init);
astgen.extra.items[extra_index] = @intFromEnum(elem_inst);
extra_index += 1;
}
}
const tag: Zir.Inst.Tag = if (is_ref) .array_init_ref else .array_init;
return try gz.addPlNodePayloadIndex(tag, node, payload_index);
}
/// An array initialization expression using element pointers.
fn arrayInitExprPtr(
gz: *GenZir,
scope: *Scope,
node: Ast.Node.Index,
elements: []const Ast.Node.Index,
ptr_inst: Zir.Inst.Ref,
) InnerError!void {
const astgen = gz.astgen;
const array_ptr_inst = try gz.addUnNode(.opt_eu_base_ptr_init, ptr_inst, node);
const payload_index = try addExtra(astgen, Zir.Inst.Block{
.body_len = @intCast(elements.len),
});
var extra_index = try reserveExtra(astgen, elements.len);
for (elements, 0..) |elem_init, i| {
const elem_ptr_inst = try gz.addPlNode(.array_init_elem_ptr, elem_init, Zir.Inst.ElemPtrImm{
.ptr = array_ptr_inst,
.index = @intCast(i),
});
astgen.extra.items[extra_index] = @intFromEnum(elem_ptr_inst.toIndex().?);
extra_index += 1;
_ = try expr(gz, scope, .{ .rl = .{ .ptr = .{ .inst = elem_ptr_inst } } }, elem_init);
}
_ = try gz.addPlNodePayloadIndex(.validate_ptr_array_init, node, payload_index);
}
fn structInitExpr(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
struct_init: Ast.full.StructInit,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
if (struct_init.ast.type_expr == 0) {
if (struct_init.ast.fields.len == 0) {
// Anonymous init with no fields.
switch (ri.rl) {
.discard => return .void_value,
.ref_coerced_ty => |ptr_ty_inst| return gz.addUnNode(.struct_init_empty_ref_result, ptr_ty_inst, node),
.ty, .coerced_ty => |ty_inst| return gz.addUnNode(.struct_init_empty_result, ty_inst, node),
.ptr => {
// TODO: should we modify this to use RLS for the field stores here?
const ty_inst = (try ri.rl.resultType(gz, node)).?;
const val = try gz.addUnNode(.struct_init_empty_result, ty_inst, node);
return rvalue(gz, ri, val, node);
},
.none, .ref, .inferred_ptr => {
return rvalue(gz, ri, .empty_struct, node);
},
.destructure => |destructure| {
return astgen.failNodeNotes(node, "empty initializer cannot be destructured", .{}, &.{
try astgen.errNoteNode(destructure.src_node, "result destructured here", .{}),
});
},
}
}
} else array: {
const node_tags = tree.nodes.items(.tag);
const main_tokens = tree.nodes.items(.main_token);
const array_type: Ast.full.ArrayType = tree.fullArrayType(struct_init.ast.type_expr) orelse {
if (struct_init.ast.fields.len == 0) {
const ty_inst = try typeExpr(gz, scope, struct_init.ast.type_expr);
const result = try gz.addUnNode(.struct_init_empty, ty_inst, node);
return rvalue(gz, ri, result, node);
}
break :array;
};
const is_inferred_array_len = node_tags[array_type.ast.elem_count] == .identifier and
// This intentionally does not support `@"_"` syntax.
mem.eql(u8, tree.tokenSlice(main_tokens[array_type.ast.elem_count]), "_");
if (struct_init.ast.fields.len == 0) {
if (is_inferred_array_len) {
const elem_type = try typeExpr(gz, scope, array_type.ast.elem_type);
const array_type_inst = if (array_type.ast.sentinel == 0) blk: {
break :blk try gz.addPlNode(.array_type, struct_init.ast.type_expr, Zir.Inst.Bin{
.lhs = .zero_usize,
.rhs = elem_type,
});
} else blk: {
const sentinel = try comptimeExpr(gz, scope, .{ .rl = .{ .ty = elem_type } }, array_type.ast.sentinel);
break :blk try gz.addPlNode(
.array_type_sentinel,
struct_init.ast.type_expr,
Zir.Inst.ArrayTypeSentinel{
.len = .zero_usize,
.elem_type = elem_type,
.sentinel = sentinel,
},
);
};
const result = try gz.addUnNode(.struct_init_empty, array_type_inst, node);
return rvalue(gz, ri, result, node);
}
const ty_inst = try typeExpr(gz, scope, struct_init.ast.type_expr);
const result = try gz.addUnNode(.struct_init_empty, ty_inst, node);
return rvalue(gz, ri, result, node);
} else {
return astgen.failNode(
struct_init.ast.type_expr,
"initializing array with struct syntax",
.{},
);
}
}
{
var sfba = std.heap.stackFallback(256, astgen.arena);
const sfba_allocator = sfba.get();
var duplicate_names = std.AutoArrayHashMap(Zir.NullTerminatedString, ArrayListUnmanaged(Ast.TokenIndex)).init(sfba_allocator);
try duplicate_names.ensureTotalCapacity(@intCast(struct_init.ast.fields.len));
// When there aren't errors, use this to avoid a second iteration.
var any_duplicate = false;
for (struct_init.ast.fields) |field| {
const name_token = tree.firstToken(field) - 2;
const name_index = try astgen.identAsString(name_token);
const gop = try duplicate_names.getOrPut(name_index);
if (gop.found_existing) {
try gop.value_ptr.append(sfba_allocator, name_token);
any_duplicate = true;
} else {
gop.value_ptr.* = .{};
try gop.value_ptr.append(sfba_allocator, name_token);
}
}
if (any_duplicate) {
var it = duplicate_names.iterator();
while (it.next()) |entry| {
const record = entry.value_ptr.*;
if (record.items.len > 1) {
var error_notes = std.ArrayList(u32).init(astgen.arena);
for (record.items[1..]) |duplicate| {
try error_notes.append(try astgen.errNoteTok(duplicate, "duplicate name here", .{}));
}
try error_notes.append(try astgen.errNoteNode(node, "struct declared here", .{}));
try astgen.appendErrorTokNotes(
record.items[0],
"duplicate struct field name",
.{},
error_notes.items,
);
}
}
return error.AnalysisFail;
}
}
if (struct_init.ast.type_expr != 0) {
// Typed inits do not use RLS for language simplicity.
const ty_inst = try typeExpr(gz, scope, struct_init.ast.type_expr);
_ = try gz.addUnNode(.validate_struct_init_ty, ty_inst, node);
switch (ri.rl) {
.ref => return structInitExprTyped(gz, scope, node, struct_init, ty_inst, true),
else => {
const struct_inst = try structInitExprTyped(gz, scope, node, struct_init, ty_inst, false);
return rvalue(gz, ri, struct_inst, node);
},
}
}
switch (ri.rl) {
.none => return structInitExprAnon(gz, scope, node, struct_init),
.discard => {
// Even if discarding we must perform side-effects.
for (struct_init.ast.fields) |field_init| {
_ = try expr(gz, scope, .{ .rl = .discard }, field_init);
}
return .void_value;
},
.ref => {
const result = try structInitExprAnon(gz, scope, node, struct_init);
return gz.addUnTok(.ref, result, tree.firstToken(node));
},
.ref_coerced_ty => |ptr_ty_inst| {
const result_ty_inst = try gz.addUnNode(.elem_type, ptr_ty_inst, node);
_ = try gz.addUnNode(.validate_struct_init_result_ty, result_ty_inst, node);
return structInitExprTyped(gz, scope, node, struct_init, result_ty_inst, true);
},
.ty, .coerced_ty => |result_ty_inst| {
_ = try gz.addUnNode(.validate_struct_init_result_ty, result_ty_inst, node);
return structInitExprTyped(gz, scope, node, struct_init, result_ty_inst, false);
},
.ptr => |ptr| {
try structInitExprPtr(gz, scope, node, struct_init, ptr.inst);
return .void_value;
},
.inferred_ptr => {
// We can't get field pointers of an untyped inferred alloc, so must perform a
// standard anonymous initialization followed by an rvalue store.
// See corresponding logic in arrayInitExpr.
const struct_inst = try structInitExprAnon(gz, scope, node, struct_init);
return rvalue(gz, ri, struct_inst, node);
},
.destructure => |destructure| {
// This is an untyped init, so is an actual struct, which does
// not support destructuring.
return astgen.failNodeNotes(node, "struct value cannot be destructured", .{}, &.{
try astgen.errNoteNode(destructure.src_node, "result destructured here", .{}),
});
},
}
}
/// A struct initialization expression using a `struct_init_anon` instruction.
fn structInitExprAnon(
gz: *GenZir,
scope: *Scope,
node: Ast.Node.Index,
struct_init: Ast.full.StructInit,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const payload_index = try addExtra(astgen, Zir.Inst.StructInitAnon{
.fields_len = @intCast(struct_init.ast.fields.len),
});
const field_size = @typeInfo(Zir.Inst.StructInitAnon.Item).Struct.fields.len;
var extra_index: usize = try reserveExtra(astgen, struct_init.ast.fields.len * field_size);
for (struct_init.ast.fields) |field_init| {
const name_token = tree.firstToken(field_init) - 2;
const str_index = try astgen.identAsString(name_token);
setExtra(astgen, extra_index, Zir.Inst.StructInitAnon.Item{
.field_name = str_index,
.init = try expr(gz, scope, .{ .rl = .none }, field_init),
});
extra_index += field_size;
}
return gz.addPlNodePayloadIndex(.struct_init_anon, node, payload_index);
}
/// A struct initialization expression using a `struct_init` or `struct_init_ref` instruction.
fn structInitExprTyped(
gz: *GenZir,
scope: *Scope,
node: Ast.Node.Index,
struct_init: Ast.full.StructInit,
ty_inst: Zir.Inst.Ref,
is_ref: bool,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const payload_index = try addExtra(astgen, Zir.Inst.StructInit{
.fields_len = @intCast(struct_init.ast.fields.len),
});
const field_size = @typeInfo(Zir.Inst.StructInit.Item).Struct.fields.len;
var extra_index: usize = try reserveExtra(astgen, struct_init.ast.fields.len * field_size);
for (struct_init.ast.fields) |field_init| {
const name_token = tree.firstToken(field_init) - 2;
const str_index = try astgen.identAsString(name_token);
const field_ty_inst = try gz.addPlNode(.struct_init_field_type, field_init, Zir.Inst.FieldType{
.container_type = ty_inst,
.name_start = str_index,
});
setExtra(astgen, extra_index, Zir.Inst.StructInit.Item{
.field_type = field_ty_inst.toIndex().?,
.init = try expr(gz, scope, .{ .rl = .{ .coerced_ty = field_ty_inst } }, field_init),
});
extra_index += field_size;
}
const tag: Zir.Inst.Tag = if (is_ref) .struct_init_ref else .struct_init;
return gz.addPlNodePayloadIndex(tag, node, payload_index);
}
/// A struct initialization expression using field pointers.
fn structInitExprPtr(
gz: *GenZir,
scope: *Scope,
node: Ast.Node.Index,
struct_init: Ast.full.StructInit,
ptr_inst: Zir.Inst.Ref,
) InnerError!void {
const astgen = gz.astgen;
const tree = astgen.tree;
const struct_ptr_inst = try gz.addUnNode(.opt_eu_base_ptr_init, ptr_inst, node);
const payload_index = try addExtra(astgen, Zir.Inst.Block{
.body_len = @intCast(struct_init.ast.fields.len),
});
var extra_index = try reserveExtra(astgen, struct_init.ast.fields.len);
for (struct_init.ast.fields) |field_init| {
const name_token = tree.firstToken(field_init) - 2;
const str_index = try astgen.identAsString(name_token);
const field_ptr = try gz.addPlNode(.struct_init_field_ptr, field_init, Zir.Inst.Field{
.lhs = struct_ptr_inst,
.field_name_start = str_index,
});
astgen.extra.items[extra_index] = @intFromEnum(field_ptr.toIndex().?);
extra_index += 1;
_ = try expr(gz, scope, .{ .rl = .{ .ptr = .{ .inst = field_ptr } } }, field_init);
}
_ = try gz.addPlNodePayloadIndex(.validate_ptr_struct_init, node, payload_index);
}
/// This explicitly calls expr in a comptime scope by wrapping it in a `block_comptime` if
/// necessary. It should be used whenever we need to force compile-time evaluation of something,
/// such as a type.
/// The function corresponding to `comptime` expression syntax is `comptimeExprAst`.
fn comptimeExpr(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
if (gz.is_comptime) {
// No need to change anything!
return expr(gz, scope, ri, node);
}
// There's an optimization here: if the body will be evaluated at comptime regardless, there's
// no need to wrap it in a block. This is hard to determine in general, but we can identify a
// common subset of trivially comptime expressions to take down the size of the ZIR a bit.
const tree = gz.astgen.tree;
const main_tokens = tree.nodes.items(.main_token);
const node_tags = tree.nodes.items(.tag);
switch (node_tags[node]) {
// Any identifier in `primitive_instrs` is trivially comptime. In particular, this includes
// some common types, so we can elide `block_comptime` for a few common type annotations.
.identifier => {
const ident_token = main_tokens[node];
const ident_name_raw = tree.tokenSlice(ident_token);
if (primitive_instrs.get(ident_name_raw)) |zir_const_ref| {
// No need to worry about result location here, we're not creating a comptime block!
return rvalue(gz, ri, zir_const_ref, node);
}
},
// We can also avoid the block for a few trivial AST tags which are always comptime-known.
.number_literal, .string_literal, .multiline_string_literal, .enum_literal, .error_value => {
// No need to worry about result location here, we're not creating a comptime block!
return expr(gz, scope, ri, node);
},
// Lastly, for labelled blocks, avoid emitting a labelled block directly inside this
// comptime block, because that would be silly! Note that we don't bother doing this for
// unlabelled blocks, since they don't generate blocks at comptime anyway (see `blockExpr`).
.block_two, .block_two_semicolon, .block, .block_semicolon => {
const token_tags = tree.tokens.items(.tag);
const lbrace = main_tokens[node];
// Careful! We can't pass in the real result location here, since it may
// refer to runtime memory. A runtime-to-comptime boundary has to remove
// result location information, compute the result, and copy it to the true
// result location at runtime. We do this below as well.
const ty_only_ri: ResultInfo = .{
.ctx = ri.ctx,
.rl = if (try ri.rl.resultType(gz, node)) |res_ty|
.{ .coerced_ty = res_ty }
else
.none,
};
if (token_tags[lbrace - 1] == .colon and
token_tags[lbrace - 2] == .identifier)
{
const node_datas = tree.nodes.items(.data);
switch (node_tags[node]) {
.block_two, .block_two_semicolon => {
const stmts: [2]Ast.Node.Index = .{ node_datas[node].lhs, node_datas[node].rhs };
const stmt_slice = if (stmts[0] == 0)
stmts[0..0]
else if (stmts[1] == 0)
stmts[0..1]
else
stmts[0..2];
const block_ref = try labeledBlockExpr(gz, scope, ty_only_ri, node, stmt_slice, true);
return rvalue(gz, ri, block_ref, node);
},
.block, .block_semicolon => {
const stmts = tree.extra_data[node_datas[node].lhs..node_datas[node].rhs];
// Replace result location and copy back later - see above.
const block_ref = try labeledBlockExpr(gz, scope, ty_only_ri, node, stmts, true);
return rvalue(gz, ri, block_ref, node);
},
else => unreachable,
}
}
},
// In other cases, we don't optimize anything - we need a wrapper comptime block.
else => {},
}
var block_scope = gz.makeSubBlock(scope);
block_scope.is_comptime = true;
defer block_scope.unstack();
const block_inst = try gz.makeBlockInst(.block_comptime, node);
// Replace result location and copy back later - see above.
const ty_only_ri: ResultInfo = .{
.ctx = ri.ctx,
.rl = if (try ri.rl.resultType(gz, node)) |res_ty|
.{ .coerced_ty = res_ty }
else
.none,
};
const block_result = try fullBodyExpr(&block_scope, scope, ty_only_ri, node);
if (!gz.refIsNoReturn(block_result)) {
_ = try block_scope.addBreak(.@"break", block_inst, block_result);
}
try block_scope.setBlockBody(block_inst);
try gz.instructions.append(gz.astgen.gpa, block_inst);
return rvalue(gz, ri, block_inst.toRef(), node);
}
/// This one is for an actual `comptime` syntax, and will emit a compile error if
/// the scope is already known to be comptime-evaluated.
/// See `comptimeExpr` for the helper function for calling expr in a comptime scope.
fn comptimeExprAst(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
if (gz.is_comptime) {
return astgen.failNode(node, "redundant comptime keyword in already comptime scope", .{});
}
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const body_node = node_datas[node].lhs;
return comptimeExpr(gz, scope, ri, body_node);
}
/// Restore the error return trace index. Performs the restore only if the result is a non-error or
/// if the result location is a non-error-handling expression.
fn restoreErrRetIndex(
gz: *GenZir,
bt: GenZir.BranchTarget,
ri: ResultInfo,
node: Ast.Node.Index,
result: Zir.Inst.Ref,
) !void {
const op = switch (nodeMayEvalToError(gz.astgen.tree, node)) {
.always => return, // never restore/pop
.never => .none, // always restore/pop
.maybe => switch (ri.ctx) {
.error_handling_expr, .@"return", .fn_arg, .const_init => switch (ri.rl) {
.ptr => |ptr_res| try gz.addUnNode(.load, ptr_res.inst, node),
.inferred_ptr => blk: {
// This is a terrible workaround for Sema's inability to load from a .alloc_inferred ptr
// before its type has been resolved. There is no valid operand to use here, so error
// traces will be popped prematurely.
// TODO: Update this to do a proper load from the rl_ptr, once Sema can support it.
break :blk .none;
},
.destructure => return, // value must be a tuple or array, so never restore/pop
else => result,
},
else => .none, // always restore/pop
},
};
_ = try gz.addRestoreErrRetIndex(bt, .{ .if_non_error = op }, node);
}
fn breakExpr(parent_gz: *GenZir, parent_scope: *Scope, node: Ast.Node.Index) InnerError!Zir.Inst.Ref {
const astgen = parent_gz.astgen;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const break_label = node_datas[node].lhs;
const rhs = node_datas[node].rhs;
// Look for the label in the scope.
var scope = parent_scope;
while (true) {
switch (scope.tag) {
.gen_zir => {
const block_gz = scope.cast(GenZir).?;
if (block_gz.cur_defer_node != 0) {
// We are breaking out of a `defer` block.
return astgen.failNodeNotes(node, "cannot break out of defer expression", .{}, &.{
try astgen.errNoteNode(
block_gz.cur_defer_node,
"defer expression here",
.{},
),
});
}
const block_inst = blk: {
if (break_label != 0) {
if (block_gz.label) |*label| {
if (try astgen.tokenIdentEql(label.token, break_label)) {
label.used = true;
break :blk label.block_inst;
}
}
} else if (block_gz.break_block.unwrap()) |i| {
break :blk i;
}
// If not the target, start over with the parent
scope = block_gz.parent;
continue;
};
// If we made it here, this block is the target of the break expr
const break_tag: Zir.Inst.Tag = if (block_gz.is_inline)
.break_inline
else
.@"break";
if (rhs == 0) {
_ = try rvalue(parent_gz, block_gz.break_result_info, .void_value, node);
try genDefers(parent_gz, scope, parent_scope, .normal_only);
// As our last action before the break, "pop" the error trace if needed
if (!block_gz.is_comptime)
_ = try parent_gz.addRestoreErrRetIndex(.{ .block = block_inst }, .always, node);
_ = try parent_gz.addBreak(break_tag, block_inst, .void_value);
return Zir.Inst.Ref.unreachable_value;
}
const operand = try reachableExpr(parent_gz, parent_scope, block_gz.break_result_info, rhs, node);
try genDefers(parent_gz, scope, parent_scope, .normal_only);
// As our last action before the break, "pop" the error trace if needed
if (!block_gz.is_comptime)
try restoreErrRetIndex(parent_gz, .{ .block = block_inst }, block_gz.break_result_info, rhs, operand);
switch (block_gz.break_result_info.rl) {
.ptr => {
// In this case we don't have any mechanism to intercept it;
// we assume the result location is written, and we break with void.
_ = try parent_gz.addBreak(break_tag, block_inst, .void_value);
},
.discard => {
_ = try parent_gz.addBreak(break_tag, block_inst, .void_value);
},
else => {
_ = try parent_gz.addBreakWithSrcNode(break_tag, block_inst, operand, rhs);
},
}
return Zir.Inst.Ref.unreachable_value;
},
.local_val => scope = scope.cast(Scope.LocalVal).?.parent,
.local_ptr => scope = scope.cast(Scope.LocalPtr).?.parent,
.namespace => break,
.defer_normal, .defer_error => scope = scope.cast(Scope.Defer).?.parent,
.top => unreachable,
}
}
if (break_label != 0) {
const label_name = try astgen.identifierTokenString(break_label);
return astgen.failTok(break_label, "label not found: '{s}'", .{label_name});
} else {
return astgen.failNode(node, "break expression outside loop", .{});
}
}
fn continueExpr(parent_gz: *GenZir, parent_scope: *Scope, node: Ast.Node.Index) InnerError!Zir.Inst.Ref {
const astgen = parent_gz.astgen;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const break_label = node_datas[node].lhs;
// Look for the label in the scope.
var scope = parent_scope;
while (true) {
switch (scope.tag) {
.gen_zir => {
const gen_zir = scope.cast(GenZir).?;
if (gen_zir.cur_defer_node != 0) {
return astgen.failNodeNotes(node, "cannot continue out of defer expression", .{}, &.{
try astgen.errNoteNode(
gen_zir.cur_defer_node,
"defer expression here",
.{},
),
});
}
const continue_block = gen_zir.continue_block.unwrap() orelse {
scope = gen_zir.parent;
continue;
};
if (break_label != 0) blk: {
if (gen_zir.label) |*label| {
if (try astgen.tokenIdentEql(label.token, break_label)) {
label.used = true;
break :blk;
}
}
// found continue but either it has a different label, or no label
scope = gen_zir.parent;
continue;
}
const break_tag: Zir.Inst.Tag = if (gen_zir.is_inline)
.break_inline
else
.@"break";
if (break_tag == .break_inline) {
_ = try parent_gz.addUnNode(.check_comptime_control_flow, continue_block.toRef(), node);
}
// As our last action before the continue, "pop" the error trace if needed
if (!gen_zir.is_comptime)
_ = try parent_gz.addRestoreErrRetIndex(.{ .block = continue_block }, .always, node);
_ = try parent_gz.addBreak(break_tag, continue_block, .void_value);
return Zir.Inst.Ref.unreachable_value;
},
.local_val => scope = scope.cast(Scope.LocalVal).?.parent,
.local_ptr => scope = scope.cast(Scope.LocalPtr).?.parent,
.defer_normal => {
const defer_scope = scope.cast(Scope.Defer).?;
scope = defer_scope.parent;
try parent_gz.addDefer(defer_scope.index, defer_scope.len);
},
.defer_error => scope = scope.cast(Scope.Defer).?.parent,
.namespace => break,
.top => unreachable,
}
}
if (break_label != 0) {
const label_name = try astgen.identifierTokenString(break_label);
return astgen.failTok(break_label, "label not found: '{s}'", .{label_name});
} else {
return astgen.failNode(node, "continue expression outside loop", .{});
}
}
/// Similar to `expr`, but intended for use when `gz` corresponds to a body
/// which will contain only this node's code. Differs from `expr` in that if the
/// root expression is an unlabeled block, does not emit an actual block.
/// Instead, the block contents are emitted directly into `gz`.
fn fullBodyExpr(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const tree = gz.astgen.tree;
const node_tags = tree.nodes.items(.tag);
const node_datas = tree.nodes.items(.data);
const main_tokens = tree.nodes.items(.main_token);
const token_tags = tree.tokens.items(.tag);
var stmt_buf: [2]Ast.Node.Index = undefined;
const statements: []const Ast.Node.Index = switch (node_tags[node]) {
else => return expr(gz, scope, ri, node),
.block_two, .block_two_semicolon => if (node_datas[node].lhs == 0) s: {
break :s &.{};
} else if (node_datas[node].rhs == 0) s: {
stmt_buf[0] = node_datas[node].lhs;
break :s stmt_buf[0..1];
} else s: {
stmt_buf[0] = node_datas[node].lhs;
stmt_buf[1] = node_datas[node].rhs;
break :s stmt_buf[0..2];
},
.block, .block_semicolon => tree.extra_data[node_datas[node].lhs..node_datas[node].rhs],
};
const lbrace = main_tokens[node];
if (token_tags[lbrace - 1] == .colon and
token_tags[lbrace - 2] == .identifier)
{
// Labeled blocks are tricky - forwarding result location information properly is non-trivial,
// plus if this block is exited with a `break_inline` we aren't allowed multiple breaks. This
// case is rare, so just treat it as a normal expression and create a nested block.
return expr(gz, scope, ri, node);
}
var sub_gz = gz.makeSubBlock(scope);
try blockExprStmts(&sub_gz, &sub_gz.base, statements);
return rvalue(gz, ri, .void_value, node);
}
fn blockExpr(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
block_node: Ast.Node.Index,
statements: []const Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const main_tokens = tree.nodes.items(.main_token);
const token_tags = tree.tokens.items(.tag);
const lbrace = main_tokens[block_node];
if (token_tags[lbrace - 1] == .colon and
token_tags[lbrace - 2] == .identifier)
{
return labeledBlockExpr(gz, scope, ri, block_node, statements, false);
}
if (!gz.is_comptime) {
// Since this block is unlabeled, its control flow is effectively linear and we
// can *almost* get away with inlining the block here. However, we actually need
// to preserve the .block for Sema, to properly pop the error return trace.
const block_tag: Zir.Inst.Tag = .block;
const block_inst = try gz.makeBlockInst(block_tag, block_node);
try gz.instructions.append(astgen.gpa, block_inst);
var block_scope = gz.makeSubBlock(scope);
defer block_scope.unstack();
try blockExprStmts(&block_scope, &block_scope.base, statements);
if (!block_scope.endsWithNoReturn()) {
// As our last action before the break, "pop" the error trace if needed
_ = try gz.addRestoreErrRetIndex(.{ .block = block_inst }, .always, block_node);
_ = try block_scope.addBreak(.@"break", block_inst, .void_value);
}
try block_scope.setBlockBody(block_inst);
} else {
var sub_gz = gz.makeSubBlock(scope);
try blockExprStmts(&sub_gz, &sub_gz.base, statements);
}
return rvalue(gz, ri, .void_value, block_node);
}
fn checkLabelRedefinition(astgen: *AstGen, parent_scope: *Scope, label: Ast.TokenIndex) !void {
// Look for the label in the scope.
var scope = parent_scope;
while (true) {
switch (scope.tag) {
.gen_zir => {
const gen_zir = scope.cast(GenZir).?;
if (gen_zir.label) |prev_label| {
if (try astgen.tokenIdentEql(label, prev_label.token)) {
const label_name = try astgen.identifierTokenString(label);
return astgen.failTokNotes(label, "redefinition of label '{s}'", .{
label_name,
}, &[_]u32{
try astgen.errNoteTok(
prev_label.token,
"previous definition here",
.{},
),
});
}
}
scope = gen_zir.parent;
},
.local_val => scope = scope.cast(Scope.LocalVal).?.parent,
.local_ptr => scope = scope.cast(Scope.LocalPtr).?.parent,
.defer_normal, .defer_error => scope = scope.cast(Scope.Defer).?.parent,
.namespace => break,
.top => unreachable,
}
}
}
fn labeledBlockExpr(
gz: *GenZir,
parent_scope: *Scope,
ri: ResultInfo,
block_node: Ast.Node.Index,
statements: []const Ast.Node.Index,
force_comptime: bool,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const main_tokens = tree.nodes.items(.main_token);
const token_tags = tree.tokens.items(.tag);
const lbrace = main_tokens[block_node];
const label_token = lbrace - 2;
assert(token_tags[label_token] == .identifier);
try astgen.checkLabelRedefinition(parent_scope, label_token);
const need_rl = astgen.nodes_need_rl.contains(block_node);
const block_ri: ResultInfo = if (need_rl) ri else .{
.rl = switch (ri.rl) {
.ptr => .{ .ty = (try ri.rl.resultType(gz, block_node)).? },
.inferred_ptr => .none,
else => ri.rl,
},
.ctx = ri.ctx,
};
// We need to call `rvalue` to write through to the pointer only if we had a
// result pointer and aren't forwarding it.
const LocTag = @typeInfo(ResultInfo.Loc).Union.tag_type.?;
const need_result_rvalue = @as(LocTag, block_ri.rl) != @as(LocTag, ri.rl);
// Reserve the Block ZIR instruction index so that we can put it into the GenZir struct
// so that break statements can reference it.
const block_tag: Zir.Inst.Tag = if (force_comptime) .block_comptime else .block;
const block_inst = try gz.makeBlockInst(block_tag, block_node);
try gz.instructions.append(astgen.gpa, block_inst);
var block_scope = gz.makeSubBlock(parent_scope);
block_scope.label = GenZir.Label{
.token = label_token,
.block_inst = block_inst,
};
block_scope.setBreakResultInfo(block_ri);
if (force_comptime) block_scope.is_comptime = true;
defer block_scope.unstack();
try blockExprStmts(&block_scope, &block_scope.base, statements);
if (!block_scope.endsWithNoReturn()) {
// As our last action before the return, "pop" the error trace if needed
_ = try gz.addRestoreErrRetIndex(.{ .block = block_inst }, .always, block_node);
_ = try block_scope.addBreak(.@"break", block_inst, .void_value);
}
if (!block_scope.label.?.used) {
try astgen.appendErrorTok(label_token, "unused block label", .{});
}
try block_scope.setBlockBody(block_inst);
if (need_result_rvalue) {
return rvalue(gz, ri, block_inst.toRef(), block_node);
} else {
return block_inst.toRef();
}
}
fn blockExprStmts(gz: *GenZir, parent_scope: *Scope, statements: []const Ast.Node.Index) !void {
const astgen = gz.astgen;
const tree = astgen.tree;
const node_tags = tree.nodes.items(.tag);
const node_data = tree.nodes.items(.data);
if (statements.len == 0) return;
var block_arena = std.heap.ArenaAllocator.init(gz.astgen.gpa);
defer block_arena.deinit();
const block_arena_allocator = block_arena.allocator();
var noreturn_src_node: Ast.Node.Index = 0;
var scope = parent_scope;
for (statements) |statement| {
if (noreturn_src_node != 0) {
try astgen.appendErrorNodeNotes(
statement,
"unreachable code",
.{},
&[_]u32{
try astgen.errNoteNode(
noreturn_src_node,
"control flow is diverted here",
.{},
),
},
);
}
var inner_node = statement;
while (true) {
switch (node_tags[inner_node]) {
// zig fmt: off
.global_var_decl,
.local_var_decl,
.simple_var_decl,
.aligned_var_decl, => scope = try varDecl(gz, scope, statement, block_arena_allocator, tree.fullVarDecl(statement).?),
.assign_destructure => scope = try assignDestructureMaybeDecls(gz, scope, statement, block_arena_allocator),
.@"defer" => scope = try deferStmt(gz, scope, statement, block_arena_allocator, .defer_normal),
.@"errdefer" => scope = try deferStmt(gz, scope, statement, block_arena_allocator, .defer_error),
.assign => try assign(gz, scope, statement),
.assign_shl => try assignShift(gz, scope, statement, .shl),
.assign_shr => try assignShift(gz, scope, statement, .shr),
.assign_bit_and => try assignOp(gz, scope, statement, .bit_and),
.assign_bit_or => try assignOp(gz, scope, statement, .bit_or),
.assign_bit_xor => try assignOp(gz, scope, statement, .xor),
.assign_div => try assignOp(gz, scope, statement, .div),
.assign_sub => try assignOp(gz, scope, statement, .sub),
.assign_sub_wrap => try assignOp(gz, scope, statement, .subwrap),
.assign_mod => try assignOp(gz, scope, statement, .mod_rem),
.assign_add => try assignOp(gz, scope, statement, .add),
.assign_add_wrap => try assignOp(gz, scope, statement, .addwrap),
.assign_mul => try assignOp(gz, scope, statement, .mul),
.assign_mul_wrap => try assignOp(gz, scope, statement, .mulwrap),
.grouped_expression => {
inner_node = node_data[statement].lhs;
continue;
},
.while_simple,
.while_cont,
.@"while", => _ = try whileExpr(gz, scope, .{ .rl = .none }, inner_node, tree.fullWhile(inner_node).?, true),
.for_simple,
.@"for", => _ = try forExpr(gz, scope, .{ .rl = .none }, inner_node, tree.fullFor(inner_node).?, true),
else => noreturn_src_node = try unusedResultExpr(gz, scope, inner_node),
// zig fmt: on
}
break;
}
}
if (noreturn_src_node == 0) {
try genDefers(gz, parent_scope, scope, .normal_only);
}
try checkUsed(gz, parent_scope, scope);
}
/// Returns AST source node of the thing that is noreturn if the statement is
/// definitely `noreturn`. Otherwise returns 0.
fn unusedResultExpr(gz: *GenZir, scope: *Scope, statement: Ast.Node.Index) InnerError!Ast.Node.Index {
try emitDbgNode(gz, statement);
// We need to emit an error if the result is not `noreturn` or `void`, but
// we want to avoid adding the ZIR instruction if possible for performance.
const maybe_unused_result = try expr(gz, scope, .{ .rl = .none }, statement);
return addEnsureResult(gz, maybe_unused_result, statement);
}
fn addEnsureResult(gz: *GenZir, maybe_unused_result: Zir.Inst.Ref, statement: Ast.Node.Index) InnerError!Ast.Node.Index {
var noreturn_src_node: Ast.Node.Index = 0;
const elide_check = if (maybe_unused_result.toIndex()) |inst| b: {
// Note that this array becomes invalid after appending more items to it
// in the above while loop.
const zir_tags = gz.astgen.instructions.items(.tag);
switch (zir_tags[@intFromEnum(inst)]) {
// For some instructions, modify the zir data
// so we can avoid a separate ensure_result_used instruction.
.call, .field_call => {
const break_extra = gz.astgen.instructions.items(.data)[@intFromEnum(inst)].pl_node.payload_index;
comptime assert(std.meta.fieldIndex(Zir.Inst.Call, "flags") ==
std.meta.fieldIndex(Zir.Inst.FieldCall, "flags"));
const flags: *Zir.Inst.Call.Flags = @ptrCast(&gz.astgen.extra.items[
break_extra + std.meta.fieldIndex(Zir.Inst.Call, "flags").?
]);
flags.ensure_result_used = true;
break :b true;
},
.builtin_call => {
const break_extra = gz.astgen.instructions.items(.data)[@intFromEnum(inst)].pl_node.payload_index;
const flags: *Zir.Inst.BuiltinCall.Flags = @ptrCast(&gz.astgen.extra.items[
break_extra + std.meta.fieldIndex(Zir.Inst.BuiltinCall, "flags").?
]);
flags.ensure_result_used = true;
break :b true;
},
// ZIR instructions that might be a type other than `noreturn` or `void`.
.add,
.addwrap,
.add_sat,
.add_unsafe,
.param,
.param_comptime,
.param_anytype,
.param_anytype_comptime,
.alloc,
.alloc_mut,
.alloc_comptime_mut,
.alloc_inferred,
.alloc_inferred_mut,
.alloc_inferred_comptime,
.alloc_inferred_comptime_mut,
.make_ptr_const,
.array_cat,
.array_mul,
.array_type,
.array_type_sentinel,
.elem_type,
.indexable_ptr_elem_type,
.vector_elem_type,
.vector_type,
.indexable_ptr_len,
.anyframe_type,
.as_node,
.as_shift_operand,
.bit_and,
.bitcast,
.bit_or,
.block,
.block_comptime,
.block_inline,
.declaration,
.suspend_block,
.loop,
.bool_br_and,
.bool_br_or,
.bool_not,
.cmp_lt,
.cmp_lte,
.cmp_eq,
.cmp_gte,
.cmp_gt,
.cmp_neq,
.decl_ref,
.decl_val,
.load,
.div,
.elem_ptr,
.elem_val,
.elem_ptr_node,
.elem_val_node,
.elem_val_imm,
.field_ptr,
.field_val,
.field_ptr_named,
.field_val_named,
.func,
.func_inferred,
.func_fancy,
.int,
.int_big,
.float,
.float128,
.int_type,
.is_non_null,
.is_non_null_ptr,
.is_non_err,
.is_non_err_ptr,
.ret_is_non_err,
.mod_rem,
.mul,
.mulwrap,
.mul_sat,
.ref,
.shl,
.shl_sat,
.shr,
.str,
.sub,
.subwrap,
.sub_sat,
.negate,
.negate_wrap,
.typeof,
.typeof_builtin,
.xor,
.optional_type,
.optional_payload_safe,
.optional_payload_unsafe,
.optional_payload_safe_ptr,
.optional_payload_unsafe_ptr,
.err_union_payload_unsafe,
.err_union_payload_unsafe_ptr,
.err_union_code,
.err_union_code_ptr,
.ptr_type,
.enum_literal,
.merge_error_sets,
.error_union_type,
.bit_not,
.error_value,
.slice_start,
.slice_end,
.slice_sentinel,
.slice_length,
.import,
.switch_block,
.switch_block_ref,
.switch_block_err_union,
.union_init,
.field_type_ref,
.error_set_decl,
.error_set_decl_anon,
.error_set_decl_func,
.enum_from_int,
.int_from_enum,
.type_info,
.size_of,
.bit_size_of,
.typeof_log2_int_type,
.int_from_ptr,
.align_of,
.int_from_bool,
.embed_file,
.error_name,
.sqrt,
.sin,
.cos,
.tan,
.exp,
.exp2,
.log,
.log2,
.log10,
.abs,
.floor,
.ceil,
.trunc,
.round,
.tag_name,
.type_name,
.frame_type,
.frame_size,
.int_from_float,
.float_from_int,
.ptr_from_int,
.float_cast,
.int_cast,
.ptr_cast,
.truncate,
.has_decl,
.has_field,
.clz,
.ctz,
.pop_count,
.byte_swap,
.bit_reverse,
.div_exact,
.div_floor,
.div_trunc,
.mod,
.rem,
.shl_exact,
.shr_exact,
.bit_offset_of,
.offset_of,
.splat,
.reduce,
.shuffle,
.atomic_load,
.atomic_rmw,
.mul_add,
.field_parent_ptr,
.max,
.min,
.c_import,
.@"resume",
.@"await",
.ret_err_value_code,
.ret_ptr,
.ret_type,
.for_len,
.@"try",
.try_ptr,
.opt_eu_base_ptr_init,
.coerce_ptr_elem_ty,
.struct_init_empty,
.struct_init_empty_result,
.struct_init_empty_ref_result,
.struct_init_anon,
.struct_init,
.struct_init_ref,
.struct_init_field_type,
.struct_init_field_ptr,
.array_init_anon,
.array_init,
.array_init_ref,
.validate_array_init_ref_ty,
.array_init_elem_type,
.array_init_elem_ptr,
=> break :b false,
.extended => switch (gz.astgen.instructions.items(.data)[@intFromEnum(inst)].extended.opcode) {
.breakpoint,
.fence,
.set_float_mode,
.set_align_stack,
.set_cold,
=> break :b true,
else => break :b false,
},
// ZIR instructions that are always `noreturn`.
.@"break",
.break_inline,
.condbr,
.condbr_inline,
.compile_error,
.ret_node,
.ret_load,
.ret_implicit,
.ret_err_value,
.@"unreachable",
.repeat,
.repeat_inline,
.panic,
.trap,
.check_comptime_control_flow,
=> {
noreturn_src_node = statement;
break :b true;
},
// ZIR instructions that are always `void`.
.dbg_stmt,
.dbg_var_ptr,
.dbg_var_val,
.ensure_result_used,
.ensure_result_non_error,
.ensure_err_union_payload_void,
.@"export",
.export_value,
.set_eval_branch_quota,
.atomic_store,
.store_node,
.store_to_inferred_ptr,
.resolve_inferred_alloc,
.set_runtime_safety,
.memcpy,
.memset,
.validate_deref,
.validate_destructure,
.save_err_ret_index,
.restore_err_ret_index_unconditional,
.restore_err_ret_index_fn_entry,
.validate_struct_init_ty,
.validate_struct_init_result_ty,
.validate_ptr_struct_init,
.validate_array_init_ty,
.validate_array_init_result_ty,
.validate_ptr_array_init,
.validate_ref_ty,
=> break :b true,
.@"defer" => unreachable,
.defer_err_code => unreachable,
}
} else switch (maybe_unused_result) {
.none => unreachable,
.unreachable_value => b: {
noreturn_src_node = statement;
break :b true;
},
.void_value => true,
else => false,
};
if (!elide_check) {
_ = try gz.addUnNode(.ensure_result_used, maybe_unused_result, statement);
}
return noreturn_src_node;
}
fn countDefers(outer_scope: *Scope, inner_scope: *Scope) struct {
have_any: bool,
have_normal: bool,
have_err: bool,
need_err_code: bool,
} {
var have_normal = false;
var have_err = false;
var need_err_code = false;
var scope = inner_scope;
while (scope != outer_scope) {
switch (scope.tag) {
.gen_zir => scope = scope.cast(GenZir).?.parent,
.local_val => scope = scope.cast(Scope.LocalVal).?.parent,
.local_ptr => scope = scope.cast(Scope.LocalPtr).?.parent,
.defer_normal => {
const defer_scope = scope.cast(Scope.Defer).?;
scope = defer_scope.parent;
have_normal = true;
},
.defer_error => {
const defer_scope = scope.cast(Scope.Defer).?;
scope = defer_scope.parent;
have_err = true;
const have_err_payload = defer_scope.remapped_err_code != .none;
need_err_code = need_err_code or have_err_payload;
},
.namespace => unreachable,
.top => unreachable,
}
}
return .{
.have_any = have_normal or have_err,
.have_normal = have_normal,
.have_err = have_err,
.need_err_code = need_err_code,
};
}
const DefersToEmit = union(enum) {
both: Zir.Inst.Ref, // err code
both_sans_err,
normal_only,
};
fn genDefers(
gz: *GenZir,
outer_scope: *Scope,
inner_scope: *Scope,
which_ones: DefersToEmit,
) InnerError!void {
const gpa = gz.astgen.gpa;
var scope = inner_scope;
while (scope != outer_scope) {
switch (scope.tag) {
.gen_zir => scope = scope.cast(GenZir).?.parent,
.local_val => scope = scope.cast(Scope.LocalVal).?.parent,
.local_ptr => scope = scope.cast(Scope.LocalPtr).?.parent,
.defer_normal => {
const defer_scope = scope.cast(Scope.Defer).?;
scope = defer_scope.parent;
try gz.addDefer(defer_scope.index, defer_scope.len);
},
.defer_error => {
const defer_scope = scope.cast(Scope.Defer).?;
scope = defer_scope.parent;
switch (which_ones) {
.both_sans_err => {
try gz.addDefer(defer_scope.index, defer_scope.len);
},
.both => |err_code| {
if (defer_scope.remapped_err_code.unwrap()) |remapped_err_code| {
try gz.instructions.ensureUnusedCapacity(gpa, 1);
try gz.astgen.instructions.ensureUnusedCapacity(gpa, 1);
const payload_index = try gz.astgen.addExtra(Zir.Inst.DeferErrCode{
.remapped_err_code = remapped_err_code,
.index = defer_scope.index,
.len = defer_scope.len,
});
const new_index: Zir.Inst.Index = @enumFromInt(gz.astgen.instructions.len);
gz.astgen.instructions.appendAssumeCapacity(.{
.tag = .defer_err_code,
.data = .{ .defer_err_code = .{
.err_code = err_code,
.payload_index = payload_index,
} },
});
gz.instructions.appendAssumeCapacity(new_index);
} else {
try gz.addDefer(defer_scope.index, defer_scope.len);
}
},
.normal_only => continue,
}
},
.namespace => unreachable,
.top => unreachable,
}
}
}
fn checkUsed(gz: *GenZir, outer_scope: *Scope, inner_scope: *Scope) InnerError!void {
const astgen = gz.astgen;
var scope = inner_scope;
while (scope != outer_scope) {
switch (scope.tag) {
.gen_zir => scope = scope.cast(GenZir).?.parent,
.local_val => {
const s = scope.cast(Scope.LocalVal).?;
if (s.used == 0 and s.discarded == 0) {
try astgen.appendErrorTok(s.token_src, "unused {s}", .{@tagName(s.id_cat)});
} else if (s.used != 0 and s.discarded != 0) {
try astgen.appendErrorTokNotes(s.discarded, "pointless discard of {s}", .{@tagName(s.id_cat)}, &[_]u32{
try gz.astgen.errNoteTok(s.used, "used here", .{}),
});
}
scope = s.parent;
},
.local_ptr => {
const s = scope.cast(Scope.LocalPtr).?;
if (s.used == 0 and s.discarded == 0) {
try astgen.appendErrorTok(s.token_src, "unused {s}", .{@tagName(s.id_cat)});
} else {
if (s.used != 0 and s.discarded != 0) {
try astgen.appendErrorTokNotes(s.discarded, "pointless discard of {s}", .{@tagName(s.id_cat)}, &[_]u32{
try astgen.errNoteTok(s.used, "used here", .{}),
});
}
if (s.id_cat == .@"local variable" and !s.used_as_lvalue) {
try astgen.appendErrorTokNotes(s.token_src, "local variable is never mutated", .{}, &.{
try astgen.errNoteTok(s.token_src, "consider using 'const'", .{}),
});
}
}
scope = s.parent;
},
.defer_normal, .defer_error => scope = scope.cast(Scope.Defer).?.parent,
.namespace => unreachable,
.top => unreachable,
}
}
}
fn deferStmt(
gz: *GenZir,
scope: *Scope,
node: Ast.Node.Index,
block_arena: Allocator,
scope_tag: Scope.Tag,
) InnerError!*Scope {
var defer_gen = gz.makeSubBlock(scope);
defer_gen.cur_defer_node = node;
defer_gen.any_defer_node = node;
defer defer_gen.unstack();
const tree = gz.astgen.tree;
const node_datas = tree.nodes.items(.data);
const expr_node = node_datas[node].rhs;
const payload_token = node_datas[node].lhs;
var local_val_scope: Scope.LocalVal = undefined;
var opt_remapped_err_code: Zir.Inst.OptionalIndex = .none;
const have_err_code = scope_tag == .defer_error and payload_token != 0;
const sub_scope = if (!have_err_code) &defer_gen.base else blk: {
const ident_name = try gz.astgen.identAsString(payload_token);
const remapped_err_code: Zir.Inst.Index = @enumFromInt(gz.astgen.instructions.len);
opt_remapped_err_code = remapped_err_code.toOptional();
try gz.astgen.instructions.append(gz.astgen.gpa, .{
.tag = .extended,
.data = .{ .extended = .{
.opcode = .value_placeholder,
.small = undefined,
.operand = undefined,
} },
});
const remapped_err_code_ref = remapped_err_code.toRef();
local_val_scope = .{
.parent = &defer_gen.base,
.gen_zir = gz,
.name = ident_name,
.inst = remapped_err_code_ref,
.token_src = payload_token,
.id_cat = .capture,
};
try gz.addDbgVar(.dbg_var_val, ident_name, remapped_err_code_ref);
break :blk &local_val_scope.base;
};
_ = try unusedResultExpr(&defer_gen, sub_scope, expr_node);
try checkUsed(gz, scope, sub_scope);
_ = try defer_gen.addBreak(.break_inline, @enumFromInt(0), .void_value);
// We must handle ref_table for remapped_err_code manually.
const body = defer_gen.instructionsSlice();
const body_len = blk: {
var refs: u32 = 0;
if (opt_remapped_err_code.unwrap()) |remapped_err_code| {
var cur_inst = remapped_err_code;
while (gz.astgen.ref_table.get(cur_inst)) |ref_inst| {
refs += 1;
cur_inst = ref_inst;
}
}
break :blk gz.astgen.countBodyLenAfterFixups(body) + refs;
};
const index: u32 = @intCast(gz.astgen.extra.items.len);
try gz.astgen.extra.ensureUnusedCapacity(gz.astgen.gpa, body_len);
if (opt_remapped_err_code.unwrap()) |remapped_err_code| {
if (gz.astgen.ref_table.fetchRemove(remapped_err_code)) |kv| {
gz.astgen.appendPossiblyRefdBodyInst(&gz.astgen.extra, kv.value);
}
}
gz.astgen.appendBodyWithFixups(body);
const defer_scope = try block_arena.create(Scope.Defer);
defer_scope.* = .{
.base = .{ .tag = scope_tag },
.parent = scope,
.index = index,
.len = body_len,
.remapped_err_code = opt_remapped_err_code,
};
return &defer_scope.base;
}
fn varDecl(
gz: *GenZir,
scope: *Scope,
node: Ast.Node.Index,
block_arena: Allocator,
var_decl: Ast.full.VarDecl,
) InnerError!*Scope {
try emitDbgNode(gz, node);
const astgen = gz.astgen;
const tree = astgen.tree;
const token_tags = tree.tokens.items(.tag);
const main_tokens = tree.nodes.items(.main_token);
const name_token = var_decl.ast.mut_token + 1;
const ident_name_raw = tree.tokenSlice(name_token);
if (mem.eql(u8, ident_name_raw, "_")) {
return astgen.failTok(name_token, "'_' used as an identifier without @\"_\" syntax", .{});
}
const ident_name = try astgen.identAsString(name_token);
try astgen.detectLocalShadowing(
scope,
ident_name,
name_token,
ident_name_raw,
if (token_tags[var_decl.ast.mut_token] == .keyword_const) .@"local constant" else .@"local variable",
);
if (var_decl.ast.init_node == 0) {
return astgen.failNode(node, "variables must be initialized", .{});
}
if (var_decl.ast.addrspace_node != 0) {
return astgen.failTok(main_tokens[var_decl.ast.addrspace_node], "cannot set address space of local variable '{s}'", .{ident_name_raw});
}
if (var_decl.ast.section_node != 0) {
return astgen.failTok(main_tokens[var_decl.ast.section_node], "cannot set section of local variable '{s}'", .{ident_name_raw});
}
const align_inst: Zir.Inst.Ref = if (var_decl.ast.align_node != 0)
try expr(gz, scope, coerced_align_ri, var_decl.ast.align_node)
else
.none;
switch (token_tags[var_decl.ast.mut_token]) {
.keyword_const => {
if (var_decl.comptime_token) |comptime_token| {
try astgen.appendErrorTok(comptime_token, "'comptime const' is redundant; instead wrap the initialization expression with 'comptime'", .{});
}
// Depending on the type of AST the initialization expression is, we may need an lvalue
// or an rvalue as a result location. If it is an rvalue, we can use the instruction as
// the variable, no memory location needed.
const type_node = var_decl.ast.type_node;
if (align_inst == .none and
!astgen.nodes_need_rl.contains(node))
{
const result_info: ResultInfo = if (type_node != 0) .{
.rl = .{ .ty = try typeExpr(gz, scope, type_node) },
.ctx = .const_init,
} else .{ .rl = .none, .ctx = .const_init };
const prev_anon_name_strategy = gz.anon_name_strategy;
gz.anon_name_strategy = .dbg_var;
const init_inst = try reachableExpr(gz, scope, result_info, var_decl.ast.init_node, node);
gz.anon_name_strategy = prev_anon_name_strategy;
try gz.addDbgVar(.dbg_var_val, ident_name, init_inst);
// The const init expression may have modified the error return trace, so signal
// to Sema that it should save the new index for restoring later.
if (nodeMayAppendToErrorTrace(tree, var_decl.ast.init_node))
_ = try gz.addSaveErrRetIndex(.{ .if_of_error_type = init_inst });
const sub_scope = try block_arena.create(Scope.LocalVal);
sub_scope.* = .{
.parent = scope,
.gen_zir = gz,
.name = ident_name,
.inst = init_inst,
.token_src = name_token,
.id_cat = .@"local constant",
};
return &sub_scope.base;
}
const is_comptime = gz.is_comptime or
tree.nodes.items(.tag)[var_decl.ast.init_node] == .@"comptime";
var resolve_inferred_alloc: Zir.Inst.Ref = .none;
var opt_type_inst: Zir.Inst.Ref = .none;
const init_rl: ResultInfo.Loc = if (type_node != 0) init_rl: {
const type_inst = try typeExpr(gz, scope, type_node);
opt_type_inst = type_inst;
if (align_inst == .none) {
break :init_rl .{ .ptr = .{ .inst = try gz.addUnNode(.alloc, type_inst, node) } };
} else {
break :init_rl .{ .ptr = .{ .inst = try gz.addAllocExtended(.{
.node = node,
.type_inst = type_inst,
.align_inst = align_inst,
.is_const = true,
.is_comptime = is_comptime,
}) } };
}
} else init_rl: {
const alloc_inst = if (align_inst == .none) ptr: {
const tag: Zir.Inst.Tag = if (is_comptime)
.alloc_inferred_comptime
else
.alloc_inferred;
break :ptr try gz.addNode(tag, node);
} else ptr: {
break :ptr try gz.addAllocExtended(.{
.node = node,
.type_inst = .none,
.align_inst = align_inst,
.is_const = true,
.is_comptime = is_comptime,
});
};
resolve_inferred_alloc = alloc_inst;
break :init_rl .{ .inferred_ptr = alloc_inst };
};
const var_ptr = switch (init_rl) {
.ptr => |ptr| ptr.inst,
.inferred_ptr => |inst| inst,
else => unreachable,
};
const init_result_info: ResultInfo = .{ .rl = init_rl, .ctx = .const_init };
const prev_anon_name_strategy = gz.anon_name_strategy;
gz.anon_name_strategy = .dbg_var;
defer gz.anon_name_strategy = prev_anon_name_strategy;
const init_inst = try reachableExpr(gz, scope, init_result_info, var_decl.ast.init_node, node);
// The const init expression may have modified the error return trace, so signal
// to Sema that it should save the new index for restoring later.
if (nodeMayAppendToErrorTrace(tree, var_decl.ast.init_node))
_ = try gz.addSaveErrRetIndex(.{ .if_of_error_type = init_inst });
const const_ptr = if (resolve_inferred_alloc != .none) p: {
_ = try gz.addUnNode(.resolve_inferred_alloc, resolve_inferred_alloc, node);
break :p var_ptr;
} else try gz.addUnNode(.make_ptr_const, var_ptr, node);
try gz.addDbgVar(.dbg_var_ptr, ident_name, const_ptr);
const sub_scope = try block_arena.create(Scope.LocalPtr);
sub_scope.* = .{
.parent = scope,
.gen_zir = gz,
.name = ident_name,
.ptr = const_ptr,
.token_src = name_token,
.maybe_comptime = true,
.id_cat = .@"local constant",
};
return &sub_scope.base;
},
.keyword_var => {
if (var_decl.comptime_token != null and gz.is_comptime)
return astgen.failTok(var_decl.comptime_token.?, "'comptime var' is redundant in comptime scope", .{});
const is_comptime = var_decl.comptime_token != null or gz.is_comptime;
var resolve_inferred_alloc: Zir.Inst.Ref = .none;
const alloc: Zir.Inst.Ref, const result_info: ResultInfo = if (var_decl.ast.type_node != 0) a: {
const type_inst = try typeExpr(gz, scope, var_decl.ast.type_node);
const alloc = alloc: {
if (align_inst == .none) {
const tag: Zir.Inst.Tag = if (is_comptime)
.alloc_comptime_mut
else
.alloc_mut;
break :alloc try gz.addUnNode(tag, type_inst, node);
} else {
break :alloc try gz.addAllocExtended(.{
.node = node,
.type_inst = type_inst,
.align_inst = align_inst,
.is_const = false,
.is_comptime = is_comptime,
});
}
};
break :a .{ alloc, .{ .rl = .{ .ptr = .{ .inst = alloc } } } };
} else a: {
const alloc = alloc: {
if (align_inst == .none) {
const tag: Zir.Inst.Tag = if (is_comptime)
.alloc_inferred_comptime_mut
else
.alloc_inferred_mut;
break :alloc try gz.addNode(tag, node);
} else {
break :alloc try gz.addAllocExtended(.{
.node = node,
.type_inst = .none,
.align_inst = align_inst,
.is_const = false,
.is_comptime = is_comptime,
});
}
};
resolve_inferred_alloc = alloc;
break :a .{ alloc, .{ .rl = .{ .inferred_ptr = alloc } } };
};
const prev_anon_name_strategy = gz.anon_name_strategy;
gz.anon_name_strategy = .dbg_var;
_ = try reachableExprComptime(gz, scope, result_info, var_decl.ast.init_node, node, is_comptime);
gz.anon_name_strategy = prev_anon_name_strategy;
if (resolve_inferred_alloc != .none) {
_ = try gz.addUnNode(.resolve_inferred_alloc, resolve_inferred_alloc, node);
}
try gz.addDbgVar(.dbg_var_ptr, ident_name, alloc);
const sub_scope = try block_arena.create(Scope.LocalPtr);
sub_scope.* = .{
.parent = scope,
.gen_zir = gz,
.name = ident_name,
.ptr = alloc,
.token_src = name_token,
.maybe_comptime = is_comptime,
.id_cat = .@"local variable",
};
return &sub_scope.base;
},
else => unreachable,
}
}
fn emitDbgNode(gz: *GenZir, node: Ast.Node.Index) !void {
// The instruction emitted here is for debugging runtime code.
// If the current block will be evaluated only during semantic analysis
// then no dbg_stmt ZIR instruction is needed.
if (gz.is_comptime) return;
const astgen = gz.astgen;
astgen.advanceSourceCursorToNode(node);
const line = astgen.source_line - gz.decl_line;
const column = astgen.source_column;
try emitDbgStmt(gz, .{ line, column });
}
fn assign(gz: *GenZir, scope: *Scope, infix_node: Ast.Node.Index) InnerError!void {
try emitDbgNode(gz, infix_node);
const astgen = gz.astgen;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const main_tokens = tree.nodes.items(.main_token);
const node_tags = tree.nodes.items(.tag);
const lhs = node_datas[infix_node].lhs;
const rhs = node_datas[infix_node].rhs;
if (node_tags[lhs] == .identifier) {
// This intentionally does not support `@"_"` syntax.
const ident_name = tree.tokenSlice(main_tokens[lhs]);
if (mem.eql(u8, ident_name, "_")) {
_ = try expr(gz, scope, .{ .rl = .discard, .ctx = .assignment }, rhs);
return;
}
}
const lvalue = try lvalExpr(gz, scope, lhs);
_ = try expr(gz, scope, .{ .rl = .{ .ptr = .{
.inst = lvalue,
.src_node = infix_node,
} } }, rhs);
}
/// Handles destructure assignments where no LHS is a `const` or `var` decl.
fn assignDestructure(gz: *GenZir, scope: *Scope, node: Ast.Node.Index) InnerError!void {
try emitDbgNode(gz, node);
const astgen = gz.astgen;
const tree = astgen.tree;
const main_tokens = tree.nodes.items(.main_token);
const node_tags = tree.nodes.items(.tag);
const full = tree.assignDestructure(node);
if (full.comptime_token != null and gz.is_comptime) {
return astgen.failNode(node, "redundant comptime keyword in already comptime scope", .{});
}
// If this expression is marked comptime, we must wrap the whole thing in a comptime block.
var gz_buf: GenZir = undefined;
const inner_gz = if (full.comptime_token) |_| bs: {
gz_buf = gz.makeSubBlock(scope);
gz_buf.is_comptime = true;
break :bs &gz_buf;
} else gz;
defer if (full.comptime_token) |_| inner_gz.unstack();
const rl_components = try astgen.arena.alloc(ResultInfo.Loc.DestructureComponent, full.ast.variables.len);
for (rl_components, full.ast.variables) |*variable_rl, variable_node| {
if (node_tags[variable_node] == .identifier) {
// This intentionally does not support `@"_"` syntax.
const ident_name = tree.tokenSlice(main_tokens[variable_node]);
if (mem.eql(u8, ident_name, "_")) {
variable_rl.* = .discard;
continue;
}
}
variable_rl.* = .{ .typed_ptr = .{
.inst = try lvalExpr(inner_gz, scope, variable_node),
.src_node = variable_node,
} };
}
const ri: ResultInfo = .{ .rl = .{ .destructure = .{
.src_node = node,
.components = rl_components,
} } };
_ = try expr(inner_gz, scope, ri, full.ast.value_expr);
if (full.comptime_token) |_| {
const comptime_block_inst = try gz.makeBlockInst(.block_comptime, node);
_ = try inner_gz.addBreak(.@"break", comptime_block_inst, .void_value);
try inner_gz.setBlockBody(comptime_block_inst);
try gz.instructions.append(gz.astgen.gpa, comptime_block_inst);
}
}
/// Handles destructure assignments where the LHS may contain `const` or `var` decls.
fn assignDestructureMaybeDecls(
gz: *GenZir,
scope: *Scope,
node: Ast.Node.Index,
block_arena: Allocator,
) InnerError!*Scope {
try emitDbgNode(gz, node);
const astgen = gz.astgen;
const tree = astgen.tree;
const token_tags = tree.tokens.items(.tag);
const main_tokens = tree.nodes.items(.main_token);
const node_tags = tree.nodes.items(.tag);
const full = tree.assignDestructure(node);
if (full.comptime_token != null and gz.is_comptime) {
return astgen.failNode(node, "redundant comptime keyword in already comptime scope", .{});
}
const is_comptime = full.comptime_token != null or gz.is_comptime;
const value_is_comptime = node_tags[full.ast.value_expr] == .@"comptime";
// When declaring consts via a destructure, we always use a result pointer.
// This avoids the need to create tuple types, and is also likely easier to
// optimize, since it's a bit tricky for the optimizer to "split up" the
// value into individual pointer writes down the line.
// We know this rl information won't live past the evaluation of this
// expression, so it may as well go in the block arena.
const rl_components = try block_arena.alloc(ResultInfo.Loc.DestructureComponent, full.ast.variables.len);
var any_non_const_variables = false;
var any_lvalue_expr = false;
for (rl_components, full.ast.variables) |*variable_rl, variable_node| {
switch (node_tags[variable_node]) {
.identifier => {
// This intentionally does not support `@"_"` syntax.
const ident_name = tree.tokenSlice(main_tokens[variable_node]);
if (mem.eql(u8, ident_name, "_")) {
any_non_const_variables = true;
variable_rl.* = .discard;
continue;
}
},
.global_var_decl, .local_var_decl, .simple_var_decl, .aligned_var_decl => {
const full_var_decl = tree.fullVarDecl(variable_node).?;
const name_token = full_var_decl.ast.mut_token + 1;
const ident_name_raw = tree.tokenSlice(name_token);
if (mem.eql(u8, ident_name_raw, "_")) {
return astgen.failTok(name_token, "'_' used as an identifier without @\"_\" syntax", .{});
}
// We detect shadowing in the second pass over these, while we're creating scopes.
if (full_var_decl.ast.addrspace_node != 0) {
return astgen.failTok(main_tokens[full_var_decl.ast.addrspace_node], "cannot set address space of local variable '{s}'", .{ident_name_raw});
}
if (full_var_decl.ast.section_node != 0) {
return astgen.failTok(main_tokens[full_var_decl.ast.section_node], "cannot set section of local variable '{s}'", .{ident_name_raw});
}
const is_const = switch (token_tags[full_var_decl.ast.mut_token]) {
.keyword_var => false,
.keyword_const => true,
else => unreachable,
};
if (!is_const) any_non_const_variables = true;
// We also mark `const`s as comptime if the RHS is definitely comptime-known.
const this_variable_comptime = is_comptime or (is_const and value_is_comptime);
const align_inst: Zir.Inst.Ref = if (full_var_decl.ast.align_node != 0)
try expr(gz, scope, coerced_align_ri, full_var_decl.ast.align_node)
else
.none;
if (full_var_decl.ast.type_node != 0) {
// Typed alloc
const type_inst = try typeExpr(gz, scope, full_var_decl.ast.type_node);
const ptr = if (align_inst == .none) ptr: {
const tag: Zir.Inst.Tag = if (is_const)
.alloc
else if (this_variable_comptime)
.alloc_comptime_mut
else
.alloc_mut;
break :ptr try gz.addUnNode(tag, type_inst, node);
} else try gz.addAllocExtended(.{
.node = node,
.type_inst = type_inst,
.align_inst = align_inst,
.is_const = is_const,
.is_comptime = this_variable_comptime,
});
variable_rl.* = .{ .typed_ptr = .{ .inst = ptr } };
} else {
// Inferred alloc
const ptr = if (align_inst == .none) ptr: {
const tag: Zir.Inst.Tag = if (is_const) tag: {
break :tag if (this_variable_comptime) .alloc_inferred_comptime else .alloc_inferred;
} else tag: {
break :tag if (this_variable_comptime) .alloc_inferred_comptime_mut else .alloc_inferred_mut;
};
break :ptr try gz.addNode(tag, node);
} else try gz.addAllocExtended(.{
.node = node,
.type_inst = .none,
.align_inst = align_inst,
.is_const = is_const,
.is_comptime = this_variable_comptime,
});
variable_rl.* = .{ .inferred_ptr = ptr };
}
continue;
},
else => {},
}
// This variable is just an lvalue expression.
// We will fill in its result pointer later, inside a comptime block.
any_non_const_variables = true;
any_lvalue_expr = true;
variable_rl.* = .{ .typed_ptr = .{
.inst = undefined,
.src_node = variable_node,
} };
}
if (full.comptime_token != null and !any_non_const_variables) {
try astgen.appendErrorTok(full.comptime_token.?, "'comptime const' is redundant; instead wrap the initialization expression with 'comptime'", .{});
}
// If this expression is marked comptime, we must wrap it in a comptime block.
var gz_buf: GenZir = undefined;
const inner_gz = if (full.comptime_token) |_| bs: {
gz_buf = gz.makeSubBlock(scope);
gz_buf.is_comptime = true;
break :bs &gz_buf;
} else gz;
defer if (full.comptime_token) |_| inner_gz.unstack();
if (any_lvalue_expr) {
// At least one variable was an lvalue expr. Iterate again in order to
// evaluate the lvalues from within the possible block_comptime.
for (rl_components, full.ast.variables) |*variable_rl, variable_node| {
if (variable_rl.* != .typed_ptr) continue;
switch (node_tags[variable_node]) {
.global_var_decl, .local_var_decl, .simple_var_decl, .aligned_var_decl => continue,
else => {},
}
variable_rl.typed_ptr.inst = try lvalExpr(inner_gz, scope, variable_node);
}
}
// We can't give a reasonable anon name strategy for destructured inits, so
// leave it at its default of `.anon`.
_ = try reachableExpr(inner_gz, scope, .{ .rl = .{ .destructure = .{
.src_node = node,
.components = rl_components,
} } }, full.ast.value_expr, node);
if (full.comptime_token) |_| {
// Finish the block_comptime. Inferred alloc resolution etc will occur
// in the parent block.
const comptime_block_inst = try gz.makeBlockInst(.block_comptime, node);
_ = try inner_gz.addBreak(.@"break", comptime_block_inst, .void_value);
try inner_gz.setBlockBody(comptime_block_inst);
try gz.instructions.append(gz.astgen.gpa, comptime_block_inst);
}
// Now, iterate over the variable exprs to construct any new scopes.
// If there were any inferred allocations, resolve them.
// If there were any `const` decls, make the pointer constant.
var cur_scope = scope;
for (rl_components, full.ast.variables) |variable_rl, variable_node| {
switch (node_tags[variable_node]) {
.local_var_decl, .simple_var_decl, .aligned_var_decl => {},
else => continue, // We were mutating an existing lvalue - nothing to do
}
const full_var_decl = tree.fullVarDecl(variable_node).?;
const raw_ptr = switch (variable_rl) {
.discard => unreachable,
.typed_ptr => |typed_ptr| typed_ptr.inst,
.inferred_ptr => |ptr_inst| ptr_inst,
};
// If the alloc was inferred, resolve it.
if (full_var_decl.ast.type_node == 0) {
_ = try gz.addUnNode(.resolve_inferred_alloc, raw_ptr, variable_node);
}
const is_const = switch (token_tags[full_var_decl.ast.mut_token]) {
.keyword_var => false,
.keyword_const => true,
else => unreachable,
};
// If the alloc was const, make it const.
const var_ptr = if (is_const and full_var_decl.ast.type_node != 0) make_const: {
// Note that we don't do this if type_node == 0 since `resolve_inferred_alloc`
// handles it for us.
break :make_const try gz.addUnNode(.make_ptr_const, raw_ptr, node);
} else raw_ptr;
const name_token = full_var_decl.ast.mut_token + 1;
const ident_name_raw = tree.tokenSlice(name_token);
const ident_name = try astgen.identAsString(name_token);
try astgen.detectLocalShadowing(
cur_scope,
ident_name,
name_token,
ident_name_raw,
if (is_const) .@"local constant" else .@"local variable",
);
try gz.addDbgVar(.dbg_var_ptr, ident_name, var_ptr);
// Finally, create the scope.
const sub_scope = try block_arena.create(Scope.LocalPtr);
sub_scope.* = .{
.parent = cur_scope,
.gen_zir = gz,
.name = ident_name,
.ptr = var_ptr,
.token_src = name_token,
.maybe_comptime = is_const or is_comptime,
.id_cat = if (is_const) .@"local constant" else .@"local variable",
};
cur_scope = &sub_scope.base;
}
return cur_scope;
}
fn assignOp(
gz: *GenZir,
scope: *Scope,
infix_node: Ast.Node.Index,
op_inst_tag: Zir.Inst.Tag,
) InnerError!void {
try emitDbgNode(gz, infix_node);
const astgen = gz.astgen;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const lhs_ptr = try lvalExpr(gz, scope, node_datas[infix_node].lhs);
const cursor = switch (op_inst_tag) {
.add, .sub, .mul, .div, .mod_rem => maybeAdvanceSourceCursorToMainToken(gz, infix_node),
else => undefined,
};
const lhs = try gz.addUnNode(.load, lhs_ptr, infix_node);
const lhs_type = try gz.addUnNode(.typeof, lhs, infix_node);
const rhs = try expr(gz, scope, .{ .rl = .{ .coerced_ty = lhs_type } }, node_datas[infix_node].rhs);
switch (op_inst_tag) {
.add, .sub, .mul, .div, .mod_rem => {
try emitDbgStmt(gz, cursor);
},
else => {},
}
const result = try gz.addPlNode(op_inst_tag, infix_node, Zir.Inst.Bin{
.lhs = lhs,
.rhs = rhs,
});
_ = try gz.addPlNode(.store_node, infix_node, Zir.Inst.Bin{
.lhs = lhs_ptr,
.rhs = result,
});
}
fn assignShift(
gz: *GenZir,
scope: *Scope,
infix_node: Ast.Node.Index,
op_inst_tag: Zir.Inst.Tag,
) InnerError!void {
try emitDbgNode(gz, infix_node);
const astgen = gz.astgen;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const lhs_ptr = try lvalExpr(gz, scope, node_datas[infix_node].lhs);
const lhs = try gz.addUnNode(.load, lhs_ptr, infix_node);
const rhs_type = try gz.addUnNode(.typeof_log2_int_type, lhs, infix_node);
const rhs = try expr(gz, scope, .{ .rl = .{ .ty = rhs_type } }, node_datas[infix_node].rhs);
const result = try gz.addPlNode(op_inst_tag, infix_node, Zir.Inst.Bin{
.lhs = lhs,
.rhs = rhs,
});
_ = try gz.addPlNode(.store_node, infix_node, Zir.Inst.Bin{
.lhs = lhs_ptr,
.rhs = result,
});
}
fn assignShiftSat(gz: *GenZir, scope: *Scope, infix_node: Ast.Node.Index) InnerError!void {
try emitDbgNode(gz, infix_node);
const astgen = gz.astgen;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const lhs_ptr = try lvalExpr(gz, scope, node_datas[infix_node].lhs);
const lhs = try gz.addUnNode(.load, lhs_ptr, infix_node);
// Saturating shift-left allows any integer type for both the LHS and RHS.
const rhs = try expr(gz, scope, .{ .rl = .none }, node_datas[infix_node].rhs);
const result = try gz.addPlNode(.shl_sat, infix_node, Zir.Inst.Bin{
.lhs = lhs,
.rhs = rhs,
});
_ = try gz.addPlNode(.store_node, infix_node, Zir.Inst.Bin{
.lhs = lhs_ptr,
.rhs = result,
});
}
fn ptrType(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
ptr_info: Ast.full.PtrType,
) InnerError!Zir.Inst.Ref {
if (ptr_info.size == .C and ptr_info.allowzero_token != null) {
return gz.astgen.failTok(ptr_info.allowzero_token.?, "C pointers always allow address zero", .{});
}
const source_offset = gz.astgen.source_offset;
const source_line = gz.astgen.source_line;
const source_column = gz.astgen.source_column;
const elem_type = try typeExpr(gz, scope, ptr_info.ast.child_type);
var sentinel_ref: Zir.Inst.Ref = .none;
var align_ref: Zir.Inst.Ref = .none;
var addrspace_ref: Zir.Inst.Ref = .none;
var bit_start_ref: Zir.Inst.Ref = .none;
var bit_end_ref: Zir.Inst.Ref = .none;
var trailing_count: u32 = 0;
if (ptr_info.ast.sentinel != 0) {
// These attributes can appear in any order and they all come before the
// element type so we need to reset the source cursor before generating them.
gz.astgen.source_offset = source_offset;
gz.astgen.source_line = source_line;
gz.astgen.source_column = source_column;
sentinel_ref = try comptimeExpr(gz, scope, .{ .rl = .{ .ty = elem_type } }, ptr_info.ast.sentinel);
trailing_count += 1;
}
if (ptr_info.ast.addrspace_node != 0) {
gz.astgen.source_offset = source_offset;
gz.astgen.source_line = source_line;
gz.astgen.source_column = source_column;
addrspace_ref = try expr(gz, scope, coerced_addrspace_ri, ptr_info.ast.addrspace_node);
trailing_count += 1;
}
if (ptr_info.ast.align_node != 0) {
gz.astgen.source_offset = source_offset;
gz.astgen.source_line = source_line;
gz.astgen.source_column = source_column;
align_ref = try expr(gz, scope, coerced_align_ri, ptr_info.ast.align_node);
trailing_count += 1;
}
if (ptr_info.ast.bit_range_start != 0) {
assert(ptr_info.ast.bit_range_end != 0);
bit_start_ref = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .u16_type } }, ptr_info.ast.bit_range_start);
bit_end_ref = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .u16_type } }, ptr_info.ast.bit_range_end);
trailing_count += 2;
}
const gpa = gz.astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
try gz.astgen.instructions.ensureUnusedCapacity(gpa, 1);
try gz.astgen.extra.ensureUnusedCapacity(gpa, @typeInfo(Zir.Inst.PtrType).Struct.fields.len +
trailing_count);
const payload_index = gz.astgen.addExtraAssumeCapacity(Zir.Inst.PtrType{
.elem_type = elem_type,
.src_node = gz.nodeIndexToRelative(node),
});
if (sentinel_ref != .none) {
gz.astgen.extra.appendAssumeCapacity(@intFromEnum(sentinel_ref));
}
if (align_ref != .none) {
gz.astgen.extra.appendAssumeCapacity(@intFromEnum(align_ref));
}
if (addrspace_ref != .none) {
gz.astgen.extra.appendAssumeCapacity(@intFromEnum(addrspace_ref));
}
if (bit_start_ref != .none) {
gz.astgen.extra.appendAssumeCapacity(@intFromEnum(bit_start_ref));
gz.astgen.extra.appendAssumeCapacity(@intFromEnum(bit_end_ref));
}
const new_index: Zir.Inst.Index = @enumFromInt(gz.astgen.instructions.len);
const result = new_index.toRef();
gz.astgen.instructions.appendAssumeCapacity(.{ .tag = .ptr_type, .data = .{
.ptr_type = .{
.flags = .{
.is_allowzero = ptr_info.allowzero_token != null,
.is_mutable = ptr_info.const_token == null,
.is_volatile = ptr_info.volatile_token != null,
.has_sentinel = sentinel_ref != .none,
.has_align = align_ref != .none,
.has_addrspace = addrspace_ref != .none,
.has_bit_range = bit_start_ref != .none,
},
.size = ptr_info.size,
.payload_index = payload_index,
},
} });
gz.instructions.appendAssumeCapacity(new_index);
return rvalue(gz, ri, result, node);
}
fn arrayType(gz: *GenZir, scope: *Scope, ri: ResultInfo, node: Ast.Node.Index) !Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const node_tags = tree.nodes.items(.tag);
const main_tokens = tree.nodes.items(.main_token);
const len_node = node_datas[node].lhs;
if (node_tags[len_node] == .identifier and
mem.eql(u8, tree.tokenSlice(main_tokens[len_node]), "_"))
{
return astgen.failNode(len_node, "unable to infer array size", .{});
}
const len = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, len_node);
const elem_type = try typeExpr(gz, scope, node_datas[node].rhs);
const result = try gz.addPlNode(.array_type, node, Zir.Inst.Bin{
.lhs = len,
.rhs = elem_type,
});
return rvalue(gz, ri, result, node);
}
fn arrayTypeSentinel(gz: *GenZir, scope: *Scope, ri: ResultInfo, node: Ast.Node.Index) !Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const node_tags = tree.nodes.items(.tag);
const main_tokens = tree.nodes.items(.main_token);
const extra = tree.extraData(node_datas[node].rhs, Ast.Node.ArrayTypeSentinel);
const len_node = node_datas[node].lhs;
if (node_tags[len_node] == .identifier and
mem.eql(u8, tree.tokenSlice(main_tokens[len_node]), "_"))
{
return astgen.failNode(len_node, "unable to infer array size", .{});
}
const len = try reachableExpr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, len_node, node);
const elem_type = try typeExpr(gz, scope, extra.elem_type);
const sentinel = try reachableExprComptime(gz, scope, .{ .rl = .{ .coerced_ty = elem_type } }, extra.sentinel, node, true);
const result = try gz.addPlNode(.array_type_sentinel, node, Zir.Inst.ArrayTypeSentinel{
.len = len,
.elem_type = elem_type,
.sentinel = sentinel,
});
return rvalue(gz, ri, result, node);
}
const WipMembers = struct {
payload: *ArrayListUnmanaged(u32),
payload_top: usize,
field_bits_start: u32,
fields_start: u32,
fields_end: u32,
decl_index: u32 = 0,
field_index: u32 = 0,
const Self = @This();
fn init(gpa: Allocator, payload: *ArrayListUnmanaged(u32), decl_count: u32, field_count: u32, comptime bits_per_field: u32, comptime max_field_size: u32) Allocator.Error!Self {
const payload_top: u32 = @intCast(payload.items.len);
const field_bits_start = payload_top + decl_count;
const fields_start = field_bits_start + if (bits_per_field > 0) blk: {
const fields_per_u32 = 32 / bits_per_field;
break :blk (field_count + fields_per_u32 - 1) / fields_per_u32;
} else 0;
const payload_end = fields_start + field_count * max_field_size;
try payload.resize(gpa, payload_end);
return .{
.payload = payload,
.payload_top = payload_top,
.field_bits_start = field_bits_start,
.fields_start = fields_start,
.fields_end = fields_start,
};
}
fn nextDecl(self: *Self, decl_inst: Zir.Inst.Index) void {
self.payload.items[self.payload_top + self.decl_index] = @intFromEnum(decl_inst);
self.decl_index += 1;
}
fn nextField(self: *Self, comptime bits_per_field: u32, bits: [bits_per_field]bool) void {
const fields_per_u32 = 32 / bits_per_field;
const index = self.field_bits_start + self.field_index / fields_per_u32;
assert(index < self.fields_start);
var bit_bag: u32 = if (self.field_index % fields_per_u32 == 0) 0 else self.payload.items[index];
bit_bag >>= bits_per_field;
comptime var i = 0;
inline while (i < bits_per_field) : (i += 1) {
bit_bag |= @as(u32, @intFromBool(bits[i])) << (32 - bits_per_field + i);
}
self.payload.items[index] = bit_bag;
self.field_index += 1;
}
fn appendToField(self: *Self, data: u32) void {
assert(self.fields_end < self.payload.items.len);
self.payload.items[self.fields_end] = data;
self.fields_end += 1;
}
fn finishBits(self: *Self, comptime bits_per_field: u32) void {
if (bits_per_field > 0) {
const fields_per_u32 = 32 / bits_per_field;
const empty_field_slots = fields_per_u32 - (self.field_index % fields_per_u32);
if (self.field_index > 0 and empty_field_slots < fields_per_u32) {
const index = self.field_bits_start + self.field_index / fields_per_u32;
self.payload.items[index] >>= @intCast(empty_field_slots * bits_per_field);
}
}
}
fn declsSlice(self: *Self) []u32 {
return self.payload.items[self.payload_top..][0..self.decl_index];
}
fn fieldsSlice(self: *Self) []u32 {
return self.payload.items[self.field_bits_start..self.fields_end];
}
fn deinit(self: *Self) void {
self.payload.items.len = self.payload_top;
}
};
fn fnDecl(
astgen: *AstGen,
gz: *GenZir,
scope: *Scope,
wip_members: *WipMembers,
decl_node: Ast.Node.Index,
body_node: Ast.Node.Index,
fn_proto: Ast.full.FnProto,
) InnerError!void {
const tree = astgen.tree;
const token_tags = tree.tokens.items(.tag);
// missing function name already happened in scanDecls()
const fn_name_token = fn_proto.name_token orelse return error.AnalysisFail;
// We insert this at the beginning so that its instruction index marks the
// start of the top level declaration.
const decl_inst = try gz.makeBlockInst(.declaration, fn_proto.ast.proto_node);
astgen.advanceSourceCursorToNode(decl_node);
var decl_gz: GenZir = .{
.is_comptime = true,
.decl_node_index = fn_proto.ast.proto_node,
.decl_line = astgen.source_line,
.parent = scope,
.astgen = astgen,
.instructions = gz.instructions,
.instructions_top = gz.instructions.items.len,
};
defer decl_gz.unstack();
var fn_gz: GenZir = .{
.is_comptime = false,
.decl_node_index = fn_proto.ast.proto_node,
.decl_line = decl_gz.decl_line,
.parent = &decl_gz.base,
.astgen = astgen,
.instructions = gz.instructions,
.instructions_top = GenZir.unstacked_top,
};
defer fn_gz.unstack();
// Set this now, since parameter types, return type, etc may be generic.
const prev_within_fn = astgen.within_fn;
defer astgen.within_fn = prev_within_fn;
astgen.within_fn = true;
const is_pub = fn_proto.visib_token != null;
const is_export = blk: {
const maybe_export_token = fn_proto.extern_export_inline_token orelse break :blk false;
break :blk token_tags[maybe_export_token] == .keyword_export;
};
const is_extern = blk: {
const maybe_extern_token = fn_proto.extern_export_inline_token orelse break :blk false;
break :blk token_tags[maybe_extern_token] == .keyword_extern;
};
const has_inline_keyword = blk: {
const maybe_inline_token = fn_proto.extern_export_inline_token orelse break :blk false;
break :blk token_tags[maybe_inline_token] == .keyword_inline;
};
const is_noinline = blk: {
const maybe_noinline_token = fn_proto.extern_export_inline_token orelse break :blk false;
break :blk token_tags[maybe_noinline_token] == .keyword_noinline;
};
const doc_comment_index = try astgen.docCommentAsString(fn_proto.firstToken());
wip_members.nextDecl(decl_inst);
var noalias_bits: u32 = 0;
var params_scope = &fn_gz.base;
const is_var_args = is_var_args: {
var param_type_i: usize = 0;
var it = fn_proto.iterate(tree);
while (it.next()) |param| : (param_type_i += 1) {
const is_comptime = if (param.comptime_noalias) |token| switch (token_tags[token]) {
.keyword_noalias => is_comptime: {
noalias_bits |= @as(u32, 1) << (std.math.cast(u5, param_type_i) orelse
return astgen.failTok(token, "this compiler implementation only supports 'noalias' on the first 32 parameters", .{}));
break :is_comptime false;
},
.keyword_comptime => true,
else => false,
} else false;
const is_anytype = if (param.anytype_ellipsis3) |token| blk: {
switch (token_tags[token]) {
.keyword_anytype => break :blk true,
.ellipsis3 => break :is_var_args true,
else => unreachable,
}
} else false;
const param_name: Zir.NullTerminatedString = if (param.name_token) |name_token| blk: {
const name_bytes = tree.tokenSlice(name_token);
if (mem.eql(u8, "_", name_bytes))
break :blk .empty;
const param_name = try astgen.identAsString(name_token);
if (!is_extern) {
try astgen.detectLocalShadowing(params_scope, param_name, name_token, name_bytes, .@"function parameter");
}
break :blk param_name;
} else if (!is_extern) {
if (param.anytype_ellipsis3) |tok| {
return astgen.failTok(tok, "missing parameter name", .{});
} else {
ambiguous: {
if (tree.nodes.items(.tag)[param.type_expr] != .identifier) break :ambiguous;
const main_token = tree.nodes.items(.main_token)[param.type_expr];
const identifier_str = tree.tokenSlice(main_token);
if (isPrimitive(identifier_str)) break :ambiguous;
return astgen.failNodeNotes(
param.type_expr,
"missing parameter name or type",
.{},
&[_]u32{
try astgen.errNoteNode(
param.type_expr,
"if this is a name, annotate its type '{s}: T'",
.{identifier_str},
),
try astgen.errNoteNode(
param.type_expr,
"if this is a type, give it a name '<name>: {s}'",
.{identifier_str},
),
},
);
}
return astgen.failNode(param.type_expr, "missing parameter name", .{});
}
} else .empty;
const param_inst = if (is_anytype) param: {
const name_token = param.name_token orelse param.anytype_ellipsis3.?;
const tag: Zir.Inst.Tag = if (is_comptime)
.param_anytype_comptime
else
.param_anytype;
break :param try decl_gz.addStrTok(tag, param_name, name_token);
} else param: {
const param_type_node = param.type_expr;
assert(param_type_node != 0);
var param_gz = decl_gz.makeSubBlock(scope);
defer param_gz.unstack();
const param_type = try fullBodyExpr(&param_gz, params_scope, coerced_type_ri, param_type_node);
const param_inst_expected: Zir.Inst.Index = @enumFromInt(astgen.instructions.len + 1);
_ = try param_gz.addBreakWithSrcNode(.break_inline, param_inst_expected, param_type, param_type_node);
const main_tokens = tree.nodes.items(.main_token);
const name_token = param.name_token orelse main_tokens[param_type_node];
const tag: Zir.Inst.Tag = if (is_comptime) .param_comptime else .param;
const param_inst = try decl_gz.addParam(&param_gz, tag, name_token, param_name, param.first_doc_comment);
assert(param_inst_expected == param_inst);
break :param param_inst.toRef();
};
if (param_name == .empty or is_extern) continue;
const sub_scope = try astgen.arena.create(Scope.LocalVal);
sub_scope.* = .{
.parent = params_scope,
.gen_zir = &decl_gz,
.name = param_name,
.inst = param_inst,
.token_src = param.name_token.?,
.id_cat = .@"function parameter",
};
params_scope = &sub_scope.base;
}
break :is_var_args false;
};
const lib_name = if (fn_proto.lib_name) |lib_name_token| blk: {
const lib_name_str = try astgen.strLitAsString(lib_name_token);
const lib_name_slice = astgen.string_bytes.items[@intFromEnum(lib_name_str.index)..][0..lib_name_str.len];
if (mem.indexOfScalar(u8, lib_name_slice, 0) != null) {
return astgen.failTok(lib_name_token, "library name cannot contain null bytes", .{});
} else if (lib_name_str.len == 0) {
return astgen.failTok(lib_name_token, "library name cannot be empty", .{});
}
break :blk lib_name_str.index;
} else .empty;
const maybe_bang = tree.firstToken(fn_proto.ast.return_type) - 1;
const is_inferred_error = token_tags[maybe_bang] == .bang;
// After creating the function ZIR instruction, it will need to update the break
// instructions inside the expression blocks for align, addrspace, cc, and ret_ty
// to use the function instruction as the "block" to break from.
var align_gz = decl_gz.makeSubBlock(params_scope);
defer align_gz.unstack();
const align_ref: Zir.Inst.Ref = if (fn_proto.ast.align_expr == 0) .none else inst: {
const inst = try expr(&decl_gz, params_scope, coerced_align_ri, fn_proto.ast.align_expr);
if (align_gz.instructionsSlice().len == 0) {
// In this case we will send a len=0 body which can be encoded more efficiently.
break :inst inst;
}
_ = try align_gz.addBreak(.break_inline, @enumFromInt(0), inst);
break :inst inst;
};
var addrspace_gz = decl_gz.makeSubBlock(params_scope);
defer addrspace_gz.unstack();
const addrspace_ref: Zir.Inst.Ref = if (fn_proto.ast.addrspace_expr == 0) .none else inst: {
const inst = try expr(&decl_gz, params_scope, coerced_addrspace_ri, fn_proto.ast.addrspace_expr);
if (addrspace_gz.instructionsSlice().len == 0) {
// In this case we will send a len=0 body which can be encoded more efficiently.
break :inst inst;
}
_ = try addrspace_gz.addBreak(.break_inline, @enumFromInt(0), inst);
break :inst inst;
};
var section_gz = decl_gz.makeSubBlock(params_scope);
defer section_gz.unstack();
const section_ref: Zir.Inst.Ref = if (fn_proto.ast.section_expr == 0) .none else inst: {
const inst = try expr(&decl_gz, params_scope, .{ .rl = .{ .coerced_ty = .slice_const_u8_type } }, fn_proto.ast.section_expr);
if (section_gz.instructionsSlice().len == 0) {
// In this case we will send a len=0 body which can be encoded more efficiently.
break :inst inst;
}
_ = try section_gz.addBreak(.break_inline, @enumFromInt(0), inst);
break :inst inst;
};
var cc_gz = decl_gz.makeSubBlock(params_scope);
defer cc_gz.unstack();
const cc_ref: Zir.Inst.Ref = blk: {
if (fn_proto.ast.callconv_expr != 0) {
if (has_inline_keyword) {
return astgen.failNode(
fn_proto.ast.callconv_expr,
"explicit callconv incompatible with inline keyword",
.{},
);
}
const inst = try expr(
&decl_gz,
params_scope,
.{ .rl = .{ .coerced_ty = .calling_convention_type } },
fn_proto.ast.callconv_expr,
);
if (cc_gz.instructionsSlice().len == 0) {
// In this case we will send a len=0 body which can be encoded more efficiently.
break :blk inst;
}
_ = try cc_gz.addBreak(.break_inline, @enumFromInt(0), inst);
break :blk inst;
} else if (is_extern) {
// note: https://github.com/ziglang/zig/issues/5269
break :blk .calling_convention_c;
} else if (has_inline_keyword) {
break :blk .calling_convention_inline;
} else {
break :blk .none;
}
};
var ret_gz = decl_gz.makeSubBlock(params_scope);
defer ret_gz.unstack();
const ret_ref: Zir.Inst.Ref = inst: {
const inst = try fullBodyExpr(&ret_gz, params_scope, coerced_type_ri, fn_proto.ast.return_type);
if (ret_gz.instructionsSlice().len == 0) {
// In this case we will send a len=0 body which can be encoded more efficiently.
break :inst inst;
}
_ = try ret_gz.addBreak(.break_inline, @enumFromInt(0), inst);
break :inst inst;
};
const func_inst: Zir.Inst.Ref = if (body_node == 0) func: {
if (!is_extern) {
return astgen.failTok(fn_proto.ast.fn_token, "non-extern function has no body", .{});
}
if (is_inferred_error) {
return astgen.failTok(maybe_bang, "function prototype may not have inferred error set", .{});
}
break :func try decl_gz.addFunc(.{
.src_node = decl_node,
.cc_ref = cc_ref,
.cc_gz = &cc_gz,
.align_ref = align_ref,
.align_gz = &align_gz,
.ret_ref = ret_ref,
.ret_gz = &ret_gz,
.section_ref = section_ref,
.section_gz = &section_gz,
.addrspace_ref = addrspace_ref,
.addrspace_gz = &addrspace_gz,
.param_block = decl_inst,
.body_gz = null,
.lib_name = lib_name,
.is_var_args = is_var_args,
.is_inferred_error = false,
.is_test = false,
.is_extern = true,
.is_noinline = is_noinline,
.noalias_bits = noalias_bits,
});
} else func: {
// as a scope, fn_gz encloses ret_gz, but for instruction list, fn_gz stacks on ret_gz
fn_gz.instructions_top = ret_gz.instructions.items.len;
const prev_fn_block = astgen.fn_block;
const prev_fn_ret_ty = astgen.fn_ret_ty;
defer {
astgen.fn_block = prev_fn_block;
astgen.fn_ret_ty = prev_fn_ret_ty;
}
astgen.fn_block = &fn_gz;
astgen.fn_ret_ty = if (is_inferred_error or ret_ref.toIndex() != null) r: {
// We're essentially guaranteed to need the return type at some point,
// since the return type is likely not `void` or `noreturn` so there
// will probably be an explicit return requiring RLS. Fetch this
// return type now so the rest of the function can use it.
break :r try fn_gz.addNode(.ret_type, decl_node);
} else ret_ref;
const prev_var_args = astgen.fn_var_args;
astgen.fn_var_args = is_var_args;
defer astgen.fn_var_args = prev_var_args;
astgen.advanceSourceCursorToNode(body_node);
const lbrace_line = astgen.source_line - decl_gz.decl_line;
const lbrace_column = astgen.source_column;
_ = try fullBodyExpr(&fn_gz, params_scope, .{ .rl = .none }, body_node);
try checkUsed(gz, &fn_gz.base, params_scope);
if (!fn_gz.endsWithNoReturn()) {
// As our last action before the return, "pop" the error trace if needed
_ = try fn_gz.addRestoreErrRetIndex(.ret, .always, decl_node);
// Add implicit return at end of function.
_ = try fn_gz.addUnTok(.ret_implicit, .void_value, tree.lastToken(body_node));
}
break :func try decl_gz.addFunc(.{
.src_node = decl_node,
.cc_ref = cc_ref,
.cc_gz = &cc_gz,
.align_ref = align_ref,
.align_gz = &align_gz,
.ret_ref = ret_ref,
.ret_gz = &ret_gz,
.section_ref = section_ref,
.section_gz = &section_gz,
.addrspace_ref = addrspace_ref,
.addrspace_gz = &addrspace_gz,
.lbrace_line = lbrace_line,
.lbrace_column = lbrace_column,
.param_block = decl_inst,
.body_gz = &fn_gz,
.lib_name = lib_name,
.is_var_args = is_var_args,
.is_inferred_error = is_inferred_error,
.is_test = false,
.is_extern = false,
.is_noinline = is_noinline,
.noalias_bits = noalias_bits,
});
};
// We add this at the end so that its instruction index marks the end range
// of the top level declaration. addFunc already unstacked fn_gz and ret_gz.
_ = try decl_gz.addBreak(.break_inline, decl_inst, func_inst);
try setDeclaration(
decl_inst,
std.zig.hashSrc(tree.getNodeSource(decl_node)),
.{ .named = fn_name_token },
decl_gz.decl_line - gz.decl_line,
is_pub,
is_export,
doc_comment_index,
&decl_gz,
// align, linksection, and addrspace are passed in the func instruction in this case.
// TODO: move them from the function instruction to the declaration instruction?
null,
);
}
fn globalVarDecl(
astgen: *AstGen,
gz: *GenZir,
scope: *Scope,
wip_members: *WipMembers,
node: Ast.Node.Index,
var_decl: Ast.full.VarDecl,
) InnerError!void {
const tree = astgen.tree;
const token_tags = tree.tokens.items(.tag);
const is_mutable = token_tags[var_decl.ast.mut_token] == .keyword_var;
// We do this at the beginning so that the instruction index marks the range start
// of the top level declaration.
const decl_inst = try gz.makeBlockInst(.declaration, node);
const name_token = var_decl.ast.mut_token + 1;
astgen.advanceSourceCursorToNode(node);
var block_scope: GenZir = .{
.parent = scope,
.decl_node_index = node,
.decl_line = astgen.source_line,
.astgen = astgen,
.is_comptime = true,
.anon_name_strategy = .parent,
.instructions = gz.instructions,
.instructions_top = gz.instructions.items.len,
};
defer block_scope.unstack();
const is_pub = var_decl.visib_token != null;
const is_export = blk: {
const maybe_export_token = var_decl.extern_export_token orelse break :blk false;
break :blk token_tags[maybe_export_token] == .keyword_export;
};
const is_extern = blk: {
const maybe_extern_token = var_decl.extern_export_token orelse break :blk false;
break :blk token_tags[maybe_extern_token] == .keyword_extern;
};
wip_members.nextDecl(decl_inst);
const is_threadlocal = if (var_decl.threadlocal_token) |tok| blk: {
if (!is_mutable) {
return astgen.failTok(tok, "threadlocal variable cannot be constant", .{});
}
break :blk true;
} else false;
const lib_name = if (var_decl.lib_name) |lib_name_token| blk: {
const lib_name_str = try astgen.strLitAsString(lib_name_token);
const lib_name_slice = astgen.string_bytes.items[@intFromEnum(lib_name_str.index)..][0..lib_name_str.len];
if (mem.indexOfScalar(u8, lib_name_slice, 0) != null) {
return astgen.failTok(lib_name_token, "library name cannot contain null bytes", .{});
} else if (lib_name_str.len == 0) {
return astgen.failTok(lib_name_token, "library name cannot be empty", .{});
}
break :blk lib_name_str.index;
} else .empty;
const doc_comment_index = try astgen.docCommentAsString(var_decl.firstToken());
assert(var_decl.comptime_token == null); // handled by parser
const var_inst: Zir.Inst.Ref = if (var_decl.ast.init_node != 0) vi: {
if (is_extern) {
return astgen.failNode(
var_decl.ast.init_node,
"extern variables have no initializers",
.{},
);
}
const type_inst: Zir.Inst.Ref = if (var_decl.ast.type_node != 0)
try expr(
&block_scope,
&block_scope.base,
coerced_type_ri,
var_decl.ast.type_node,
)
else
.none;
const init_inst = try expr(
&block_scope,
&block_scope.base,
if (type_inst != .none) .{ .rl = .{ .ty = type_inst } } else .{ .rl = .none },
var_decl.ast.init_node,
);
if (is_mutable) {
const var_inst = try block_scope.addVar(.{
.var_type = type_inst,
.lib_name = .empty,
.align_inst = .none, // passed via the decls data
.init = init_inst,
.is_extern = false,
.is_const = !is_mutable,
.is_threadlocal = is_threadlocal,
});
break :vi var_inst;
} else {
break :vi init_inst;
}
} else if (!is_extern) {
return astgen.failNode(node, "variables must be initialized", .{});
} else if (var_decl.ast.type_node != 0) vi: {
// Extern variable which has an explicit type.
const type_inst = try typeExpr(&block_scope, &block_scope.base, var_decl.ast.type_node);
const var_inst = try block_scope.addVar(.{
.var_type = type_inst,
.lib_name = lib_name,
.align_inst = .none, // passed via the decls data
.init = .none,
.is_extern = true,
.is_const = !is_mutable,
.is_threadlocal = is_threadlocal,
});
break :vi var_inst;
} else {
return astgen.failNode(node, "unable to infer variable type", .{});
};
// We do this at the end so that the instruction index marks the end
// range of a top level declaration.
_ = try block_scope.addBreakWithSrcNode(.break_inline, decl_inst, var_inst, node);
var align_gz = block_scope.makeSubBlock(scope);
if (var_decl.ast.align_node != 0) {
const align_inst = try fullBodyExpr(&align_gz, &align_gz.base, coerced_align_ri, var_decl.ast.align_node);
_ = try align_gz.addBreakWithSrcNode(.break_inline, decl_inst, align_inst, node);
}
var linksection_gz = align_gz.makeSubBlock(scope);
if (var_decl.ast.section_node != 0) {
const linksection_inst = try fullBodyExpr(&linksection_gz, &linksection_gz.base, coerced_linksection_ri, var_decl.ast.section_node);
_ = try linksection_gz.addBreakWithSrcNode(.break_inline, decl_inst, linksection_inst, node);
}
var addrspace_gz = linksection_gz.makeSubBlock(scope);
if (var_decl.ast.addrspace_node != 0) {
const addrspace_inst = try fullBodyExpr(&addrspace_gz, &addrspace_gz.base, coerced_addrspace_ri, var_decl.ast.addrspace_node);
_ = try addrspace_gz.addBreakWithSrcNode(.break_inline, decl_inst, addrspace_inst, node);
}
try setDeclaration(
decl_inst,
std.zig.hashSrc(tree.getNodeSource(node)),
.{ .named = name_token },
block_scope.decl_line - gz.decl_line,
is_pub,
is_export,
doc_comment_index,
&block_scope,
.{
.align_gz = &align_gz,
.linksection_gz = &linksection_gz,
.addrspace_gz = &addrspace_gz,
},
);
}
fn comptimeDecl(
astgen: *AstGen,
gz: *GenZir,
scope: *Scope,
wip_members: *WipMembers,
node: Ast.Node.Index,
) InnerError!void {
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const body_node = node_datas[node].lhs;
// Up top so the ZIR instruction index marks the start range of this
// top-level declaration.
const decl_inst = try gz.makeBlockInst(.declaration, node);
wip_members.nextDecl(decl_inst);
astgen.advanceSourceCursorToNode(node);
var decl_block: GenZir = .{
.is_comptime = true,
.decl_node_index = node,
.decl_line = astgen.source_line,
.parent = scope,
.astgen = astgen,
.instructions = gz.instructions,
.instructions_top = gz.instructions.items.len,
};
defer decl_block.unstack();
const block_result = try fullBodyExpr(&decl_block, &decl_block.base, .{ .rl = .none }, body_node);
if (decl_block.isEmpty() or !decl_block.refIsNoReturn(block_result)) {
_ = try decl_block.addBreak(.break_inline, decl_inst, .void_value);
}
try setDeclaration(
decl_inst,
std.zig.hashSrc(tree.getNodeSource(node)),
.@"comptime",
decl_block.decl_line - gz.decl_line,
false,
false,
.empty,
&decl_block,
null,
);
}
fn usingnamespaceDecl(
astgen: *AstGen,
gz: *GenZir,
scope: *Scope,
wip_members: *WipMembers,
node: Ast.Node.Index,
) InnerError!void {
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const type_expr = node_datas[node].lhs;
const is_pub = blk: {
const main_tokens = tree.nodes.items(.main_token);
const token_tags = tree.tokens.items(.tag);
const main_token = main_tokens[node];
break :blk (main_token > 0 and token_tags[main_token - 1] == .keyword_pub);
};
// Up top so the ZIR instruction index marks the start range of this
// top-level declaration.
const decl_inst = try gz.makeBlockInst(.declaration, node);
wip_members.nextDecl(decl_inst);
astgen.advanceSourceCursorToNode(node);
var decl_block: GenZir = .{
.is_comptime = true,
.decl_node_index = node,
.decl_line = astgen.source_line,
.parent = scope,
.astgen = astgen,
.instructions = gz.instructions,
.instructions_top = gz.instructions.items.len,
};
defer decl_block.unstack();
const namespace_inst = try typeExpr(&decl_block, &decl_block.base, type_expr);
_ = try decl_block.addBreak(.break_inline, decl_inst, namespace_inst);
try setDeclaration(
decl_inst,
std.zig.hashSrc(tree.getNodeSource(node)),
.@"usingnamespace",
decl_block.decl_line - gz.decl_line,
is_pub,
false,
.empty,
&decl_block,
null,
);
}
fn testDecl(
astgen: *AstGen,
gz: *GenZir,
scope: *Scope,
wip_members: *WipMembers,
node: Ast.Node.Index,
) InnerError!void {
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const body_node = node_datas[node].rhs;
// Up top so the ZIR instruction index marks the start range of this
// top-level declaration.
const decl_inst = try gz.makeBlockInst(.declaration, node);
wip_members.nextDecl(decl_inst);
astgen.advanceSourceCursorToNode(node);
var decl_block: GenZir = .{
.is_comptime = true,
.decl_node_index = node,
.decl_line = astgen.source_line,
.parent = scope,
.astgen = astgen,
.instructions = gz.instructions,
.instructions_top = gz.instructions.items.len,
};
defer decl_block.unstack();
const main_tokens = tree.nodes.items(.main_token);
const token_tags = tree.tokens.items(.tag);
const test_token = main_tokens[node];
const test_name_token = test_token + 1;
const test_name: DeclarationName = switch (token_tags[test_name_token]) {
else => .unnamed_test,
.string_literal => .{ .named_test = test_name_token },
.identifier => blk: {
const ident_name_raw = tree.tokenSlice(test_name_token);
if (mem.eql(u8, ident_name_raw, "_")) return astgen.failTok(test_name_token, "'_' used as an identifier without @\"_\" syntax", .{});
// if not @"" syntax, just use raw token slice
if (ident_name_raw[0] != '@') {
if (isPrimitive(ident_name_raw)) return astgen.failTok(test_name_token, "cannot test a primitive", .{});
}
// Local variables, including function parameters.
const name_str_index = try astgen.identAsString(test_name_token);
var s = scope;
var found_already: ?Ast.Node.Index = null; // we have found a decl with the same name already
var num_namespaces_out: u32 = 0;
var capturing_namespace: ?*Scope.Namespace = null;
while (true) switch (s.tag) {
.local_val => {
const local_val = s.cast(Scope.LocalVal).?;
if (local_val.name == name_str_index) {
local_val.used = test_name_token;
return astgen.failTokNotes(test_name_token, "cannot test a {s}", .{
@tagName(local_val.id_cat),
}, &[_]u32{
try astgen.errNoteTok(local_val.token_src, "{s} declared here", .{
@tagName(local_val.id_cat),
}),
});
}
s = local_val.parent;
},
.local_ptr => {
const local_ptr = s.cast(Scope.LocalPtr).?;
if (local_ptr.name == name_str_index) {
local_ptr.used = test_name_token;
return astgen.failTokNotes(test_name_token, "cannot test a {s}", .{
@tagName(local_ptr.id_cat),
}, &[_]u32{
try astgen.errNoteTok(local_ptr.token_src, "{s} declared here", .{
@tagName(local_ptr.id_cat),
}),
});
}
s = local_ptr.parent;
},
.gen_zir => s = s.cast(GenZir).?.parent,
.defer_normal, .defer_error => s = s.cast(Scope.Defer).?.parent,
.namespace => {
const ns = s.cast(Scope.Namespace).?;
if (ns.decls.get(name_str_index)) |i| {
if (found_already) |f| {
return astgen.failTokNotes(test_name_token, "ambiguous reference", .{}, &.{
try astgen.errNoteNode(f, "declared here", .{}),
try astgen.errNoteNode(i, "also declared here", .{}),
});
}
// We found a match but must continue looking for ambiguous references to decls.
found_already = i;
}
num_namespaces_out += 1;
capturing_namespace = ns;
s = ns.parent;
},
.top => break,
};
if (found_already == null) {
const ident_name = try astgen.identifierTokenString(test_name_token);
return astgen.failTok(test_name_token, "use of undeclared identifier '{s}'", .{ident_name});
}
break :blk .{ .decltest = name_str_index };
},
};
var fn_block: GenZir = .{
.is_comptime = false,
.decl_node_index = node,
.decl_line = decl_block.decl_line,
.parent = &decl_block.base,
.astgen = astgen,
.instructions = decl_block.instructions,
.instructions_top = decl_block.instructions.items.len,
};
defer fn_block.unstack();
const prev_within_fn = astgen.within_fn;
const prev_fn_block = astgen.fn_block;
const prev_fn_ret_ty = astgen.fn_ret_ty;
astgen.within_fn = true;
astgen.fn_block = &fn_block;
astgen.fn_ret_ty = .anyerror_void_error_union_type;
defer {
astgen.within_fn = prev_within_fn;
astgen.fn_block = prev_fn_block;
astgen.fn_ret_ty = prev_fn_ret_ty;
}
astgen.advanceSourceCursorToNode(body_node);
const lbrace_line = astgen.source_line - decl_block.decl_line;
const lbrace_column = astgen.source_column;
const block_result = try fullBodyExpr(&fn_block, &fn_block.base, .{ .rl = .none }, body_node);
if (fn_block.isEmpty() or !fn_block.refIsNoReturn(block_result)) {
// As our last action before the return, "pop" the error trace if needed
_ = try fn_block.addRestoreErrRetIndex(.ret, .always, node);
// Add implicit return at end of function.
_ = try fn_block.addUnTok(.ret_implicit, .void_value, tree.lastToken(body_node));
}
const func_inst = try decl_block.addFunc(.{
.src_node = node,
.cc_ref = .none,
.cc_gz = null,
.align_ref = .none,
.align_gz = null,
.ret_ref = .anyerror_void_error_union_type,
.ret_gz = null,
.section_ref = .none,
.section_gz = null,
.addrspace_ref = .none,
.addrspace_gz = null,
.lbrace_line = lbrace_line,
.lbrace_column = lbrace_column,
.param_block = decl_inst,
.body_gz = &fn_block,
.lib_name = .empty,
.is_var_args = false,
.is_inferred_error = false,
.is_test = true,
.is_extern = false,
.is_noinline = false,
.noalias_bits = 0,
});
_ = try decl_block.addBreak(.break_inline, decl_inst, func_inst);
try setDeclaration(
decl_inst,
std.zig.hashSrc(tree.getNodeSource(node)),
test_name,
decl_block.decl_line - gz.decl_line,
false,
false,
.empty,
&decl_block,
null,
);
}
fn structDeclInner(
gz: *GenZir,
scope: *Scope,
node: Ast.Node.Index,
container_decl: Ast.full.ContainerDecl,
layout: std.builtin.Type.ContainerLayout,
backing_int_node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const decl_inst = try gz.reserveInstructionIndex();
if (container_decl.ast.members.len == 0 and backing_int_node == 0) {
try gz.setStruct(decl_inst, .{
.src_node = node,
.layout = layout,
.captures_len = 0,
.fields_len = 0,
.decls_len = 0,
.has_backing_int = false,
.known_non_opv = false,
.known_comptime_only = false,
.is_tuple = false,
.any_comptime_fields = false,
.any_default_inits = false,
.any_aligned_fields = false,
.fields_hash = std.zig.hashSrc(@tagName(layout)),
});
return decl_inst.toRef();
}
const astgen = gz.astgen;
const gpa = astgen.gpa;
const tree = astgen.tree;
var namespace: Scope.Namespace = .{
.parent = scope,
.node = node,
.inst = decl_inst,
.declaring_gz = gz,
.maybe_generic = astgen.within_fn,
};
defer namespace.deinit(gpa);
// The struct_decl instruction introduces a scope in which the decls of the struct
// are in scope, so that field types, alignments, and default value expressions
// can refer to decls within the struct itself.
astgen.advanceSourceCursorToNode(node);
var block_scope: GenZir = .{
.parent = &namespace.base,
.decl_node_index = node,
.decl_line = gz.decl_line,
.astgen = astgen,
.is_comptime = true,
.instructions = gz.instructions,
.instructions_top = gz.instructions.items.len,
};
defer block_scope.unstack();
const scratch_top = astgen.scratch.items.len;
defer astgen.scratch.items.len = scratch_top;
var backing_int_body_len: usize = 0;
const backing_int_ref: Zir.Inst.Ref = blk: {
if (backing_int_node != 0) {
if (layout != .@"packed") {
return astgen.failNode(backing_int_node, "non-packed struct does not support backing integer type", .{});
} else {
const backing_int_ref = try typeExpr(&block_scope, &namespace.base, backing_int_node);
if (!block_scope.isEmpty()) {
if (!block_scope.endsWithNoReturn()) {
_ = try block_scope.addBreak(.break_inline, decl_inst, backing_int_ref);
}
const body = block_scope.instructionsSlice();
const old_scratch_len = astgen.scratch.items.len;
try astgen.scratch.ensureUnusedCapacity(gpa, countBodyLenAfterFixups(astgen, body));
appendBodyWithFixupsArrayList(astgen, &astgen.scratch, body);
backing_int_body_len = astgen.scratch.items.len - old_scratch_len;
block_scope.instructions.items.len = block_scope.instructions_top;
}
break :blk backing_int_ref;
}
} else {
break :blk .none;
}
};
const decl_count = try astgen.scanDecls(&namespace, container_decl.ast.members);
const field_count: u32 = @intCast(container_decl.ast.members.len - decl_count);
const bits_per_field = 4;
const max_field_size = 5;
var wip_members = try WipMembers.init(gpa, &astgen.scratch, decl_count, field_count, bits_per_field, max_field_size);
defer wip_members.deinit();
// We will use the scratch buffer, starting here, for the bodies:
// bodies: { // for every fields_len
// field_type_body_inst: Inst, // for each field_type_body_len
// align_body_inst: Inst, // for each align_body_len
// init_body_inst: Inst, // for each init_body_len
// }
// Note that the scratch buffer is simultaneously being used by WipMembers, however
// it will not access any elements beyond this point in the ArrayList. It also
// accesses via the ArrayList items field so it can handle the scratch buffer being
// reallocated.
// No defer needed here because it is handled by `wip_members.deinit()` above.
const bodies_start = astgen.scratch.items.len;
const node_tags = tree.nodes.items(.tag);
const is_tuple = for (container_decl.ast.members) |member_node| {
const container_field = tree.fullContainerField(member_node) orelse continue;
if (container_field.ast.tuple_like) break true;
} else false;
if (is_tuple) switch (layout) {
.auto => {},
.@"extern" => return astgen.failNode(node, "extern tuples are not supported", .{}),
.@"packed" => return astgen.failNode(node, "packed tuples are not supported", .{}),
};
if (is_tuple) for (container_decl.ast.members) |member_node| {
switch (node_tags[member_node]) {
.container_field_init,
.container_field_align,
.container_field,
.@"comptime",
.test_decl,
=> continue,
else => {
const tuple_member = for (container_decl.ast.members) |maybe_tuple| switch (node_tags[maybe_tuple]) {
.container_field_init,
.container_field_align,
.container_field,
=> break maybe_tuple,
else => {},
} else unreachable;
return astgen.failNodeNotes(
member_node,
"tuple declarations cannot contain declarations",
.{},
&[_]u32{
try astgen.errNoteNode(tuple_member, "tuple field here", .{}),
},
);
},
}
};
var fields_hasher = std.zig.SrcHasher.init(.{});
fields_hasher.update(@tagName(layout));
if (backing_int_node != 0) {
fields_hasher.update(tree.getNodeSource(backing_int_node));
}
var sfba = std.heap.stackFallback(256, astgen.arena);
const sfba_allocator = sfba.get();
var duplicate_names = std.AutoArrayHashMap(Zir.NullTerminatedString, std.ArrayListUnmanaged(Ast.TokenIndex)).init(sfba_allocator);
try duplicate_names.ensureTotalCapacity(field_count);
// When there aren't errors, use this to avoid a second iteration.
var any_duplicate = false;
var known_non_opv = false;
var known_comptime_only = false;
var any_comptime_fields = false;
var any_aligned_fields = false;
var any_default_inits = false;
for (container_decl.ast.members) |member_node| {
var member = switch (try containerMember(&block_scope, &namespace.base, &wip_members, member_node)) {
.decl => continue,
.field => |field| field,
};
fields_hasher.update(tree.getNodeSource(member_node));
if (!is_tuple) {
const field_name = try astgen.identAsString(member.ast.main_token);
member.convertToNonTupleLike(astgen.tree.nodes);
assert(!member.ast.tuple_like);
wip_members.appendToField(@intFromEnum(field_name));
const gop = try duplicate_names.getOrPut(field_name);
if (gop.found_existing) {
try gop.value_ptr.append(sfba_allocator, member.ast.main_token);
any_duplicate = true;
} else {
gop.value_ptr.* = .{};
try gop.value_ptr.append(sfba_allocator, member.ast.main_token);
}
} else if (!member.ast.tuple_like) {
return astgen.failTok(member.ast.main_token, "tuple field has a name", .{});
}
const doc_comment_index = try astgen.docCommentAsString(member.firstToken());
wip_members.appendToField(@intFromEnum(doc_comment_index));
if (member.ast.type_expr == 0) {
return astgen.failTok(member.ast.main_token, "struct field missing type", .{});
}
const field_type = try typeExpr(&block_scope, &namespace.base, member.ast.type_expr);
const have_type_body = !block_scope.isEmpty();
const have_align = member.ast.align_expr != 0;
const have_value = member.ast.value_expr != 0;
const is_comptime = member.comptime_token != null;
if (is_comptime) {
switch (layout) {
.@"packed" => return astgen.failTok(member.comptime_token.?, "packed struct fields cannot be marked comptime", .{}),
.@"extern" => return astgen.failTok(member.comptime_token.?, "extern struct fields cannot be marked comptime", .{}),
.auto => any_comptime_fields = true,
}
} else {
known_non_opv = known_non_opv or
nodeImpliesMoreThanOnePossibleValue(tree, member.ast.type_expr);
known_comptime_only = known_comptime_only or
nodeImpliesComptimeOnly(tree, member.ast.type_expr);
}
wip_members.nextField(bits_per_field, .{ have_align, have_value, is_comptime, have_type_body });
if (have_type_body) {
if (!block_scope.endsWithNoReturn()) {
_ = try block_scope.addBreak(.break_inline, decl_inst, field_type);
}
const body = block_scope.instructionsSlice();
const old_scratch_len = astgen.scratch.items.len;
try astgen.scratch.ensureUnusedCapacity(gpa, countBodyLenAfterFixups(astgen, body));
appendBodyWithFixupsArrayList(astgen, &astgen.scratch, body);
wip_members.appendToField(@intCast(astgen.scratch.items.len - old_scratch_len));
block_scope.instructions.items.len = block_scope.instructions_top;
} else {
wip_members.appendToField(@intFromEnum(field_type));
}
if (have_align) {
if (layout == .@"packed") {
try astgen.appendErrorNode(member.ast.align_expr, "unable to override alignment of packed struct fields", .{});
}
any_aligned_fields = true;
const align_ref = try expr(&block_scope, &namespace.base, coerced_align_ri, member.ast.align_expr);
if (!block_scope.endsWithNoReturn()) {
_ = try block_scope.addBreak(.break_inline, decl_inst, align_ref);
}
const body = block_scope.instructionsSlice();
const old_scratch_len = astgen.scratch.items.len;
try astgen.scratch.ensureUnusedCapacity(gpa, countBodyLenAfterFixups(astgen, body));
appendBodyWithFixupsArrayList(astgen, &astgen.scratch, body);
wip_members.appendToField(@intCast(astgen.scratch.items.len - old_scratch_len));
block_scope.instructions.items.len = block_scope.instructions_top;
}
if (have_value) {
any_default_inits = true;
// The decl_inst is used as here so that we can easily reconstruct a mapping
// between it and the field type when the fields inits are analzyed.
const ri: ResultInfo = .{ .rl = if (field_type == .none) .none else .{ .coerced_ty = decl_inst.toRef() } };
const default_inst = try expr(&block_scope, &namespace.base, ri, member.ast.value_expr);
if (!block_scope.endsWithNoReturn()) {
_ = try block_scope.addBreak(.break_inline, decl_inst, default_inst);
}
const body = block_scope.instructionsSlice();
const old_scratch_len = astgen.scratch.items.len;
try astgen.scratch.ensureUnusedCapacity(gpa, countBodyLenAfterFixups(astgen, body));
appendBodyWithFixupsArrayList(astgen, &astgen.scratch, body);
wip_members.appendToField(@intCast(astgen.scratch.items.len - old_scratch_len));
block_scope.instructions.items.len = block_scope.instructions_top;
} else if (member.comptime_token) |comptime_token| {
return astgen.failTok(comptime_token, "comptime field without default initialization value", .{});
}
}
if (any_duplicate) {
var it = duplicate_names.iterator();
while (it.next()) |entry| {
const record = entry.value_ptr.*;
if (record.items.len > 1) {
var error_notes = std.ArrayList(u32).init(astgen.arena);
for (record.items[1..]) |duplicate| {
try error_notes.append(try astgen.errNoteTok(duplicate, "duplicate field here", .{}));
}
try error_notes.append(try astgen.errNoteNode(node, "struct declared here", .{}));
try astgen.appendErrorTokNotes(
record.items[0],
"duplicate struct field name",
.{},
error_notes.items,
);
}
}
return error.AnalysisFail;
}
var fields_hash: std.zig.SrcHash = undefined;
fields_hasher.final(&fields_hash);
try gz.setStruct(decl_inst, .{
.src_node = node,
.layout = layout,
.captures_len = @intCast(namespace.captures.count()),
.fields_len = field_count,
.decls_len = decl_count,
.has_backing_int = backing_int_ref != .none,
.known_non_opv = known_non_opv,
.known_comptime_only = known_comptime_only,
.is_tuple = is_tuple,
.any_comptime_fields = any_comptime_fields,
.any_default_inits = any_default_inits,
.any_aligned_fields = any_aligned_fields,
.fields_hash = fields_hash,
});
wip_members.finishBits(bits_per_field);
const decls_slice = wip_members.declsSlice();
const fields_slice = wip_members.fieldsSlice();
const bodies_slice = astgen.scratch.items[bodies_start..];
try astgen.extra.ensureUnusedCapacity(gpa, backing_int_body_len + 2 +
decls_slice.len + namespace.captures.count() + fields_slice.len + bodies_slice.len);
astgen.extra.appendSliceAssumeCapacity(@ptrCast(namespace.captures.keys()));
if (backing_int_ref != .none) {
astgen.extra.appendAssumeCapacity(@intCast(backing_int_body_len));
if (backing_int_body_len == 0) {
astgen.extra.appendAssumeCapacity(@intFromEnum(backing_int_ref));
} else {
astgen.extra.appendSliceAssumeCapacity(astgen.scratch.items[scratch_top..][0..backing_int_body_len]);
}
}
astgen.extra.appendSliceAssumeCapacity(decls_slice);
astgen.extra.appendSliceAssumeCapacity(fields_slice);
astgen.extra.appendSliceAssumeCapacity(bodies_slice);
block_scope.unstack();
return decl_inst.toRef();
}
fn unionDeclInner(
gz: *GenZir,
scope: *Scope,
node: Ast.Node.Index,
members: []const Ast.Node.Index,
layout: std.builtin.Type.ContainerLayout,
arg_node: Ast.Node.Index,
auto_enum_tok: ?Ast.TokenIndex,
) InnerError!Zir.Inst.Ref {
const decl_inst = try gz.reserveInstructionIndex();
const astgen = gz.astgen;
const gpa = astgen.gpa;
var namespace: Scope.Namespace = .{
.parent = scope,
.node = node,
.inst = decl_inst,
.declaring_gz = gz,
.maybe_generic = astgen.within_fn,
};
defer namespace.deinit(gpa);
// The union_decl instruction introduces a scope in which the decls of the union
// are in scope, so that field types, alignments, and default value expressions
// can refer to decls within the union itself.
astgen.advanceSourceCursorToNode(node);
var block_scope: GenZir = .{
.parent = &namespace.base,
.decl_node_index = node,
.decl_line = gz.decl_line,
.astgen = astgen,
.is_comptime = true,
.instructions = gz.instructions,
.instructions_top = gz.instructions.items.len,
};
defer block_scope.unstack();
const decl_count = try astgen.scanDecls(&namespace, members);
const field_count: u32 = @intCast(members.len - decl_count);
if (layout != .auto and (auto_enum_tok != null or arg_node != 0)) {
if (arg_node != 0) {
return astgen.failNode(arg_node, "{s} union does not support enum tag type", .{@tagName(layout)});
} else {
return astgen.failTok(auto_enum_tok.?, "{s} union does not support enum tag type", .{@tagName(layout)});
}
}
const arg_inst: Zir.Inst.Ref = if (arg_node != 0)
try typeExpr(&block_scope, &namespace.base, arg_node)
else
.none;
const bits_per_field = 4;
const max_field_size = 5;
var any_aligned_fields = false;
var wip_members = try WipMembers.init(gpa, &astgen.scratch, decl_count, field_count, bits_per_field, max_field_size);
defer wip_members.deinit();
var fields_hasher = std.zig.SrcHasher.init(.{});
fields_hasher.update(@tagName(layout));
fields_hasher.update(&.{@intFromBool(auto_enum_tok != null)});
if (arg_node != 0) {
fields_hasher.update(astgen.tree.getNodeSource(arg_node));
}
var sfba = std.heap.stackFallback(256, astgen.arena);
const sfba_allocator = sfba.get();
var duplicate_names = std.AutoArrayHashMap(Zir.NullTerminatedString, std.ArrayListUnmanaged(Ast.TokenIndex)).init(sfba_allocator);
try duplicate_names.ensureTotalCapacity(field_count);
// When there aren't errors, use this to avoid a second iteration.
var any_duplicate = false;
for (members) |member_node| {
var member = switch (try containerMember(&block_scope, &namespace.base, &wip_members, member_node)) {
.decl => continue,
.field => |field| field,
};
fields_hasher.update(astgen.tree.getNodeSource(member_node));
member.convertToNonTupleLike(astgen.tree.nodes);
if (member.ast.tuple_like) {
return astgen.failTok(member.ast.main_token, "union field missing name", .{});
}
if (member.comptime_token) |comptime_token| {
return astgen.failTok(comptime_token, "union fields cannot be marked comptime", .{});
}
const field_name = try astgen.identAsString(member.ast.main_token);
wip_members.appendToField(@intFromEnum(field_name));
const gop = try duplicate_names.getOrPut(field_name);
if (gop.found_existing) {
try gop.value_ptr.append(sfba_allocator, member.ast.main_token);
any_duplicate = true;
} else {
gop.value_ptr.* = .{};
try gop.value_ptr.append(sfba_allocator, member.ast.main_token);
}
const doc_comment_index = try astgen.docCommentAsString(member.firstToken());
wip_members.appendToField(@intFromEnum(doc_comment_index));
const have_type = member.ast.type_expr != 0;
const have_align = member.ast.align_expr != 0;
const have_value = member.ast.value_expr != 0;
const unused = false;
wip_members.nextField(bits_per_field, .{ have_type, have_align, have_value, unused });
if (have_type) {
const field_type = try typeExpr(&block_scope, &namespace.base, member.ast.type_expr);
wip_members.appendToField(@intFromEnum(field_type));
} else if (arg_inst == .none and auto_enum_tok == null) {
return astgen.failNode(member_node, "union field missing type", .{});
}
if (have_align) {
const align_inst = try expr(&block_scope, &block_scope.base, coerced_align_ri, member.ast.align_expr);
wip_members.appendToField(@intFromEnum(align_inst));
any_aligned_fields = true;
}
if (have_value) {
if (arg_inst == .none) {
return astgen.failNodeNotes(
node,
"explicitly valued tagged union missing integer tag type",
.{},
&[_]u32{
try astgen.errNoteNode(
member.ast.value_expr,
"tag value specified here",
.{},
),
},
);
}
if (auto_enum_tok == null) {
return astgen.failNodeNotes(
node,
"explicitly valued tagged union requires inferred enum tag type",
.{},
&[_]u32{
try astgen.errNoteNode(
member.ast.value_expr,
"tag value specified here",
.{},
),
},
);
}
const tag_value = try expr(&block_scope, &block_scope.base, .{ .rl = .{ .ty = arg_inst } }, member.ast.value_expr);
wip_members.appendToField(@intFromEnum(tag_value));
}
}
if (any_duplicate) {
var it = duplicate_names.iterator();
while (it.next()) |entry| {
const record = entry.value_ptr.*;
if (record.items.len > 1) {
var error_notes = std.ArrayList(u32).init(astgen.arena);
for (record.items[1..]) |duplicate| {
try error_notes.append(try astgen.errNoteTok(duplicate, "duplicate field here", .{}));
}
try error_notes.append(try astgen.errNoteNode(node, "union declared here", .{}));
try astgen.appendErrorTokNotes(
record.items[0],
"duplicate union field name",
.{},
error_notes.items,
);
}
}
return error.AnalysisFail;
}
var fields_hash: std.zig.SrcHash = undefined;
fields_hasher.final(&fields_hash);
if (!block_scope.isEmpty()) {
_ = try block_scope.addBreak(.break_inline, decl_inst, .void_value);
}
const body = block_scope.instructionsSlice();
const body_len = astgen.countBodyLenAfterFixups(body);
try gz.setUnion(decl_inst, .{
.src_node = node,
.layout = layout,
.tag_type = arg_inst,
.captures_len = @intCast(namespace.captures.count()),
.body_len = body_len,
.fields_len = field_count,
.decls_len = decl_count,
.auto_enum_tag = auto_enum_tok != null,
.any_aligned_fields = any_aligned_fields,
.fields_hash = fields_hash,
});
wip_members.finishBits(bits_per_field);
const decls_slice = wip_members.declsSlice();
const fields_slice = wip_members.fieldsSlice();
try astgen.extra.ensureUnusedCapacity(gpa, namespace.captures.count() + decls_slice.len + body_len + fields_slice.len);
astgen.extra.appendSliceAssumeCapacity(@ptrCast(namespace.captures.keys()));
astgen.extra.appendSliceAssumeCapacity(decls_slice);
astgen.appendBodyWithFixups(body);
astgen.extra.appendSliceAssumeCapacity(fields_slice);
block_scope.unstack();
return decl_inst.toRef();
}
fn containerDecl(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
container_decl: Ast.full.ContainerDecl,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const gpa = astgen.gpa;
const tree = astgen.tree;
const token_tags = tree.tokens.items(.tag);
const prev_fn_block = astgen.fn_block;
astgen.fn_block = null;
defer astgen.fn_block = prev_fn_block;
// We must not create any types until Sema. Here the goal is only to generate
// ZIR for all the field types, alignments, and default value expressions.
switch (token_tags[container_decl.ast.main_token]) {
.keyword_struct => {
const layout: std.builtin.Type.ContainerLayout = if (container_decl.layout_token) |t| switch (token_tags[t]) {
.keyword_packed => .@"packed",
.keyword_extern => .@"extern",
else => unreachable,
} else .auto;
const result = try structDeclInner(gz, scope, node, container_decl, layout, container_decl.ast.arg);
return rvalue(gz, ri, result, node);
},
.keyword_union => {
const layout: std.builtin.Type.ContainerLayout = if (container_decl.layout_token) |t| switch (token_tags[t]) {
.keyword_packed => .@"packed",
.keyword_extern => .@"extern",
else => unreachable,
} else .auto;
const result = try unionDeclInner(gz, scope, node, container_decl.ast.members, layout, container_decl.ast.arg, container_decl.ast.enum_token);
return rvalue(gz, ri, result, node);
},
.keyword_enum => {
if (container_decl.layout_token) |t| {
return astgen.failTok(t, "enums do not support 'packed' or 'extern'; instead provide an explicit integer tag type", .{});
}
// Count total fields as well as how many have explicitly provided tag values.
const counts = blk: {
var values: usize = 0;
var total_fields: usize = 0;
var decls: usize = 0;
var nonexhaustive_node: Ast.Node.Index = 0;
var nonfinal_nonexhaustive = false;
for (container_decl.ast.members) |member_node| {
var member = tree.fullContainerField(member_node) orelse {
decls += 1;
continue;
};
member.convertToNonTupleLike(astgen.tree.nodes);
if (member.ast.tuple_like) {
return astgen.failTok(member.ast.main_token, "enum field missing name", .{});
}
if (member.comptime_token) |comptime_token| {
return astgen.failTok(comptime_token, "enum fields cannot be marked comptime", .{});
}
if (member.ast.type_expr != 0) {
return astgen.failNodeNotes(
member.ast.type_expr,
"enum fields do not have types",
.{},
&[_]u32{
try astgen.errNoteNode(
node,
"consider 'union(enum)' here to make it a tagged union",
.{},
),
},
);
}
if (member.ast.align_expr != 0) {
return astgen.failNode(member.ast.align_expr, "enum fields cannot be aligned", .{});
}
const name_token = member.ast.main_token;
if (mem.eql(u8, tree.tokenSlice(name_token), "_")) {
if (nonexhaustive_node != 0) {
return astgen.failNodeNotes(
member_node,
"redundant non-exhaustive enum mark",
.{},
&[_]u32{
try astgen.errNoteNode(
nonexhaustive_node,
"other mark here",
.{},
),
},
);
}
nonexhaustive_node = member_node;
if (member.ast.value_expr != 0) {
return astgen.failNode(member.ast.value_expr, "'_' is used to mark an enum as non-exhaustive and cannot be assigned a value", .{});
}
continue;
} else if (nonexhaustive_node != 0) {
nonfinal_nonexhaustive = true;
}
total_fields += 1;
if (member.ast.value_expr != 0) {
if (container_decl.ast.arg == 0) {
return astgen.failNode(member.ast.value_expr, "value assigned to enum tag with inferred tag type", .{});
}
values += 1;
}
}
if (nonfinal_nonexhaustive) {
return astgen.failNode(nonexhaustive_node, "'_' field of non-exhaustive enum must be last", .{});
}
break :blk .{
.total_fields = total_fields,
.values = values,
.decls = decls,
.nonexhaustive_node = nonexhaustive_node,
};
};
if (counts.nonexhaustive_node != 0 and container_decl.ast.arg == 0) {
try astgen.appendErrorNodeNotes(
node,
"non-exhaustive enum missing integer tag type",
.{},
&[_]u32{
try astgen.errNoteNode(
counts.nonexhaustive_node,
"marked non-exhaustive here",
.{},
),
},
);
}
// In this case we must generate ZIR code for the tag values, similar to
// how structs are handled above.
const nonexhaustive = counts.nonexhaustive_node != 0;
const decl_inst = try gz.reserveInstructionIndex();
var namespace: Scope.Namespace = .{
.parent = scope,
.node = node,
.inst = decl_inst,
.declaring_gz = gz,
.maybe_generic = astgen.within_fn,
};
defer namespace.deinit(gpa);
// The enum_decl instruction introduces a scope in which the decls of the enum
// are in scope, so that tag values can refer to decls within the enum itself.
astgen.advanceSourceCursorToNode(node);
var block_scope: GenZir = .{
.parent = &namespace.base,
.decl_node_index = node,
.decl_line = gz.decl_line,
.astgen = astgen,
.is_comptime = true,
.instructions = gz.instructions,
.instructions_top = gz.instructions.items.len,
};
defer block_scope.unstack();
_ = try astgen.scanDecls(&namespace, container_decl.ast.members);
namespace.base.tag = .namespace;
const arg_inst: Zir.Inst.Ref = if (container_decl.ast.arg != 0)
try comptimeExpr(&block_scope, &namespace.base, coerced_type_ri, container_decl.ast.arg)
else
.none;
const bits_per_field = 1;
const max_field_size = 3;
var wip_members = try WipMembers.init(gpa, &astgen.scratch, @intCast(counts.decls), @intCast(counts.total_fields), bits_per_field, max_field_size);
defer wip_members.deinit();
var fields_hasher = std.zig.SrcHasher.init(.{});
if (container_decl.ast.arg != 0) {
fields_hasher.update(tree.getNodeSource(container_decl.ast.arg));
}
fields_hasher.update(&.{@intFromBool(nonexhaustive)});
var sfba = std.heap.stackFallback(256, astgen.arena);
const sfba_allocator = sfba.get();
var duplicate_names = std.AutoArrayHashMap(Zir.NullTerminatedString, std.ArrayListUnmanaged(Ast.TokenIndex)).init(sfba_allocator);
try duplicate_names.ensureTotalCapacity(counts.total_fields);
// When there aren't errors, use this to avoid a second iteration.
var any_duplicate = false;
for (container_decl.ast.members) |member_node| {
if (member_node == counts.nonexhaustive_node)
continue;
fields_hasher.update(tree.getNodeSource(member_node));
var member = switch (try containerMember(&block_scope, &namespace.base, &wip_members, member_node)) {
.decl => continue,
.field => |field| field,
};
member.convertToNonTupleLike(astgen.tree.nodes);
assert(member.comptime_token == null);
assert(member.ast.type_expr == 0);
assert(member.ast.align_expr == 0);
const field_name = try astgen.identAsString(member.ast.main_token);
wip_members.appendToField(@intFromEnum(field_name));
const gop = try duplicate_names.getOrPut(field_name);
if (gop.found_existing) {
try gop.value_ptr.append(sfba_allocator, member.ast.main_token);
any_duplicate = true;
} else {
gop.value_ptr.* = .{};
try gop.value_ptr.append(sfba_allocator, member.ast.main_token);
}
const doc_comment_index = try astgen.docCommentAsString(member.firstToken());
wip_members.appendToField(@intFromEnum(doc_comment_index));
const have_value = member.ast.value_expr != 0;
wip_members.nextField(bits_per_field, .{have_value});
if (have_value) {
if (arg_inst == .none) {
return astgen.failNodeNotes(
node,
"explicitly valued enum missing integer tag type",
.{},
&[_]u32{
try astgen.errNoteNode(
member.ast.value_expr,
"tag value specified here",
.{},
),
},
);
}
const tag_value_inst = try expr(&block_scope, &namespace.base, .{ .rl = .{ .ty = arg_inst } }, member.ast.value_expr);
wip_members.appendToField(@intFromEnum(tag_value_inst));
}
}
if (any_duplicate) {
var it = duplicate_names.iterator();
while (it.next()) |entry| {
const record = entry.value_ptr.*;
if (record.items.len > 1) {
var error_notes = std.ArrayList(u32).init(astgen.arena);
for (record.items[1..]) |duplicate| {
try error_notes.append(try astgen.errNoteTok(duplicate, "duplicate field here", .{}));
}
try error_notes.append(try astgen.errNoteNode(node, "enum declared here", .{}));
try astgen.appendErrorTokNotes(
record.items[0],
"duplicate enum field name",
.{},
error_notes.items,
);
}
}
return error.AnalysisFail;
}
if (!block_scope.isEmpty()) {
_ = try block_scope.addBreak(.break_inline, decl_inst, .void_value);
}
var fields_hash: std.zig.SrcHash = undefined;
fields_hasher.final(&fields_hash);
const body = block_scope.instructionsSlice();
const body_len = astgen.countBodyLenAfterFixups(body);
try gz.setEnum(decl_inst, .{
.src_node = node,
.nonexhaustive = nonexhaustive,
.tag_type = arg_inst,
.captures_len = @intCast(namespace.captures.count()),
.body_len = body_len,
.fields_len = @intCast(counts.total_fields),
.decls_len = @intCast(counts.decls),
.fields_hash = fields_hash,
});
wip_members.finishBits(bits_per_field);
const decls_slice = wip_members.declsSlice();
const fields_slice = wip_members.fieldsSlice();
try astgen.extra.ensureUnusedCapacity(gpa, namespace.captures.count() + decls_slice.len + body_len + fields_slice.len);
astgen.extra.appendSliceAssumeCapacity(@ptrCast(namespace.captures.keys()));
astgen.extra.appendSliceAssumeCapacity(decls_slice);
astgen.appendBodyWithFixups(body);
astgen.extra.appendSliceAssumeCapacity(fields_slice);
block_scope.unstack();
return rvalue(gz, ri, decl_inst.toRef(), node);
},
.keyword_opaque => {
assert(container_decl.ast.arg == 0);
const decl_inst = try gz.reserveInstructionIndex();
var namespace: Scope.Namespace = .{
.parent = scope,
.node = node,
.inst = decl_inst,
.declaring_gz = gz,
.maybe_generic = astgen.within_fn,
};
defer namespace.deinit(gpa);
astgen.advanceSourceCursorToNode(node);
var block_scope: GenZir = .{
.parent = &namespace.base,
.decl_node_index = node,
.decl_line = gz.decl_line,
.astgen = astgen,
.is_comptime = true,
.instructions = gz.instructions,
.instructions_top = gz.instructions.items.len,
};
defer block_scope.unstack();
const decl_count = try astgen.scanDecls(&namespace, container_decl.ast.members);
var wip_members = try WipMembers.init(gpa, &astgen.scratch, decl_count, 0, 0, 0);
defer wip_members.deinit();
for (container_decl.ast.members) |member_node| {
const res = try containerMember(&block_scope, &namespace.base, &wip_members, member_node);
if (res == .field) {
return astgen.failNode(member_node, "opaque types cannot have fields", .{});
}
}
try gz.setOpaque(decl_inst, .{
.src_node = node,
.captures_len = @intCast(namespace.captures.count()),
.decls_len = decl_count,
});
wip_members.finishBits(0);
const decls_slice = wip_members.declsSlice();
try astgen.extra.ensureUnusedCapacity(gpa, namespace.captures.count() + decls_slice.len);
astgen.extra.appendSliceAssumeCapacity(@ptrCast(namespace.captures.keys()));
astgen.extra.appendSliceAssumeCapacity(decls_slice);
block_scope.unstack();
return rvalue(gz, ri, decl_inst.toRef(), node);
},
else => unreachable,
}
}
const ContainerMemberResult = union(enum) { decl, field: Ast.full.ContainerField };
fn containerMember(
gz: *GenZir,
scope: *Scope,
wip_members: *WipMembers,
member_node: Ast.Node.Index,
) InnerError!ContainerMemberResult {
const astgen = gz.astgen;
const tree = astgen.tree;
const node_tags = tree.nodes.items(.tag);
const node_datas = tree.nodes.items(.data);
switch (node_tags[member_node]) {
.container_field_init,
.container_field_align,
.container_field,
=> return ContainerMemberResult{ .field = tree.fullContainerField(member_node).? },
.fn_proto,
.fn_proto_multi,
.fn_proto_one,
.fn_proto_simple,
.fn_decl,
=> {
var buf: [1]Ast.Node.Index = undefined;
const full = tree.fullFnProto(&buf, member_node).?;
const body = if (node_tags[member_node] == .fn_decl) node_datas[member_node].rhs else 0;
astgen.fnDecl(gz, scope, wip_members, member_node, body, full) catch |err| switch (err) {
error.OutOfMemory => return error.OutOfMemory,
error.AnalysisFail => {},
};
},
.global_var_decl,
.local_var_decl,
.simple_var_decl,
.aligned_var_decl,
=> {
astgen.globalVarDecl(gz, scope, wip_members, member_node, tree.fullVarDecl(member_node).?) catch |err| switch (err) {
error.OutOfMemory => return error.OutOfMemory,
error.AnalysisFail => {},
};
},
.@"comptime" => {
astgen.comptimeDecl(gz, scope, wip_members, member_node) catch |err| switch (err) {
error.OutOfMemory => return error.OutOfMemory,
error.AnalysisFail => {},
};
},
.@"usingnamespace" => {
astgen.usingnamespaceDecl(gz, scope, wip_members, member_node) catch |err| switch (err) {
error.OutOfMemory => return error.OutOfMemory,
error.AnalysisFail => {},
};
},
.test_decl => {
astgen.testDecl(gz, scope, wip_members, member_node) catch |err| switch (err) {
error.OutOfMemory => return error.OutOfMemory,
error.AnalysisFail => {},
};
},
else => unreachable,
}
return .decl;
}
fn errorSetDecl(gz: *GenZir, ri: ResultInfo, node: Ast.Node.Index) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const gpa = astgen.gpa;
const tree = astgen.tree;
const main_tokens = tree.nodes.items(.main_token);
const token_tags = tree.tokens.items(.tag);
const payload_index = try reserveExtra(astgen, @typeInfo(Zir.Inst.ErrorSetDecl).Struct.fields.len);
var fields_len: usize = 0;
{
var idents: std.AutoHashMapUnmanaged(Zir.NullTerminatedString, Ast.TokenIndex) = .{};
defer idents.deinit(gpa);
const error_token = main_tokens[node];
var tok_i = error_token + 2;
while (true) : (tok_i += 1) {
switch (token_tags[tok_i]) {
.doc_comment, .comma => {},
.identifier => {
const str_index = try astgen.identAsString(tok_i);
const gop = try idents.getOrPut(gpa, str_index);
if (gop.found_existing) {
const name = try gpa.dupe(u8, mem.span(astgen.nullTerminatedString(str_index)));
defer gpa.free(name);
return astgen.failTokNotes(
tok_i,
"duplicate error set field '{s}'",
.{name},
&[_]u32{
try astgen.errNoteTok(
gop.value_ptr.*,
"previous declaration here",
.{},
),
},
);
}
gop.value_ptr.* = tok_i;
try astgen.extra.ensureUnusedCapacity(gpa, 2);
astgen.extra.appendAssumeCapacity(@intFromEnum(str_index));
const doc_comment_index = try astgen.docCommentAsString(tok_i);
astgen.extra.appendAssumeCapacity(@intFromEnum(doc_comment_index));
fields_len += 1;
},
.r_brace => break,
else => unreachable,
}
}
}
setExtra(astgen, payload_index, Zir.Inst.ErrorSetDecl{
.fields_len = @intCast(fields_len),
});
const result = try gz.addPlNodePayloadIndex(.error_set_decl, node, payload_index);
return rvalue(gz, ri, result, node);
}
fn tryExpr(
parent_gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
operand_node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = parent_gz.astgen;
const fn_block = astgen.fn_block orelse {
return astgen.failNode(node, "'try' outside function scope", .{});
};
if (parent_gz.any_defer_node != 0) {
return astgen.failNodeNotes(node, "'try' not allowed inside defer expression", .{}, &.{
try astgen.errNoteNode(
parent_gz.any_defer_node,
"defer expression here",
.{},
),
});
}
// Ensure debug line/column information is emitted for this try expression.
// Then we will save the line/column so that we can emit another one that goes
// "backwards" because we want to evaluate the operand, but then put the debug
// info back at the try keyword for error return tracing.
if (!parent_gz.is_comptime) {
try emitDbgNode(parent_gz, node);
}
const try_lc = LineColumn{ astgen.source_line - parent_gz.decl_line, astgen.source_column };
const operand_ri: ResultInfo = switch (ri.rl) {
.ref, .ref_coerced_ty => .{ .rl = .ref, .ctx = .error_handling_expr },
else => .{ .rl = .none, .ctx = .error_handling_expr },
};
// This could be a pointer or value depending on the `ri` parameter.
const operand = try reachableExpr(parent_gz, scope, operand_ri, operand_node, node);
const block_tag: Zir.Inst.Tag = if (operand_ri.rl == .ref) .try_ptr else .@"try";
const try_inst = try parent_gz.makeBlockInst(block_tag, node);
try parent_gz.instructions.append(astgen.gpa, try_inst);
var else_scope = parent_gz.makeSubBlock(scope);
defer else_scope.unstack();
const err_tag = switch (ri.rl) {
.ref, .ref_coerced_ty => Zir.Inst.Tag.err_union_code_ptr,
else => Zir.Inst.Tag.err_union_code,
};
const err_code = try else_scope.addUnNode(err_tag, operand, node);
try genDefers(&else_scope, &fn_block.base, scope, .{ .both = err_code });
try emitDbgStmt(&else_scope, try_lc);
_ = try else_scope.addUnNode(.ret_node, err_code, node);
try else_scope.setTryBody(try_inst, operand);
const result = try_inst.toRef();
switch (ri.rl) {
.ref, .ref_coerced_ty => return result,
else => return rvalue(parent_gz, ri, result, node),
}
}
fn orelseCatchExpr(
parent_gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
lhs: Ast.Node.Index,
cond_op: Zir.Inst.Tag,
unwrap_op: Zir.Inst.Tag,
unwrap_code_op: Zir.Inst.Tag,
rhs: Ast.Node.Index,
payload_token: ?Ast.TokenIndex,
) InnerError!Zir.Inst.Ref {
const astgen = parent_gz.astgen;
const tree = astgen.tree;
const need_rl = astgen.nodes_need_rl.contains(node);
const block_ri: ResultInfo = if (need_rl) ri else .{
.rl = switch (ri.rl) {
.ptr => .{ .ty = (try ri.rl.resultType(parent_gz, node)).? },
.inferred_ptr => .none,
else => ri.rl,
},
.ctx = ri.ctx,
};
// We need to call `rvalue` to write through to the pointer only if we had a
// result pointer and aren't forwarding it.
const LocTag = @typeInfo(ResultInfo.Loc).Union.tag_type.?;
const need_result_rvalue = @as(LocTag, block_ri.rl) != @as(LocTag, ri.rl);
const do_err_trace = astgen.fn_block != null and (cond_op == .is_non_err or cond_op == .is_non_err_ptr);
var block_scope = parent_gz.makeSubBlock(scope);
block_scope.setBreakResultInfo(block_ri);
defer block_scope.unstack();
const operand_ri: ResultInfo = switch (block_scope.break_result_info.rl) {
.ref, .ref_coerced_ty => .{ .rl = .ref, .ctx = if (do_err_trace) .error_handling_expr else .none },
else => .{ .rl = .none, .ctx = if (do_err_trace) .error_handling_expr else .none },
};
// This could be a pointer or value depending on the `operand_ri` parameter.
// We cannot use `block_scope.break_result_info` because that has the bare
// type, whereas this expression has the optional type. Later we make
// up for this fact by calling rvalue on the else branch.
const operand = try reachableExpr(&block_scope, &block_scope.base, operand_ri, lhs, rhs);
const cond = try block_scope.addUnNode(cond_op, operand, node);
const condbr = try block_scope.addCondBr(.condbr, node);
const block = try parent_gz.makeBlockInst(.block, node);
try block_scope.setBlockBody(block);
// block_scope unstacked now, can add new instructions to parent_gz
try parent_gz.instructions.append(astgen.gpa, block);
var then_scope = block_scope.makeSubBlock(scope);
defer then_scope.unstack();
// This could be a pointer or value depending on `unwrap_op`.
const unwrapped_payload = try then_scope.addUnNode(unwrap_op, operand, node);
const then_result = switch (ri.rl) {
.ref, .ref_coerced_ty => unwrapped_payload,
else => try rvalue(&then_scope, block_scope.break_result_info, unwrapped_payload, node),
};
_ = try then_scope.addBreakWithSrcNode(.@"break", block, then_result, node);
var else_scope = block_scope.makeSubBlock(scope);
defer else_scope.unstack();
// We know that the operand (almost certainly) modified the error return trace,
// so signal to Sema that it should save the new index for restoring later.
if (do_err_trace and nodeMayAppendToErrorTrace(tree, lhs))
_ = try else_scope.addSaveErrRetIndex(.always);
var err_val_scope: Scope.LocalVal = undefined;
const else_sub_scope = blk: {
const payload = payload_token orelse break :blk &else_scope.base;
const err_str = tree.tokenSlice(payload);
if (mem.eql(u8, err_str, "_")) {
return astgen.failTok(payload, "discard of error capture; omit it instead", .{});
}
const err_name = try astgen.identAsString(payload);
try astgen.detectLocalShadowing(scope, err_name, payload, err_str, .capture);
err_val_scope = .{
.parent = &else_scope.base,
.gen_zir = &else_scope,
.name = err_name,
.inst = try else_scope.addUnNode(unwrap_code_op, operand, node),
.token_src = payload,
.id_cat = .capture,
};
break :blk &err_val_scope.base;
};
const else_result = try fullBodyExpr(&else_scope, else_sub_scope, block_scope.break_result_info, rhs);
if (!else_scope.endsWithNoReturn()) {
// As our last action before the break, "pop" the error trace if needed
if (do_err_trace)
try restoreErrRetIndex(&else_scope, .{ .block = block }, block_scope.break_result_info, rhs, else_result);
_ = try else_scope.addBreakWithSrcNode(.@"break", block, else_result, rhs);
}
try checkUsed(parent_gz, &else_scope.base, else_sub_scope);
try setCondBrPayload(condbr, cond, &then_scope, &else_scope);
if (need_result_rvalue) {
return rvalue(parent_gz, ri, block.toRef(), node);
} else {
return block.toRef();
}
}
/// Return whether the identifier names of two tokens are equal. Resolves @""
/// tokens without allocating.
/// OK in theory it could do it without allocating. This implementation
/// allocates when the @"" form is used.
fn tokenIdentEql(astgen: *AstGen, token1: Ast.TokenIndex, token2: Ast.TokenIndex) !bool {
const ident_name_1 = try astgen.identifierTokenString(token1);
const ident_name_2 = try astgen.identifierTokenString(token2);
return mem.eql(u8, ident_name_1, ident_name_2);
}
fn fieldAccess(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
switch (ri.rl) {
.ref, .ref_coerced_ty => return addFieldAccess(.field_ptr, gz, scope, .{ .rl = .ref }, node),
else => {
const access = try addFieldAccess(.field_val, gz, scope, .{ .rl = .none }, node);
return rvalue(gz, ri, access, node);
},
}
}
fn addFieldAccess(
tag: Zir.Inst.Tag,
gz: *GenZir,
scope: *Scope,
lhs_ri: ResultInfo,
node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const main_tokens = tree.nodes.items(.main_token);
const node_datas = tree.nodes.items(.data);
const object_node = node_datas[node].lhs;
const dot_token = main_tokens[node];
const field_ident = dot_token + 1;
const str_index = try astgen.identAsString(field_ident);
const lhs = try expr(gz, scope, lhs_ri, object_node);
const cursor = maybeAdvanceSourceCursorToMainToken(gz, node);
try emitDbgStmt(gz, cursor);
return gz.addPlNode(tag, node, Zir.Inst.Field{
.lhs = lhs,
.field_name_start = str_index,
});
}
fn arrayAccess(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const tree = gz.astgen.tree;
const node_datas = tree.nodes.items(.data);
switch (ri.rl) {
.ref, .ref_coerced_ty => {
const lhs = try expr(gz, scope, .{ .rl = .ref }, node_datas[node].lhs);
const cursor = maybeAdvanceSourceCursorToMainToken(gz, node);
const rhs = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, node_datas[node].rhs);
try emitDbgStmt(gz, cursor);
return gz.addPlNode(.elem_ptr_node, node, Zir.Inst.Bin{ .lhs = lhs, .rhs = rhs });
},
else => {
const lhs = try expr(gz, scope, .{ .rl = .none }, node_datas[node].lhs);
const cursor = maybeAdvanceSourceCursorToMainToken(gz, node);
const rhs = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .usize_type } }, node_datas[node].rhs);
try emitDbgStmt(gz, cursor);
return rvalue(gz, ri, try gz.addPlNode(.elem_val_node, node, Zir.Inst.Bin{ .lhs = lhs, .rhs = rhs }), node);
},
}
}
fn simpleBinOp(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
op_inst_tag: Zir.Inst.Tag,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
if (op_inst_tag == .cmp_neq or op_inst_tag == .cmp_eq) {
const node_tags = tree.nodes.items(.tag);
const str = if (op_inst_tag == .cmp_eq) "==" else "!=";
if (node_tags[node_datas[node].lhs] == .string_literal or
node_tags[node_datas[node].rhs] == .string_literal)
return astgen.failNode(node, "cannot compare strings with {s}", .{str});
}
const lhs = try reachableExpr(gz, scope, .{ .rl = .none }, node_datas[node].lhs, node);
const cursor = switch (op_inst_tag) {
.add, .sub, .mul, .div, .mod_rem => maybeAdvanceSourceCursorToMainToken(gz, node),
else => undefined,
};
const rhs = try reachableExpr(gz, scope, .{ .rl = .none }, node_datas[node].rhs, node);
switch (op_inst_tag) {
.add, .sub, .mul, .div, .mod_rem => {
try emitDbgStmt(gz, cursor);
},
else => {},
}
const result = try gz.addPlNode(op_inst_tag, node, Zir.Inst.Bin{ .lhs = lhs, .rhs = rhs });
return rvalue(gz, ri, result, node);
}
fn simpleStrTok(
gz: *GenZir,
ri: ResultInfo,
ident_token: Ast.TokenIndex,
node: Ast.Node.Index,
op_inst_tag: Zir.Inst.Tag,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const str_index = try astgen.identAsString(ident_token);
const result = try gz.addStrTok(op_inst_tag, str_index, ident_token);
return rvalue(gz, ri, result, node);
}
fn boolBinOp(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
zir_tag: Zir.Inst.Tag,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const lhs = try expr(gz, scope, coerced_bool_ri, node_datas[node].lhs);
const bool_br = (try gz.addPlNodePayloadIndex(zir_tag, node, undefined)).toIndex().?;
var rhs_scope = gz.makeSubBlock(scope);
defer rhs_scope.unstack();
const rhs = try fullBodyExpr(&rhs_scope, &rhs_scope.base, coerced_bool_ri, node_datas[node].rhs);
if (!gz.refIsNoReturn(rhs)) {
_ = try rhs_scope.addBreakWithSrcNode(.break_inline, bool_br, rhs, node_datas[node].rhs);
}
try rhs_scope.setBoolBrBody(bool_br, lhs);
const block_ref = bool_br.toRef();
return rvalue(gz, ri, block_ref, node);
}
fn ifExpr(
parent_gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
if_full: Ast.full.If,
) InnerError!Zir.Inst.Ref {
const astgen = parent_gz.astgen;
const tree = astgen.tree;
const token_tags = tree.tokens.items(.tag);
const do_err_trace = astgen.fn_block != null and if_full.error_token != null;
const need_rl = astgen.nodes_need_rl.contains(node);
const block_ri: ResultInfo = if (need_rl) ri else .{
.rl = switch (ri.rl) {
.ptr => .{ .ty = (try ri.rl.resultType(parent_gz, node)).? },
.inferred_ptr => .none,
else => ri.rl,
},
.ctx = ri.ctx,
};
// We need to call `rvalue` to write through to the pointer only if we had a
// result pointer and aren't forwarding it.
const LocTag = @typeInfo(ResultInfo.Loc).Union.tag_type.?;
const need_result_rvalue = @as(LocTag, block_ri.rl) != @as(LocTag, ri.rl);
var block_scope = parent_gz.makeSubBlock(scope);
block_scope.setBreakResultInfo(block_ri);
defer block_scope.unstack();
const payload_is_ref = if (if_full.payload_token) |payload_token|
token_tags[payload_token] == .asterisk
else
false;
try emitDbgNode(parent_gz, if_full.ast.cond_expr);
const cond: struct {
inst: Zir.Inst.Ref,
bool_bit: Zir.Inst.Ref,
} = c: {
if (if_full.error_token) |_| {
const cond_ri: ResultInfo = .{ .rl = if (payload_is_ref) .ref else .none, .ctx = .error_handling_expr };
const err_union = try expr(&block_scope, &block_scope.base, cond_ri, if_full.ast.cond_expr);
const tag: Zir.Inst.Tag = if (payload_is_ref) .is_non_err_ptr else .is_non_err;
break :c .{
.inst = err_union,
.bool_bit = try block_scope.addUnNode(tag, err_union, if_full.ast.cond_expr),
};
} else if (if_full.payload_token) |_| {
const cond_ri: ResultInfo = .{ .rl = if (payload_is_ref) .ref else .none };
const optional = try expr(&block_scope, &block_scope.base, cond_ri, if_full.ast.cond_expr);
const tag: Zir.Inst.Tag = if (payload_is_ref) .is_non_null_ptr else .is_non_null;
break :c .{
.inst = optional,
.bool_bit = try block_scope.addUnNode(tag, optional, if_full.ast.cond_expr),
};
} else {
const cond = try expr(&block_scope, &block_scope.base, coerced_bool_ri, if_full.ast.cond_expr);
break :c .{
.inst = cond,
.bool_bit = cond,
};
}
};
const condbr = try block_scope.addCondBr(.condbr, node);
const block = try parent_gz.makeBlockInst(.block, node);
try block_scope.setBlockBody(block);
// block_scope unstacked now, can add new instructions to parent_gz
try parent_gz.instructions.append(astgen.gpa, block);
var then_scope = parent_gz.makeSubBlock(scope);
defer then_scope.unstack();
var payload_val_scope: Scope.LocalVal = undefined;
const then_node = if_full.ast.then_expr;
const then_sub_scope = s: {
if (if_full.error_token != null) {
if (if_full.payload_token) |payload_token| {
const tag: Zir.Inst.Tag = if (payload_is_ref)
.err_union_payload_unsafe_ptr
else
.err_union_payload_unsafe;
const payload_inst = try then_scope.addUnNode(tag, cond.inst, then_node);
const token_name_index = payload_token + @intFromBool(payload_is_ref);
const ident_name = try astgen.identAsString(token_name_index);
const token_name_str = tree.tokenSlice(token_name_index);
if (mem.eql(u8, "_", token_name_str))
break :s &then_scope.base;
try astgen.detectLocalShadowing(&then_scope.base, ident_name, token_name_index, token_name_str, .capture);
payload_val_scope = .{
.parent = &then_scope.base,
.gen_zir = &then_scope,
.name = ident_name,
.inst = payload_inst,
.token_src = token_name_index,
.id_cat = .capture,
};
try then_scope.addDbgVar(.dbg_var_val, ident_name, payload_inst);
break :s &payload_val_scope.base;
} else {
_ = try then_scope.addUnNode(.ensure_err_union_payload_void, cond.inst, node);
break :s &then_scope.base;
}
} else if (if_full.payload_token) |payload_token| {
const ident_token = if (payload_is_ref) payload_token + 1 else payload_token;
const tag: Zir.Inst.Tag = if (payload_is_ref)
.optional_payload_unsafe_ptr
else
.optional_payload_unsafe;
const ident_bytes = tree.tokenSlice(ident_token);
if (mem.eql(u8, "_", ident_bytes))
break :s &then_scope.base;
const payload_inst = try then_scope.addUnNode(tag, cond.inst, then_node);
const ident_name = try astgen.identAsString(ident_token);
try astgen.detectLocalShadowing(&then_scope.base, ident_name, ident_token, ident_bytes, .capture);
payload_val_scope = .{
.parent = &then_scope.base,
.gen_zir = &then_scope,
.name = ident_name,
.inst = payload_inst,
.token_src = ident_token,
.id_cat = .capture,
};
try then_scope.addDbgVar(.dbg_var_val, ident_name, payload_inst);
break :s &payload_val_scope.base;
} else {
break :s &then_scope.base;
}
};
const then_result = try fullBodyExpr(&then_scope, then_sub_scope, block_scope.break_result_info, then_node);
try checkUsed(parent_gz, &then_scope.base, then_sub_scope);
if (!then_scope.endsWithNoReturn()) {
_ = try then_scope.addBreakWithSrcNode(.@"break", block, then_result, then_node);
}
var else_scope = parent_gz.makeSubBlock(scope);
defer else_scope.unstack();
// We know that the operand (almost certainly) modified the error return trace,
// so signal to Sema that it should save the new index for restoring later.
if (do_err_trace and nodeMayAppendToErrorTrace(tree, if_full.ast.cond_expr))
_ = try else_scope.addSaveErrRetIndex(.always);
const else_node = if_full.ast.else_expr;
if (else_node != 0) {
const sub_scope = s: {
if (if_full.error_token) |error_token| {
const tag: Zir.Inst.Tag = if (payload_is_ref)
.err_union_code_ptr
else
.err_union_code;
const payload_inst = try else_scope.addUnNode(tag, cond.inst, if_full.ast.cond_expr);
const ident_name = try astgen.identAsString(error_token);
const error_token_str = tree.tokenSlice(error_token);
if (mem.eql(u8, "_", error_token_str))
break :s &else_scope.base;
try astgen.detectLocalShadowing(&else_scope.base, ident_name, error_token, error_token_str, .capture);
payload_val_scope = .{
.parent = &else_scope.base,
.gen_zir = &else_scope,
.name = ident_name,
.inst = payload_inst,
.token_src = error_token,
.id_cat = .capture,
};
try else_scope.addDbgVar(.dbg_var_val, ident_name, payload_inst);
break :s &payload_val_scope.base;
} else {
break :s &else_scope.base;
}
};
const else_result = try fullBodyExpr(&else_scope, sub_scope, block_scope.break_result_info, else_node);
if (!else_scope.endsWithNoReturn()) {
// As our last action before the break, "pop" the error trace if needed
if (do_err_trace)
try restoreErrRetIndex(&else_scope, .{ .block = block }, block_scope.break_result_info, else_node, else_result);
_ = try else_scope.addBreakWithSrcNode(.@"break", block, else_result, else_node);
}
try checkUsed(parent_gz, &else_scope.base, sub_scope);
} else {
const result = try rvalue(&else_scope, ri, .void_value, node);
_ = try else_scope.addBreak(.@"break", block, result);
}
try setCondBrPayload(condbr, cond.bool_bit, &then_scope, &else_scope);
if (need_result_rvalue) {
return rvalue(parent_gz, ri, block.toRef(), node);
} else {
return block.toRef();
}
}
/// Supports `else_scope` stacked on `then_scope`. Unstacks `else_scope` then `then_scope`.
fn setCondBrPayload(
condbr: Zir.Inst.Index,
cond: Zir.Inst.Ref,
then_scope: *GenZir,
else_scope: *GenZir,
) !void {
defer then_scope.unstack();
defer else_scope.unstack();
const astgen = then_scope.astgen;
const then_body = then_scope.instructionsSliceUpto(else_scope);
const else_body = else_scope.instructionsSlice();
const then_body_len = astgen.countBodyLenAfterFixups(then_body);
const else_body_len = astgen.countBodyLenAfterFixups(else_body);
try astgen.extra.ensureUnusedCapacity(
astgen.gpa,
@typeInfo(Zir.Inst.CondBr).Struct.fields.len + then_body_len + else_body_len,
);
const zir_datas = astgen.instructions.items(.data);
zir_datas[@intFromEnum(condbr)].pl_node.payload_index = astgen.addExtraAssumeCapacity(Zir.Inst.CondBr{
.condition = cond,
.then_body_len = then_body_len,
.else_body_len = else_body_len,
});
astgen.appendBodyWithFixups(then_body);
astgen.appendBodyWithFixups(else_body);
}
fn whileExpr(
parent_gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
while_full: Ast.full.While,
is_statement: bool,
) InnerError!Zir.Inst.Ref {
const astgen = parent_gz.astgen;
const tree = astgen.tree;
const token_tags = tree.tokens.items(.tag);
const need_rl = astgen.nodes_need_rl.contains(node);
const block_ri: ResultInfo = if (need_rl) ri else .{
.rl = switch (ri.rl) {
.ptr => .{ .ty = (try ri.rl.resultType(parent_gz, node)).? },
.inferred_ptr => .none,
else => ri.rl,
},
.ctx = ri.ctx,
};
// We need to call `rvalue` to write through to the pointer only if we had a
// result pointer and aren't forwarding it.
const LocTag = @typeInfo(ResultInfo.Loc).Union.tag_type.?;
const need_result_rvalue = @as(LocTag, block_ri.rl) != @as(LocTag, ri.rl);
if (while_full.label_token) |label_token| {
try astgen.checkLabelRedefinition(scope, label_token);
}
const is_inline = while_full.inline_token != null;
if (parent_gz.is_comptime and is_inline) {
return astgen.failTok(while_full.inline_token.?, "redundant inline keyword in comptime scope", .{});
}
const loop_tag: Zir.Inst.Tag = if (is_inline) .block_inline else .loop;
const loop_block = try parent_gz.makeBlockInst(loop_tag, node);
try parent_gz.instructions.append(astgen.gpa, loop_block);
var loop_scope = parent_gz.makeSubBlock(scope);
loop_scope.is_inline = is_inline;
loop_scope.setBreakResultInfo(block_ri);
defer loop_scope.unstack();
var cond_scope = parent_gz.makeSubBlock(&loop_scope.base);
defer cond_scope.unstack();
const payload_is_ref = if (while_full.payload_token) |payload_token|
token_tags[payload_token] == .asterisk
else
false;
try emitDbgNode(parent_gz, while_full.ast.cond_expr);
const cond: struct {
inst: Zir.Inst.Ref,
bool_bit: Zir.Inst.Ref,
} = c: {
if (while_full.error_token) |_| {
const cond_ri: ResultInfo = .{ .rl = if (payload_is_ref) .ref else .none };
const err_union = try fullBodyExpr(&cond_scope, &cond_scope.base, cond_ri, while_full.ast.cond_expr);
const tag: Zir.Inst.Tag = if (payload_is_ref) .is_non_err_ptr else .is_non_err;
break :c .{
.inst = err_union,
.bool_bit = try cond_scope.addUnNode(tag, err_union, while_full.ast.cond_expr),
};
} else if (while_full.payload_token) |_| {
const cond_ri: ResultInfo = .{ .rl = if (payload_is_ref) .ref else .none };
const optional = try fullBodyExpr(&cond_scope, &cond_scope.base, cond_ri, while_full.ast.cond_expr);
const tag: Zir.Inst.Tag = if (payload_is_ref) .is_non_null_ptr else .is_non_null;
break :c .{
.inst = optional,
.bool_bit = try cond_scope.addUnNode(tag, optional, while_full.ast.cond_expr),
};
} else {
const cond = try fullBodyExpr(&cond_scope, &cond_scope.base, coerced_bool_ri, while_full.ast.cond_expr);
break :c .{
.inst = cond,
.bool_bit = cond,
};
}
};
const condbr_tag: Zir.Inst.Tag = if (is_inline) .condbr_inline else .condbr;
const condbr = try cond_scope.addCondBr(condbr_tag, node);
const block_tag: Zir.Inst.Tag = if (is_inline) .block_inline else .block;
const cond_block = try loop_scope.makeBlockInst(block_tag, node);
try cond_scope.setBlockBody(cond_block);
// cond_scope unstacked now, can add new instructions to loop_scope
try loop_scope.instructions.append(astgen.gpa, cond_block);
// make scope now but don't stack on parent_gz until loop_scope
// gets unstacked after cont_expr is emitted and added below
var then_scope = parent_gz.makeSubBlock(&cond_scope.base);
then_scope.instructions_top = GenZir.unstacked_top;
defer then_scope.unstack();
var dbg_var_name: Zir.NullTerminatedString = .empty;
var dbg_var_inst: Zir.Inst.Ref = undefined;
var opt_payload_inst: Zir.Inst.OptionalIndex = .none;
var payload_val_scope: Scope.LocalVal = undefined;
const then_sub_scope = s: {
if (while_full.error_token != null) {
if (while_full.payload_token) |payload_token| {
const tag: Zir.Inst.Tag = if (payload_is_ref)
.err_union_payload_unsafe_ptr
else
.err_union_payload_unsafe;
// will add this instruction to then_scope.instructions below
const payload_inst = try then_scope.makeUnNode(tag, cond.inst, while_full.ast.cond_expr);
opt_payload_inst = payload_inst.toOptional();
const ident_token = payload_token + @intFromBool(payload_is_ref);
const ident_bytes = tree.tokenSlice(ident_token);
if (mem.eql(u8, "_", ident_bytes))
break :s &then_scope.base;
const ident_name = try astgen.identAsString(ident_token);
try astgen.detectLocalShadowing(&then_scope.base, ident_name, ident_token, ident_bytes, .capture);
payload_val_scope = .{
.parent = &then_scope.base,
.gen_zir = &then_scope,
.name = ident_name,
.inst = payload_inst.toRef(),
.token_src = ident_token,
.id_cat = .capture,
};
dbg_var_name = ident_name;
dbg_var_inst = payload_inst.toRef();
break :s &payload_val_scope.base;
} else {
_ = try then_scope.addUnNode(.ensure_err_union_payload_void, cond.inst, node);
break :s &then_scope.base;
}
} else if (while_full.payload_token) |payload_token| {
const ident_token = if (payload_is_ref) payload_token + 1 else payload_token;
const tag: Zir.Inst.Tag = if (payload_is_ref)
.optional_payload_unsafe_ptr
else
.optional_payload_unsafe;
// will add this instruction to then_scope.instructions below
const payload_inst = try then_scope.makeUnNode(tag, cond.inst, while_full.ast.cond_expr);
opt_payload_inst = payload_inst.toOptional();
const ident_name = try astgen.identAsString(ident_token);
const ident_bytes = tree.tokenSlice(ident_token);
if (mem.eql(u8, "_", ident_bytes))
break :s &then_scope.base;
try astgen.detectLocalShadowing(&then_scope.base, ident_name, ident_token, ident_bytes, .capture);
payload_val_scope = .{
.parent = &then_scope.base,
.gen_zir = &then_scope,
.name = ident_name,
.inst = payload_inst.toRef(),
.token_src = ident_token,
.id_cat = .capture,
};
dbg_var_name = ident_name;
dbg_var_inst = payload_inst.toRef();
break :s &payload_val_scope.base;
} else {
break :s &then_scope.base;
}
};
var continue_scope = parent_gz.makeSubBlock(then_sub_scope);
continue_scope.instructions_top = GenZir.unstacked_top;
defer continue_scope.unstack();
const continue_block = try then_scope.makeBlockInst(block_tag, node);
const repeat_tag: Zir.Inst.Tag = if (is_inline) .repeat_inline else .repeat;
_ = try loop_scope.addNode(repeat_tag, node);
try loop_scope.setBlockBody(loop_block);
loop_scope.break_block = loop_block.toOptional();
loop_scope.continue_block = continue_block.toOptional();
if (while_full.label_token) |label_token| {
loop_scope.label = .{
.token = label_token,
.block_inst = loop_block,
};
}
// done adding instructions to loop_scope, can now stack then_scope
then_scope.instructions_top = then_scope.instructions.items.len;
const then_node = while_full.ast.then_expr;
if (opt_payload_inst.unwrap()) |payload_inst| {
try then_scope.instructions.append(astgen.gpa, payload_inst);
}
if (dbg_var_name != .empty) try then_scope.addDbgVar(.dbg_var_val, dbg_var_name, dbg_var_inst);
try then_scope.instructions.append(astgen.gpa, continue_block);
// This code could be improved to avoid emitting the continue expr when there
// are no jumps to it. This happens when the last statement of a while body is noreturn
// and there are no `continue` statements.
// Tracking issue: https://github.com/ziglang/zig/issues/9185
if (while_full.ast.cont_expr != 0) {
_ = try unusedResultExpr(&then_scope, then_sub_scope, while_full.ast.cont_expr);
}
continue_scope.instructions_top = continue_scope.instructions.items.len;
{
try emitDbgNode(&continue_scope, then_node);
const unused_result = try fullBodyExpr(&continue_scope, &continue_scope.base, .{ .rl = .none }, then_node);
_ = try addEnsureResult(&continue_scope, unused_result, then_node);
}
try checkUsed(parent_gz, &then_scope.base, then_sub_scope);
const break_tag: Zir.Inst.Tag = if (is_inline) .break_inline else .@"break";
if (!continue_scope.endsWithNoReturn()) {
_ = try continue_scope.addBreak(break_tag, continue_block, .void_value);
}
try continue_scope.setBlockBody(continue_block);
_ = try then_scope.addBreak(break_tag, cond_block, .void_value);
var else_scope = parent_gz.makeSubBlock(&cond_scope.base);
defer else_scope.unstack();
const else_node = while_full.ast.else_expr;
if (else_node != 0) {
const sub_scope = s: {
if (while_full.error_token) |error_token| {
const tag: Zir.Inst.Tag = if (payload_is_ref)
.err_union_code_ptr
else
.err_union_code;
const else_payload_inst = try else_scope.addUnNode(tag, cond.inst, while_full.ast.cond_expr);
const ident_name = try astgen.identAsString(error_token);
const ident_bytes = tree.tokenSlice(error_token);
if (mem.eql(u8, ident_bytes, "_"))
break :s &else_scope.base;
try astgen.detectLocalShadowing(&else_scope.base, ident_name, error_token, ident_bytes, .capture);
payload_val_scope = .{
.parent = &else_scope.base,
.gen_zir = &else_scope,
.name = ident_name,
.inst = else_payload_inst,
.token_src = error_token,
.id_cat = .capture,
};
try else_scope.addDbgVar(.dbg_var_val, ident_name, else_payload_inst);
break :s &payload_val_scope.base;
} else {
break :s &else_scope.base;
}
};
// Remove the continue block and break block so that `continue` and `break`
// control flow apply to outer loops; not this one.
loop_scope.continue_block = .none;
loop_scope.break_block = .none;
const else_result = try fullBodyExpr(&else_scope, sub_scope, loop_scope.break_result_info, else_node);
if (is_statement) {
_ = try addEnsureResult(&else_scope, else_result, else_node);
}
try checkUsed(parent_gz, &else_scope.base, sub_scope);
if (!else_scope.endsWithNoReturn()) {
_ = try else_scope.addBreakWithSrcNode(break_tag, loop_block, else_result, else_node);
}
} else {
const result = try rvalue(&else_scope, ri, .void_value, node);
_ = try else_scope.addBreak(break_tag, loop_block, result);
}
if (loop_scope.label) |some| {
if (!some.used) {
try astgen.appendErrorTok(some.token, "unused while loop label", .{});
}
}
try setCondBrPayload(condbr, cond.bool_bit, &then_scope, &else_scope);
const result = if (need_result_rvalue)
try rvalue(parent_gz, ri, loop_block.toRef(), node)
else
loop_block.toRef();
if (is_statement) {
_ = try parent_gz.addUnNode(.ensure_result_used, result, node);
}
return result;
}
fn forExpr(
parent_gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
for_full: Ast.full.For,
is_statement: bool,
) InnerError!Zir.Inst.Ref {
const astgen = parent_gz.astgen;
if (for_full.label_token) |label_token| {
try astgen.checkLabelRedefinition(scope, label_token);
}
const need_rl = astgen.nodes_need_rl.contains(node);
const block_ri: ResultInfo = if (need_rl) ri else .{
.rl = switch (ri.rl) {
.ptr => .{ .ty = (try ri.rl.resultType(parent_gz, node)).? },
.inferred_ptr => .none,
else => ri.rl,
},
.ctx = ri.ctx,
};
// We need to call `rvalue` to write through to the pointer only if we had a
// result pointer and aren't forwarding it.
const LocTag = @typeInfo(ResultInfo.Loc).Union.tag_type.?;
const need_result_rvalue = @as(LocTag, block_ri.rl) != @as(LocTag, ri.rl);
const is_inline = for_full.inline_token != null;
if (parent_gz.is_comptime and is_inline) {
return astgen.failTok(for_full.inline_token.?, "redundant inline keyword in comptime scope", .{});
}
const tree = astgen.tree;
const token_tags = tree.tokens.items(.tag);
const node_tags = tree.nodes.items(.tag);
const node_data = tree.nodes.items(.data);
const gpa = astgen.gpa;
// For counters, this is the start value; for indexables, this is the base
// pointer that can be used with elem_ptr and similar instructions.
// Special value `none` means that this is a counter and its start value is
// zero, indicating that the main index counter can be used directly.
const indexables = try gpa.alloc(Zir.Inst.Ref, for_full.ast.inputs.len);
defer gpa.free(indexables);
// elements of this array can be `none`, indicating no length check.
const lens = try gpa.alloc(Zir.Inst.Ref, for_full.ast.inputs.len);
defer gpa.free(lens);
// We will use a single zero-based counter no matter how many indexables there are.
const index_ptr = blk: {
const alloc_tag: Zir.Inst.Tag = if (is_inline) .alloc_comptime_mut else .alloc;
const index_ptr = try parent_gz.addUnNode(alloc_tag, .usize_type, node);
// initialize to zero
_ = try parent_gz.addPlNode(.store_node, node, Zir.Inst.Bin{
.lhs = index_ptr,
.rhs = .zero_usize,
});
break :blk index_ptr;
};
var any_len_checks = false;
{
var capture_token = for_full.payload_token;
for (for_full.ast.inputs, indexables, lens) |input, *indexable_ref, *len_ref| {
const capture_is_ref = token_tags[capture_token] == .asterisk;
const ident_tok = capture_token + @intFromBool(capture_is_ref);
const is_discard = mem.eql(u8, tree.tokenSlice(ident_tok), "_");
if (is_discard and capture_is_ref) {
return astgen.failTok(capture_token, "pointer modifier invalid on discard", .{});
}
// Skip over the comma, and on to the next capture (or the ending pipe character).
capture_token = ident_tok + 2;
try emitDbgNode(parent_gz, input);
if (node_tags[input] == .for_range) {
if (capture_is_ref) {
return astgen.failTok(ident_tok, "cannot capture reference to range", .{});
}
const start_node = node_data[input].lhs;
const start_val = try expr(parent_gz, scope, .{ .rl = .{ .ty = .usize_type } }, start_node);
const end_node = node_data[input].rhs;
const end_val = if (end_node != 0)
try expr(parent_gz, scope, .{ .rl = .{ .ty = .usize_type } }, node_data[input].rhs)
else
.none;
if (end_val == .none and is_discard) {
return astgen.failTok(ident_tok, "discard of unbounded counter", .{});
}
const start_is_zero = nodeIsTriviallyZero(tree, start_node);
const range_len = if (end_val == .none or start_is_zero)
end_val
else
try parent_gz.addPlNode(.sub, input, Zir.Inst.Bin{
.lhs = end_val,
.rhs = start_val,
});
any_len_checks = any_len_checks or range_len != .none;
indexable_ref.* = if (start_is_zero) .none else start_val;
len_ref.* = range_len;
} else {
const indexable = try expr(parent_gz, scope, .{ .rl = .none }, input);
any_len_checks = true;
indexable_ref.* = indexable;
len_ref.* = indexable;
}
}
}
if (!any_len_checks) {
return astgen.failNode(node, "unbounded for loop", .{});
}
// We use a dedicated ZIR instruction to assert the lengths to assist with
// nicer error reporting as well as fewer ZIR bytes emitted.
const len: Zir.Inst.Ref = len: {
const lens_len: u32 = @intCast(lens.len);
try astgen.extra.ensureUnusedCapacity(gpa, @typeInfo(Zir.Inst.MultiOp).Struct.fields.len + lens_len);
const len = try parent_gz.addPlNode(.for_len, node, Zir.Inst.MultiOp{
.operands_len = lens_len,
});
appendRefsAssumeCapacity(astgen, lens);
break :len len;
};
const loop_tag: Zir.Inst.Tag = if (is_inline) .block_inline else .loop;
const loop_block = try parent_gz.makeBlockInst(loop_tag, node);
try parent_gz.instructions.append(gpa, loop_block);
var loop_scope = parent_gz.makeSubBlock(scope);
loop_scope.is_inline = is_inline;
loop_scope.setBreakResultInfo(block_ri);
defer loop_scope.unstack();
// We need to finish loop_scope later once we have the deferred refs from then_scope. However, the
// load must be removed from instructions in the meantime or it appears to be part of parent_gz.
const index = try loop_scope.addUnNode(.load, index_ptr, node);
_ = loop_scope.instructions.pop();
var cond_scope = parent_gz.makeSubBlock(&loop_scope.base);
defer cond_scope.unstack();
// Check the condition.
const cond = try cond_scope.addPlNode(.cmp_lt, node, Zir.Inst.Bin{
.lhs = index,
.rhs = len,
});
const condbr_tag: Zir.Inst.Tag = if (is_inline) .condbr_inline else .condbr;
const condbr = try cond_scope.addCondBr(condbr_tag, node);
const block_tag: Zir.Inst.Tag = if (is_inline) .block_inline else .block;
const cond_block = try loop_scope.makeBlockInst(block_tag, node);
try cond_scope.setBlockBody(cond_block);
loop_scope.break_block = loop_block.toOptional();
loop_scope.continue_block = cond_block.toOptional();
if (for_full.label_token) |label_token| {
loop_scope.label = .{
.token = label_token,
.block_inst = loop_block,
};
}
const then_node = for_full.ast.then_expr;
var then_scope = parent_gz.makeSubBlock(&cond_scope.base);
defer then_scope.unstack();
const capture_scopes = try gpa.alloc(Scope.LocalVal, for_full.ast.inputs.len);
defer gpa.free(capture_scopes);
const then_sub_scope = blk: {
var capture_token = for_full.payload_token;
var capture_sub_scope: *Scope = &then_scope.base;
for (for_full.ast.inputs, indexables, capture_scopes) |input, indexable_ref, *capture_scope| {
const capture_is_ref = token_tags[capture_token] == .asterisk;
const ident_tok = capture_token + @intFromBool(capture_is_ref);
const capture_name = tree.tokenSlice(ident_tok);
// Skip over the comma, and on to the next capture (or the ending pipe character).
capture_token = ident_tok + 2;
if (mem.eql(u8, capture_name, "_")) continue;
const name_str_index = try astgen.identAsString(ident_tok);
try astgen.detectLocalShadowing(capture_sub_scope, name_str_index, ident_tok, capture_name, .capture);
const capture_inst = inst: {
const is_counter = node_tags[input] == .for_range;
if (indexable_ref == .none) {
// Special case: the main index can be used directly.
assert(is_counter);
assert(!capture_is_ref);
break :inst index;
}
// For counters, we add the index variable to the start value; for
// indexables, we use it as an element index. This is so similar
// that they can share the same code paths, branching only on the
// ZIR tag.
const switch_cond = (@as(u2, @intFromBool(capture_is_ref)) << 1) | @intFromBool(is_counter);
const tag: Zir.Inst.Tag = switch (switch_cond) {
0b00 => .elem_val,
0b01 => .add,
0b10 => .elem_ptr,
0b11 => unreachable, // compile error emitted already
};
break :inst try then_scope.addPlNode(tag, input, Zir.Inst.Bin{
.lhs = indexable_ref,
.rhs = index,
});
};
capture_scope.* = .{
.parent = capture_sub_scope,
.gen_zir = &then_scope,
.name = name_str_index,
.inst = capture_inst,
.token_src = ident_tok,
.id_cat = .capture,
};
try then_scope.addDbgVar(.dbg_var_val, name_str_index, capture_inst);
capture_sub_scope = &capture_scope.base;
}
break :blk capture_sub_scope;
};
const then_result = try fullBodyExpr(&then_scope, then_sub_scope, .{ .rl = .none }, then_node);
_ = try addEnsureResult(&then_scope, then_result, then_node);
try checkUsed(parent_gz, &then_scope.base, then_sub_scope);
const break_tag: Zir.Inst.Tag = if (is_inline) .break_inline else .@"break";
_ = try then_scope.addBreak(break_tag, cond_block, .void_value);
var else_scope = parent_gz.makeSubBlock(&cond_scope.base);
defer else_scope.unstack();
const else_node = for_full.ast.else_expr;
if (else_node != 0) {
const sub_scope = &else_scope.base;
// Remove the continue block and break block so that `continue` and `break`
// control flow apply to outer loops; not this one.
loop_scope.continue_block = .none;
loop_scope.break_block = .none;
const else_result = try fullBodyExpr(&else_scope, sub_scope, loop_scope.break_result_info, else_node);
if (is_statement) {
_ = try addEnsureResult(&else_scope, else_result, else_node);
}
if (!else_scope.endsWithNoReturn()) {
_ = try else_scope.addBreakWithSrcNode(break_tag, loop_block, else_result, else_node);
}
} else {
const result = try rvalue(&else_scope, ri, .void_value, node);
_ = try else_scope.addBreak(break_tag, loop_block, result);
}
if (loop_scope.label) |some| {
if (!some.used) {
try astgen.appendErrorTok(some.token, "unused for loop label", .{});
}
}
try setCondBrPayload(condbr, cond, &then_scope, &else_scope);
// then_block and else_block unstacked now, can resurrect loop_scope to finally finish it
{
loop_scope.instructions_top = loop_scope.instructions.items.len;
try loop_scope.instructions.appendSlice(gpa, &.{ index.toIndex().?, cond_block });
// Increment the index variable.
const index_plus_one = try loop_scope.addPlNode(.add_unsafe, node, Zir.Inst.Bin{
.lhs = index,
.rhs = .one_usize,
});
_ = try loop_scope.addPlNode(.store_node, node, Zir.Inst.Bin{
.lhs = index_ptr,
.rhs = index_plus_one,
});
const repeat_tag: Zir.Inst.Tag = if (is_inline) .repeat_inline else .repeat;
_ = try loop_scope.addNode(repeat_tag, node);
try loop_scope.setBlockBody(loop_block);
}
const result = if (need_result_rvalue)
try rvalue(parent_gz, ri, loop_block.toRef(), node)
else
loop_block.toRef();
if (is_statement) {
_ = try parent_gz.addUnNode(.ensure_result_used, result, node);
}
return result;
}
fn switchExprErrUnion(
parent_gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
catch_or_if_node: Ast.Node.Index,
node_ty: enum { @"catch", @"if" },
) InnerError!Zir.Inst.Ref {
const astgen = parent_gz.astgen;
const gpa = astgen.gpa;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const node_tags = tree.nodes.items(.tag);
const main_tokens = tree.nodes.items(.main_token);
const token_tags = tree.tokens.items(.tag);
const if_full = switch (node_ty) {
.@"catch" => undefined,
.@"if" => tree.fullIf(catch_or_if_node).?,
};
const switch_node, const operand_node, const error_payload = switch (node_ty) {
.@"catch" => .{
node_datas[catch_or_if_node].rhs,
node_datas[catch_or_if_node].lhs,
main_tokens[catch_or_if_node] + 2,
},
.@"if" => .{
if_full.ast.else_expr,
if_full.ast.cond_expr,
if_full.error_token.?,
},
};
assert(node_tags[switch_node] == .@"switch" or node_tags[switch_node] == .switch_comma);
const do_err_trace = astgen.fn_block != null;
const extra = tree.extraData(node_datas[switch_node].rhs, Ast.Node.SubRange);
const case_nodes = tree.extra_data[extra.start..extra.end];
const need_rl = astgen.nodes_need_rl.contains(catch_or_if_node);
const block_ri: ResultInfo = if (need_rl) ri else .{
.rl = switch (ri.rl) {
.ptr => .{ .ty = (try ri.rl.resultType(parent_gz, catch_or_if_node)).? },
.inferred_ptr => .none,
else => ri.rl,
},
.ctx = ri.ctx,
};
const payload_is_ref = node_ty == .@"if" and
if_full.payload_token != null and token_tags[if_full.payload_token.?] == .asterisk;
// We need to call `rvalue` to write through to the pointer only if we had a
// result pointer and aren't forwarding it.
const LocTag = @typeInfo(ResultInfo.Loc).Union.tag_type.?;
const need_result_rvalue = @as(LocTag, block_ri.rl) != @as(LocTag, ri.rl);
var scalar_cases_len: u32 = 0;
var multi_cases_len: u32 = 0;
var inline_cases_len: u32 = 0;
var has_else = false;
var else_node: Ast.Node.Index = 0;
var else_src: ?Ast.TokenIndex = null;
for (case_nodes) |case_node| {
const case = tree.fullSwitchCase(case_node).?;
if (case.ast.values.len == 0) {
const case_src = case.ast.arrow_token - 1;
if (else_src) |src| {
return astgen.failTokNotes(
case_src,
"multiple else prongs in switch expression",
.{},
&[_]u32{
try astgen.errNoteTok(
src,
"previous else prong here",
.{},
),
},
);
}
has_else = true;
else_node = case_node;
else_src = case_src;
continue;
} else if (case.ast.values.len == 1 and
node_tags[case.ast.values[0]] == .identifier and
mem.eql(u8, tree.tokenSlice(main_tokens[case.ast.values[0]]), "_"))
{
const case_src = case.ast.arrow_token - 1;
return astgen.failTokNotes(
case_src,
"'_' prong is not allowed when switching on errors",
.{},
&[_]u32{
try astgen.errNoteTok(
case_src,
"consider using 'else'",
.{},
),
},
);
}
for (case.ast.values) |val| {
if (node_tags[val] == .string_literal)
return astgen.failNode(val, "cannot switch on strings", .{});
}
if (case.ast.values.len == 1 and node_tags[case.ast.values[0]] != .switch_range) {
scalar_cases_len += 1;
} else {
multi_cases_len += 1;
}
if (case.inline_token != null) {
inline_cases_len += 1;
}
}
const operand_ri: ResultInfo = .{
.rl = if (payload_is_ref) .ref else .none,
.ctx = .error_handling_expr,
};
astgen.advanceSourceCursorToNode(operand_node);
const operand_lc = LineColumn{ astgen.source_line - parent_gz.decl_line, astgen.source_column };
const raw_operand = try reachableExpr(parent_gz, scope, operand_ri, operand_node, switch_node);
const item_ri: ResultInfo = .{ .rl = .none };
// This contains the data that goes into the `extra` array for the SwitchBlockErrUnion, except
// the first cases_nodes.len slots are a table that indexes payloads later in the array,
// with the non-error and else case indices coming first, then scalar_cases_len indexes, then
// multi_cases_len indexes
const payloads = &astgen.scratch;
const scratch_top = astgen.scratch.items.len;
const case_table_start = scratch_top;
const scalar_case_table = case_table_start + 1 + @intFromBool(has_else);
const multi_case_table = scalar_case_table + scalar_cases_len;
const case_table_end = multi_case_table + multi_cases_len;
try astgen.scratch.resize(gpa, case_table_end);
defer astgen.scratch.items.len = scratch_top;
var block_scope = parent_gz.makeSubBlock(scope);
// block_scope not used for collecting instructions
block_scope.instructions_top = GenZir.unstacked_top;
block_scope.setBreakResultInfo(block_ri);
// Sema expects a dbg_stmt immediately before switch_block_err_union
try emitDbgStmtForceCurrentIndex(parent_gz, operand_lc);
// This gets added to the parent block later, after the item expressions.
const switch_block = try parent_gz.makeBlockInst(.switch_block_err_union, switch_node);
// We re-use this same scope for all cases, including the special prong, if any.
var case_scope = parent_gz.makeSubBlock(&block_scope.base);
case_scope.instructions_top = GenZir.unstacked_top;
{
const body_len_index: u32 = @intCast(payloads.items.len);
payloads.items[case_table_start] = body_len_index;
try payloads.resize(gpa, body_len_index + 1); // body_len
case_scope.instructions_top = parent_gz.instructions.items.len;
defer case_scope.unstack();
const unwrap_payload_tag: Zir.Inst.Tag = if (payload_is_ref)
.err_union_payload_unsafe_ptr
else
.err_union_payload_unsafe;
const unwrapped_payload = try case_scope.addUnNode(
unwrap_payload_tag,
raw_operand,
catch_or_if_node,
);
switch (node_ty) {
.@"catch" => {
const case_result = switch (ri.rl) {
.ref, .ref_coerced_ty => unwrapped_payload,
else => try rvalue(
&case_scope,
block_scope.break_result_info,
unwrapped_payload,
catch_or_if_node,
),
};
_ = try case_scope.addBreakWithSrcNode(
.@"break",
switch_block,
case_result,
catch_or_if_node,
);
},
.@"if" => {
var payload_val_scope: Scope.LocalVal = undefined;
const then_node = if_full.ast.then_expr;
const then_sub_scope = s: {
assert(if_full.error_token != null);
if (if_full.payload_token) |payload_token| {
const token_name_index = payload_token + @intFromBool(payload_is_ref);
const ident_name = try astgen.identAsString(token_name_index);
const token_name_str = tree.tokenSlice(token_name_index);
if (mem.eql(u8, "_", token_name_str))
break :s &case_scope.base;
try astgen.detectLocalShadowing(
&case_scope.base,
ident_name,
token_name_index,
token_name_str,
.capture,
);
payload_val_scope = .{
.parent = &case_scope.base,
.gen_zir = &case_scope,
.name = ident_name,
.inst = unwrapped_payload,
.token_src = token_name_index,
.id_cat = .capture,
};
try case_scope.addDbgVar(.dbg_var_val, ident_name, unwrapped_payload);
break :s &payload_val_scope.base;
} else {
_ = try case_scope.addUnNode(
.ensure_err_union_payload_void,
raw_operand,
catch_or_if_node,
);
break :s &case_scope.base;
}
};
const then_result = try expr(
&case_scope,
then_sub_scope,
block_scope.break_result_info,
then_node,
);
try checkUsed(parent_gz, &case_scope.base, then_sub_scope);
if (!case_scope.endsWithNoReturn()) {
_ = try case_scope.addBreakWithSrcNode(
.@"break",
switch_block,
then_result,
then_node,
);
}
},
}
const case_slice = case_scope.instructionsSlice();
// Since we use the switch_block_err_union instruction itself to refer
// to the capture, which will not be added to the child block, we need
// to handle ref_table manually.
const refs_len = refs: {
var n: usize = 0;
var check_inst = switch_block;
while (astgen.ref_table.get(check_inst)) |ref_inst| {
n += 1;
check_inst = ref_inst;
}
break :refs n;
};
const body_len = refs_len + astgen.countBodyLenAfterFixups(case_slice);
try payloads.ensureUnusedCapacity(gpa, body_len);
const capture: Zir.Inst.SwitchBlock.ProngInfo.Capture = switch (node_ty) {
.@"catch" => .none,
.@"if" => if (if_full.payload_token == null)
.none
else if (payload_is_ref)
.by_ref
else
.by_val,
};
payloads.items[body_len_index] = @bitCast(Zir.Inst.SwitchBlock.ProngInfo{
.body_len = @intCast(body_len),
.capture = capture,
.is_inline = false,
.has_tag_capture = false,
});
if (astgen.ref_table.fetchRemove(switch_block)) |kv| {
appendPossiblyRefdBodyInst(astgen, payloads, kv.value);
}
appendBodyWithFixupsArrayList(astgen, payloads, case_slice);
}
const err_name = blk: {
const err_str = tree.tokenSlice(error_payload);
if (mem.eql(u8, err_str, "_")) {
return astgen.failTok(error_payload, "discard of error capture; omit it instead", .{});
}
const err_name = try astgen.identAsString(error_payload);
try astgen.detectLocalShadowing(scope, err_name, error_payload, err_str, .capture);
break :blk err_name;
};
// allocate a shared dummy instruction for the error capture
const err_inst = err_inst: {
const inst: Zir.Inst.Index = @enumFromInt(astgen.instructions.len);
try astgen.instructions.append(astgen.gpa, .{
.tag = .extended,
.data = .{ .extended = .{
.opcode = .value_placeholder,
.small = undefined,
.operand = undefined,
} },
});
break :err_inst inst;
};
// In this pass we generate all the item and prong expressions for error cases.
var multi_case_index: u32 = 0;
var scalar_case_index: u32 = 0;
var any_uses_err_capture = false;
for (case_nodes) |case_node| {
const case = tree.fullSwitchCase(case_node).?;
const is_multi_case = case.ast.values.len > 1 or
(case.ast.values.len == 1 and node_tags[case.ast.values[0]] == .switch_range);
var dbg_var_name: Zir.NullTerminatedString = .empty;
var dbg_var_inst: Zir.Inst.Ref = undefined;
var err_scope: Scope.LocalVal = undefined;
var capture_scope: Scope.LocalVal = undefined;
const sub_scope = blk: {
err_scope = .{
.parent = &case_scope.base,
.gen_zir = &case_scope,
.name = err_name,
.inst = err_inst.toRef(),
.token_src = error_payload,
.id_cat = .capture,
};
const capture_token = case.payload_token orelse break :blk &err_scope.base;
if (token_tags[capture_token] != .identifier) {
return astgen.failTok(capture_token + 1, "error set cannot be captured by reference", .{});
}
const capture_slice = tree.tokenSlice(capture_token);
if (mem.eql(u8, capture_slice, "_")) {
return astgen.failTok(capture_token, "discard of error capture; omit it instead", .{});
}
const tag_name = try astgen.identAsString(capture_token);
try astgen.detectLocalShadowing(&case_scope.base, tag_name, capture_token, capture_slice, .capture);
capture_scope = .{
.parent = &case_scope.base,
.gen_zir = &case_scope,
.name = tag_name,
.inst = switch_block.toRef(),
.token_src = capture_token,
.id_cat = .capture,
};
dbg_var_name = tag_name;
dbg_var_inst = switch_block.toRef();
err_scope.parent = &capture_scope.base;
break :blk &err_scope.base;
};
const header_index: u32 = @intCast(payloads.items.len);
const body_len_index = if (is_multi_case) blk: {
payloads.items[multi_case_table + multi_case_index] = header_index;
multi_case_index += 1;
try payloads.resize(gpa, header_index + 3); // items_len, ranges_len, body_len
// items
var items_len: u32 = 0;
for (case.ast.values) |item_node| {
if (node_tags[item_node] == .switch_range) continue;
items_len += 1;
const item_inst = try comptimeExpr(parent_gz, scope, item_ri, item_node);
try payloads.append(gpa, @intFromEnum(item_inst));
}
// ranges
var ranges_len: u32 = 0;
for (case.ast.values) |range| {
if (node_tags[range] != .switch_range) continue;
ranges_len += 1;
const first = try comptimeExpr(parent_gz, scope, item_ri, node_datas[range].lhs);
const last = try comptimeExpr(parent_gz, scope, item_ri, node_datas[range].rhs);
try payloads.appendSlice(gpa, &[_]u32{
@intFromEnum(first), @intFromEnum(last),
});
}
payloads.items[header_index] = items_len;
payloads.items[header_index + 1] = ranges_len;
break :blk header_index + 2;
} else if (case_node == else_node) blk: {
payloads.items[case_table_start + 1] = header_index;
try payloads.resize(gpa, header_index + 1); // body_len
break :blk header_index;
} else blk: {
payloads.items[scalar_case_table + scalar_case_index] = header_index;
scalar_case_index += 1;
try payloads.resize(gpa, header_index + 2); // item, body_len
const item_node = case.ast.values[0];
const item_inst = try comptimeExpr(parent_gz, scope, item_ri, item_node);
payloads.items[header_index] = @intFromEnum(item_inst);
break :blk header_index + 1;
};
{
// temporarily stack case_scope on parent_gz
case_scope.instructions_top = parent_gz.instructions.items.len;
defer case_scope.unstack();
if (do_err_trace and nodeMayAppendToErrorTrace(tree, operand_node))
_ = try case_scope.addSaveErrRetIndex(.always);
if (dbg_var_name != .empty) {
try case_scope.addDbgVar(.dbg_var_val, dbg_var_name, dbg_var_inst);
}
const target_expr_node = case.ast.target_expr;
const case_result = try fullBodyExpr(&case_scope, sub_scope, block_scope.break_result_info, target_expr_node);
// check capture_scope, not err_scope to avoid false positive unused error capture
try checkUsed(parent_gz, &case_scope.base, err_scope.parent);
const uses_err = err_scope.used != 0 or err_scope.discarded != 0;
if (uses_err) {
try case_scope.addDbgVar(.dbg_var_val, err_name, err_inst.toRef());
any_uses_err_capture = true;
}
if (!parent_gz.refIsNoReturn(case_result)) {
if (do_err_trace)
try restoreErrRetIndex(
&case_scope,
.{ .block = switch_block },
block_scope.break_result_info,
target_expr_node,
case_result,
);
_ = try case_scope.addBreakWithSrcNode(.@"break", switch_block, case_result, target_expr_node);
}
const case_slice = case_scope.instructionsSlice();
// Since we use the switch_block_err_union instruction itself to refer
// to the capture, which will not be added to the child block, we need
// to handle ref_table manually.
const refs_len = refs: {
var n: usize = 0;
var check_inst = switch_block;
while (astgen.ref_table.get(check_inst)) |ref_inst| {
n += 1;
check_inst = ref_inst;
}
if (uses_err) {
check_inst = err_inst;
while (astgen.ref_table.get(check_inst)) |ref_inst| {
n += 1;
check_inst = ref_inst;
}
}
break :refs n;
};
const body_len = refs_len + astgen.countBodyLenAfterFixups(case_slice);
try payloads.ensureUnusedCapacity(gpa, body_len);
payloads.items[body_len_index] = @bitCast(Zir.Inst.SwitchBlock.ProngInfo{
.body_len = @intCast(body_len),
.capture = if (case.payload_token != null) .by_val else .none,
.is_inline = case.inline_token != null,
.has_tag_capture = false,
});
if (astgen.ref_table.fetchRemove(switch_block)) |kv| {
appendPossiblyRefdBodyInst(astgen, payloads, kv.value);
}
if (uses_err) {
if (astgen.ref_table.fetchRemove(err_inst)) |kv| {
appendPossiblyRefdBodyInst(astgen, payloads, kv.value);
}
}
appendBodyWithFixupsArrayList(astgen, payloads, case_slice);
}
}
// Now that the item expressions are generated we can add this.
try parent_gz.instructions.append(gpa, switch_block);
try astgen.extra.ensureUnusedCapacity(gpa, @typeInfo(Zir.Inst.SwitchBlockErrUnion).Struct.fields.len +
@intFromBool(multi_cases_len != 0) +
payloads.items.len - case_table_end +
(case_table_end - case_table_start) * @typeInfo(Zir.Inst.As).Struct.fields.len);
const payload_index = astgen.addExtraAssumeCapacity(Zir.Inst.SwitchBlockErrUnion{
.operand = raw_operand,
.bits = Zir.Inst.SwitchBlockErrUnion.Bits{
.has_multi_cases = multi_cases_len != 0,
.has_else = has_else,
.scalar_cases_len = @intCast(scalar_cases_len),
.any_uses_err_capture = any_uses_err_capture,
.payload_is_ref = payload_is_ref,
},
.main_src_node_offset = parent_gz.nodeIndexToRelative(catch_or_if_node),
});
if (multi_cases_len != 0) {
astgen.extra.appendAssumeCapacity(multi_cases_len);
}
if (any_uses_err_capture) {
astgen.extra.appendAssumeCapacity(@intFromEnum(err_inst));
}
const zir_datas = astgen.instructions.items(.data);
zir_datas[@intFromEnum(switch_block)].pl_node.payload_index = payload_index;
for (payloads.items[case_table_start..case_table_end], 0..) |start_index, i| {
var body_len_index = start_index;
var end_index = start_index;
const table_index = case_table_start + i;
if (table_index < scalar_case_table) {
end_index += 1;
} else if (table_index < multi_case_table) {
body_len_index += 1;
end_index += 2;
} else {
body_len_index += 2;
const items_len = payloads.items[start_index];
const ranges_len = payloads.items[start_index + 1];
end_index += 3 + items_len + 2 * ranges_len;
}
const prong_info: Zir.Inst.SwitchBlock.ProngInfo = @bitCast(payloads.items[body_len_index]);
end_index += prong_info.body_len;
astgen.extra.appendSliceAssumeCapacity(payloads.items[start_index..end_index]);
}
if (need_result_rvalue) {
return rvalue(parent_gz, ri, switch_block.toRef(), switch_node);
} else {
return switch_block.toRef();
}
}
fn switchExpr(
parent_gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
switch_node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = parent_gz.astgen;
const gpa = astgen.gpa;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const node_tags = tree.nodes.items(.tag);
const main_tokens = tree.nodes.items(.main_token);
const token_tags = tree.tokens.items(.tag);
const operand_node = node_datas[switch_node].lhs;
const extra = tree.extraData(node_datas[switch_node].rhs, Ast.Node.SubRange);
const case_nodes = tree.extra_data[extra.start..extra.end];
const need_rl = astgen.nodes_need_rl.contains(switch_node);
const block_ri: ResultInfo = if (need_rl) ri else .{
.rl = switch (ri.rl) {
.ptr => .{ .ty = (try ri.rl.resultType(parent_gz, switch_node)).? },
.inferred_ptr => .none,
else => ri.rl,
},
.ctx = ri.ctx,
};
// We need to call `rvalue` to write through to the pointer only if we had a
// result pointer and aren't forwarding it.
const LocTag = @typeInfo(ResultInfo.Loc).Union.tag_type.?;
const need_result_rvalue = @as(LocTag, block_ri.rl) != @as(LocTag, ri.rl);
// We perform two passes over the AST. This first pass is to collect information
// for the following variables, make note of the special prong AST node index,
// and bail out with a compile error if there are multiple special prongs present.
var any_payload_is_ref = false;
var any_has_tag_capture = false;
var scalar_cases_len: u32 = 0;
var multi_cases_len: u32 = 0;
var inline_cases_len: u32 = 0;
var special_prong: Zir.SpecialProng = .none;
var special_node: Ast.Node.Index = 0;
var else_src: ?Ast.TokenIndex = null;
var underscore_src: ?Ast.TokenIndex = null;
for (case_nodes) |case_node| {
const case = tree.fullSwitchCase(case_node).?;
if (case.payload_token) |payload_token| {
const ident = if (token_tags[payload_token] == .asterisk) blk: {
any_payload_is_ref = true;
break :blk payload_token + 1;
} else payload_token;
if (token_tags[ident + 1] == .comma) {
any_has_tag_capture = true;
}
}
// Check for else/`_` prong.
if (case.ast.values.len == 0) {
const case_src = case.ast.arrow_token - 1;
if (else_src) |src| {
return astgen.failTokNotes(
case_src,
"multiple else prongs in switch expression",
.{},
&[_]u32{
try astgen.errNoteTok(
src,
"previous else prong here",
.{},
),
},
);
} else if (underscore_src) |some_underscore| {
return astgen.failNodeNotes(
switch_node,
"else and '_' prong in switch expression",
.{},
&[_]u32{
try astgen.errNoteTok(
case_src,
"else prong here",
.{},
),
try astgen.errNoteTok(
some_underscore,
"'_' prong here",
.{},
),
},
);
}
special_node = case_node;
special_prong = .@"else";
else_src = case_src;
continue;
} else if (case.ast.values.len == 1 and
node_tags[case.ast.values[0]] == .identifier and
mem.eql(u8, tree.tokenSlice(main_tokens[case.ast.values[0]]), "_"))
{
const case_src = case.ast.arrow_token - 1;
if (underscore_src) |src| {
return astgen.failTokNotes(
case_src,
"multiple '_' prongs in switch expression",
.{},
&[_]u32{
try astgen.errNoteTok(
src,
"previous '_' prong here",
.{},
),
},
);
} else if (else_src) |some_else| {
return astgen.failNodeNotes(
switch_node,
"else and '_' prong in switch expression",
.{},
&[_]u32{
try astgen.errNoteTok(
some_else,
"else prong here",
.{},
),
try astgen.errNoteTok(
case_src,
"'_' prong here",
.{},
),
},
);
}
if (case.inline_token != null) {
return astgen.failTok(case_src, "cannot inline '_' prong", .{});
}
special_node = case_node;
special_prong = .under;
underscore_src = case_src;
continue;
}
for (case.ast.values) |val| {
if (node_tags[val] == .string_literal)
return astgen.failNode(val, "cannot switch on strings", .{});
}
if (case.ast.values.len == 1 and node_tags[case.ast.values[0]] != .switch_range) {
scalar_cases_len += 1;
} else {
multi_cases_len += 1;
}
if (case.inline_token != null) {
inline_cases_len += 1;
}
}
const operand_ri: ResultInfo = .{ .rl = if (any_payload_is_ref) .ref else .none };
astgen.advanceSourceCursorToNode(operand_node);
const operand_lc = LineColumn{ astgen.source_line - parent_gz.decl_line, astgen.source_column };
const raw_operand = try expr(parent_gz, scope, operand_ri, operand_node);
const item_ri: ResultInfo = .{ .rl = .none };
// This contains the data that goes into the `extra` array for the SwitchBlock/SwitchBlockMulti,
// except the first cases_nodes.len slots are a table that indexes payloads later in the array, with
// the special case index coming first, then scalar_case_len indexes, then multi_cases_len indexes
const payloads = &astgen.scratch;
const scratch_top = astgen.scratch.items.len;
const case_table_start = scratch_top;
const scalar_case_table = case_table_start + @intFromBool(special_prong != .none);
const multi_case_table = scalar_case_table + scalar_cases_len;
const case_table_end = multi_case_table + multi_cases_len;
try astgen.scratch.resize(gpa, case_table_end);
defer astgen.scratch.items.len = scratch_top;
var block_scope = parent_gz.makeSubBlock(scope);
// block_scope not used for collecting instructions
block_scope.instructions_top = GenZir.unstacked_top;
block_scope.setBreakResultInfo(block_ri);
// Sema expects a dbg_stmt immediately before switch_block(_ref)
try emitDbgStmtForceCurrentIndex(parent_gz, operand_lc);
// This gets added to the parent block later, after the item expressions.
const switch_tag: Zir.Inst.Tag = if (any_payload_is_ref) .switch_block_ref else .switch_block;
const switch_block = try parent_gz.makeBlockInst(switch_tag, switch_node);
// We re-use this same scope for all cases, including the special prong, if any.
var case_scope = parent_gz.makeSubBlock(&block_scope.base);
case_scope.instructions_top = GenZir.unstacked_top;
// If any prong has an inline tag capture, allocate a shared dummy instruction for it
const tag_inst = if (any_has_tag_capture) tag_inst: {
const inst: Zir.Inst.Index = @enumFromInt(astgen.instructions.len);
try astgen.instructions.append(astgen.gpa, .{
.tag = .extended,
.data = .{ .extended = .{
.opcode = .value_placeholder,
.small = undefined,
.operand = undefined,
} },
});
break :tag_inst inst;
} else undefined;
// In this pass we generate all the item and prong expressions.
var multi_case_index: u32 = 0;
var scalar_case_index: u32 = 0;
for (case_nodes) |case_node| {
const case = tree.fullSwitchCase(case_node).?;
const is_multi_case = case.ast.values.len > 1 or
(case.ast.values.len == 1 and node_tags[case.ast.values[0]] == .switch_range);
var dbg_var_name: Zir.NullTerminatedString = .empty;
var dbg_var_inst: Zir.Inst.Ref = undefined;
var dbg_var_tag_name: Zir.NullTerminatedString = .empty;
var dbg_var_tag_inst: Zir.Inst.Ref = undefined;
var has_tag_capture = false;
var capture_val_scope: Scope.LocalVal = undefined;
var tag_scope: Scope.LocalVal = undefined;
var capture: Zir.Inst.SwitchBlock.ProngInfo.Capture = .none;
const sub_scope = blk: {
const payload_token = case.payload_token orelse break :blk &case_scope.base;
const ident = if (token_tags[payload_token] == .asterisk)
payload_token + 1
else
payload_token;
const is_ptr = ident != payload_token;
capture = if (is_ptr) .by_ref else .by_val;
const ident_slice = tree.tokenSlice(ident);
var payload_sub_scope: *Scope = undefined;
if (mem.eql(u8, ident_slice, "_")) {
if (is_ptr) {
return astgen.failTok(payload_token, "pointer modifier invalid on discard", .{});
}
payload_sub_scope = &case_scope.base;
} else {
const capture_name = try astgen.identAsString(ident);
try astgen.detectLocalShadowing(&case_scope.base, capture_name, ident, ident_slice, .capture);
capture_val_scope = .{
.parent = &case_scope.base,
.gen_zir = &case_scope,
.name = capture_name,
.inst = switch_block.toRef(),
.token_src = ident,
.id_cat = .capture,
};
dbg_var_name = capture_name;
dbg_var_inst = switch_block.toRef();
payload_sub_scope = &capture_val_scope.base;
}
const tag_token = if (token_tags[ident + 1] == .comma)
ident + 2
else
break :blk payload_sub_scope;
const tag_slice = tree.tokenSlice(tag_token);
if (mem.eql(u8, tag_slice, "_")) {
return astgen.failTok(tag_token, "discard of tag capture; omit it instead", .{});
} else if (case.inline_token == null) {
return astgen.failTok(tag_token, "tag capture on non-inline prong", .{});
}
const tag_name = try astgen.identAsString(tag_token);
try astgen.detectLocalShadowing(payload_sub_scope, tag_name, tag_token, tag_slice, .@"switch tag capture");
assert(any_has_tag_capture);
has_tag_capture = true;
tag_scope = .{
.parent = payload_sub_scope,
.gen_zir = &case_scope,
.name = tag_name,
.inst = tag_inst.toRef(),
.token_src = tag_token,
.id_cat = .@"switch tag capture",
};
dbg_var_tag_name = tag_name;
dbg_var_tag_inst = tag_inst.toRef();
break :blk &tag_scope.base;
};
const header_index: u32 = @intCast(payloads.items.len);
const body_len_index = if (is_multi_case) blk: {
payloads.items[multi_case_table + multi_case_index] = header_index;
multi_case_index += 1;
try payloads.resize(gpa, header_index + 3); // items_len, ranges_len, body_len
// items
var items_len: u32 = 0;
for (case.ast.values) |item_node| {
if (node_tags[item_node] == .switch_range) continue;
items_len += 1;
const item_inst = try comptimeExpr(parent_gz, scope, item_ri, item_node);
try payloads.append(gpa, @intFromEnum(item_inst));
}
// ranges
var ranges_len: u32 = 0;
for (case.ast.values) |range| {
if (node_tags[range] != .switch_range) continue;
ranges_len += 1;
const first = try comptimeExpr(parent_gz, scope, item_ri, node_datas[range].lhs);
const last = try comptimeExpr(parent_gz, scope, item_ri, node_datas[range].rhs);
try payloads.appendSlice(gpa, &[_]u32{
@intFromEnum(first), @intFromEnum(last),
});
}
payloads.items[header_index] = items_len;
payloads.items[header_index + 1] = ranges_len;
break :blk header_index + 2;
} else if (case_node == special_node) blk: {
payloads.items[case_table_start] = header_index;
try payloads.resize(gpa, header_index + 1); // body_len
break :blk header_index;
} else blk: {
payloads.items[scalar_case_table + scalar_case_index] = header_index;
scalar_case_index += 1;
try payloads.resize(gpa, header_index + 2); // item, body_len
const item_node = case.ast.values[0];
const item_inst = try comptimeExpr(parent_gz, scope, item_ri, item_node);
payloads.items[header_index] = @intFromEnum(item_inst);
break :blk header_index + 1;
};
{
// temporarily stack case_scope on parent_gz
case_scope.instructions_top = parent_gz.instructions.items.len;
defer case_scope.unstack();
if (dbg_var_name != .empty) {
try case_scope.addDbgVar(.dbg_var_val, dbg_var_name, dbg_var_inst);
}
if (dbg_var_tag_name != .empty) {
try case_scope.addDbgVar(.dbg_var_val, dbg_var_tag_name, dbg_var_tag_inst);
}
const target_expr_node = case.ast.target_expr;
const case_result = try fullBodyExpr(&case_scope, sub_scope, block_scope.break_result_info, target_expr_node);
try checkUsed(parent_gz, &case_scope.base, sub_scope);
if (!parent_gz.refIsNoReturn(case_result)) {
_ = try case_scope.addBreakWithSrcNode(.@"break", switch_block, case_result, target_expr_node);
}
const case_slice = case_scope.instructionsSlice();
// Since we use the switch_block instruction itself to refer to the
// capture, which will not be added to the child block, we need to
// handle ref_table manually, and the same for the inline tag
// capture instruction.
const refs_len = refs: {
var n: usize = 0;
var check_inst = switch_block;
while (astgen.ref_table.get(check_inst)) |ref_inst| {
n += 1;
check_inst = ref_inst;
}
if (has_tag_capture) {
check_inst = tag_inst;
while (astgen.ref_table.get(check_inst)) |ref_inst| {
n += 1;
check_inst = ref_inst;
}
}
break :refs n;
};
const body_len = refs_len + astgen.countBodyLenAfterFixups(case_slice);
try payloads.ensureUnusedCapacity(gpa, body_len);
payloads.items[body_len_index] = @bitCast(Zir.Inst.SwitchBlock.ProngInfo{
.body_len = @intCast(body_len),
.capture = capture,
.is_inline = case.inline_token != null,
.has_tag_capture = has_tag_capture,
});
if (astgen.ref_table.fetchRemove(switch_block)) |kv| {
appendPossiblyRefdBodyInst(astgen, payloads, kv.value);
}
if (has_tag_capture) {
if (astgen.ref_table.fetchRemove(tag_inst)) |kv| {
appendPossiblyRefdBodyInst(astgen, payloads, kv.value);
}
}
appendBodyWithFixupsArrayList(astgen, payloads, case_slice);
}
}
// Now that the item expressions are generated we can add this.
try parent_gz.instructions.append(gpa, switch_block);
try astgen.extra.ensureUnusedCapacity(gpa, @typeInfo(Zir.Inst.SwitchBlock).Struct.fields.len +
@intFromBool(multi_cases_len != 0) +
@intFromBool(any_has_tag_capture) +
payloads.items.len - case_table_end +
(case_table_end - case_table_start) * @typeInfo(Zir.Inst.As).Struct.fields.len);
const payload_index = astgen.addExtraAssumeCapacity(Zir.Inst.SwitchBlock{
.operand = raw_operand,
.bits = Zir.Inst.SwitchBlock.Bits{
.has_multi_cases = multi_cases_len != 0,
.has_else = special_prong == .@"else",
.has_under = special_prong == .under,
.any_has_tag_capture = any_has_tag_capture,
.scalar_cases_len = @intCast(scalar_cases_len),
},
});
if (multi_cases_len != 0) {
astgen.extra.appendAssumeCapacity(multi_cases_len);
}
if (any_has_tag_capture) {
astgen.extra.appendAssumeCapacity(@intFromEnum(tag_inst));
}
const zir_datas = astgen.instructions.items(.data);
zir_datas[@intFromEnum(switch_block)].pl_node.payload_index = payload_index;
for (payloads.items[case_table_start..case_table_end], 0..) |start_index, i| {
var body_len_index = start_index;
var end_index = start_index;
const table_index = case_table_start + i;
if (table_index < scalar_case_table) {
end_index += 1;
} else if (table_index < multi_case_table) {
body_len_index += 1;
end_index += 2;
} else {
body_len_index += 2;
const items_len = payloads.items[start_index];
const ranges_len = payloads.items[start_index + 1];
end_index += 3 + items_len + 2 * ranges_len;
}
const prong_info: Zir.Inst.SwitchBlock.ProngInfo = @bitCast(payloads.items[body_len_index]);
end_index += prong_info.body_len;
astgen.extra.appendSliceAssumeCapacity(payloads.items[start_index..end_index]);
}
if (need_result_rvalue) {
return rvalue(parent_gz, ri, switch_block.toRef(), switch_node);
} else {
return switch_block.toRef();
}
}
fn ret(gz: *GenZir, scope: *Scope, node: Ast.Node.Index) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const node_tags = tree.nodes.items(.tag);
if (astgen.fn_block == null) {
return astgen.failNode(node, "'return' outside function scope", .{});
}
if (gz.any_defer_node != 0) {
return astgen.failNodeNotes(node, "cannot return from defer expression", .{}, &.{
try astgen.errNoteNode(
gz.any_defer_node,
"defer expression here",
.{},
),
});
}
// Ensure debug line/column information is emitted for this return expression.
// Then we will save the line/column so that we can emit another one that goes
// "backwards" because we want to evaluate the operand, but then put the debug
// info back at the return keyword for error return tracing.
if (!gz.is_comptime) {
try emitDbgNode(gz, node);
}
const ret_lc = LineColumn{ astgen.source_line - gz.decl_line, astgen.source_column };
const defer_outer = &astgen.fn_block.?.base;
const operand_node = node_datas[node].lhs;
if (operand_node == 0) {
// Returning a void value; skip error defers.
try genDefers(gz, defer_outer, scope, .normal_only);
// As our last action before the return, "pop" the error trace if needed
_ = try gz.addRestoreErrRetIndex(.ret, .always, node);
_ = try gz.addUnNode(.ret_node, .void_value, node);
return Zir.Inst.Ref.unreachable_value;
}
if (node_tags[operand_node] == .error_value) {
// Hot path for `return error.Foo`. This bypasses result location logic as well as logic
// for detecting whether to add something to the function's inferred error set.
const ident_token = node_datas[operand_node].rhs;
const err_name_str_index = try astgen.identAsString(ident_token);
const defer_counts = countDefers(defer_outer, scope);
if (!defer_counts.need_err_code) {
try genDefers(gz, defer_outer, scope, .both_sans_err);
try emitDbgStmt(gz, ret_lc);
_ = try gz.addStrTok(.ret_err_value, err_name_str_index, ident_token);
return Zir.Inst.Ref.unreachable_value;
}
const err_code = try gz.addStrTok(.ret_err_value_code, err_name_str_index, ident_token);
try genDefers(gz, defer_outer, scope, .{ .both = err_code });
try emitDbgStmt(gz, ret_lc);
_ = try gz.addUnNode(.ret_node, err_code, node);
return Zir.Inst.Ref.unreachable_value;
}
const ri: ResultInfo = if (astgen.nodes_need_rl.contains(node)) .{
.rl = .{ .ptr = .{ .inst = try gz.addNode(.ret_ptr, node) } },
.ctx = .@"return",
} else .{
.rl = .{ .coerced_ty = astgen.fn_ret_ty },
.ctx = .@"return",
};
const prev_anon_name_strategy = gz.anon_name_strategy;
gz.anon_name_strategy = .func;
const operand = try reachableExpr(gz, scope, ri, operand_node, node);
gz.anon_name_strategy = prev_anon_name_strategy;
switch (nodeMayEvalToError(tree, operand_node)) {
.never => {
// Returning a value that cannot be an error; skip error defers.
try genDefers(gz, defer_outer, scope, .normal_only);
// As our last action before the return, "pop" the error trace if needed
_ = try gz.addRestoreErrRetIndex(.ret, .always, node);
try emitDbgStmt(gz, ret_lc);
try gz.addRet(ri, operand, node);
return Zir.Inst.Ref.unreachable_value;
},
.always => {
// Value is always an error. Emit both error defers and regular defers.
const err_code = if (ri.rl == .ptr) try gz.addUnNode(.load, ri.rl.ptr.inst, node) else operand;
try genDefers(gz, defer_outer, scope, .{ .both = err_code });
try emitDbgStmt(gz, ret_lc);
try gz.addRet(ri, operand, node);
return Zir.Inst.Ref.unreachable_value;
},
.maybe => {
const defer_counts = countDefers(defer_outer, scope);
if (!defer_counts.have_err) {
// Only regular defers; no branch needed.
try genDefers(gz, defer_outer, scope, .normal_only);
try emitDbgStmt(gz, ret_lc);
// As our last action before the return, "pop" the error trace if needed
const result = if (ri.rl == .ptr) try gz.addUnNode(.load, ri.rl.ptr.inst, node) else operand;
_ = try gz.addRestoreErrRetIndex(.ret, .{ .if_non_error = result }, node);
try gz.addRet(ri, operand, node);
return Zir.Inst.Ref.unreachable_value;
}
// Emit conditional branch for generating errdefers.
const result = if (ri.rl == .ptr) try gz.addUnNode(.load, ri.rl.ptr.inst, node) else operand;
const is_non_err = try gz.addUnNode(.ret_is_non_err, result, node);
const condbr = try gz.addCondBr(.condbr, node);
var then_scope = gz.makeSubBlock(scope);
defer then_scope.unstack();
try genDefers(&then_scope, defer_outer, scope, .normal_only);
// As our last action before the return, "pop" the error trace if needed
_ = try then_scope.addRestoreErrRetIndex(.ret, .always, node);
try emitDbgStmt(&then_scope, ret_lc);
try then_scope.addRet(ri, operand, node);
var else_scope = gz.makeSubBlock(scope);
defer else_scope.unstack();
const which_ones: DefersToEmit = if (!defer_counts.need_err_code) .both_sans_err else .{
.both = try else_scope.addUnNode(.err_union_code, result, node),
};
try genDefers(&else_scope, defer_outer, scope, which_ones);
try emitDbgStmt(&else_scope, ret_lc);
try else_scope.addRet(ri, operand, node);
try setCondBrPayload(condbr, is_non_err, &then_scope, &else_scope);
return Zir.Inst.Ref.unreachable_value;
},
}
}
/// Parses the string `buf` as a base 10 integer of type `u16`.
///
/// Unlike std.fmt.parseInt, does not allow the '_' character in `buf`.
fn parseBitCount(buf: []const u8) std.fmt.ParseIntError!u16 {
if (buf.len == 0) return error.InvalidCharacter;
var x: u16 = 0;
for (buf) |c| {
const digit = switch (c) {
'0'...'9' => c - '0',
else => return error.InvalidCharacter,
};
if (x != 0) x = try std.math.mul(u16, x, 10);
x = try std.math.add(u16, x, digit);
}
return x;
}
fn identifier(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
ident: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const main_tokens = tree.nodes.items(.main_token);
const ident_token = main_tokens[ident];
const ident_name_raw = tree.tokenSlice(ident_token);
if (mem.eql(u8, ident_name_raw, "_")) {
return astgen.failNode(ident, "'_' used as an identifier without @\"_\" syntax", .{});
}
// if not @"" syntax, just use raw token slice
if (ident_name_raw[0] != '@') {
if (primitive_instrs.get(ident_name_raw)) |zir_const_ref| {
return rvalue(gz, ri, zir_const_ref, ident);
}
if (ident_name_raw.len >= 2) integer: {
const first_c = ident_name_raw[0];
if (first_c == 'i' or first_c == 'u') {
const signedness: std.builtin.Signedness = switch (first_c == 'i') {
true => .signed,
false => .unsigned,
};
if (ident_name_raw.len >= 3 and ident_name_raw[1] == '0') {
return astgen.failNode(
ident,
"primitive integer type '{s}' has leading zero",
.{ident_name_raw},
);
}
const bit_count = parseBitCount(ident_name_raw[1..]) catch |err| switch (err) {
error.Overflow => return astgen.failNode(
ident,
"primitive integer type '{s}' exceeds maximum bit width of 65535",
.{ident_name_raw},
),
error.InvalidCharacter => break :integer,
};
const result = try gz.add(.{
.tag = .int_type,
.data = .{ .int_type = .{
.src_node = gz.nodeIndexToRelative(ident),
.signedness = signedness,
.bit_count = bit_count,
} },
});
return rvalue(gz, ri, result, ident);
}
}
}
// Local variables, including function parameters.
return localVarRef(gz, scope, ri, ident, ident_token);
}
fn localVarRef(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
ident: Ast.Node.Index,
ident_token: Ast.TokenIndex,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const name_str_index = try astgen.identAsString(ident_token);
var s = scope;
var found_already: ?Ast.Node.Index = null; // we have found a decl with the same name already
var found_needs_tunnel: bool = undefined; // defined when `found_already != null`
var found_namespaces_out: u32 = undefined; // defined when `found_already != null`
// The number of namespaces above `gz` we currently are
var num_namespaces_out: u32 = 0;
// defined by `num_namespaces_out != 0`
var capturing_namespace: *Scope.Namespace = undefined;
while (true) switch (s.tag) {
.local_val => {
const local_val = s.cast(Scope.LocalVal).?;
if (local_val.name == name_str_index) {
// Locals cannot shadow anything, so we do not need to look for ambiguous
// references in this case.
if (ri.rl == .discard and ri.ctx == .assignment) {
local_val.discarded = ident_token;
} else {
local_val.used = ident_token;
}
const value_inst = if (num_namespaces_out != 0) try tunnelThroughClosure(
gz,
ident,
num_namespaces_out,
.{ .ref = local_val.inst },
.{ .token = local_val.token_src },
) else local_val.inst;
return rvalueNoCoercePreRef(gz, ri, value_inst, ident);
}
s = local_val.parent;
},
.local_ptr => {
const local_ptr = s.cast(Scope.LocalPtr).?;
if (local_ptr.name == name_str_index) {
if (ri.rl == .discard and ri.ctx == .assignment) {
local_ptr.discarded = ident_token;
} else {
local_ptr.used = ident_token;
}
// Can't close over a runtime variable
if (num_namespaces_out != 0 and !local_ptr.maybe_comptime and !gz.is_typeof) {
const ident_name = try astgen.identifierTokenString(ident_token);
return astgen.failNodeNotes(ident, "mutable '{s}' not accessible from here", .{ident_name}, &.{
try astgen.errNoteTok(local_ptr.token_src, "declared mutable here", .{}),
try astgen.errNoteNode(capturing_namespace.node, "crosses namespace boundary here", .{}),
});
}
switch (ri.rl) {
.ref, .ref_coerced_ty => {
const ptr_inst = if (num_namespaces_out != 0) try tunnelThroughClosure(
gz,
ident,
num_namespaces_out,
.{ .ref = local_ptr.ptr },
.{ .token = local_ptr.token_src },
) else local_ptr.ptr;
local_ptr.used_as_lvalue = true;
return ptr_inst;
},
else => {
const val_inst = if (num_namespaces_out != 0) try tunnelThroughClosure(
gz,
ident,
num_namespaces_out,
.{ .ref_load = local_ptr.ptr },
.{ .token = local_ptr.token_src },
) else try gz.addUnNode(.load, local_ptr.ptr, ident);
return rvalueNoCoercePreRef(gz, ri, val_inst, ident);
},
}
}
s = local_ptr.parent;
},
.gen_zir => s = s.cast(GenZir).?.parent,
.defer_normal, .defer_error => s = s.cast(Scope.Defer).?.parent,
.namespace => {
const ns = s.cast(Scope.Namespace).?;
if (ns.decls.get(name_str_index)) |i| {
if (found_already) |f| {
return astgen.failNodeNotes(ident, "ambiguous reference", .{}, &.{
try astgen.errNoteNode(f, "declared here", .{}),
try astgen.errNoteNode(i, "also declared here", .{}),
});
}
// We found a match but must continue looking for ambiguous references to decls.
found_already = i;
found_needs_tunnel = ns.maybe_generic;
found_namespaces_out = num_namespaces_out;
}
num_namespaces_out += 1;
capturing_namespace = ns;
s = ns.parent;
},
.top => break,
};
if (found_already == null) {
const ident_name = try astgen.identifierTokenString(ident_token);
return astgen.failNode(ident, "use of undeclared identifier '{s}'", .{ident_name});
}
// Decl references happen by name rather than ZIR index so that when unrelated
// decls are modified, ZIR code containing references to them can be unmodified.
if (found_namespaces_out > 0 and found_needs_tunnel) {
switch (ri.rl) {
.ref, .ref_coerced_ty => return tunnelThroughClosure(
gz,
ident,
found_namespaces_out,
.{ .decl_ref = name_str_index },
.{ .node = found_already.? },
),
else => {
const result = try tunnelThroughClosure(
gz,
ident,
found_namespaces_out,
.{ .decl_val = name_str_index },
.{ .node = found_already.? },
);
return rvalueNoCoercePreRef(gz, ri, result, ident);
},
}
}
switch (ri.rl) {
.ref, .ref_coerced_ty => return gz.addStrTok(.decl_ref, name_str_index, ident_token),
else => {
const result = try gz.addStrTok(.decl_val, name_str_index, ident_token);
return rvalueNoCoercePreRef(gz, ri, result, ident);
},
}
}
/// Access a ZIR instruction through closure. May tunnel through arbitrarily
/// many namespaces, adding closure captures as required.
/// Returns the index of the `closure_get` instruction added to `gz`.
fn tunnelThroughClosure(
gz: *GenZir,
/// The node which references the value to be captured.
inner_ref_node: Ast.Node.Index,
/// The number of namespaces being tunnelled through. At least 1.
num_tunnels: u32,
/// The value being captured.
value: union(enum) {
ref: Zir.Inst.Ref,
ref_load: Zir.Inst.Ref,
decl_val: Zir.NullTerminatedString,
decl_ref: Zir.NullTerminatedString,
},
/// The location of the value's declaration.
decl_src: union(enum) {
token: Ast.TokenIndex,
node: Ast.Node.Index,
},
) !Zir.Inst.Ref {
switch (value) {
.ref => |v| if (v.toIndex() == null) return v, // trivial value; do not need tunnel
.ref_load => |v| assert(v.toIndex() != null), // there are no constant pointer refs
.decl_val, .decl_ref => {},
}
const astgen = gz.astgen;
const gpa = astgen.gpa;
// Otherwise we need a tunnel. First, figure out the path of namespaces we
// are tunneling through. This is usually only going to be one or two, so
// use an SFBA to optimize for the common case.
var sfba = std.heap.stackFallback(@sizeOf(usize) * 2, astgen.arena);
var intermediate_tunnels = try sfba.get().alloc(*Scope.Namespace, num_tunnels - 1);
const root_ns = ns: {
var i: usize = num_tunnels - 1;
var scope: *Scope = gz.parent;
while (i > 0) {
if (scope.cast(Scope.Namespace)) |mid_ns| {
i -= 1;
intermediate_tunnels[i] = mid_ns;
}
scope = scope.parent().?;
}
while (true) {
if (scope.cast(Scope.Namespace)) |ns| break :ns ns;
scope = scope.parent().?;
}
};
// Now that we know the scopes we're tunneling through, begin adding
// captures as required, starting with the outermost namespace.
const root_capture = Zir.Inst.Capture.wrap(switch (value) {
.ref => |v| .{ .instruction = v.toIndex().? },
.ref_load => |v| .{ .instruction_load = v.toIndex().? },
.decl_val => |str| .{ .decl_val = str },
.decl_ref => |str| .{ .decl_ref = str },
});
var cur_capture_index = std.math.cast(
u16,
(try root_ns.captures.getOrPut(gpa, root_capture)).index,
) orelse return astgen.failNodeNotes(root_ns.node, "this compiler implementation only supports up to 65536 captures per namespace", .{}, &.{
switch (decl_src) {
.token => |t| try astgen.errNoteTok(t, "captured value here", .{}),
.node => |n| try astgen.errNoteNode(n, "captured value here", .{}),
},
try astgen.errNoteNode(inner_ref_node, "value used here", .{}),
});
for (intermediate_tunnels) |tunnel_ns| {
cur_capture_index = std.math.cast(
u16,
(try tunnel_ns.captures.getOrPut(gpa, Zir.Inst.Capture.wrap(.{ .nested = cur_capture_index }))).index,
) orelse return astgen.failNodeNotes(tunnel_ns.node, "this compiler implementation only supports up to 65536 captures per namespace", .{}, &.{
switch (decl_src) {
.token => |t| try astgen.errNoteTok(t, "captured value here", .{}),
.node => |n| try astgen.errNoteNode(n, "captured value here", .{}),
},
try astgen.errNoteNode(inner_ref_node, "value used here", .{}),
});
}
// Add an instruction to get the value from the closure.
return gz.addExtendedNodeSmall(.closure_get, inner_ref_node, cur_capture_index);
}
fn stringLiteral(
gz: *GenZir,
ri: ResultInfo,
node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const main_tokens = tree.nodes.items(.main_token);
const str_lit_token = main_tokens[node];
const str = try astgen.strLitAsString(str_lit_token);
const result = try gz.add(.{
.tag = .str,
.data = .{ .str = .{
.start = str.index,
.len = str.len,
} },
});
return rvalue(gz, ri, result, node);
}
fn multilineStringLiteral(
gz: *GenZir,
ri: ResultInfo,
node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const str = try astgen.strLitNodeAsString(node);
const result = try gz.add(.{
.tag = .str,
.data = .{ .str = .{
.start = str.index,
.len = str.len,
} },
});
return rvalue(gz, ri, result, node);
}
fn charLiteral(gz: *GenZir, ri: ResultInfo, node: Ast.Node.Index) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const main_tokens = tree.nodes.items(.main_token);
const main_token = main_tokens[node];
const slice = tree.tokenSlice(main_token);
switch (std.zig.parseCharLiteral(slice)) {
.success => |codepoint| {
const result = try gz.addInt(codepoint);
return rvalue(gz, ri, result, node);
},
.failure => |err| return astgen.failWithStrLitError(err, main_token, slice, 0),
}
}
const Sign = enum { negative, positive };
fn numberLiteral(gz: *GenZir, ri: ResultInfo, node: Ast.Node.Index, source_node: Ast.Node.Index, sign: Sign) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const main_tokens = tree.nodes.items(.main_token);
const num_token = main_tokens[node];
const bytes = tree.tokenSlice(num_token);
const result: Zir.Inst.Ref = switch (std.zig.parseNumberLiteral(bytes)) {
.int => |num| switch (num) {
0 => if (sign == .positive) .zero else return astgen.failTokNotes(
num_token,
"integer literal '-0' is ambiguous",
.{},
&.{
try astgen.errNoteTok(num_token, "use '0' for an integer zero", .{}),
try astgen.errNoteTok(num_token, "use '-0.0' for a floating-point signed zero", .{}),
},
),
1 => {
// Handle the negation here!
const result: Zir.Inst.Ref = switch (sign) {
.positive => .one,
.negative => .negative_one,
};
return rvalue(gz, ri, result, source_node);
},
else => try gz.addInt(num),
},
.big_int => |base| big: {
const gpa = astgen.gpa;
var big_int = try std.math.big.int.Managed.init(gpa);
defer big_int.deinit();
const prefix_offset: usize = if (base == .decimal) 0 else 2;
big_int.setString(@intFromEnum(base), bytes[prefix_offset..]) catch |err| switch (err) {
error.InvalidCharacter => unreachable, // caught in `parseNumberLiteral`
error.InvalidBase => unreachable, // we only pass 16, 8, 2, see above
error.OutOfMemory => return error.OutOfMemory,
};
const limbs = big_int.limbs[0..big_int.len()];
assert(big_int.isPositive());
break :big try gz.addIntBig(limbs);
},
.float => {
const unsigned_float_number = std.fmt.parseFloat(f128, bytes) catch |err| switch (err) {
error.InvalidCharacter => unreachable, // validated by tokenizer
};
const float_number = switch (sign) {
.negative => -unsigned_float_number,
.positive => unsigned_float_number,
};
// If the value fits into a f64 without losing any precision, store it that way.
@setFloatMode(.strict);
const smaller_float: f64 = @floatCast(float_number);
const bigger_again: f128 = smaller_float;
if (bigger_again == float_number) {
const result = try gz.addFloat(smaller_float);
return rvalue(gz, ri, result, source_node);
}
// We need to use 128 bits. Break the float into 4 u32 values so we can
// put it into the `extra` array.
const int_bits: u128 = @bitCast(float_number);
const result = try gz.addPlNode(.float128, node, Zir.Inst.Float128{
.piece0 = @truncate(int_bits),
.piece1 = @truncate(int_bits >> 32),
.piece2 = @truncate(int_bits >> 64),
.piece3 = @truncate(int_bits >> 96),
});
return rvalue(gz, ri, result, source_node);
},
.failure => |err| return astgen.failWithNumberError(err, num_token, bytes),
};
if (sign == .positive) {
return rvalue(gz, ri, result, source_node);
} else {
const negated = try gz.addUnNode(.negate, result, source_node);
return rvalue(gz, ri, negated, source_node);
}
}
fn failWithNumberError(astgen: *AstGen, err: std.zig.number_literal.Error, token: Ast.TokenIndex, bytes: []const u8) InnerError {
const is_float = std.mem.indexOfScalar(u8, bytes, '.') != null;
switch (err) {
.leading_zero => if (is_float) {
return astgen.failTok(token, "number '{s}' has leading zero", .{bytes});
} else {
return astgen.failTokNotes(token, "number '{s}' has leading zero", .{bytes}, &.{
try astgen.errNoteTok(token, "use '0o' prefix for octal literals", .{}),
});
},
.digit_after_base => return astgen.failTok(token, "expected a digit after base prefix", .{}),
.upper_case_base => |i| return astgen.failOff(token, @intCast(i), "base prefix must be lowercase", .{}),
.invalid_float_base => |i| return astgen.failOff(token, @intCast(i), "invalid base for float literal", .{}),
.repeated_underscore => |i| return astgen.failOff(token, @intCast(i), "repeated digit separator", .{}),
.invalid_underscore_after_special => |i| return astgen.failOff(token, @intCast(i), "expected digit before digit separator", .{}),
.invalid_digit => |info| return astgen.failOff(token, @intCast(info.i), "invalid digit '{c}' for {s} base", .{ bytes[info.i], @tagName(info.base) }),
.invalid_digit_exponent => |i| return astgen.failOff(token, @intCast(i), "invalid digit '{c}' in exponent", .{bytes[i]}),
.duplicate_exponent => |i| return astgen.failOff(token, @intCast(i), "duplicate exponent", .{}),
.exponent_after_underscore => |i| return astgen.failOff(token, @intCast(i), "expected digit before exponent", .{}),
.special_after_underscore => |i| return astgen.failOff(token, @intCast(i), "expected digit before '{c}'", .{bytes[i]}),
.trailing_special => |i| return astgen.failOff(token, @intCast(i), "expected digit after '{c}'", .{bytes[i - 1]}),
.trailing_underscore => |i| return astgen.failOff(token, @intCast(i), "trailing digit separator", .{}),
.duplicate_period => unreachable, // Validated by tokenizer
.invalid_character => unreachable, // Validated by tokenizer
.invalid_exponent_sign => |i| {
assert(bytes.len >= 2 and bytes[0] == '0' and bytes[1] == 'x'); // Validated by tokenizer
return astgen.failOff(token, @intCast(i), "sign '{c}' cannot follow digit '{c}' in hex base", .{ bytes[i], bytes[i - 1] });
},
}
}
fn asmExpr(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
full: Ast.full.Asm,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const main_tokens = tree.nodes.items(.main_token);
const node_datas = tree.nodes.items(.data);
const node_tags = tree.nodes.items(.tag);
const token_tags = tree.tokens.items(.tag);
const TagAndTmpl = struct { tag: Zir.Inst.Extended, tmpl: Zir.NullTerminatedString };
const tag_and_tmpl: TagAndTmpl = switch (node_tags[full.ast.template]) {
.string_literal => .{
.tag = .@"asm",
.tmpl = (try astgen.strLitAsString(main_tokens[full.ast.template])).index,
},
.multiline_string_literal => .{
.tag = .@"asm",
.tmpl = (try astgen.strLitNodeAsString(full.ast.template)).index,
},
else => .{
.tag = .asm_expr,
.tmpl = @enumFromInt(@intFromEnum(try comptimeExpr(gz, scope, .{ .rl = .none }, full.ast.template))),
},
};
// See https://github.com/ziglang/zig/issues/215 and related issues discussing
// possible inline assembly improvements. Until then here is status quo AstGen
// for assembly syntax. It's used by std lib crypto aesni.zig.
const is_container_asm = astgen.fn_block == null;
if (is_container_asm) {
if (full.volatile_token) |t|
return astgen.failTok(t, "volatile is meaningless on global assembly", .{});
if (full.outputs.len != 0 or full.inputs.len != 0 or full.first_clobber != null)
return astgen.failNode(node, "global assembly cannot have inputs, outputs, or clobbers", .{});
} else {
if (full.outputs.len == 0 and full.volatile_token == null) {
return astgen.failNode(node, "assembly expression with no output must be marked volatile", .{});
}
}
if (full.outputs.len > 32) {
return astgen.failNode(full.outputs[32], "too many asm outputs", .{});
}
var outputs_buffer: [32]Zir.Inst.Asm.Output = undefined;
const outputs = outputs_buffer[0..full.outputs.len];
var output_type_bits: u32 = 0;
for (full.outputs, 0..) |output_node, i| {
const symbolic_name = main_tokens[output_node];
const name = try astgen.identAsString(symbolic_name);
const constraint_token = symbolic_name + 2;
const constraint = (try astgen.strLitAsString(constraint_token)).index;
const has_arrow = token_tags[symbolic_name + 4] == .arrow;
if (has_arrow) {
if (output_type_bits != 0) {
return astgen.failNode(output_node, "inline assembly allows up to one output value", .{});
}
output_type_bits |= @as(u32, 1) << @intCast(i);
const out_type_node = node_datas[output_node].lhs;
const out_type_inst = try typeExpr(gz, scope, out_type_node);
outputs[i] = .{
.name = name,
.constraint = constraint,
.operand = out_type_inst,
};
} else {
const ident_token = symbolic_name + 4;
// TODO have a look at #215 and related issues and decide how to
// handle outputs. Do we want this to be identifiers?
// Or maybe we want to force this to be expressions with a pointer type.
outputs[i] = .{
.name = name,
.constraint = constraint,
.operand = try localVarRef(gz, scope, .{ .rl = .ref }, node, ident_token),
};
}
}
if (full.inputs.len > 32) {
return astgen.failNode(full.inputs[32], "too many asm inputs", .{});
}
var inputs_buffer: [32]Zir.Inst.Asm.Input = undefined;
const inputs = inputs_buffer[0..full.inputs.len];
for (full.inputs, 0..) |input_node, i| {
const symbolic_name = main_tokens[input_node];
const name = try astgen.identAsString(symbolic_name);
const constraint_token = symbolic_name + 2;
const constraint = (try astgen.strLitAsString(constraint_token)).index;
const operand = try expr(gz, scope, .{ .rl = .none }, node_datas[input_node].lhs);
inputs[i] = .{
.name = name,
.constraint = constraint,
.operand = operand,
};
}
var clobbers_buffer: [32]u32 = undefined;
var clobber_i: usize = 0;
if (full.first_clobber) |first_clobber| clobbers: {
// asm ("foo" ::: "a", "b")
// asm ("foo" ::: "a", "b",)
var tok_i = first_clobber;
while (true) : (tok_i += 1) {
if (clobber_i >= clobbers_buffer.len) {
return astgen.failTok(tok_i, "too many asm clobbers", .{});
}
clobbers_buffer[clobber_i] = @intFromEnum((try astgen.strLitAsString(tok_i)).index);
clobber_i += 1;
tok_i += 1;
switch (token_tags[tok_i]) {
.r_paren => break :clobbers,
.comma => {
if (token_tags[tok_i + 1] == .r_paren) {
break :clobbers;
} else {
continue;
}
},
else => unreachable,
}
}
}
const result = try gz.addAsm(.{
.tag = tag_and_tmpl.tag,
.node = node,
.asm_source = tag_and_tmpl.tmpl,
.is_volatile = full.volatile_token != null,
.output_type_bits = output_type_bits,
.outputs = outputs,
.inputs = inputs,
.clobbers = clobbers_buffer[0..clobber_i],
});
return rvalue(gz, ri, result, node);
}
fn as(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
lhs: Ast.Node.Index,
rhs: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const dest_type = try typeExpr(gz, scope, lhs);
const result = try reachableExpr(gz, scope, .{ .rl = .{ .ty = dest_type } }, rhs, node);
return rvalue(gz, ri, result, node);
}
fn unionInit(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
params: []const Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const union_type = try typeExpr(gz, scope, params[0]);
const field_name = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .slice_const_u8_type } }, params[1]);
const field_type = try gz.addPlNode(.field_type_ref, node, Zir.Inst.FieldTypeRef{
.container_type = union_type,
.field_name = field_name,
});
const init = try reachableExpr(gz, scope, .{ .rl = .{ .ty = field_type } }, params[2], node);
const result = try gz.addPlNode(.union_init, node, Zir.Inst.UnionInit{
.union_type = union_type,
.init = init,
.field_name = field_name,
});
return rvalue(gz, ri, result, node);
}
fn bitCast(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
operand_node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const dest_type = try ri.rl.resultTypeForCast(gz, node, "@bitCast");
const operand = try reachableExpr(gz, scope, .{ .rl = .none }, operand_node, node);
const result = try gz.addPlNode(.bitcast, node, Zir.Inst.Bin{
.lhs = dest_type,
.rhs = operand,
});
return rvalue(gz, ri, result, node);
}
/// Handle one or more nested pointer cast builtins:
/// * @ptrCast
/// * @alignCast
/// * @addrSpaceCast
/// * @constCast
/// * @volatileCast
/// Any sequence of such builtins is treated as a single operation. This allowed
/// for sequences like `@ptrCast(@alignCast(ptr))` to work correctly despite the
/// intermediate result type being unknown.
fn ptrCast(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
root_node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const main_tokens = tree.nodes.items(.main_token);
const node_datas = tree.nodes.items(.data);
const node_tags = tree.nodes.items(.tag);
var flags: Zir.Inst.FullPtrCastFlags = .{};
// Note that all pointer cast builtins have one parameter, so we only need
// to handle `builtin_call_two`.
var node = root_node;
while (true) {
switch (node_tags[node]) {
.builtin_call_two, .builtin_call_two_comma => {},
.grouped_expression => {
// Handle the chaining even with redundant parentheses
node = node_datas[node].lhs;
continue;
},
else => break,
}
if (node_datas[node].lhs == 0) break; // 0 args
if (node_datas[node].rhs != 0) break; // 2 args
const builtin_token = main_tokens[node];
const builtin_name = tree.tokenSlice(builtin_token);
const info = BuiltinFn.list.get(builtin_name) orelse break;
if (info.param_count != 1) break;
switch (info.tag) {
else => break,
inline .ptr_cast,
.align_cast,
.addrspace_cast,
.const_cast,
.volatile_cast,
=> |tag| {
if (@field(flags, @tagName(tag))) {
return astgen.failNode(node, "redundant {s}", .{builtin_name});
}
@field(flags, @tagName(tag)) = true;
},
}
node = node_datas[node].lhs;
}
const flags_i: u5 = @bitCast(flags);
assert(flags_i != 0);
const ptr_only: Zir.Inst.FullPtrCastFlags = .{ .ptr_cast = true };
if (flags_i == @as(u5, @bitCast(ptr_only))) {
// Special case: simpler representation
return typeCast(gz, scope, ri, root_node, node, .ptr_cast, "@ptrCast");
}
const no_result_ty_flags: Zir.Inst.FullPtrCastFlags = .{
.const_cast = true,
.volatile_cast = true,
};
if ((flags_i & ~@as(u5, @bitCast(no_result_ty_flags))) == 0) {
// Result type not needed
const cursor = maybeAdvanceSourceCursorToMainToken(gz, root_node);
const operand = try expr(gz, scope, .{ .rl = .none }, node);
try emitDbgStmt(gz, cursor);
const result = try gz.addExtendedPayloadSmall(.ptr_cast_no_dest, flags_i, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(root_node),
.operand = operand,
});
return rvalue(gz, ri, result, root_node);
}
// Full cast including result type
const cursor = maybeAdvanceSourceCursorToMainToken(gz, root_node);
const result_type = try ri.rl.resultTypeForCast(gz, root_node, flags.needResultTypeBuiltinName());
const operand = try expr(gz, scope, .{ .rl = .none }, node);
try emitDbgStmt(gz, cursor);
const result = try gz.addExtendedPayloadSmall(.ptr_cast_full, flags_i, Zir.Inst.BinNode{
.node = gz.nodeIndexToRelative(root_node),
.lhs = result_type,
.rhs = operand,
});
return rvalue(gz, ri, result, root_node);
}
fn typeOf(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
args: []const Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
if (args.len < 1) {
return astgen.failNode(node, "expected at least 1 argument, found 0", .{});
}
const gpa = astgen.gpa;
if (args.len == 1) {
const typeof_inst = try gz.makeBlockInst(.typeof_builtin, node);
var typeof_scope = gz.makeSubBlock(scope);
typeof_scope.is_comptime = false;
typeof_scope.is_typeof = true;
typeof_scope.c_import = false;
defer typeof_scope.unstack();
const ty_expr = try reachableExpr(&typeof_scope, &typeof_scope.base, .{ .rl = .none }, args[0], node);
if (!gz.refIsNoReturn(ty_expr)) {
_ = try typeof_scope.addBreak(.break_inline, typeof_inst, ty_expr);
}
try typeof_scope.setBlockBody(typeof_inst);
// typeof_scope unstacked now, can add new instructions to gz
try gz.instructions.append(gpa, typeof_inst);
return rvalue(gz, ri, typeof_inst.toRef(), node);
}
const payload_size: u32 = std.meta.fields(Zir.Inst.TypeOfPeer).len;
const payload_index = try reserveExtra(astgen, payload_size + args.len);
const args_index = payload_index + payload_size;
const typeof_inst = try gz.addExtendedMultiOpPayloadIndex(.typeof_peer, payload_index, args.len);
var typeof_scope = gz.makeSubBlock(scope);
typeof_scope.is_comptime = false;
for (args, 0..) |arg, i| {
const param_ref = try reachableExpr(&typeof_scope, &typeof_scope.base, .{ .rl = .none }, arg, node);
astgen.extra.items[args_index + i] = @intFromEnum(param_ref);
}
_ = try typeof_scope.addBreak(.break_inline, typeof_inst.toIndex().?, .void_value);
const body = typeof_scope.instructionsSlice();
const body_len = astgen.countBodyLenAfterFixups(body);
astgen.setExtra(payload_index, Zir.Inst.TypeOfPeer{
.body_len = @intCast(body_len),
.body_index = @intCast(astgen.extra.items.len),
.src_node = gz.nodeIndexToRelative(node),
});
try astgen.extra.ensureUnusedCapacity(gpa, body_len);
astgen.appendBodyWithFixups(body);
typeof_scope.unstack();
return rvalue(gz, ri, typeof_inst, node);
}
fn minMax(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
args: []const Ast.Node.Index,
comptime op: enum { min, max },
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
if (args.len < 2) {
return astgen.failNode(node, "expected at least 2 arguments, found 0", .{});
}
if (args.len == 2) {
const tag: Zir.Inst.Tag = switch (op) {
.min => .min,
.max => .max,
};
const a = try expr(gz, scope, .{ .rl = .none }, args[0]);
const b = try expr(gz, scope, .{ .rl = .none }, args[1]);
const result = try gz.addPlNode(tag, node, Zir.Inst.Bin{
.lhs = a,
.rhs = b,
});
return rvalue(gz, ri, result, node);
}
const payload_index = try addExtra(astgen, Zir.Inst.NodeMultiOp{
.src_node = gz.nodeIndexToRelative(node),
});
var extra_index = try reserveExtra(gz.astgen, args.len);
for (args) |arg| {
const arg_ref = try expr(gz, scope, .{ .rl = .none }, arg);
astgen.extra.items[extra_index] = @intFromEnum(arg_ref);
extra_index += 1;
}
const tag: Zir.Inst.Extended = switch (op) {
.min => .min_multi,
.max => .max_multi,
};
const result = try gz.addExtendedMultiOpPayloadIndex(tag, payload_index, args.len);
return rvalue(gz, ri, result, node);
}
fn builtinCall(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
params: []const Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const main_tokens = tree.nodes.items(.main_token);
const builtin_token = main_tokens[node];
const builtin_name = tree.tokenSlice(builtin_token);
// We handle the different builtins manually because they have different semantics depending
// on the function. For example, `@as` and others participate in result location semantics,
// and `@cImport` creates a special scope that collects a .c source code text buffer.
// Also, some builtins have a variable number of parameters.
const info = BuiltinFn.list.get(builtin_name) orelse {
return astgen.failNode(node, "invalid builtin function: '{s}'", .{
builtin_name,
});
};
if (info.param_count) |expected| {
if (expected != params.len) {
const s = if (expected == 1) "" else "s";
return astgen.failNode(node, "expected {d} argument{s}, found {d}", .{
expected, s, params.len,
});
}
}
// Check function scope-only builtins
if (astgen.fn_block == null and info.illegal_outside_function)
return astgen.failNode(node, "'{s}' outside function scope", .{builtin_name});
switch (info.tag) {
.import => {
const node_tags = tree.nodes.items(.tag);
const operand_node = params[0];
if (node_tags[operand_node] != .string_literal) {
// Spec reference: https://github.com/ziglang/zig/issues/2206
return astgen.failNode(operand_node, "@import operand must be a string literal", .{});
}
const str_lit_token = main_tokens[operand_node];
const str = try astgen.strLitAsString(str_lit_token);
const str_slice = astgen.string_bytes.items[@intFromEnum(str.index)..][0..str.len];
if (mem.indexOfScalar(u8, str_slice, 0) != null) {
return astgen.failTok(str_lit_token, "import path cannot contain null bytes", .{});
} else if (str.len == 0) {
return astgen.failTok(str_lit_token, "import path cannot be empty", .{});
}
const result = try gz.addStrTok(.import, str.index, str_lit_token);
const gop = try astgen.imports.getOrPut(astgen.gpa, str.index);
if (!gop.found_existing) {
gop.value_ptr.* = str_lit_token;
}
return rvalue(gz, ri, result, node);
},
.compile_log => {
const payload_index = try addExtra(gz.astgen, Zir.Inst.NodeMultiOp{
.src_node = gz.nodeIndexToRelative(node),
});
var extra_index = try reserveExtra(gz.astgen, params.len);
for (params) |param| {
const param_ref = try expr(gz, scope, .{ .rl = .none }, param);
astgen.extra.items[extra_index] = @intFromEnum(param_ref);
extra_index += 1;
}
const result = try gz.addExtendedMultiOpPayloadIndex(.compile_log, payload_index, params.len);
return rvalue(gz, ri, result, node);
},
.field => {
if (ri.rl == .ref or ri.rl == .ref_coerced_ty) {
return gz.addPlNode(.field_ptr_named, node, Zir.Inst.FieldNamed{
.lhs = try expr(gz, scope, .{ .rl = .ref }, params[0]),
.field_name = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .slice_const_u8_type } }, params[1]),
});
}
const result = try gz.addPlNode(.field_val_named, node, Zir.Inst.FieldNamed{
.lhs = try expr(gz, scope, .{ .rl = .none }, params[0]),
.field_name = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .slice_const_u8_type } }, params[1]),
});
return rvalue(gz, ri, result, node);
},
// zig fmt: off
.as => return as( gz, scope, ri, node, params[0], params[1]),
.bit_cast => return bitCast( gz, scope, ri, node, params[0]),
.TypeOf => return typeOf( gz, scope, ri, node, params),
.union_init => return unionInit(gz, scope, ri, node, params),
.c_import => return cImport( gz, scope, node, params[0]),
.min => return minMax( gz, scope, ri, node, params, .min),
.max => return minMax( gz, scope, ri, node, params, .max),
// zig fmt: on
.@"export" => {
const node_tags = tree.nodes.items(.tag);
const node_datas = tree.nodes.items(.data);
// This function causes a Decl to be exported. The first parameter is not an expression,
// but an identifier of the Decl to be exported.
var namespace: Zir.Inst.Ref = .none;
var decl_name: Zir.NullTerminatedString = .empty;
switch (node_tags[params[0]]) {
.identifier => {
const ident_token = main_tokens[params[0]];
if (isPrimitive(tree.tokenSlice(ident_token))) {
return astgen.failTok(ident_token, "unable to export primitive value", .{});
}
decl_name = try astgen.identAsString(ident_token);
var s = scope;
var found_already: ?Ast.Node.Index = null; // we have found a decl with the same name already
while (true) switch (s.tag) {
.local_val => {
const local_val = s.cast(Scope.LocalVal).?;
if (local_val.name == decl_name) {
local_val.used = ident_token;
_ = try gz.addPlNode(.export_value, node, Zir.Inst.ExportValue{
.operand = local_val.inst,
.options = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .export_options_type } }, params[1]),
});
return rvalue(gz, ri, .void_value, node);
}
s = local_val.parent;
},
.local_ptr => {
const local_ptr = s.cast(Scope.LocalPtr).?;
if (local_ptr.name == decl_name) {
if (!local_ptr.maybe_comptime)
return astgen.failNode(params[0], "unable to export runtime-known value", .{});
local_ptr.used = ident_token;
const loaded = try gz.addUnNode(.load, local_ptr.ptr, node);
_ = try gz.addPlNode(.export_value, node, Zir.Inst.ExportValue{
.operand = loaded,
.options = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .export_options_type } }, params[1]),
});
return rvalue(gz, ri, .void_value, node);
}
s = local_ptr.parent;
},
.gen_zir => s = s.cast(GenZir).?.parent,
.defer_normal, .defer_error => s = s.cast(Scope.Defer).?.parent,
.namespace => {
const ns = s.cast(Scope.Namespace).?;
if (ns.decls.get(decl_name)) |i| {
if (found_already) |f| {
return astgen.failNodeNotes(node, "ambiguous reference", .{}, &.{
try astgen.errNoteNode(f, "declared here", .{}),
try astgen.errNoteNode(i, "also declared here", .{}),
});
}
// We found a match but must continue looking for ambiguous references to decls.
found_already = i;
}
s = ns.parent;
},
.top => break,
};
if (found_already == null) {
const ident_name = try astgen.identifierTokenString(ident_token);
return astgen.failNode(params[0], "use of undeclared identifier '{s}'", .{ident_name});
}
},
.field_access => {
const namespace_node = node_datas[params[0]].lhs;
namespace = try typeExpr(gz, scope, namespace_node);
const dot_token = main_tokens[params[0]];
const field_ident = dot_token + 1;
decl_name = try astgen.identAsString(field_ident);
},
else => return astgen.failNode(params[0], "symbol to export must identify a declaration", .{}),
}
const options = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .export_options_type } }, params[1]);
_ = try gz.addPlNode(.@"export", node, Zir.Inst.Export{
.namespace = namespace,
.decl_name = decl_name,
.options = options,
});
return rvalue(gz, ri, .void_value, node);
},
.@"extern" => {
const type_inst = try typeExpr(gz, scope, params[0]);
const options = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .extern_options_type } }, params[1]);
const result = try gz.addExtendedPayload(.builtin_extern, Zir.Inst.BinNode{
.node = gz.nodeIndexToRelative(node),
.lhs = type_inst,
.rhs = options,
});
return rvalue(gz, ri, result, node);
},
.fence => {
const order = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .atomic_order_type } }, params[0]);
_ = try gz.addExtendedPayload(.fence, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = order,
});
return rvalue(gz, ri, .void_value, node);
},
.set_float_mode => {
const order = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .float_mode_type } }, params[0]);
_ = try gz.addExtendedPayload(.set_float_mode, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = order,
});
return rvalue(gz, ri, .void_value, node);
},
.set_align_stack => {
const order = try expr(gz, scope, coerced_align_ri, params[0]);
_ = try gz.addExtendedPayload(.set_align_stack, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = order,
});
return rvalue(gz, ri, .void_value, node);
},
.set_cold => {
const order = try expr(gz, scope, ri, params[0]);
_ = try gz.addExtendedPayload(.set_cold, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = order,
});
return rvalue(gz, ri, .void_value, node);
},
.src => {
const token_starts = tree.tokens.items(.start);
const node_start = token_starts[tree.firstToken(node)];
astgen.advanceSourceCursor(node_start);
const result = try gz.addExtendedPayload(.builtin_src, Zir.Inst.Src{
.node = gz.nodeIndexToRelative(node),
.line = astgen.source_line,
.column = astgen.source_column,
});
return rvalue(gz, ri, result, node);
},
// zig fmt: off
.This => return rvalue(gz, ri, try gz.addNodeExtended(.this, node), node),
.return_address => return rvalue(gz, ri, try gz.addNodeExtended(.ret_addr, node), node),
.error_return_trace => return rvalue(gz, ri, try gz.addNodeExtended(.error_return_trace, node), node),
.frame => return rvalue(gz, ri, try gz.addNodeExtended(.frame, node), node),
.frame_address => return rvalue(gz, ri, try gz.addNodeExtended(.frame_address, node), node),
.breakpoint => return rvalue(gz, ri, try gz.addNodeExtended(.breakpoint, node), node),
.in_comptime => return rvalue(gz, ri, try gz.addNodeExtended(.in_comptime, node), node),
.type_info => return simpleUnOpType(gz, scope, ri, node, params[0], .type_info),
.size_of => return simpleUnOpType(gz, scope, ri, node, params[0], .size_of),
.bit_size_of => return simpleUnOpType(gz, scope, ri, node, params[0], .bit_size_of),
.align_of => return simpleUnOpType(gz, scope, ri, node, params[0], .align_of),
.int_from_ptr => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .int_from_ptr),
.compile_error => return simpleUnOp(gz, scope, ri, node, .{ .rl = .{ .coerced_ty = .slice_const_u8_type } }, params[0], .compile_error),
.set_eval_branch_quota => return simpleUnOp(gz, scope, ri, node, .{ .rl = .{ .coerced_ty = .u32_type } }, params[0], .set_eval_branch_quota),
.int_from_enum => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .int_from_enum),
.int_from_bool => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .int_from_bool),
.embed_file => return simpleUnOp(gz, scope, ri, node, .{ .rl = .{ .coerced_ty = .slice_const_u8_type } }, params[0], .embed_file),
.error_name => return simpleUnOp(gz, scope, ri, node, .{ .rl = .{ .coerced_ty = .anyerror_type } }, params[0], .error_name),
.set_runtime_safety => return simpleUnOp(gz, scope, ri, node, coerced_bool_ri, params[0], .set_runtime_safety),
.sqrt => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .sqrt),
.sin => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .sin),
.cos => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .cos),
.tan => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .tan),
.exp => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .exp),
.exp2 => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .exp2),
.log => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .log),
.log2 => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .log2),
.log10 => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .log10),
.abs => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .abs),
.floor => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .floor),
.ceil => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .ceil),
.trunc => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .trunc),
.round => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .round),
.tag_name => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .tag_name),
.type_name => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .type_name),
.Frame => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .frame_type),
.frame_size => return simpleUnOp(gz, scope, ri, node, .{ .rl = .none }, params[0], .frame_size),
.int_from_float => return typeCast(gz, scope, ri, node, params[0], .int_from_float, builtin_name),
.float_from_int => return typeCast(gz, scope, ri, node, params[0], .float_from_int, builtin_name),
.ptr_from_int => return typeCast(gz, scope, ri, node, params[0], .ptr_from_int, builtin_name),
.enum_from_int => return typeCast(gz, scope, ri, node, params[0], .enum_from_int, builtin_name),
.float_cast => return typeCast(gz, scope, ri, node, params[0], .float_cast, builtin_name),
.int_cast => return typeCast(gz, scope, ri, node, params[0], .int_cast, builtin_name),
.truncate => return typeCast(gz, scope, ri, node, params[0], .truncate, builtin_name),
// zig fmt: on
.Type => {
const operand = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .type_info_type } }, params[0]);
const gpa = gz.astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
try gz.astgen.instructions.ensureUnusedCapacity(gpa, 1);
const payload_index = try gz.astgen.addExtra(Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = operand,
});
const new_index: Zir.Inst.Index = @enumFromInt(gz.astgen.instructions.len);
gz.astgen.instructions.appendAssumeCapacity(.{
.tag = .extended,
.data = .{ .extended = .{
.opcode = .reify,
.small = @intFromEnum(gz.anon_name_strategy),
.operand = payload_index,
} },
});
gz.instructions.appendAssumeCapacity(new_index);
const result = new_index.toRef();
return rvalue(gz, ri, result, node);
},
.panic => {
try emitDbgNode(gz, node);
return simpleUnOp(gz, scope, ri, node, .{ .rl = .{ .coerced_ty = .slice_const_u8_type } }, params[0], .panic);
},
.trap => {
try emitDbgNode(gz, node);
_ = try gz.addNode(.trap, node);
return rvalue(gz, ri, .unreachable_value, node);
},
.int_from_error => {
const operand = try expr(gz, scope, .{ .rl = .none }, params[0]);
const result = try gz.addExtendedPayload(.int_from_error, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = operand,
});
return rvalue(gz, ri, result, node);
},
.error_from_int => {
const operand = try expr(gz, scope, .{ .rl = .none }, params[0]);
const result = try gz.addExtendedPayload(.error_from_int, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = operand,
});
return rvalue(gz, ri, result, node);
},
.error_cast => {
try emitDbgNode(gz, node);
const result = try gz.addExtendedPayload(.error_cast, Zir.Inst.BinNode{
.lhs = try ri.rl.resultTypeForCast(gz, node, "@errorCast"),
.rhs = try expr(gz, scope, .{ .rl = .none }, params[0]),
.node = gz.nodeIndexToRelative(node),
});
return rvalue(gz, ri, result, node);
},
.ptr_cast,
.align_cast,
.addrspace_cast,
.const_cast,
.volatile_cast,
=> return ptrCast(gz, scope, ri, node),
// zig fmt: off
.has_decl => return hasDeclOrField(gz, scope, ri, node, params[0], params[1], .has_decl),
.has_field => return hasDeclOrField(gz, scope, ri, node, params[0], params[1], .has_field),
.clz => return bitBuiltin(gz, scope, ri, node, params[0], .clz),
.ctz => return bitBuiltin(gz, scope, ri, node, params[0], .ctz),
.pop_count => return bitBuiltin(gz, scope, ri, node, params[0], .pop_count),
.byte_swap => return bitBuiltin(gz, scope, ri, node, params[0], .byte_swap),
.bit_reverse => return bitBuiltin(gz, scope, ri, node, params[0], .bit_reverse),
.div_exact => return divBuiltin(gz, scope, ri, node, params[0], params[1], .div_exact),
.div_floor => return divBuiltin(gz, scope, ri, node, params[0], params[1], .div_floor),
.div_trunc => return divBuiltin(gz, scope, ri, node, params[0], params[1], .div_trunc),
.mod => return divBuiltin(gz, scope, ri, node, params[0], params[1], .mod),
.rem => return divBuiltin(gz, scope, ri, node, params[0], params[1], .rem),
.shl_exact => return shiftOp(gz, scope, ri, node, params[0], params[1], .shl_exact),
.shr_exact => return shiftOp(gz, scope, ri, node, params[0], params[1], .shr_exact),
.bit_offset_of => return offsetOf(gz, scope, ri, node, params[0], params[1], .bit_offset_of),
.offset_of => return offsetOf(gz, scope, ri, node, params[0], params[1], .offset_of),
.c_undef => return simpleCBuiltin(gz, scope, ri, node, params[0], .c_undef),
.c_include => return simpleCBuiltin(gz, scope, ri, node, params[0], .c_include),
.cmpxchg_strong => return cmpxchg(gz, scope, ri, node, params, 1),
.cmpxchg_weak => return cmpxchg(gz, scope, ri, node, params, 0),
// zig fmt: on
.wasm_memory_size => {
const operand = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .u32_type } }, params[0]);
const result = try gz.addExtendedPayload(.wasm_memory_size, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = operand,
});
return rvalue(gz, ri, result, node);
},
.wasm_memory_grow => {
const index_arg = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .u32_type } }, params[0]);
const delta_arg = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .u32_type } }, params[1]);
const result = try gz.addExtendedPayload(.wasm_memory_grow, Zir.Inst.BinNode{
.node = gz.nodeIndexToRelative(node),
.lhs = index_arg,
.rhs = delta_arg,
});
return rvalue(gz, ri, result, node);
},
.c_define => {
if (!gz.c_import) return gz.astgen.failNode(node, "C define valid only inside C import block", .{});
const name = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .slice_const_u8_type } }, params[0]);
const value = try comptimeExpr(gz, scope, .{ .rl = .none }, params[1]);
const result = try gz.addExtendedPayload(.c_define, Zir.Inst.BinNode{
.node = gz.nodeIndexToRelative(node),
.lhs = name,
.rhs = value,
});
return rvalue(gz, ri, result, node);
},
.splat => {
const result_type = try ri.rl.resultTypeForCast(gz, node, "@splat");
const elem_type = try gz.addUnNode(.vector_elem_type, result_type, node);
const scalar = try expr(gz, scope, .{ .rl = .{ .ty = elem_type } }, params[0]);
const result = try gz.addPlNode(.splat, node, Zir.Inst.Bin{
.lhs = result_type,
.rhs = scalar,
});
return rvalue(gz, ri, result, node);
},
.reduce => {
const op = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .reduce_op_type } }, params[0]);
const scalar = try expr(gz, scope, .{ .rl = .none }, params[1]);
const result = try gz.addPlNode(.reduce, node, Zir.Inst.Bin{
.lhs = op,
.rhs = scalar,
});
return rvalue(gz, ri, result, node);
},
.add_with_overflow => return overflowArithmetic(gz, scope, ri, node, params, .add_with_overflow),
.sub_with_overflow => return overflowArithmetic(gz, scope, ri, node, params, .sub_with_overflow),
.mul_with_overflow => return overflowArithmetic(gz, scope, ri, node, params, .mul_with_overflow),
.shl_with_overflow => return overflowArithmetic(gz, scope, ri, node, params, .shl_with_overflow),
.atomic_load => {
const result = try gz.addPlNode(.atomic_load, node, Zir.Inst.AtomicLoad{
// zig fmt: off
.elem_type = try typeExpr(gz, scope, params[0]),
.ptr = try expr (gz, scope, .{ .rl = .none }, params[1]),
.ordering = try expr (gz, scope, .{ .rl = .{ .coerced_ty = .atomic_order_type } }, params[2]),
// zig fmt: on
});
return rvalue(gz, ri, result, node);
},
.atomic_rmw => {
const int_type = try typeExpr(gz, scope, params[0]);
const result = try gz.addPlNode(.atomic_rmw, node, Zir.Inst.AtomicRmw{
// zig fmt: off
.ptr = try expr(gz, scope, .{ .rl = .none }, params[1]),
.operation = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .atomic_rmw_op_type } }, params[2]),
.operand = try expr(gz, scope, .{ .rl = .{ .ty = int_type } }, params[3]),
.ordering = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .atomic_order_type } }, params[4]),
// zig fmt: on
});
return rvalue(gz, ri, result, node);
},
.atomic_store => {
const int_type = try typeExpr(gz, scope, params[0]);
_ = try gz.addPlNode(.atomic_store, node, Zir.Inst.AtomicStore{
// zig fmt: off
.ptr = try expr(gz, scope, .{ .rl = .none }, params[1]),
.operand = try expr(gz, scope, .{ .rl = .{ .ty = int_type } }, params[2]),
.ordering = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .atomic_order_type } }, params[3]),
// zig fmt: on
});
return rvalue(gz, ri, .void_value, node);
},
.mul_add => {
const float_type = try typeExpr(gz, scope, params[0]);
const mulend1 = try expr(gz, scope, .{ .rl = .{ .coerced_ty = float_type } }, params[1]);
const mulend2 = try expr(gz, scope, .{ .rl = .{ .coerced_ty = float_type } }, params[2]);
const addend = try expr(gz, scope, .{ .rl = .{ .ty = float_type } }, params[3]);
const result = try gz.addPlNode(.mul_add, node, Zir.Inst.MulAdd{
.mulend1 = mulend1,
.mulend2 = mulend2,
.addend = addend,
});
return rvalue(gz, ri, result, node);
},
.call => {
const modifier = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .call_modifier_type } }, params[0]);
const callee = try expr(gz, scope, .{ .rl = .none }, params[1]);
const args = try expr(gz, scope, .{ .rl = .none }, params[2]);
const result = try gz.addPlNode(.builtin_call, node, Zir.Inst.BuiltinCall{
.modifier = modifier,
.callee = callee,
.args = args,
.flags = .{
.is_nosuspend = gz.nosuspend_node != 0,
.ensure_result_used = false,
},
});
return rvalue(gz, ri, result, node);
},
.field_parent_ptr => {
const parent_type = try typeExpr(gz, scope, params[0]);
const field_name = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .slice_const_u8_type } }, params[1]);
const result = try gz.addPlNode(.field_parent_ptr, node, Zir.Inst.FieldParentPtr{
.parent_type = parent_type,
.field_name = field_name,
.field_ptr = try expr(gz, scope, .{ .rl = .none }, params[2]),
});
return rvalue(gz, ri, result, node);
},
.memcpy => {
_ = try gz.addPlNode(.memcpy, node, Zir.Inst.Bin{
.lhs = try expr(gz, scope, .{ .rl = .none }, params[0]),
.rhs = try expr(gz, scope, .{ .rl = .none }, params[1]),
});
return rvalue(gz, ri, .void_value, node);
},
.memset => {
const lhs = try expr(gz, scope, .{ .rl = .none }, params[0]);
const lhs_ty = try gz.addUnNode(.typeof, lhs, params[0]);
const elem_ty = try gz.addUnNode(.indexable_ptr_elem_type, lhs_ty, params[0]);
_ = try gz.addPlNode(.memset, node, Zir.Inst.Bin{
.lhs = lhs,
.rhs = try expr(gz, scope, .{ .rl = .{ .coerced_ty = elem_ty } }, params[1]),
});
return rvalue(gz, ri, .void_value, node);
},
.shuffle => {
const result = try gz.addPlNode(.shuffle, node, Zir.Inst.Shuffle{
.elem_type = try typeExpr(gz, scope, params[0]),
.a = try expr(gz, scope, .{ .rl = .none }, params[1]),
.b = try expr(gz, scope, .{ .rl = .none }, params[2]),
.mask = try comptimeExpr(gz, scope, .{ .rl = .none }, params[3]),
});
return rvalue(gz, ri, result, node);
},
.select => {
const result = try gz.addExtendedPayload(.select, Zir.Inst.Select{
.node = gz.nodeIndexToRelative(node),
.elem_type = try typeExpr(gz, scope, params[0]),
.pred = try expr(gz, scope, .{ .rl = .none }, params[1]),
.a = try expr(gz, scope, .{ .rl = .none }, params[2]),
.b = try expr(gz, scope, .{ .rl = .none }, params[3]),
});
return rvalue(gz, ri, result, node);
},
.async_call => {
const result = try gz.addExtendedPayload(.builtin_async_call, Zir.Inst.AsyncCall{
.node = gz.nodeIndexToRelative(node),
.frame_buffer = try expr(gz, scope, .{ .rl = .none }, params[0]),
.result_ptr = try expr(gz, scope, .{ .rl = .none }, params[1]),
.fn_ptr = try expr(gz, scope, .{ .rl = .none }, params[2]),
.args = try expr(gz, scope, .{ .rl = .none }, params[3]),
});
return rvalue(gz, ri, result, node);
},
.Vector => {
const result = try gz.addPlNode(.vector_type, node, Zir.Inst.Bin{
.lhs = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .u32_type } }, params[0]),
.rhs = try typeExpr(gz, scope, params[1]),
});
return rvalue(gz, ri, result, node);
},
.prefetch => {
const ptr = try expr(gz, scope, .{ .rl = .none }, params[0]);
const options = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .prefetch_options_type } }, params[1]);
_ = try gz.addExtendedPayload(.prefetch, Zir.Inst.BinNode{
.node = gz.nodeIndexToRelative(node),
.lhs = ptr,
.rhs = options,
});
return rvalue(gz, ri, .void_value, node);
},
.c_va_arg => {
const result = try gz.addExtendedPayload(.c_va_arg, Zir.Inst.BinNode{
.node = gz.nodeIndexToRelative(node),
.lhs = try expr(gz, scope, .{ .rl = .none }, params[0]),
.rhs = try typeExpr(gz, scope, params[1]),
});
return rvalue(gz, ri, result, node);
},
.c_va_copy => {
const result = try gz.addExtendedPayload(.c_va_copy, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = try expr(gz, scope, .{ .rl = .none }, params[0]),
});
return rvalue(gz, ri, result, node);
},
.c_va_end => {
const result = try gz.addExtendedPayload(.c_va_end, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = try expr(gz, scope, .{ .rl = .none }, params[0]),
});
return rvalue(gz, ri, result, node);
},
.c_va_start => {
if (!astgen.fn_var_args) {
return astgen.failNode(node, "'@cVaStart' in a non-variadic function", .{});
}
return rvalue(gz, ri, try gz.addNodeExtended(.c_va_start, node), node);
},
.work_item_id => {
const operand = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .u32_type } }, params[0]);
const result = try gz.addExtendedPayload(.work_item_id, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = operand,
});
return rvalue(gz, ri, result, node);
},
.work_group_size => {
const operand = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .u32_type } }, params[0]);
const result = try gz.addExtendedPayload(.work_group_size, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = operand,
});
return rvalue(gz, ri, result, node);
},
.work_group_id => {
const operand = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .u32_type } }, params[0]);
const result = try gz.addExtendedPayload(.work_group_id, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = operand,
});
return rvalue(gz, ri, result, node);
},
}
}
fn hasDeclOrField(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
lhs_node: Ast.Node.Index,
rhs_node: Ast.Node.Index,
tag: Zir.Inst.Tag,
) InnerError!Zir.Inst.Ref {
const container_type = try typeExpr(gz, scope, lhs_node);
const name = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .slice_const_u8_type } }, rhs_node);
const result = try gz.addPlNode(tag, node, Zir.Inst.Bin{
.lhs = container_type,
.rhs = name,
});
return rvalue(gz, ri, result, node);
}
fn typeCast(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
operand_node: Ast.Node.Index,
tag: Zir.Inst.Tag,
builtin_name: []const u8,
) InnerError!Zir.Inst.Ref {
const cursor = maybeAdvanceSourceCursorToMainToken(gz, node);
const result_type = try ri.rl.resultTypeForCast(gz, node, builtin_name);
const operand = try expr(gz, scope, .{ .rl = .none }, operand_node);
try emitDbgStmt(gz, cursor);
const result = try gz.addPlNode(tag, node, Zir.Inst.Bin{
.lhs = result_type,
.rhs = operand,
});
return rvalue(gz, ri, result, node);
}
fn simpleUnOpType(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
operand_node: Ast.Node.Index,
tag: Zir.Inst.Tag,
) InnerError!Zir.Inst.Ref {
const operand = try typeExpr(gz, scope, operand_node);
const result = try gz.addUnNode(tag, operand, node);
return rvalue(gz, ri, result, node);
}
fn simpleUnOp(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
operand_ri: ResultInfo,
operand_node: Ast.Node.Index,
tag: Zir.Inst.Tag,
) InnerError!Zir.Inst.Ref {
const cursor = maybeAdvanceSourceCursorToMainToken(gz, node);
const operand = if (tag == .compile_error)
try comptimeExpr(gz, scope, operand_ri, operand_node)
else
try expr(gz, scope, operand_ri, operand_node);
switch (tag) {
.tag_name, .error_name, .int_from_ptr => try emitDbgStmt(gz, cursor),
else => {},
}
const result = try gz.addUnNode(tag, operand, node);
return rvalue(gz, ri, result, node);
}
fn negation(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const tree = astgen.tree;
const node_tags = tree.nodes.items(.tag);
const node_datas = tree.nodes.items(.data);
// Check for float literal as the sub-expression because we want to preserve
// its negativity rather than having it go through comptime subtraction.
const operand_node = node_datas[node].lhs;
if (node_tags[operand_node] == .number_literal) {
return numberLiteral(gz, ri, operand_node, node, .negative);
}
const operand = try expr(gz, scope, .{ .rl = .none }, operand_node);
const result = try gz.addUnNode(.negate, operand, node);
return rvalue(gz, ri, result, node);
}
fn cmpxchg(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
params: []const Ast.Node.Index,
small: u16,
) InnerError!Zir.Inst.Ref {
const int_type = try typeExpr(gz, scope, params[0]);
const result = try gz.addExtendedPayloadSmall(.cmpxchg, small, Zir.Inst.Cmpxchg{
// zig fmt: off
.node = gz.nodeIndexToRelative(node),
.ptr = try expr(gz, scope, .{ .rl = .none }, params[1]),
.expected_value = try expr(gz, scope, .{ .rl = .{ .ty = int_type } }, params[2]),
.new_value = try expr(gz, scope, .{ .rl = .{ .coerced_ty = int_type } }, params[3]),
.success_order = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .atomic_order_type } }, params[4]),
.failure_order = try expr(gz, scope, .{ .rl = .{ .coerced_ty = .atomic_order_type } }, params[5]),
// zig fmt: on
});
return rvalue(gz, ri, result, node);
}
fn bitBuiltin(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
operand_node: Ast.Node.Index,
tag: Zir.Inst.Tag,
) InnerError!Zir.Inst.Ref {
const operand = try expr(gz, scope, .{ .rl = .none }, operand_node);
const result = try gz.addUnNode(tag, operand, node);
return rvalue(gz, ri, result, node);
}
fn divBuiltin(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
lhs_node: Ast.Node.Index,
rhs_node: Ast.Node.Index,
tag: Zir.Inst.Tag,
) InnerError!Zir.Inst.Ref {
const cursor = maybeAdvanceSourceCursorToMainToken(gz, node);
const lhs = try expr(gz, scope, .{ .rl = .none }, lhs_node);
const rhs = try expr(gz, scope, .{ .rl = .none }, rhs_node);
try emitDbgStmt(gz, cursor);
const result = try gz.addPlNode(tag, node, Zir.Inst.Bin{ .lhs = lhs, .rhs = rhs });
return rvalue(gz, ri, result, node);
}
fn simpleCBuiltin(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
operand_node: Ast.Node.Index,
tag: Zir.Inst.Extended,
) InnerError!Zir.Inst.Ref {
const name: []const u8 = if (tag == .c_undef) "C undef" else "C include";
if (!gz.c_import) return gz.astgen.failNode(node, "{s} valid only inside C import block", .{name});
const operand = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .slice_const_u8_type } }, operand_node);
_ = try gz.addExtendedPayload(tag, Zir.Inst.UnNode{
.node = gz.nodeIndexToRelative(node),
.operand = operand,
});
return rvalue(gz, ri, .void_value, node);
}
fn offsetOf(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
lhs_node: Ast.Node.Index,
rhs_node: Ast.Node.Index,
tag: Zir.Inst.Tag,
) InnerError!Zir.Inst.Ref {
const type_inst = try typeExpr(gz, scope, lhs_node);
const field_name = try comptimeExpr(gz, scope, .{ .rl = .{ .coerced_ty = .slice_const_u8_type } }, rhs_node);
const result = try gz.addPlNode(tag, node, Zir.Inst.Bin{
.lhs = type_inst,
.rhs = field_name,
});
return rvalue(gz, ri, result, node);
}
fn shiftOp(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
lhs_node: Ast.Node.Index,
rhs_node: Ast.Node.Index,
tag: Zir.Inst.Tag,
) InnerError!Zir.Inst.Ref {
const lhs = try expr(gz, scope, .{ .rl = .none }, lhs_node);
const cursor = switch (gz.astgen.tree.nodes.items(.tag)[node]) {
.shl, .shr => maybeAdvanceSourceCursorToMainToken(gz, node),
else => undefined,
};
const log2_int_type = try gz.addUnNode(.typeof_log2_int_type, lhs, lhs_node);
const rhs = try expr(gz, scope, .{ .rl = .{ .ty = log2_int_type }, .ctx = .shift_op }, rhs_node);
switch (gz.astgen.tree.nodes.items(.tag)[node]) {
.shl, .shr => try emitDbgStmt(gz, cursor),
else => undefined,
}
const result = try gz.addPlNode(tag, node, Zir.Inst.Bin{
.lhs = lhs,
.rhs = rhs,
});
return rvalue(gz, ri, result, node);
}
fn cImport(
gz: *GenZir,
scope: *Scope,
node: Ast.Node.Index,
body_node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const gpa = astgen.gpa;
if (gz.c_import) return gz.astgen.failNode(node, "cannot nest @cImport", .{});
var block_scope = gz.makeSubBlock(scope);
block_scope.is_comptime = true;
block_scope.c_import = true;
defer block_scope.unstack();
const block_inst = try gz.makeBlockInst(.c_import, node);
const block_result = try fullBodyExpr(&block_scope, &block_scope.base, .{ .rl = .none }, body_node);
_ = try gz.addUnNode(.ensure_result_used, block_result, node);
if (!gz.refIsNoReturn(block_result)) {
_ = try block_scope.addBreak(.break_inline, block_inst, .void_value);
}
try block_scope.setBlockBody(block_inst);
// block_scope unstacked now, can add new instructions to gz
try gz.instructions.append(gpa, block_inst);
return block_inst.toRef();
}
fn overflowArithmetic(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
params: []const Ast.Node.Index,
tag: Zir.Inst.Extended,
) InnerError!Zir.Inst.Ref {
const lhs = try expr(gz, scope, .{ .rl = .none }, params[0]);
const rhs = try expr(gz, scope, .{ .rl = .none }, params[1]);
const result = try gz.addExtendedPayload(tag, Zir.Inst.BinNode{
.node = gz.nodeIndexToRelative(node),
.lhs = lhs,
.rhs = rhs,
});
return rvalue(gz, ri, result, node);
}
fn callExpr(
gz: *GenZir,
scope: *Scope,
ri: ResultInfo,
node: Ast.Node.Index,
call: Ast.full.Call,
) InnerError!Zir.Inst.Ref {
const astgen = gz.astgen;
const callee = try calleeExpr(gz, scope, call.ast.fn_expr);
const modifier: std.builtin.CallModifier = blk: {
if (gz.is_comptime) {
break :blk .compile_time;
}
if (call.async_token != null) {
break :blk .async_kw;
}
if (gz.nosuspend_node != 0) {
break :blk .no_async;
}
break :blk .auto;
};
{
astgen.advanceSourceCursor(astgen.tree.tokens.items(.start)[call.ast.lparen]);
const line = astgen.source_line - gz.decl_line;
const column = astgen.source_column;
// Sema expects a dbg_stmt immediately before call,
try emitDbgStmtForceCurrentIndex(gz, .{ line, column });
}
switch (callee) {
.direct => |obj| assert(obj != .none),
.field => |field| assert(field.obj_ptr != .none),
}
assert(node != 0);
const call_index: Zir.Inst.Index = @enumFromInt(astgen.instructions.len);
const call_inst = call_index.toRef();
try gz.astgen.instructions.append(astgen.gpa, undefined);
try gz.instructions.append(astgen.gpa, call_index);
const scratch_top = astgen.scratch.items.len;
defer astgen.scratch.items.len = scratch_top;
var scratch_index = scratch_top;
try astgen.scratch.resize(astgen.gpa, scratch_top + call.ast.params.len);
for (call.ast.params) |param_node| {
var arg_block = gz.makeSubBlock(scope);
defer arg_block.unstack();
// `call_inst` is reused to provide the param type.
const arg_ref = try fullBodyExpr(&arg_block, &arg_block.base, .{ .rl = .{ .coerced_ty = call_inst }, .ctx = .fn_arg }, param_node);
_ = try arg_block.addBreakWithSrcNode(.break_inline, call_index, arg_ref, param_node);
const body = arg_block.instructionsSlice();
try astgen.scratch.ensureUnusedCapacity(astgen.gpa, countBodyLenAfterFixups(astgen, body));
appendBodyWithFixupsArrayList(astgen, &astgen.scratch, body);
astgen.scratch.items[scratch_index] = @intCast(astgen.scratch.items.len - scratch_top);
scratch_index += 1;
}
// If our result location is a try/catch/error-union-if/return, a function argument,
// or an initializer for a `const` variable, the error trace propagates.
// Otherwise, it should always be popped (handled in Sema).
const propagate_error_trace = switch (ri.ctx) {
.error_handling_expr, .@"return", .fn_arg, .const_init => true,
else => false,
};
switch (callee) {
.direct => |callee_obj| {
const payload_index = try addExtra(astgen, Zir.Inst.Call{
.callee = callee_obj,
.flags = .{
.pop_error_return_trace = !propagate_error_trace,
.packed_modifier = @intCast(@intFromEnum(modifier)),
.args_len = @intCast(call.ast.params.len),
},
});
if (call.ast.params.len != 0) {
try astgen.extra.appendSlice(astgen.gpa, astgen.scratch.items[scratch_top..]);
}
gz.astgen.instructions.set(@intFromEnum(call_index), .{
.tag = .call,
.data = .{ .pl_node = .{
.src_node = gz.nodeIndexToRelative(node),
.payload_index = payload_index,
} },
});
},
.field => |callee_field| {
const payload_index = try addExtra(astgen, Zir.Inst.FieldCall{
.obj_ptr = callee_field.obj_ptr,
.field_name_start = callee_field.field_name_start,
.flags = .{
.pop_error_return_trace = !propagate_error_trace,
.packed_modifier = @intCast(@intFromEnum(modifier)),
.args_len = @intCast(call.ast.params.len),
},
});
if (call.ast.params.len != 0) {
try astgen.extra.appendSlice(astgen.gpa, astgen.scratch.items[scratch_top..]);
}
gz.astgen.instructions.set(@intFromEnum(call_index), .{
.tag = .field_call,
.data = .{ .pl_node = .{
.src_node = gz.nodeIndexToRelative(node),
.payload_index = payload_index,
} },
});
},
}
return rvalue(gz, ri, call_inst, node); // TODO function call with result location
}
const Callee = union(enum) {
field: struct {
/// A *pointer* to the object the field is fetched on, so that we can
/// promote the lvalue to an address if the first parameter requires it.
obj_ptr: Zir.Inst.Ref,
/// Offset into `string_bytes`.
field_name_start: Zir.NullTerminatedString,
},
direct: Zir.Inst.Ref,
};
/// calleeExpr generates the function part of a call expression (f in f(x)), but
/// *not* the callee argument to the @call() builtin. Its purpose is to
/// distinguish between standard calls and method call syntax `a.b()`. Thus, if
/// the lhs is a field access, we return using the `field` union field;
/// otherwise, we use the `direct` union field.
fn calleeExpr(
gz: *GenZir,
scope: *Scope,
node: Ast.Node.Index,
) InnerError!Callee {
const astgen = gz.astgen;
const tree = astgen.tree;
const tag = tree.nodes.items(.tag)[node];
switch (tag) {
.field_access => {
const main_tokens = tree.nodes.items(.main_token);
const node_datas = tree.nodes.items(.data);
const object_node = node_datas[node].lhs;
const dot_token = main_tokens[node];
const field_ident = dot_token + 1;
const str_index = try astgen.identAsString(field_ident);
// Capture the object by reference so we can promote it to an
// address in Sema if needed.
const lhs = try expr(gz, scope, .{ .rl = .ref }, object_node);
const cursor = maybeAdvanceSourceCursorToMainToken(gz, node);
try emitDbgStmt(gz, cursor);
return .{ .field = .{
.obj_ptr = lhs,
.field_name_start = str_index,
} };
},
else => return .{ .direct = try expr(gz, scope, .{ .rl = .none }, node) },
}
}
const primitive_instrs = std.ComptimeStringMap(Zir.Inst.Ref, .{
.{ "anyerror", .anyerror_type },
.{ "anyframe", .anyframe_type },
.{ "anyopaque", .anyopaque_type },
.{ "bool", .bool_type },
.{ "c_int", .c_int_type },
.{ "c_long", .c_long_type },
.{ "c_longdouble", .c_longdouble_type },
.{ "c_longlong", .c_longlong_type },
.{ "c_char", .c_char_type },
.{ "c_short", .c_short_type },
.{ "c_uint", .c_uint_type },
.{ "c_ulong", .c_ulong_type },
.{ "c_ulonglong", .c_ulonglong_type },
.{ "c_ushort", .c_ushort_type },
.{ "comptime_float", .comptime_float_type },
.{ "comptime_int", .comptime_int_type },
.{ "f128", .f128_type },
.{ "f16", .f16_type },
.{ "f32", .f32_type },
.{ "f64", .f64_type },
.{ "f80", .f80_type },
.{ "false", .bool_false },
.{ "i16", .i16_type },
.{ "i32", .i32_type },
.{ "i64", .i64_type },
.{ "i128", .i128_type },
.{ "i8", .i8_type },
.{ "isize", .isize_type },
.{ "noreturn", .noreturn_type },
.{ "null", .null_value },
.{ "true", .bool_true },
.{ "type", .type_type },
.{ "u16", .u16_type },
.{ "u29", .u29_type },
.{ "u32", .u32_type },
.{ "u64", .u64_type },
.{ "u128", .u128_type },
.{ "u1", .u1_type },
.{ "u8", .u8_type },
.{ "undefined", .undef },
.{ "usize", .usize_type },
.{ "void", .void_type },
});
comptime {
// These checks ensure that std.zig.primitives stays in sync with the primitive->Zir map.
const primitives = std.zig.primitives;
for (primitive_instrs.kvs) |kv| {
if (!primitives.isPrimitive(kv.key)) {
@compileError("std.zig.isPrimitive() is not aware of Zir instr '" ++ @tagName(kv.value) ++ "'");
}
}
for (primitives.names.kvs) |kv| {
if (primitive_instrs.get(kv.key) == null) {
@compileError("std.zig.primitives entry '" ++ kv.key ++ "' does not have a corresponding Zir instr");
}
}
}
fn nodeIsTriviallyZero(tree: *const Ast, node: Ast.Node.Index) bool {
const node_tags = tree.nodes.items(.tag);
const main_tokens = tree.nodes.items(.main_token);
switch (node_tags[node]) {
.number_literal => {
const ident = main_tokens[node];
return switch (std.zig.parseNumberLiteral(tree.tokenSlice(ident))) {
.int => |number| switch (number) {
0 => true,
else => false,
},
else => false,
};
},
else => return false,
}
}
fn nodeMayAppendToErrorTrace(tree: *const Ast, start_node: Ast.Node.Index) bool {
const node_tags = tree.nodes.items(.tag);
const node_datas = tree.nodes.items(.data);
var node = start_node;
while (true) {
switch (node_tags[node]) {
// These don't have the opportunity to call any runtime functions.
.error_value,
.identifier,
.@"comptime",
=> return false,
// Forward the question to the LHS sub-expression.
.grouped_expression,
.@"try",
.@"nosuspend",
.unwrap_optional,
=> node = node_datas[node].lhs,
// Anything that does not eval to an error is guaranteed to pop any
// additions to the error trace, so it effectively does not append.
else => return nodeMayEvalToError(tree, start_node) != .never,
}
}
}
fn nodeMayEvalToError(tree: *const Ast, start_node: Ast.Node.Index) BuiltinFn.EvalToError {
const node_tags = tree.nodes.items(.tag);
const node_datas = tree.nodes.items(.data);
const main_tokens = tree.nodes.items(.main_token);
const token_tags = tree.tokens.items(.tag);
var node = start_node;
while (true) {
switch (node_tags[node]) {
.root,
.@"usingnamespace",
.test_decl,
.switch_case,
.switch_case_inline,
.switch_case_one,
.switch_case_inline_one,
.container_field_init,
.container_field_align,
.container_field,
.asm_output,
.asm_input,
=> unreachable,
.error_value => return .always,
.@"asm",
.asm_simple,
.identifier,
.field_access,
.deref,
.array_access,
.while_simple,
.while_cont,
.for_simple,
.if_simple,
.@"while",
.@"if",
.@"for",
.@"switch",
.switch_comma,
.call_one,
.call_one_comma,
.async_call_one,
.async_call_one_comma,
.call,
.call_comma,
.async_call,
.async_call_comma,
=> return .maybe,
.@"return",
.@"break",
.@"continue",
.bit_not,
.bool_not,
.global_var_decl,
.local_var_decl,
.simple_var_decl,
.aligned_var_decl,
.@"defer",
.@"errdefer",
.address_of,
.optional_type,
.negation,
.negation_wrap,
.@"resume",
.array_type,
.array_type_sentinel,
.ptr_type_aligned,
.ptr_type_sentinel,
.ptr_type,
.ptr_type_bit_range,
.@"suspend",
.fn_proto_simple,
.fn_proto_multi,
.fn_proto_one,
.fn_proto,
.fn_decl,
.anyframe_type,
.anyframe_literal,
.number_literal,
.enum_literal,
.string_literal,
.multiline_string_literal,
.char_literal,
.unreachable_literal,
.error_set_decl,
.container_decl,
.container_decl_trailing,
.container_decl_two,
.container_decl_two_trailing,
.container_decl_arg,
.container_decl_arg_trailing,
.tagged_union,
.tagged_union_trailing,
.tagged_union_two,
.tagged_union_two_trailing,
.tagged_union_enum_tag,
.tagged_union_enum_tag_trailing,
.add,
.add_wrap,
.add_sat,
.array_cat,
.array_mult,
.assign,
.assign_destructure,
.assign_bit_and,
.assign_bit_or,
.assign_shl,
.assign_shl_sat,
.assign_shr,
.assign_bit_xor,
.assign_div,
.assign_sub,
.assign_sub_wrap,
.assign_sub_sat,
.assign_mod,
.assign_add,
.assign_add_wrap,
.assign_add_sat,
.assign_mul,
.assign_mul_wrap,
.assign_mul_sat,
.bang_equal,
.bit_and,
.bit_or,
.shl,
.shl_sat,
.shr,
.bit_xor,
.bool_and,
.bool_or,
.div,
.equal_equal,
.error_union,
.greater_or_equal,
.greater_than,
.less_or_equal,
.less_than,
.merge_error_sets,
.mod,
.mul,
.mul_wrap,
.mul_sat,
.switch_range,
.for_range,
.sub,
.sub_wrap,
.sub_sat,
.slice,
.slice_open,
.slice_sentinel,
.array_init_one,
.array_init_one_comma,
.array_init_dot_two,
.array_init_dot_two_comma,
.array_init_dot,
.array_init_dot_comma,
.array_init,
.array_init_comma,
.struct_init_one,
.struct_init_one_comma,
.struct_init_dot_two,
.struct_init_dot_two_comma,
.struct_init_dot,
.struct_init_dot_comma,
.struct_init,
.struct_init_comma,
=> return .never,
// Forward the question to the LHS sub-expression.
.grouped_expression,
.@"try",
.@"await",
.@"comptime",
.@"nosuspend",
.unwrap_optional,
=> node = node_datas[node].lhs,
// LHS sub-expression may still be an error under the outer optional or error union
.@"catch",
.@"orelse",
=> return .maybe,
.block_two,
.block_two_semicolon,
.block,
.block_semicolon,
=> {
const lbrace = main_tokens[node];
if (token_tags[lbrace - 1] == .colon) {
// Labeled blocks may need a memory location to forward
// to their break statements.
return .maybe;
} else {
return .never;
}
},
.builtin_call,
.builtin_call_comma,
.builtin_call_two,
.builtin_call_two_comma,
=> {
const builtin_token = main_tokens[node];
const builtin_name = tree.tokenSlice(builtin_token);
// If the builtin is an invalid name, we don't cause an error here; instead
// let it pass, and the error will be "invalid builtin function" later.
const builtin_info = BuiltinFn.list.get(builtin_name) orelse return .maybe;
return builtin_info.eval_to_error;
},
}
}
}
/// Returns `true` if it is known the type expression has more than one possible value;
/// `false` otherwise.
fn nodeImpliesMoreThanOnePossibleValue(tree: *const Ast, start_node: Ast.Node.Index) bool {
const node_tags = tree.nodes.items(.tag);
const node_datas = tree.nodes.items(.data);
var node = start_node;
while (true) {
switch (node_tags[node]) {
.root,
.@"usingnamespace",
.test_decl,
.switch_case,
.switch_case_inline,
.switch_case_one,
.switch_case_inline_one,
.container_field_init,
.container_field_align,
.container_field,
.asm_output,
.asm_input,
.global_var_decl,
.local_var_decl,
.simple_var_decl,
.aligned_var_decl,
=> unreachable,
.@"return",
.@"break",
.@"continue",
.bit_not,
.bool_not,
.@"defer",
.@"errdefer",
.address_of,
.negation,
.negation_wrap,
.@"resume",
.array_type,
.@"suspend",
.fn_decl,
.anyframe_literal,
.number_literal,
.enum_literal,
.string_literal,
.multiline_string_literal,
.char_literal,
.unreachable_literal,
.error_set_decl,
.container_decl,
.container_decl_trailing,
.container_decl_two,
.container_decl_two_trailing,
.container_decl_arg,
.container_decl_arg_trailing,
.tagged_union,
.tagged_union_trailing,
.tagged_union_two,
.tagged_union_two_trailing,
.tagged_union_enum_tag,
.tagged_union_enum_tag_trailing,
.@"asm",
.asm_simple,
.add,
.add_wrap,
.add_sat,
.array_cat,
.array_mult,
.assign,
.assign_destructure,
.assign_bit_and,
.assign_bit_or,
.assign_shl,
.assign_shl_sat,
.assign_shr,
.assign_bit_xor,
.assign_div,
.assign_sub,
.assign_sub_wrap,
.assign_sub_sat,
.assign_mod,
.assign_add,
.assign_add_wrap,
.assign_add_sat,
.assign_mul,
.assign_mul_wrap,
.assign_mul_sat,
.bang_equal,
.bit_and,
.bit_or,
.shl,
.shl_sat,
.shr,
.bit_xor,
.bool_and,
.bool_or,
.div,
.equal_equal,
.error_union,
.greater_or_equal,
.greater_than,
.less_or_equal,
.less_than,
.merge_error_sets,
.mod,
.mul,
.mul_wrap,
.mul_sat,
.switch_range,
.for_range,
.field_access,
.sub,
.sub_wrap,
.sub_sat,
.slice,
.slice_open,
.slice_sentinel,
.deref,
.array_access,
.error_value,
.while_simple,
.while_cont,
.for_simple,
.if_simple,
.@"catch",
.@"orelse",
.array_init_one,
.array_init_one_comma,
.array_init_dot_two,
.array_init_dot_two_comma,
.array_init_dot,
.array_init_dot_comma,
.array_init,
.array_init_comma,
.struct_init_one,
.struct_init_one_comma,
.struct_init_dot_two,
.struct_init_dot_two_comma,
.struct_init_dot,
.struct_init_dot_comma,
.struct_init,
.struct_init_comma,
.@"while",
.@"if",
.@"for",
.@"switch",
.switch_comma,
.call_one,
.call_one_comma,
.async_call_one,
.async_call_one_comma,
.call,
.call_comma,
.async_call,
.async_call_comma,
.block_two,
.block_two_semicolon,
.block,
.block_semicolon,
.builtin_call,
.builtin_call_comma,
.builtin_call_two,
.builtin_call_two_comma,
// these are function bodies, not pointers
.fn_proto_simple,
.fn_proto_multi,
.fn_proto_one,
.fn_proto,
=> return false,
// Forward the question to the LHS sub-expression.
.grouped_expression,
.@"try",
.@"await",
.@"comptime",
.@"nosuspend",
.unwrap_optional,
=> node = node_datas[node].lhs,
.ptr_type_aligned,
.ptr_type_sentinel,
.ptr_type,
.ptr_type_bit_range,
.optional_type,
.anyframe_type,
.array_type_sentinel,
=> return true,
.identifier => {
const main_tokens = tree.nodes.items(.main_token);
const ident_bytes = tree.tokenSlice(main_tokens[node]);
if (primitive_instrs.get(ident_bytes)) |primitive| switch (primitive) {
.anyerror_type,
.anyframe_type,
.anyopaque_type,
.bool_type,
.c_int_type,
.c_long_type,
.c_longdouble_type,
.c_longlong_type,
.c_char_type,
.c_short_type,
.c_uint_type,
.c_ulong_type,
.c_ulonglong_type,
.c_ushort_type,
.comptime_float_type,
.comptime_int_type,
.f16_type,
.f32_type,
.f64_type,
.f80_type,
.f128_type,
.i16_type,
.i32_type,
.i64_type,
.i128_type,
.i8_type,
.isize_type,
.type_type,
.u16_type,
.u29_type,
.u32_type,
.u64_type,
.u128_type,
.u1_type,
.u8_type,
.usize_type,
=> return true,
.void_type,
.bool_false,
.bool_true,
.null_value,
.undef,
.noreturn_type,
=> return false,
else => unreachable, // that's all the values from `primitives`.
} else {
return false;
}
},
}
}
}
/// Returns `true` if it is known the expression is a type that cannot be used at runtime;
/// `false` otherwise.
fn nodeImpliesComptimeOnly(tree: *const Ast, start_node: Ast.Node.Index) bool {
const node_tags = tree.nodes.items(.tag);
const node_datas = tree.nodes.items(.data);
var node = start_node;
while (true) {
switch (node_tags[node]) {
.root,
.@"usingnamespace",
.test_decl,
.switch_case,
.switch_case_inline,
.switch_case_one,
.switch_case_inline_one,
.container_field_init,
.container_field_align,
.container_field,
.asm_output,
.asm_input,
.global_var_decl,
.local_var_decl,
.simple_var_decl,
.aligned_var_decl,
=> unreachable,
.@"return",
.@"break",
.@"continue",
.bit_not,
.bool_not,
.@"defer",
.@"errdefer",
.address_of,
.negation,
.negation_wrap,
.@"resume",
.array_type,
.@"suspend",
.fn_decl,
.anyframe_literal,
.number_literal,
.enum_literal,
.string_literal,
.multiline_string_literal,
.char_literal,
.unreachable_literal,
.error_set_decl,
.container_decl,
.container_decl_trailing,
.container_decl_two,
.container_decl_two_trailing,
.container_decl_arg,
.container_decl_arg_trailing,
.tagged_union,
.tagged_union_trailing,
.tagged_union_two,
.tagged_union_two_trailing,
.tagged_union_enum_tag,
.tagged_union_enum_tag_trailing,
.@"asm",
.asm_simple,
.add,
.add_wrap,
.add_sat,
.array_cat,
.array_mult,
.assign,
.assign_destructure,
.assign_bit_and,
.assign_bit_or,
.assign_shl,
.assign_shl_sat,
.assign_shr,
.assign_bit_xor,
.assign_div,
.assign_sub,
.assign_sub_wrap,
.assign_sub_sat,
.assign_mod,
.assign_add,
.assign_add_wrap,
.assign_add_sat,
.assign_mul,
.assign_mul_wrap,
.assign_mul_sat,
.bang_equal,
.bit_and,
.bit_or,
.shl,
.shl_sat,
.shr,
.bit_xor,
.bool_and,
.bool_or,
.div,
.equal_equal,
.error_union,
.greater_or_equal,
.greater_than,
.less_or_equal,
.less_than,
.merge_error_sets,
.mod,
.mul,
.mul_wrap,
.mul_sat,
.switch_range,
.for_range,
.field_access,
.sub,
.sub_wrap,
.sub_sat,
.slice,
.slice_open,
.slice_sentinel,
.deref,
.array_access,
.error_value,
.while_simple,
.while_cont,
.for_simple,
.if_simple,
.@"catch",
.@"orelse",
.array_init_one,
.array_init_one_comma,
.array_init_dot_two,
.array_init_dot_two_comma,
.array_init_dot,
.array_init_dot_comma,
.array_init,
.array_init_comma,
.struct_init_one,
.struct_init_one_comma,
.struct_init_dot_two,
.struct_init_dot_two_comma,
.struct_init_dot,
.struct_init_dot_comma,
.struct_init,
.struct_init_comma,
.@"while",
.@"if",
.@"for",
.@"switch",
.switch_comma,
.call_one,
.call_one_comma,
.async_call_one,
.async_call_one_comma,
.call,
.call_comma,
.async_call,
.async_call_comma,
.block_two,
.block_two_semicolon,
.block,
.block_semicolon,
.builtin_call,
.builtin_call_comma,
.builtin_call_two,
.builtin_call_two_comma,
.ptr_type_aligned,
.ptr_type_sentinel,
.ptr_type,
.ptr_type_bit_range,
.optional_type,
.anyframe_type,
.array_type_sentinel,
=> return false,
// these are function bodies, not pointers
.fn_proto_simple,
.fn_proto_multi,
.fn_proto_one,
.fn_proto,
=> return true,
// Forward the question to the LHS sub-expression.
.grouped_expression,
.@"try",
.@"await",
.@"comptime",
.@"nosuspend",
.unwrap_optional,
=> node = node_datas[node].lhs,
.identifier => {
const main_tokens = tree.nodes.items(.main_token);
const ident_bytes = tree.tokenSlice(main_tokens[node]);
if (primitive_instrs.get(ident_bytes)) |primitive| switch (primitive) {
.anyerror_type,
.anyframe_type,
.anyopaque_type,
.bool_type,
.c_int_type,
.c_long_type,
.c_longdouble_type,
.c_longlong_type,
.c_char_type,
.c_short_type,
.c_uint_type,
.c_ulong_type,
.c_ulonglong_type,
.c_ushort_type,
.f16_type,
.f32_type,
.f64_type,
.f80_type,
.f128_type,
.i16_type,
.i32_type,
.i64_type,
.i128_type,
.i8_type,
.isize_type,
.u16_type,
.u29_type,
.u32_type,
.u64_type,
.u128_type,
.u1_type,
.u8_type,
.usize_type,
.void_type,
.bool_false,
.bool_true,
.null_value,
.undef,
.noreturn_type,
=> return false,
.comptime_float_type,
.comptime_int_type,
.type_type,
=> return true,
else => unreachable, // that's all the values from `primitives`.
} else {
return false;
}
},
}
}
}
/// Returns `true` if the node uses `gz.anon_name_strategy`.
fn nodeUsesAnonNameStrategy(tree: *const Ast, node: Ast.Node.Index) bool {
const node_tags = tree.nodes.items(.tag);
switch (node_tags[node]) {
.container_decl,
.container_decl_trailing,
.container_decl_two,
.container_decl_two_trailing,
.container_decl_arg,
.container_decl_arg_trailing,
.tagged_union,
.tagged_union_trailing,
.tagged_union_two,
.tagged_union_two_trailing,
.tagged_union_enum_tag,
.tagged_union_enum_tag_trailing,
=> return true,
.builtin_call_two, .builtin_call_two_comma, .builtin_call, .builtin_call_comma => {
const builtin_token = tree.nodes.items(.main_token)[node];
const builtin_name = tree.tokenSlice(builtin_token);
return std.mem.eql(u8, builtin_name, "@Type");
},
else => return false,
}
}
/// Applies `rl` semantics to `result`. Expressions which do not do their own handling of
/// result locations must call this function on their result.
/// As an example, if `ri.rl` is `.ptr`, it will write the result to the pointer.
/// If `ri.rl` is `.ty`, it will coerce the result to the type.
/// Assumes nothing stacked on `gz`.
fn rvalue(
gz: *GenZir,
ri: ResultInfo,
raw_result: Zir.Inst.Ref,
src_node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
return rvalueInner(gz, ri, raw_result, src_node, true);
}
/// Like `rvalue`, but refuses to perform coercions before taking references for
/// the `ref_coerced_ty` result type. This is used for local variables which do
/// not have `alloc`s, because we want variables to have consistent addresses,
/// i.e. we want them to act like lvalues.
fn rvalueNoCoercePreRef(
gz: *GenZir,
ri: ResultInfo,
raw_result: Zir.Inst.Ref,
src_node: Ast.Node.Index,
) InnerError!Zir.Inst.Ref {
return rvalueInner(gz, ri, raw_result, src_node, false);
}
fn rvalueInner(
gz: *GenZir,
ri: ResultInfo,
raw_result: Zir.Inst.Ref,
src_node: Ast.Node.Index,
allow_coerce_pre_ref: bool,
) InnerError!Zir.Inst.Ref {
const result = r: {
if (raw_result.toIndex()) |result_index| {
const zir_tags = gz.astgen.instructions.items(.tag);
const data = gz.astgen.instructions.items(.data)[@intFromEnum(result_index)];
if (zir_tags[@intFromEnum(result_index)].isAlwaysVoid(data)) {
break :r Zir.Inst.Ref.void_value;
}
}
break :r raw_result;
};
if (gz.endsWithNoReturn()) return result;
switch (ri.rl) {
.none, .coerced_ty => return result,
.discard => {
// Emit a compile error for discarding error values.
_ = try gz.addUnNode(.ensure_result_non_error, result, src_node);
return .void_value;
},
.ref, .ref_coerced_ty => {
const coerced_result = if (allow_coerce_pre_ref and ri.rl == .ref_coerced_ty) res: {
const ptr_ty = ri.rl.ref_coerced_ty;
break :res try gz.addPlNode(.coerce_ptr_elem_ty, src_node, Zir.Inst.Bin{
.lhs = ptr_ty,
.rhs = result,
});
} else result;
// We need a pointer but we have a value.
// Unfortunately it's not quite as simple as directly emitting a ref
// instruction here because we need subsequent address-of operator on
// const locals to return the same address.
const astgen = gz.astgen;
const tree = astgen.tree;
const src_token = tree.firstToken(src_node);
const result_index = coerced_result.toIndex() orelse
return gz.addUnTok(.ref, coerced_result, src_token);
const zir_tags = gz.astgen.instructions.items(.tag);
if (zir_tags[@intFromEnum(result_index)].isParam() or astgen.isInferred(coerced_result))
return gz.addUnTok(.ref, coerced_result, src_token);
const gop = try astgen.ref_table.getOrPut(astgen.gpa, result_index);
if (!gop.found_existing) {
gop.value_ptr.* = try gz.makeUnTok(.ref, coerced_result, src_token);
}
return gop.value_ptr.*.toRef();
},
.ty => |ty_inst| {
// Quickly eliminate some common, unnecessary type coercion.
const as_ty = @as(u64, @intFromEnum(Zir.Inst.Ref.type_type)) << 32;
const as_bool = @as(u64, @intFromEnum(Zir.Inst.Ref.bool_type)) << 32;
const as_void = @as(u64, @intFromEnum(Zir.Inst.Ref.void_type)) << 32;
const as_comptime_int = @as(u64, @intFromEnum(Zir.Inst.Ref.comptime_int_type)) << 32;
const as_usize = @as(u64, @intFromEnum(Zir.Inst.Ref.usize_type)) << 32;
const as_u8 = @as(u64, @intFromEnum(Zir.Inst.Ref.u8_type)) << 32;
switch ((@as(u64, @intFromEnum(ty_inst)) << 32) | @as(u64, @intFromEnum(result))) {
as_ty | @intFromEnum(Zir.Inst.Ref.u1_type),
as_ty | @intFromEnum(Zir.Inst.Ref.u8_type),
as_ty | @intFromEnum(Zir.Inst.Ref.i8_type),
as_ty | @intFromEnum(Zir.Inst.Ref.u16_type),
as_ty | @intFromEnum(Zir.Inst.Ref.u29_type),
as_ty | @intFromEnum(Zir.Inst.Ref.i16_type),
as_ty | @intFromEnum(Zir.Inst.Ref.u32_type),
as_ty | @intFromEnum(Zir.Inst.Ref.i32_type),
as_ty | @intFromEnum(Zir.Inst.Ref.u64_type),
as_ty | @intFromEnum(Zir.Inst.Ref.i64_type),
as_ty | @intFromEnum(Zir.Inst.Ref.u128_type),
as_ty | @intFromEnum(Zir.Inst.Ref.i128_type),
as_ty | @intFromEnum(Zir.Inst.Ref.usize_type),
as_ty | @intFromEnum(Zir.Inst.Ref.isize_type),
as_ty | @intFromEnum(Zir.Inst.Ref.c_char_type),
as_ty | @intFromEnum(Zir.Inst.Ref.c_short_type),
as_ty | @intFromEnum(Zir.Inst.Ref.c_ushort_type),
as_ty | @intFromEnum(Zir.Inst.Ref.c_int_type),
as_ty | @intFromEnum(Zir.Inst.Ref.c_uint_type),
as_ty | @intFromEnum(Zir.Inst.Ref.c_long_type),
as_ty | @intFromEnum(Zir.Inst.Ref.c_ulong_type),
as_ty | @intFromEnum(Zir.Inst.Ref.c_longlong_type),
as_ty | @intFromEnum(Zir.Inst.Ref.c_ulonglong_type),
as_ty | @intFromEnum(Zir.Inst.Ref.c_longdouble_type),
as_ty | @intFromEnum(Zir.Inst.Ref.f16_type),
as_ty | @intFromEnum(Zir.Inst.Ref.f32_type),
as_ty | @intFromEnum(Zir.Inst.Ref.f64_type),
as_ty | @intFromEnum(Zir.Inst.Ref.f80_type),
as_ty | @intFromEnum(Zir.Inst.Ref.f128_type),
as_ty | @intFromEnum(Zir.Inst.Ref.anyopaque_type),
as_ty | @intFromEnum(Zir.Inst.Ref.bool_type),
as_ty | @intFromEnum(Zir.Inst.Ref.void_type),
as_ty | @intFromEnum(Zir.Inst.Ref.type_type),
as_ty | @intFromEnum(Zir.Inst.Ref.anyerror_type),
as_ty | @intFromEnum(Zir.Inst.Ref.comptime_int_type),
as_ty | @intFromEnum(Zir.Inst.Ref.comptime_float_type),
as_ty | @intFromEnum(Zir.Inst.Ref.noreturn_type),
as_ty | @intFromEnum(Zir.Inst.Ref.anyframe_type),
as_ty | @intFromEnum(Zir.Inst.Ref.null_type),
as_ty | @intFromEnum(Zir.Inst.Ref.undefined_type),
as_ty | @intFromEnum(Zir.Inst.Ref.enum_literal_type),
as_ty | @intFromEnum(Zir.Inst.Ref.atomic_order_type),
as_ty | @intFromEnum(Zir.Inst.Ref.atomic_rmw_op_type),
as_ty | @intFromEnum(Zir.Inst.Ref.calling_convention_type),
as_ty | @intFromEnum(Zir.Inst.Ref.address_space_type),
as_ty | @intFromEnum(Zir.Inst.Ref.float_mode_type),
as_ty | @intFromEnum(Zir.Inst.Ref.reduce_op_type),
as_ty | @intFromEnum(Zir.Inst.Ref.call_modifier_type),
as_ty | @intFromEnum(Zir.Inst.Ref.prefetch_options_type),
as_ty | @intFromEnum(Zir.Inst.Ref.export_options_type),
as_ty | @intFromEnum(Zir.Inst.Ref.extern_options_type),
as_ty | @intFromEnum(Zir.Inst.Ref.type_info_type),
as_ty | @intFromEnum(Zir.Inst.Ref.manyptr_u8_type),
as_ty | @intFromEnum(Zir.Inst.Ref.manyptr_const_u8_type),
as_ty | @intFromEnum(Zir.Inst.Ref.manyptr_const_u8_sentinel_0_type),
as_ty | @intFromEnum(Zir.Inst.Ref.single_const_pointer_to_comptime_int_type),
as_ty | @intFromEnum(Zir.Inst.Ref.slice_const_u8_type),
as_ty | @intFromEnum(Zir.Inst.Ref.slice_const_u8_sentinel_0_type),
as_ty | @intFromEnum(Zir.Inst.Ref.anyerror_void_error_union_type),
as_ty | @intFromEnum(Zir.Inst.Ref.generic_poison_type),
as_ty | @intFromEnum(Zir.Inst.Ref.empty_struct_type),
as_comptime_int | @intFromEnum(Zir.Inst.Ref.zero),
as_comptime_int | @intFromEnum(Zir.Inst.Ref.one),
as_comptime_int | @intFromEnum(Zir.Inst.Ref.negative_one),
as_usize | @intFromEnum(Zir.Inst.Ref.zero_usize),
as_usize | @intFromEnum(Zir.Inst.Ref.one_usize),
as_u8 | @intFromEnum(Zir.Inst.Ref.zero_u8),
as_u8 | @intFromEnum(Zir.Inst.Ref.one_u8),
as_u8 | @intFromEnum(Zir.Inst.Ref.four_u8),
as_bool | @intFromEnum(Zir.Inst.Ref.bool_true),
as_bool | @intFromEnum(Zir.Inst.Ref.bool_false),
as_void | @intFromEnum(Zir.Inst.Ref.void_value),
=> return result, // type of result is already correct
as_usize | @intFromEnum(Zir.Inst.Ref.zero) => return .zero_usize,
as_u8 | @intFromEnum(Zir.Inst.Ref.zero) => return .zero_u8,
as_usize | @intFromEnum(Zir.Inst.Ref.one) => return .one_usize,
as_u8 | @intFromEnum(Zir.Inst.Ref.one) => return .one_u8,
as_comptime_int | @intFromEnum(Zir.Inst.Ref.zero_usize) => return .zero,
as_u8 | @intFromEnum(Zir.Inst.Ref.zero_usize) => return .zero_u8,
as_comptime_int | @intFromEnum(Zir.Inst.Ref.one_usize) => return .one,
as_u8 | @intFromEnum(Zir.Inst.Ref.one_usize) => return .one_u8,
as_comptime_int | @intFromEnum(Zir.Inst.Ref.zero_u8) => return .zero,
as_usize | @intFromEnum(Zir.Inst.Ref.zero_u8) => return .zero_usize,
as_comptime_int | @intFromEnum(Zir.Inst.Ref.one_u8) => return .one,
as_usize | @intFromEnum(Zir.Inst.Ref.one_u8) => return .one_usize,
// Need an explicit type coercion instruction.
else => return gz.addPlNode(ri.zirTag(), src_node, Zir.Inst.As{
.dest_type = ty_inst,
.operand = result,
}),
}
},
.ptr => |ptr_res| {
_ = try gz.addPlNode(.store_node, ptr_res.src_node orelse src_node, Zir.Inst.Bin{
.lhs = ptr_res.inst,
.rhs = result,
});
return .void_value;
},
.inferred_ptr => |alloc| {
_ = try gz.addPlNode(.store_to_inferred_ptr, src_node, Zir.Inst.Bin{
.lhs = alloc,
.rhs = result,
});
return .void_value;
},
.destructure => |destructure| {
const components = destructure.components;
_ = try gz.addPlNode(.validate_destructure, src_node, Zir.Inst.ValidateDestructure{
.operand = result,
.destructure_node = gz.nodeIndexToRelative(destructure.src_node),
.expect_len = @intCast(components.len),
});
for (components, 0..) |component, i| {
if (component == .discard) continue;
const elem_val = try gz.add(.{
.tag = .elem_val_imm,
.data = .{ .elem_val_imm = .{
.operand = result,
.idx = @intCast(i),
} },
});
switch (component) {
.typed_ptr => |ptr_res| {
_ = try gz.addPlNode(.store_node, ptr_res.src_node orelse src_node, Zir.Inst.Bin{
.lhs = ptr_res.inst,
.rhs = elem_val,
});
},
.inferred_ptr => |ptr_inst| {
_ = try gz.addPlNode(.store_to_inferred_ptr, src_node, Zir.Inst.Bin{
.lhs = ptr_inst,
.rhs = elem_val,
});
},
.discard => unreachable,
}
}
return .void_value;
},
}
}
/// Given an identifier token, obtain the string for it.
/// If the token uses @"" syntax, parses as a string, reports errors if applicable,
/// and allocates the result within `astgen.arena`.
/// Otherwise, returns a reference to the source code bytes directly.
/// See also `appendIdentStr` and `parseStrLit`.
fn identifierTokenString(astgen: *AstGen, token: Ast.TokenIndex) InnerError![]const u8 {
const tree = astgen.tree;
const token_tags = tree.tokens.items(.tag);
assert(token_tags[token] == .identifier);
const ident_name = tree.tokenSlice(token);
if (!mem.startsWith(u8, ident_name, "@")) {
return ident_name;
}
var buf: ArrayListUnmanaged(u8) = .{};
defer buf.deinit(astgen.gpa);
try astgen.parseStrLit(token, &buf, ident_name, 1);
if (mem.indexOfScalar(u8, buf.items, 0) != null) {
return astgen.failTok(token, "identifier cannot contain null bytes", .{});
} else if (buf.items.len == 0) {
return astgen.failTok(token, "identifier cannot be empty", .{});
}
const duped = try astgen.arena.dupe(u8, buf.items);
return duped;
}
/// Given an identifier token, obtain the string for it (possibly parsing as a string
/// literal if it is @"" syntax), and append the string to `buf`.
/// See also `identifierTokenString` and `parseStrLit`.
fn appendIdentStr(
astgen: *AstGen,
token: Ast.TokenIndex,
buf: *ArrayListUnmanaged(u8),
) InnerError!void {
const tree = astgen.tree;
const token_tags = tree.tokens.items(.tag);
assert(token_tags[token] == .identifier);
const ident_name = tree.tokenSlice(token);
if (!mem.startsWith(u8, ident_name, "@")) {
return buf.appendSlice(astgen.gpa, ident_name);
} else {
const start = buf.items.len;
try astgen.parseStrLit(token, buf, ident_name, 1);
const slice = buf.items[start..];
if (mem.indexOfScalar(u8, slice, 0) != null) {
return astgen.failTok(token, "identifier cannot contain null bytes", .{});
} else if (slice.len == 0) {
return astgen.failTok(token, "identifier cannot be empty", .{});
}
}
}
/// Appends the result to `buf`.
fn parseStrLit(
astgen: *AstGen,
token: Ast.TokenIndex,
buf: *ArrayListUnmanaged(u8),
bytes: []const u8,
offset: u32,
) InnerError!void {
const raw_string = bytes[offset..];
var buf_managed = buf.toManaged(astgen.gpa);
const result = std.zig.string_literal.parseWrite(buf_managed.writer(), raw_string);
buf.* = buf_managed.moveToUnmanaged();
switch (try result) {
.success => return,
.failure => |err| return astgen.failWithStrLitError(err, token, bytes, offset),
}
}
fn failWithStrLitError(astgen: *AstGen, err: std.zig.string_literal.Error, token: Ast.TokenIndex, bytes: []const u8, offset: u32) InnerError {
const raw_string = bytes[offset..];
switch (err) {
.invalid_escape_character => |bad_index| {
return astgen.failOff(
token,
offset + @as(u32, @intCast(bad_index)),
"invalid escape character: '{c}'",
.{raw_string[bad_index]},
);
},
.expected_hex_digit => |bad_index| {
return astgen.failOff(
token,
offset + @as(u32, @intCast(bad_index)),
"expected hex digit, found '{c}'",
.{raw_string[bad_index]},
);
},
.empty_unicode_escape_sequence => |bad_index| {
return astgen.failOff(
token,
offset + @as(u32, @intCast(bad_index)),
"empty unicode escape sequence",
.{},
);
},
.expected_hex_digit_or_rbrace => |bad_index| {
return astgen.failOff(
token,
offset + @as(u32, @intCast(bad_index)),
"expected hex digit or '}}', found '{c}'",
.{raw_string[bad_index]},
);
},
.invalid_unicode_codepoint => |bad_index| {
return astgen.failOff(
token,
offset + @as(u32, @intCast(bad_index)),
"unicode escape does not correspond to a valid codepoint",
.{},
);
},
.expected_lbrace => |bad_index| {
return astgen.failOff(
token,
offset + @as(u32, @intCast(bad_index)),
"expected '{{', found '{c}",
.{raw_string[bad_index]},
);
},
.expected_rbrace => |bad_index| {
return astgen.failOff(
token,
offset + @as(u32, @intCast(bad_index)),
"expected '}}', found '{c}",
.{raw_string[bad_index]},
);
},
.expected_single_quote => |bad_index| {
return astgen.failOff(
token,
offset + @as(u32, @intCast(bad_index)),
"expected single quote ('), found '{c}",
.{raw_string[bad_index]},
);
},
.invalid_character => |bad_index| {
return astgen.failOff(
token,
offset + @as(u32, @intCast(bad_index)),
"invalid byte in string or character literal: '{c}'",
.{raw_string[bad_index]},
);
},
}
}
fn failNode(
astgen: *AstGen,
node: Ast.Node.Index,
comptime format: []const u8,
args: anytype,
) InnerError {
return astgen.failNodeNotes(node, format, args, &[0]u32{});
}
fn appendErrorNode(
astgen: *AstGen,
node: Ast.Node.Index,
comptime format: []const u8,
args: anytype,
) Allocator.Error!void {
try astgen.appendErrorNodeNotes(node, format, args, &[0]u32{});
}
fn appendErrorNodeNotes(
astgen: *AstGen,
node: Ast.Node.Index,
comptime format: []const u8,
args: anytype,
notes: []const u32,
) Allocator.Error!void {
@setCold(true);
const string_bytes = &astgen.string_bytes;
const msg: Zir.NullTerminatedString = @enumFromInt(string_bytes.items.len);
try string_bytes.writer(astgen.gpa).print(format ++ "\x00", args);
const notes_index: u32 = if (notes.len != 0) blk: {
const notes_start = astgen.extra.items.len;
try astgen.extra.ensureTotalCapacity(astgen.gpa, notes_start + 1 + notes.len);
astgen.extra.appendAssumeCapacity(@intCast(notes.len));
astgen.extra.appendSliceAssumeCapacity(notes);
break :blk @intCast(notes_start);
} else 0;
try astgen.compile_errors.append(astgen.gpa, .{
.msg = msg,
.node = node,
.token = 0,
.byte_offset = 0,
.notes = notes_index,
});
}
fn failNodeNotes(
astgen: *AstGen,
node: Ast.Node.Index,
comptime format: []const u8,
args: anytype,
notes: []const u32,
) InnerError {
try appendErrorNodeNotes(astgen, node, format, args, notes);
return error.AnalysisFail;
}
fn failTok(
astgen: *AstGen,
token: Ast.TokenIndex,
comptime format: []const u8,
args: anytype,
) InnerError {
return astgen.failTokNotes(token, format, args, &[0]u32{});
}
fn appendErrorTok(
astgen: *AstGen,
token: Ast.TokenIndex,
comptime format: []const u8,
args: anytype,
) !void {
try astgen.appendErrorTokNotesOff(token, 0, format, args, &[0]u32{});
}
fn failTokNotes(
astgen: *AstGen,
token: Ast.TokenIndex,
comptime format: []const u8,
args: anytype,
notes: []const u32,
) InnerError {
try appendErrorTokNotesOff(astgen, token, 0, format, args, notes);
return error.AnalysisFail;
}
fn appendErrorTokNotes(
astgen: *AstGen,
token: Ast.TokenIndex,
comptime format: []const u8,
args: anytype,
notes: []const u32,
) !void {
return appendErrorTokNotesOff(astgen, token, 0, format, args, notes);
}
/// Same as `fail`, except given a token plus an offset from its starting byte
/// offset.
fn failOff(
astgen: *AstGen,
token: Ast.TokenIndex,
byte_offset: u32,
comptime format: []const u8,
args: anytype,
) InnerError {
try appendErrorTokNotesOff(astgen, token, byte_offset, format, args, &.{});
return error.AnalysisFail;
}
fn appendErrorTokNotesOff(
astgen: *AstGen,
token: Ast.TokenIndex,
byte_offset: u32,
comptime format: []const u8,
args: anytype,
notes: []const u32,
) !void {
@setCold(true);
const gpa = astgen.gpa;
const string_bytes = &astgen.string_bytes;
const msg: Zir.NullTerminatedString = @enumFromInt(string_bytes.items.len);
try string_bytes.writer(gpa).print(format ++ "\x00", args);
const notes_index: u32 = if (notes.len != 0) blk: {
const notes_start = astgen.extra.items.len;
try astgen.extra.ensureTotalCapacity(gpa, notes_start + 1 + notes.len);
astgen.extra.appendAssumeCapacity(@intCast(notes.len));
astgen.extra.appendSliceAssumeCapacity(notes);
break :blk @intCast(notes_start);
} else 0;
try astgen.compile_errors.append(gpa, .{
.msg = msg,
.node = 0,
.token = token,
.byte_offset = byte_offset,
.notes = notes_index,
});
}
fn errNoteTok(
astgen: *AstGen,
token: Ast.TokenIndex,
comptime format: []const u8,
args: anytype,
) Allocator.Error!u32 {
return errNoteTokOff(astgen, token, 0, format, args);
}
fn errNoteTokOff(
astgen: *AstGen,
token: Ast.TokenIndex,
byte_offset: u32,
comptime format: []const u8,
args: anytype,
) Allocator.Error!u32 {
@setCold(true);
const string_bytes = &astgen.string_bytes;
const msg: Zir.NullTerminatedString = @enumFromInt(string_bytes.items.len);
try string_bytes.writer(astgen.gpa).print(format ++ "\x00", args);
return astgen.addExtra(Zir.Inst.CompileErrors.Item{
.msg = msg,
.node = 0,
.token = token,
.byte_offset = byte_offset,
.notes = 0,
});
}
fn errNoteNode(
astgen: *AstGen,
node: Ast.Node.Index,
comptime format: []const u8,
args: anytype,
) Allocator.Error!u32 {
@setCold(true);
const string_bytes = &astgen.string_bytes;
const msg: Zir.NullTerminatedString = @enumFromInt(string_bytes.items.len);
try string_bytes.writer(astgen.gpa).print(format ++ "\x00", args);
return astgen.addExtra(Zir.Inst.CompileErrors.Item{
.msg = msg,
.node = node,
.token = 0,
.byte_offset = 0,
.notes = 0,
});
}
fn identAsString(astgen: *AstGen, ident_token: Ast.TokenIndex) !Zir.NullTerminatedString {
const gpa = astgen.gpa;
const string_bytes = &astgen.string_bytes;
const str_index: u32 = @intCast(string_bytes.items.len);
try astgen.appendIdentStr(ident_token, string_bytes);
const key: []const u8 = string_bytes.items[str_index..];
const gop = try astgen.string_table.getOrPutContextAdapted(gpa, key, StringIndexAdapter{
.bytes = string_bytes,
}, StringIndexContext{
.bytes = string_bytes,
});
if (gop.found_existing) {
string_bytes.shrinkRetainingCapacity(str_index);
return @enumFromInt(gop.key_ptr.*);
} else {
gop.key_ptr.* = str_index;
try string_bytes.append(gpa, 0);
return @enumFromInt(str_index);
}
}
/// Adds a doc comment block to `string_bytes` by walking backwards from `end_token`.
/// `end_token` must point at the first token after the last doc coment line.
/// Returns 0 if no doc comment is present.
fn docCommentAsString(astgen: *AstGen, end_token: Ast.TokenIndex) !Zir.NullTerminatedString {
if (end_token == 0) return .empty;
const token_tags = astgen.tree.tokens.items(.tag);
var tok = end_token - 1;
while (token_tags[tok] == .doc_comment) {
if (tok == 0) break;
tok -= 1;
} else {
tok += 1;
}
return docCommentAsStringFromFirst(astgen, end_token, tok);
}
/// end_token must be > the index of the last doc comment.
fn docCommentAsStringFromFirst(
astgen: *AstGen,
end_token: Ast.TokenIndex,
start_token: Ast.TokenIndex,
) !Zir.NullTerminatedString {
if (start_token == end_token) return .empty;
const gpa = astgen.gpa;
const string_bytes = &astgen.string_bytes;
const str_index: u32 = @intCast(string_bytes.items.len);
const token_starts = astgen.tree.tokens.items(.start);
const token_tags = astgen.tree.tokens.items(.tag);
const total_bytes = token_starts[end_token] - token_starts[start_token];
try string_bytes.ensureUnusedCapacity(gpa, total_bytes);
var current_token = start_token;
while (current_token < end_token) : (current_token += 1) {
switch (token_tags[current_token]) {
.doc_comment => {
const tok_bytes = astgen.tree.tokenSlice(current_token)[3..];
string_bytes.appendSliceAssumeCapacity(tok_bytes);
if (current_token != end_token - 1) {
string_bytes.appendAssumeCapacity('\n');
}
},
else => break,
}
}
const key: []const u8 = string_bytes.items[str_index..];
const gop = try astgen.string_table.getOrPutContextAdapted(gpa, key, StringIndexAdapter{
.bytes = string_bytes,
}, StringIndexContext{
.bytes = string_bytes,
});
if (gop.found_existing) {
string_bytes.shrinkRetainingCapacity(str_index);
return @enumFromInt(gop.key_ptr.*);
} else {
gop.key_ptr.* = str_index;
try string_bytes.append(gpa, 0);
return @enumFromInt(str_index);
}
}
const IndexSlice = struct { index: Zir.NullTerminatedString, len: u32 };
fn strLitAsString(astgen: *AstGen, str_lit_token: Ast.TokenIndex) !IndexSlice {
const gpa = astgen.gpa;
const string_bytes = &astgen.string_bytes;
const str_index: u32 = @intCast(string_bytes.items.len);
const token_bytes = astgen.tree.tokenSlice(str_lit_token);
try astgen.parseStrLit(str_lit_token, string_bytes, token_bytes, 0);
const key: []const u8 = string_bytes.items[str_index..];
if (std.mem.indexOfScalar(u8, key, 0)) |_| return .{
.index = @enumFromInt(str_index),
.len = @intCast(key.len),
};
const gop = try astgen.string_table.getOrPutContextAdapted(gpa, key, StringIndexAdapter{
.bytes = string_bytes,
}, StringIndexContext{
.bytes = string_bytes,
});
if (gop.found_existing) {
string_bytes.shrinkRetainingCapacity(str_index);
return .{
.index = @enumFromInt(gop.key_ptr.*),
.len = @intCast(key.len),
};
} else {
gop.key_ptr.* = str_index;
// Still need a null byte because we are using the same table
// to lookup null terminated strings, so if we get a match, it has to
// be null terminated for that to work.
try string_bytes.append(gpa, 0);
return .{
.index = @enumFromInt(str_index),
.len = @intCast(key.len),
};
}
}
fn strLitNodeAsString(astgen: *AstGen, node: Ast.Node.Index) !IndexSlice {
const tree = astgen.tree;
const node_datas = tree.nodes.items(.data);
const start = node_datas[node].lhs;
const end = node_datas[node].rhs;
const gpa = astgen.gpa;
const string_bytes = &astgen.string_bytes;
const str_index = string_bytes.items.len;
// First line: do not append a newline.
var tok_i = start;
{
const slice = tree.tokenSlice(tok_i);
const carriage_return_ending: usize = if (slice[slice.len - 2] == '\r') 2 else 1;
const line_bytes = slice[2 .. slice.len - carriage_return_ending];
try string_bytes.appendSlice(gpa, line_bytes);
tok_i += 1;
}
// Following lines: each line prepends a newline.
while (tok_i <= end) : (tok_i += 1) {
const slice = tree.tokenSlice(tok_i);
const carriage_return_ending: usize = if (slice[slice.len - 2] == '\r') 2 else 1;
const line_bytes = slice[2 .. slice.len - carriage_return_ending];
try string_bytes.ensureUnusedCapacity(gpa, line_bytes.len + 1);
string_bytes.appendAssumeCapacity('\n');
string_bytes.appendSliceAssumeCapacity(line_bytes);
}
const len = string_bytes.items.len - str_index;
try string_bytes.append(gpa, 0);
return IndexSlice{
.index = @enumFromInt(str_index),
.len = @intCast(len),
};
}
fn testNameString(astgen: *AstGen, str_lit_token: Ast.TokenIndex) !Zir.NullTerminatedString {
const gpa = astgen.gpa;
const string_bytes = &astgen.string_bytes;
const str_index: u32 = @intCast(string_bytes.items.len);
const token_bytes = astgen.tree.tokenSlice(str_lit_token);
try string_bytes.append(gpa, 0); // Indicates this is a test.
try astgen.parseStrLit(str_lit_token, string_bytes, token_bytes, 0);
const slice = string_bytes.items[str_index + 1 ..];
if (mem.indexOfScalar(u8, slice, 0) != null) {
return astgen.failTok(str_lit_token, "test name cannot contain null bytes", .{});
} else if (slice.len == 0) {
return astgen.failTok(str_lit_token, "empty test name must be omitted", .{});
}
try string_bytes.append(gpa, 0);
return @enumFromInt(str_index);
}
const Scope = struct {
tag: Tag,
fn cast(base: *Scope, comptime T: type) ?*T {
if (T == Defer) {
switch (base.tag) {
.defer_normal, .defer_error => return @alignCast(@fieldParentPtr(*align(1) T, "base", base)),
else => return null,
}
}
if (T == Namespace) {
switch (base.tag) {
.namespace => return @alignCast(@fieldParentPtr(*align(1) T, "base", base)),
else => return null,
}
}
if (base.tag != T.base_tag)
return null;
return @alignCast(@fieldParentPtr(*align(1) T, "base", base));
}
fn parent(base: *Scope) ?*Scope {
return switch (base.tag) {
.gen_zir => base.cast(GenZir).?.parent,
.local_val => base.cast(LocalVal).?.parent,
.local_ptr => base.cast(LocalPtr).?.parent,
.defer_normal, .defer_error => base.cast(Defer).?.parent,
.namespace => base.cast(Namespace).?.parent,
.top => null,
};
}
const Tag = enum {
gen_zir,
local_val,
local_ptr,
defer_normal,
defer_error,
namespace,
top,
};
/// The category of identifier. These tag names are user-visible in compile errors.
const IdCat = enum {
@"function parameter",
@"local constant",
@"local variable",
@"switch tag capture",
capture,
};
/// This is always a `const` local and importantly the `inst` is a value type, not a pointer.
/// This structure lives as long as the AST generation of the Block
/// node that contains the variable.
const LocalVal = struct {
const base_tag: Tag = .local_val;
base: Scope = Scope{ .tag = base_tag },
/// Parents can be: `LocalVal`, `LocalPtr`, `GenZir`, `Defer`, `Namespace`.
parent: *Scope,
gen_zir: *GenZir,
inst: Zir.Inst.Ref,
/// Source location of the corresponding variable declaration.
token_src: Ast.TokenIndex,
/// Track the first identifer where it is referenced.
/// 0 means never referenced.
used: Ast.TokenIndex = 0,
/// Track the identifier where it is discarded, like this `_ = foo;`.
/// 0 means never discarded.
discarded: Ast.TokenIndex = 0,
/// String table index.
name: Zir.NullTerminatedString,
id_cat: IdCat,
};
/// This could be a `const` or `var` local. It has a pointer instead of a value.
/// This structure lives as long as the AST generation of the Block
/// node that contains the variable.
const LocalPtr = struct {
const base_tag: Tag = .local_ptr;
base: Scope = Scope{ .tag = base_tag },
/// Parents can be: `LocalVal`, `LocalPtr`, `GenZir`, `Defer`, `Namespace`.
parent: *Scope,
gen_zir: *GenZir,
ptr: Zir.Inst.Ref,
/// Source location of the corresponding variable declaration.
token_src: Ast.TokenIndex,
/// Track the first identifer where it is referenced.
/// 0 means never referenced.
used: Ast.TokenIndex = 0,
/// Track the identifier where it is discarded, like this `_ = foo;`.
/// 0 means never discarded.
discarded: Ast.TokenIndex = 0,
/// Whether this value is used as an lvalue after inititialization.
/// If not, we know it can be `const`, so will emit a compile error if it is `var`.
used_as_lvalue: bool = false,
/// String table index.
name: Zir.NullTerminatedString,
id_cat: IdCat,
/// true means we find out during Sema whether the value is comptime.
/// false means it is already known at AstGen the value is runtime-known.
maybe_comptime: bool,
};
const Defer = struct {
base: Scope,
/// Parents can be: `LocalVal`, `LocalPtr`, `GenZir`, `Defer`, `Namespace`.
parent: *Scope,
index: u32,
len: u32,
remapped_err_code: Zir.Inst.OptionalIndex = .none,
};
/// Represents a global scope that has any number of declarations in it.
/// Each declaration has this as the parent scope.
const Namespace = struct {
const base_tag: Tag = .namespace;
base: Scope = Scope{ .tag = base_tag },
/// Parents can be: `LocalVal`, `LocalPtr`, `GenZir`, `Defer`, `Namespace`.
parent: *Scope,
/// Maps string table index to the source location of declaration,
/// for the purposes of reporting name shadowing compile errors.
decls: std.AutoHashMapUnmanaged(Zir.NullTerminatedString, Ast.Node.Index) = .{},
node: Ast.Node.Index,
inst: Zir.Inst.Index,
maybe_generic: bool,
/// The astgen scope containing this namespace.
/// Only valid during astgen.
declaring_gz: ?*GenZir,
/// Set of captures used by this namespace.
captures: std.AutoArrayHashMapUnmanaged(Zir.Inst.Capture, void) = .{},
fn deinit(self: *Namespace, gpa: Allocator) void {
self.decls.deinit(gpa);
self.captures.deinit(gpa);
self.* = undefined;
}
};
const Top = struct {
const base_tag: Scope.Tag = .top;
base: Scope = Scope{ .tag = base_tag },
};
};
/// This is a temporary structure; references to it are valid only
/// while constructing a `Zir`.
const GenZir = struct {
const base_tag: Scope.Tag = .gen_zir;
base: Scope = Scope{ .tag = base_tag },
/// Whether we're already in a scope known to be comptime. This is set
/// whenever we know Sema will analyze the current block with `is_comptime`,
/// for instance when we're within a `struct_decl` or a `block_comptime`.
is_comptime: bool,
/// Whether we're in an expression within a `@TypeOf` operand. In this case, closure of runtime
/// variables is permitted where it is usually not.
is_typeof: bool = false,
/// This is set to true for a `GenZir` of a `block_inline`, indicating that
/// exits from this block should use `break_inline` rather than `break`.
is_inline: bool = false,
c_import: bool = false,
/// How decls created in this scope should be named.
anon_name_strategy: Zir.Inst.NameStrategy = .anon,
/// The containing decl AST node.
decl_node_index: Ast.Node.Index,
/// The containing decl line index, absolute.
decl_line: u32,
/// Parents can be: `LocalVal`, `LocalPtr`, `GenZir`, `Defer`, `Namespace`.
parent: *Scope,
/// All `GenZir` scopes for the same ZIR share this.
astgen: *AstGen,
/// Keeps track of the list of instructions in this scope. Possibly shared.
/// Indexes to instructions in `astgen`.
instructions: *ArrayListUnmanaged(Zir.Inst.Index),
/// A sub-block may share its instructions ArrayList with containing GenZir,
/// if use is strictly nested. This saves prior size of list for unstacking.
instructions_top: usize,
label: ?Label = null,
break_block: Zir.Inst.OptionalIndex = .none,
continue_block: Zir.Inst.OptionalIndex = .none,
/// Only valid when setBreakResultInfo is called.
break_result_info: AstGen.ResultInfo = undefined,
suspend_node: Ast.Node.Index = 0,
nosuspend_node: Ast.Node.Index = 0,
/// Set if this GenZir is a defer.
cur_defer_node: Ast.Node.Index = 0,
// Set if this GenZir is a defer or it is inside a defer.
any_defer_node: Ast.Node.Index = 0,
const unstacked_top = std.math.maxInt(usize);
/// Call unstack before adding any new instructions to containing GenZir.
fn unstack(self: *GenZir) void {
if (self.instructions_top != unstacked_top) {
self.instructions.items.len = self.instructions_top;
self.instructions_top = unstacked_top;
}
}
fn isEmpty(self: *const GenZir) bool {
return (self.instructions_top == unstacked_top) or
(self.instructions.items.len == self.instructions_top);
}
fn instructionsSlice(self: *const GenZir) []Zir.Inst.Index {
return if (self.instructions_top == unstacked_top)
&[0]Zir.Inst.Index{}
else
self.instructions.items[self.instructions_top..];
}
fn instructionsSliceUpto(self: *const GenZir, stacked_gz: *GenZir) []Zir.Inst.Index {
return if (self.instructions_top == unstacked_top)
&[0]Zir.Inst.Index{}
else if (self.instructions == stacked_gz.instructions and stacked_gz.instructions_top != unstacked_top)
self.instructions.items[self.instructions_top..stacked_gz.instructions_top]
else
self.instructions.items[self.instructions_top..];
}
fn makeSubBlock(gz: *GenZir, scope: *Scope) GenZir {
return .{
.is_comptime = gz.is_comptime,
.is_typeof = gz.is_typeof,
.c_import = gz.c_import,
.decl_node_index = gz.decl_node_index,
.decl_line = gz.decl_line,
.parent = scope,
.astgen = gz.astgen,
.suspend_node = gz.suspend_node,
.nosuspend_node = gz.nosuspend_node,
.any_defer_node = gz.any_defer_node,
.instructions = gz.instructions,
.instructions_top = gz.instructions.items.len,
};
}
const Label = struct {
token: Ast.TokenIndex,
block_inst: Zir.Inst.Index,
used: bool = false,
};
/// Assumes nothing stacked on `gz`.
fn endsWithNoReturn(gz: GenZir) bool {
if (gz.isEmpty()) return false;
const tags = gz.astgen.instructions.items(.tag);
const last_inst = gz.instructions.items[gz.instructions.items.len - 1];
return tags[@intFromEnum(last_inst)].isNoReturn();
}
/// TODO all uses of this should be replaced with uses of `endsWithNoReturn`.
fn refIsNoReturn(gz: GenZir, inst_ref: Zir.Inst.Ref) bool {
if (inst_ref == .unreachable_value) return true;
if (inst_ref.toIndex()) |inst_index| {
return gz.astgen.instructions.items(.tag)[@intFromEnum(inst_index)].isNoReturn();
}
return false;
}
fn nodeIndexToRelative(gz: GenZir, node_index: Ast.Node.Index) i32 {
return @as(i32, @bitCast(node_index)) - @as(i32, @bitCast(gz.decl_node_index));
}
fn tokenIndexToRelative(gz: GenZir, token: Ast.TokenIndex) u32 {
return token - gz.srcToken();
}
fn srcToken(gz: GenZir) Ast.TokenIndex {
return gz.astgen.tree.firstToken(gz.decl_node_index);
}
fn setBreakResultInfo(gz: *GenZir, parent_ri: AstGen.ResultInfo) void {
// Depending on whether the result location is a pointer or value, different
// ZIR needs to be generated. In the former case we rely on storing to the
// pointer to communicate the result, and use breakvoid; in the latter case
// the block break instructions will have the result values.
switch (parent_ri.rl) {
.coerced_ty => |ty_inst| {
// Type coercion needs to happen before breaks.
gz.break_result_info = .{ .rl = .{ .ty = ty_inst }, .ctx = parent_ri.ctx };
},
.discard => {
// We don't forward the result context here. This prevents
// "unnecessary discard" errors from being caused by expressions
// far from the actual discard, such as a `break` from a
// discarded block.
gz.break_result_info = .{ .rl = .discard };
},
else => {
gz.break_result_info = parent_ri;
},
}
}
/// Assumes nothing stacked on `gz`. Unstacks `gz`.
fn setBoolBrBody(gz: *GenZir, bool_br: Zir.Inst.Index, bool_br_lhs: Zir.Inst.Ref) !void {
const astgen = gz.astgen;
const gpa = astgen.gpa;
const body = gz.instructionsSlice();
const body_len = astgen.countBodyLenAfterFixups(body);
try astgen.extra.ensureUnusedCapacity(
gpa,
@typeInfo(Zir.Inst.BoolBr).Struct.fields.len + body_len,
);
const zir_datas = astgen.instructions.items(.data);
zir_datas[@intFromEnum(bool_br)].pl_node.payload_index = astgen.addExtraAssumeCapacity(Zir.Inst.BoolBr{
.lhs = bool_br_lhs,
.body_len = body_len,
});
astgen.appendBodyWithFixups(body);
gz.unstack();
}
/// Assumes nothing stacked on `gz`. Unstacks `gz`.
fn setBlockBody(gz: *GenZir, inst: Zir.Inst.Index) !void {
const astgen = gz.astgen;
const gpa = astgen.gpa;
const body = gz.instructionsSlice();
const body_len = astgen.countBodyLenAfterFixups(body);
try astgen.extra.ensureUnusedCapacity(
gpa,
@typeInfo(Zir.Inst.Block).Struct.fields.len + body_len,
);
const zir_datas = astgen.instructions.items(.data);
zir_datas[@intFromEnum(inst)].pl_node.payload_index = astgen.addExtraAssumeCapacity(
Zir.Inst.Block{ .body_len = body_len },
);
astgen.appendBodyWithFixups(body);
gz.unstack();
}
/// Assumes nothing stacked on `gz`. Unstacks `gz`.
fn setTryBody(gz: *GenZir, inst: Zir.Inst.Index, operand: Zir.Inst.Ref) !void {
const astgen = gz.astgen;
const gpa = astgen.gpa;
const body = gz.instructionsSlice();
const body_len = astgen.countBodyLenAfterFixups(body);
try astgen.extra.ensureUnusedCapacity(
gpa,
@typeInfo(Zir.Inst.Try).Struct.fields.len + body_len,
);
const zir_datas = astgen.instructions.items(.data);
zir_datas[@intFromEnum(inst)].pl_node.payload_index = astgen.addExtraAssumeCapacity(
Zir.Inst.Try{
.operand = operand,
.body_len = body_len,
},
);
astgen.appendBodyWithFixups(body);
gz.unstack();
}
/// Must be called with the following stack set up:
/// * gz (bottom)
/// * align_gz
/// * addrspace_gz
/// * section_gz
/// * cc_gz
/// * ret_gz
/// * body_gz (top)
/// Unstacks all of those except for `gz`.
fn addFunc(gz: *GenZir, args: struct {
src_node: Ast.Node.Index,
lbrace_line: u32 = 0,
lbrace_column: u32 = 0,
param_block: Zir.Inst.Index,
align_gz: ?*GenZir,
addrspace_gz: ?*GenZir,
section_gz: ?*GenZir,
cc_gz: ?*GenZir,
ret_gz: ?*GenZir,
body_gz: ?*GenZir,
align_ref: Zir.Inst.Ref,
addrspace_ref: Zir.Inst.Ref,
section_ref: Zir.Inst.Ref,
cc_ref: Zir.Inst.Ref,
ret_ref: Zir.Inst.Ref,
lib_name: Zir.NullTerminatedString,
noalias_bits: u32,
is_var_args: bool,
is_inferred_error: bool,
is_test: bool,
is_extern: bool,
is_noinline: bool,
}) !Zir.Inst.Ref {
assert(args.src_node != 0);
const astgen = gz.astgen;
const gpa = astgen.gpa;
const ret_ref = if (args.ret_ref == .void_type) .none else args.ret_ref;
const new_index: Zir.Inst.Index = @enumFromInt(astgen.instructions.len);
try astgen.instructions.ensureUnusedCapacity(gpa, 1);
var body: []Zir.Inst.Index = &[0]Zir.Inst.Index{};
var ret_body: []Zir.Inst.Index = &[0]Zir.Inst.Index{};
var src_locs_and_hash_buffer: [7]u32 = undefined;
var src_locs_and_hash: []u32 = src_locs_and_hash_buffer[0..0];
if (args.body_gz) |body_gz| {
const tree = astgen.tree;
const node_tags = tree.nodes.items(.tag);
const node_datas = tree.nodes.items(.data);
const token_starts = tree.tokens.items(.start);
const fn_decl = args.src_node;
assert(node_tags[fn_decl] == .fn_decl or node_tags[fn_decl] == .test_decl);
const block = node_datas[fn_decl].rhs;
const rbrace_start = token_starts[tree.lastToken(block)];
astgen.advanceSourceCursor(rbrace_start);
const rbrace_line: u32 = @intCast(astgen.source_line - gz.decl_line);
const rbrace_column: u32 = @intCast(astgen.source_column);
const columns = args.lbrace_column | (rbrace_column << 16);
const proto_hash: std.zig.SrcHash = switch (node_tags[fn_decl]) {
.fn_decl => sig_hash: {
const proto_node = node_datas[fn_decl].lhs;
break :sig_hash std.zig.hashSrc(tree.getNodeSource(proto_node));
},
.test_decl => std.zig.hashSrc(""), // tests don't have a prototype
else => unreachable,
};
const proto_hash_arr: [4]u32 = @bitCast(proto_hash);
src_locs_and_hash_buffer = .{
args.lbrace_line,
rbrace_line,
columns,
proto_hash_arr[0],
proto_hash_arr[1],
proto_hash_arr[2],
proto_hash_arr[3],
};
src_locs_and_hash = &src_locs_and_hash_buffer;
body = body_gz.instructionsSlice();
if (args.ret_gz) |ret_gz|
ret_body = ret_gz.instructionsSliceUpto(body_gz);
} else {
if (args.ret_gz) |ret_gz|
ret_body = ret_gz.instructionsSlice();
}
const body_len = astgen.countBodyLenAfterFixups(body);
if (args.cc_ref != .none or args.lib_name != .empty or args.is_var_args or args.is_test or
args.is_extern or args.align_ref != .none or args.section_ref != .none or
args.addrspace_ref != .none or args.noalias_bits != 0 or args.is_noinline)
{
var align_body: []Zir.Inst.Index = &.{};
var addrspace_body: []Zir.Inst.Index = &.{};
var section_body: []Zir.Inst.Index = &.{};
var cc_body: []Zir.Inst.Index = &.{};
if (args.ret_gz != null) {
align_body = args.align_gz.?.instructionsSliceUpto(args.addrspace_gz.?);
addrspace_body = args.addrspace_gz.?.instructionsSliceUpto(args.section_gz.?);
section_body = args.section_gz.?.instructionsSliceUpto(args.cc_gz.?);
cc_body = args.cc_gz.?.instructionsSliceUpto(args.ret_gz.?);
}
try astgen.extra.ensureUnusedCapacity(
gpa,
@typeInfo(Zir.Inst.FuncFancy).Struct.fields.len +
fancyFnExprExtraLen(astgen, align_body, args.align_ref) +
fancyFnExprExtraLen(astgen, addrspace_body, args.addrspace_ref) +
fancyFnExprExtraLen(astgen, section_body, args.section_ref) +
fancyFnExprExtraLen(astgen, cc_body, args.cc_ref) +
fancyFnExprExtraLen(astgen, ret_body, ret_ref) +
body_len + src_locs_and_hash.len +
@intFromBool(args.lib_name != .empty) +
@intFromBool(args.noalias_bits != 0),
);
const payload_index = astgen.addExtraAssumeCapacity(Zir.Inst.FuncFancy{
.param_block = args.param_block,
.body_len = body_len,
.bits = .{
.is_var_args = args.is_var_args,
.is_inferred_error = args.is_inferred_error,
.is_test = args.is_test,
.is_extern = args.is_extern,
.is_noinline = args.is_noinline,
.has_lib_name = args.lib_name != .empty,
.has_any_noalias = args.noalias_bits != 0,
.has_align_ref = args.align_ref != .none,
.has_addrspace_ref = args.addrspace_ref != .none,
.has_section_ref = args.section_ref != .none,
.has_cc_ref = args.cc_ref != .none,
.has_ret_ty_ref = ret_ref != .none,
.has_align_body = align_body.len != 0,
.has_addrspace_body = addrspace_body.len != 0,
.has_section_body = section_body.len != 0,
.has_cc_body = cc_body.len != 0,
.has_ret_ty_body = ret_body.len != 0,
},
});
if (args.lib_name != .empty) {
astgen.extra.appendAssumeCapacity(@intFromEnum(args.lib_name));
}
const zir_datas = astgen.instructions.items(.data);
if (align_body.len != 0) {
astgen.extra.appendAssumeCapacity(countBodyLenAfterFixups(astgen, align_body));
astgen.appendBodyWithFixups(align_body);
const break_extra = zir_datas[@intFromEnum(align_body[align_body.len - 1])].@"break".payload_index;
astgen.extra.items[break_extra + std.meta.fieldIndex(Zir.Inst.Break, "block_inst").?] =
@intFromEnum(new_index);
} else if (args.align_ref != .none) {
astgen.extra.appendAssumeCapacity(@intFromEnum(args.align_ref));
}
if (addrspace_body.len != 0) {
astgen.extra.appendAssumeCapacity(countBodyLenAfterFixups(astgen, addrspace_body));
astgen.appendBodyWithFixups(addrspace_body);
const break_extra =
zir_datas[@intFromEnum(addrspace_body[addrspace_body.len - 1])].@"break".payload_index;
astgen.extra.items[break_extra + std.meta.fieldIndex(Zir.Inst.Break, "block_inst").?] =
@intFromEnum(new_index);
} else if (args.addrspace_ref != .none) {
astgen.extra.appendAssumeCapacity(@intFromEnum(args.addrspace_ref));
}
if (section_body.len != 0) {
astgen.extra.appendAssumeCapacity(countBodyLenAfterFixups(astgen, section_body));
astgen.appendBodyWithFixups(section_body);
const break_extra =
zir_datas[@intFromEnum(section_body[section_body.len - 1])].@"break".payload_index;
astgen.extra.items[break_extra + std.meta.fieldIndex(Zir.Inst.Break, "block_inst").?] =
@intFromEnum(new_index);
} else if (args.section_ref != .none) {
astgen.extra.appendAssumeCapacity(@intFromEnum(args.section_ref));
}
if (cc_body.len != 0) {
astgen.extra.appendAssumeCapacity(countBodyLenAfterFixups(astgen, cc_body));
astgen.appendBodyWithFixups(cc_body);
const break_extra = zir_datas[@intFromEnum(cc_body[cc_body.len - 1])].@"break".payload_index;
astgen.extra.items[break_extra + std.meta.fieldIndex(Zir.Inst.Break, "block_inst").?] =
@intFromEnum(new_index);
} else if (args.cc_ref != .none) {
astgen.extra.appendAssumeCapacity(@intFromEnum(args.cc_ref));
}
if (ret_body.len != 0) {
astgen.extra.appendAssumeCapacity(countBodyLenAfterFixups(astgen, ret_body));
astgen.appendBodyWithFixups(ret_body);
const break_extra = zir_datas[@intFromEnum(ret_body[ret_body.len - 1])].@"break".payload_index;
astgen.extra.items[break_extra + std.meta.fieldIndex(Zir.Inst.Break, "block_inst").?] =
@intFromEnum(new_index);
} else if (ret_ref != .none) {
astgen.extra.appendAssumeCapacity(@intFromEnum(ret_ref));
}
if (args.noalias_bits != 0) {
astgen.extra.appendAssumeCapacity(args.noalias_bits);
}
astgen.appendBodyWithFixups(body);
astgen.extra.appendSliceAssumeCapacity(src_locs_and_hash);
// Order is important when unstacking.
if (args.body_gz) |body_gz| body_gz.unstack();
if (args.ret_gz != null) {
args.ret_gz.?.unstack();
args.cc_gz.?.unstack();
args.section_gz.?.unstack();
args.addrspace_gz.?.unstack();
args.align_gz.?.unstack();
}
try gz.instructions.ensureUnusedCapacity(gpa, 1);
astgen.instructions.appendAssumeCapacity(.{
.tag = .func_fancy,
.data = .{ .pl_node = .{
.src_node = gz.nodeIndexToRelative(args.src_node),
.payload_index = payload_index,
} },
});
gz.instructions.appendAssumeCapacity(new_index);
return new_index.toRef();
} else {
try astgen.extra.ensureUnusedCapacity(
gpa,
@typeInfo(Zir.Inst.Func).Struct.fields.len + 1 +
fancyFnExprExtraLen(astgen, ret_body, ret_ref) +
body_len + src_locs_and_hash.len,
);
const ret_body_len = if (ret_body.len != 0)
countBodyLenAfterFixups(astgen, ret_body)
else
@intFromBool(ret_ref != .none);
const payload_index = astgen.addExtraAssumeCapacity(Zir.Inst.Func{
.param_block = args.param_block,
.ret_body_len = ret_body_len,
.body_len = body_len,
});
const zir_datas = astgen.instructions.items(.data);
if (ret_body.len != 0) {
astgen.appendBodyWithFixups(ret_body);
const break_extra = zir_datas[@intFromEnum(ret_body[ret_body.len - 1])].@"break".payload_index;
astgen.extra.items[break_extra + std.meta.fieldIndex(Zir.Inst.Break, "block_inst").?] =
@intFromEnum(new_index);
} else if (ret_ref != .none) {
astgen.extra.appendAssumeCapacity(@intFromEnum(ret_ref));
}
astgen.appendBodyWithFixups(body);
astgen.extra.appendSliceAssumeCapacity(src_locs_and_hash);
// Order is important when unstacking.
if (args.body_gz) |body_gz| body_gz.unstack();
if (args.ret_gz) |ret_gz| ret_gz.unstack();
if (args.cc_gz) |cc_gz| cc_gz.unstack();
if (args.section_gz) |section_gz| section_gz.unstack();
if (args.addrspace_gz) |addrspace_gz| addrspace_gz.unstack();
if (args.align_gz) |align_gz| align_gz.unstack();
try gz.instructions.ensureUnusedCapacity(gpa, 1);
const tag: Zir.Inst.Tag = if (args.is_inferred_error) .func_inferred else .func;
astgen.instructions.appendAssumeCapacity(.{
.tag = tag,
.data = .{ .pl_node = .{
.src_node = gz.nodeIndexToRelative(args.src_node),
.payload_index = payload_index,
} },
});
gz.instructions.appendAssumeCapacity(new_index);
return new_index.toRef();
}
}
fn fancyFnExprExtraLen(astgen: *AstGen, body: []Zir.Inst.Index, ref: Zir.Inst.Ref) u32 {
// In the case of non-empty body, there is one for the body length,
// and then one for each instruction.
return countBodyLenAfterFixups(astgen, body) + @intFromBool(ref != .none);
}
fn addVar(gz: *GenZir, args: struct {
align_inst: Zir.Inst.Ref,
lib_name: Zir.NullTerminatedString,
var_type: Zir.Inst.Ref,
init: Zir.Inst.Ref,
is_extern: bool,
is_const: bool,
is_threadlocal: bool,
}) !Zir.Inst.Ref {
const astgen = gz.astgen;
const gpa = astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
try astgen.instructions.ensureUnusedCapacity(gpa, 1);
try astgen.extra.ensureUnusedCapacity(
gpa,
@typeInfo(Zir.Inst.ExtendedVar).Struct.fields.len +
@intFromBool(args.lib_name != .empty) +
@intFromBool(args.align_inst != .none) +
@intFromBool(args.init != .none),
);
const payload_index = astgen.addExtraAssumeCapacity(Zir.Inst.ExtendedVar{
.var_type = args.var_type,
});
if (args.lib_name != .empty) {
astgen.extra.appendAssumeCapacity(@intFromEnum(args.lib_name));
}
if (args.align_inst != .none) {
astgen.extra.appendAssumeCapacity(@intFromEnum(args.align_inst));
}
if (args.init != .none) {
astgen.extra.appendAssumeCapacity(@intFromEnum(args.init));
}
const new_index: Zir.Inst.Index = @enumFromInt(astgen.instructions.len);
astgen.instructions.appendAssumeCapacity(.{
.tag = .extended,
.data = .{ .extended = .{
.opcode = .variable,
.small = @bitCast(Zir.Inst.ExtendedVar.Small{
.has_lib_name = args.lib_name != .empty,
.has_align = args.align_inst != .none,
.has_init = args.init != .none,
.is_extern = args.is_extern,
.is_const = args.is_const,
.is_threadlocal = args.is_threadlocal,
}),
.operand = payload_index,
} },
});
gz.instructions.appendAssumeCapacity(new_index);
return new_index.toRef();
}
fn addInt(gz: *GenZir, integer: u64) !Zir.Inst.Ref {
return gz.add(.{
.tag = .int,
.data = .{ .int = integer },
});
}
fn addIntBig(gz: *GenZir, limbs: []const std.math.big.Limb) !Zir.Inst.Ref {
const astgen = gz.astgen;
const gpa = astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
try astgen.instructions.ensureUnusedCapacity(gpa, 1);
try astgen.string_bytes.ensureUnusedCapacity(gpa, @sizeOf(std.math.big.Limb) * limbs.len);
const new_index: Zir.Inst.Index = @enumFromInt(astgen.instructions.len);
astgen.instructions.appendAssumeCapacity(.{
.tag = .int_big,
.data = .{ .str = .{
.start = @enumFromInt(astgen.string_bytes.items.len),
.len = @intCast(limbs.len),
} },
});
gz.instructions.appendAssumeCapacity(new_index);
astgen.string_bytes.appendSliceAssumeCapacity(mem.sliceAsBytes(limbs));
return new_index.toRef();
}
fn addFloat(gz: *GenZir, number: f64) !Zir.Inst.Ref {
return gz.add(.{
.tag = .float,
.data = .{ .float = number },
});
}
fn addUnNode(
gz: *GenZir,
tag: Zir.Inst.Tag,
operand: Zir.Inst.Ref,
/// Absolute node index. This function does the conversion to offset from Decl.
src_node: Ast.Node.Index,
) !Zir.Inst.Ref {
assert(operand != .none);
return gz.add(.{
.tag = tag,
.data = .{ .un_node = .{
.operand = operand,
.src_node = gz.nodeIndexToRelative(src_node),
} },
});
}
fn makeUnNode(
gz: *GenZir,
tag: Zir.Inst.Tag,
operand: Zir.Inst.Ref,
/// Absolute node index. This function does the conversion to offset from Decl.
src_node: Ast.Node.Index,
) !Zir.Inst.Index {
assert(operand != .none);
const new_index: Zir.Inst.Index = @enumFromInt(gz.astgen.instructions.len);
try gz.astgen.instructions.append(gz.astgen.gpa, .{
.tag = tag,
.data = .{ .un_node = .{
.operand = operand,
.src_node = gz.nodeIndexToRelative(src_node),
} },
});
return new_index;
}
fn addPlNode(
gz: *GenZir,
tag: Zir.Inst.Tag,
/// Absolute node index. This function does the conversion to offset from Decl.
src_node: Ast.Node.Index,
extra: anytype,
) !Zir.Inst.Ref {
const gpa = gz.astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
try gz.astgen.instructions.ensureUnusedCapacity(gpa, 1);
const payload_index = try gz.astgen.addExtra(extra);
const new_index: Zir.Inst.Index = @enumFromInt(gz.astgen.instructions.len);
gz.astgen.instructions.appendAssumeCapacity(.{
.tag = tag,
.data = .{ .pl_node = .{
.src_node = gz.nodeIndexToRelative(src_node),
.payload_index = payload_index,
} },
});
gz.instructions.appendAssumeCapacity(new_index);
return new_index.toRef();
}
fn addPlNodePayloadIndex(
gz: *GenZir,
tag: Zir.Inst.Tag,
/// Absolute node index. This function does the conversion to offset from Decl.
src_node: Ast.Node.Index,
payload_index: u32,
) !Zir.Inst.Ref {
return try gz.add(.{
.tag = tag,
.data = .{ .pl_node = .{
.src_node = gz.nodeIndexToRelative(src_node),
.payload_index = payload_index,
} },
});
}
/// Supports `param_gz` stacked on `gz`. Assumes nothing stacked on `param_gz`. Unstacks `param_gz`.
fn addParam(
gz: *GenZir,
param_gz: *GenZir,
tag: Zir.Inst.Tag,
/// Absolute token index. This function does the conversion to Decl offset.
abs_tok_index: Ast.TokenIndex,
name: Zir.NullTerminatedString,
first_doc_comment: ?Ast.TokenIndex,
) !Zir.Inst.Index {
const gpa = gz.astgen.gpa;
const param_body = param_gz.instructionsSlice();
const body_len = gz.astgen.countBodyLenAfterFixups(param_body);
try gz.astgen.instructions.ensureUnusedCapacity(gpa, 1);
try gz.astgen.extra.ensureUnusedCapacity(gpa, @typeInfo(Zir.Inst.Param).Struct.fields.len + body_len);
const doc_comment_index = if (first_doc_comment) |first|
try gz.astgen.docCommentAsStringFromFirst(abs_tok_index, first)
else
.empty;
const payload_index = gz.astgen.addExtraAssumeCapacity(Zir.Inst.Param{
.name = name,
.doc_comment = doc_comment_index,
.body_len = @intCast(body_len),
});
gz.astgen.appendBodyWithFixups(param_body);
param_gz.unstack();
const new_index: Zir.Inst.Index = @enumFromInt(gz.astgen.instructions.len);
gz.astgen.instructions.appendAssumeCapacity(.{
.tag = tag,
.data = .{ .pl_tok = .{
.src_tok = gz.tokenIndexToRelative(abs_tok_index),
.payload_index = payload_index,
} },
});
gz.instructions.appendAssumeCapacity(new_index);
return new_index;
}
fn addExtendedPayload(gz: *GenZir, opcode: Zir.Inst.Extended, extra: anytype) !Zir.Inst.Ref {
return addExtendedPayloadSmall(gz, opcode, undefined, extra);
}
fn addExtendedPayloadSmall(
gz: *GenZir,
opcode: Zir.Inst.Extended,
small: u16,
extra: anytype,
) !Zir.Inst.Ref {
const gpa = gz.astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
try gz.astgen.instructions.ensureUnusedCapacity(gpa, 1);
const payload_index = try gz.astgen.addExtra(extra);
const new_index: Zir.Inst.Index = @enumFromInt(gz.astgen.instructions.len);
gz.astgen.instructions.appendAssumeCapacity(.{
.tag = .extended,
.data = .{ .extended = .{
.opcode = opcode,
.small = small,
.operand = payload_index,
} },
});
gz.instructions.appendAssumeCapacity(new_index);
return new_index.toRef();
}
fn addExtendedMultiOp(
gz: *GenZir,
opcode: Zir.Inst.Extended,
node: Ast.Node.Index,
operands: []const Zir.Inst.Ref,
) !Zir.Inst.Ref {
const astgen = gz.astgen;
const gpa = astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
try astgen.instructions.ensureUnusedCapacity(gpa, 1);
try astgen.extra.ensureUnusedCapacity(
gpa,
@typeInfo(Zir.Inst.NodeMultiOp).Struct.fields.len + operands.len,
);
const payload_index = astgen.addExtraAssumeCapacity(Zir.Inst.NodeMultiOp{
.src_node = gz.nodeIndexToRelative(node),
});
const new_index: Zir.Inst.Index = @enumFromInt(astgen.instructions.len);
astgen.instructions.appendAssumeCapacity(.{
.tag = .extended,
.data = .{ .extended = .{
.opcode = opcode,
.small = @intCast(operands.len),
.operand = payload_index,
} },
});
gz.instructions.appendAssumeCapacity(new_index);
astgen.appendRefsAssumeCapacity(operands);
return new_index.toRef();
}
fn addExtendedMultiOpPayloadIndex(
gz: *GenZir,
opcode: Zir.Inst.Extended,
payload_index: u32,
trailing_len: usize,
) !Zir.Inst.Ref {
const astgen = gz.astgen;
const gpa = astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
try astgen.instructions.ensureUnusedCapacity(gpa, 1);
const new_index: Zir.Inst.Index = @enumFromInt(astgen.instructions.len);
astgen.instructions.appendAssumeCapacity(.{
.tag = .extended,
.data = .{ .extended = .{
.opcode = opcode,
.small = @intCast(trailing_len),
.operand = payload_index,
} },
});
gz.instructions.appendAssumeCapacity(new_index);
return new_index.toRef();
}
fn addExtendedNodeSmall(
gz: *GenZir,
opcode: Zir.Inst.Extended,
src_node: Ast.Node.Index,
small: u16,
) !Zir.Inst.Ref {
const astgen = gz.astgen;
const gpa = astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
try astgen.instructions.ensureUnusedCapacity(gpa, 1);
const new_index: Zir.Inst.Index = @enumFromInt(astgen.instructions.len);
astgen.instructions.appendAssumeCapacity(.{
.tag = .extended,
.data = .{ .extended = .{
.opcode = opcode,
.small = small,
.operand = @bitCast(gz.nodeIndexToRelative(src_node)),
} },
});
gz.instructions.appendAssumeCapacity(new_index);
return new_index.toRef();
}
fn addUnTok(
gz: *GenZir,
tag: Zir.Inst.Tag,
operand: Zir.Inst.Ref,
/// Absolute token index. This function does the conversion to Decl offset.
abs_tok_index: Ast.TokenIndex,
) !Zir.Inst.Ref {
assert(operand != .none);
return gz.add(.{
.tag = tag,
.data = .{ .un_tok = .{
.operand = operand,
.src_tok = gz.tokenIndexToRelative(abs_tok_index),
} },
});
}
fn makeUnTok(
gz: *GenZir,
tag: Zir.Inst.Tag,
operand: Zir.Inst.Ref,
/// Absolute token index. This function does the conversion to Decl offset.
abs_tok_index: Ast.TokenIndex,
) !Zir.Inst.Index {
const astgen = gz.astgen;
const new_index: Zir.Inst.Index = @enumFromInt(astgen.instructions.len);
assert(operand != .none);
try astgen.instructions.append(astgen.gpa, .{
.tag = tag,
.data = .{ .un_tok = .{
.operand = operand,
.src_tok = gz.tokenIndexToRelative(abs_tok_index),
} },
});
return new_index;
}
fn addStrTok(
gz: *GenZir,
tag: Zir.Inst.Tag,
str_index: Zir.NullTerminatedString,
/// Absolute token index. This function does the conversion to Decl offset.
abs_tok_index: Ast.TokenIndex,
) !Zir.Inst.Ref {
return gz.add(.{
.tag = tag,
.data = .{ .str_tok = .{
.start = str_index,
.src_tok = gz.tokenIndexToRelative(abs_tok_index),
} },
});
}
fn addSaveErrRetIndex(
gz: *GenZir,
cond: union(enum) {
always: void,
if_of_error_type: Zir.Inst.Ref,
},
) !Zir.Inst.Index {
return gz.addAsIndex(.{
.tag = .save_err_ret_index,
.data = .{ .save_err_ret_index = .{
.operand = switch (cond) {
.if_of_error_type => |x| x,
else => .none,
},
} },
});
}
const BranchTarget = union(enum) {
ret,
block: Zir.Inst.Index,
};
fn addRestoreErrRetIndex(
gz: *GenZir,
bt: BranchTarget,
cond: union(enum) {
always: void,
if_non_error: Zir.Inst.Ref,
},
src_node: Ast.Node.Index,
) !Zir.Inst.Index {
switch (cond) {
.always => return gz.addAsIndex(.{
.tag = .restore_err_ret_index_unconditional,
.data = .{ .un_node = .{
.operand = switch (bt) {
.ret => .none,
.block => |b| b.toRef(),
},
.src_node = gz.nodeIndexToRelative(src_node),
} },
}),
.if_non_error => |operand| switch (bt) {
.ret => return gz.addAsIndex(.{
.tag = .restore_err_ret_index_fn_entry,
.data = .{ .un_node = .{
.operand = operand,
.src_node = gz.nodeIndexToRelative(src_node),
} },
}),
.block => |block| return (try gz.addExtendedPayload(
.restore_err_ret_index,
Zir.Inst.RestoreErrRetIndex{
.src_node = gz.nodeIndexToRelative(src_node),
.block = block.toRef(),
.operand = operand,
},
)).toIndex().?,
},
}
}
fn addBreak(
gz: *GenZir,
tag: Zir.Inst.Tag,
block_inst: Zir.Inst.Index,
operand: Zir.Inst.Ref,
) !Zir.Inst.Index {
const gpa = gz.astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
const new_index = try gz.makeBreak(tag, block_inst, operand);
gz.instructions.appendAssumeCapacity(new_index);
return new_index;
}
fn makeBreak(
gz: *GenZir,
tag: Zir.Inst.Tag,
block_inst: Zir.Inst.Index,
operand: Zir.Inst.Ref,
) !Zir.Inst.Index {
return gz.makeBreakCommon(tag, block_inst, operand, null);
}
fn addBreakWithSrcNode(
gz: *GenZir,
tag: Zir.Inst.Tag,
block_inst: Zir.Inst.Index,
operand: Zir.Inst.Ref,
operand_src_node: Ast.Node.Index,
) !Zir.Inst.Index {
const gpa = gz.astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
const new_index = try gz.makeBreakWithSrcNode(tag, block_inst, operand, operand_src_node);
gz.instructions.appendAssumeCapacity(new_index);
return new_index;
}
fn makeBreakWithSrcNode(
gz: *GenZir,
tag: Zir.Inst.Tag,
block_inst: Zir.Inst.Index,
operand: Zir.Inst.Ref,
operand_src_node: Ast.Node.Index,
) !Zir.Inst.Index {
return gz.makeBreakCommon(tag, block_inst, operand, operand_src_node);
}
fn makeBreakCommon(
gz: *GenZir,
tag: Zir.Inst.Tag,
block_inst: Zir.Inst.Index,
operand: Zir.Inst.Ref,
operand_src_node: ?Ast.Node.Index,
) !Zir.Inst.Index {
const gpa = gz.astgen.gpa;
try gz.astgen.instructions.ensureUnusedCapacity(gpa, 1);
try gz.astgen.extra.ensureUnusedCapacity(gpa, @typeInfo(Zir.Inst.Break).Struct.fields.len);
const new_index: Zir.Inst.Index = @enumFromInt(gz.astgen.instructions.len);
gz.astgen.instructions.appendAssumeCapacity(.{
.tag = tag,
.data = .{ .@"break" = .{
.operand = operand,
.payload_index = gz.astgen.addExtraAssumeCapacity(Zir.Inst.Break{
.operand_src_node = if (operand_src_node) |src_node|
gz.nodeIndexToRelative(src_node)
else
Zir.Inst.Break.no_src_node,
.block_inst = block_inst,
}),
} },
});
return new_index;
}
fn addBin(
gz: *GenZir,
tag: Zir.Inst.Tag,
lhs: Zir.Inst.Ref,
rhs: Zir.Inst.Ref,
) !Zir.Inst.Ref {
assert(lhs != .none);
assert(rhs != .none);
return gz.add(.{
.tag = tag,
.data = .{ .bin = .{
.lhs = lhs,
.rhs = rhs,
} },
});
}
fn addDefer(gz: *GenZir, index: u32, len: u32) !void {
_ = try gz.add(.{
.tag = .@"defer",
.data = .{ .@"defer" = .{
.index = index,
.len = len,
} },
});
}
fn addDecl(
gz: *GenZir,
tag: Zir.Inst.Tag,
decl_index: u32,
src_node: Ast.Node.Index,
) !Zir.Inst.Ref {
return gz.add(.{
.tag = tag,
.data = .{ .pl_node = .{
.src_node = gz.nodeIndexToRelative(src_node),
.payload_index = decl_index,
} },
});
}
fn addNode(
gz: *GenZir,
tag: Zir.Inst.Tag,
/// Absolute node index. This function does the conversion to offset from Decl.
src_node: Ast.Node.Index,
) !Zir.Inst.Ref {
return gz.add(.{
.tag = tag,
.data = .{ .node = gz.nodeIndexToRelative(src_node) },
});
}
fn addInstNode(
gz: *GenZir,
tag: Zir.Inst.Tag,
inst: Zir.Inst.Index,
/// Absolute node index. This function does the conversion to offset from Decl.
src_node: Ast.Node.Index,
) !Zir.Inst.Ref {
return gz.add(.{
.tag = tag,
.data = .{ .inst_node = .{
.inst = inst,
.src_node = gz.nodeIndexToRelative(src_node),
} },
});
}
fn addNodeExtended(
gz: *GenZir,
opcode: Zir.Inst.Extended,
/// Absolute node index. This function does the conversion to offset from Decl.
src_node: Ast.Node.Index,
) !Zir.Inst.Ref {
return gz.add(.{
.tag = .extended,
.data = .{ .extended = .{
.opcode = opcode,
.small = undefined,
.operand = @bitCast(gz.nodeIndexToRelative(src_node)),
} },
});
}
fn addAllocExtended(
gz: *GenZir,
args: struct {
/// Absolute node index. This function does the conversion to offset from Decl.
node: Ast.Node.Index,
type_inst: Zir.Inst.Ref,
align_inst: Zir.Inst.Ref,
is_const: bool,
is_comptime: bool,
},
) !Zir.Inst.Ref {
const astgen = gz.astgen;
const gpa = astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
try astgen.instructions.ensureUnusedCapacity(gpa, 1);
try astgen.extra.ensureUnusedCapacity(
gpa,
@typeInfo(Zir.Inst.AllocExtended).Struct.fields.len +
@intFromBool(args.type_inst != .none) +
@intFromBool(args.align_inst != .none),
);
const payload_index = gz.astgen.addExtraAssumeCapacity(Zir.Inst.AllocExtended{
.src_node = gz.nodeIndexToRelative(args.node),
});
if (args.type_inst != .none) {
astgen.extra.appendAssumeCapacity(@intFromEnum(args.type_inst));
}
if (args.align_inst != .none) {
astgen.extra.appendAssumeCapacity(@intFromEnum(args.align_inst));
}
const has_type: u4 = @intFromBool(args.type_inst != .none);
const has_align: u4 = @intFromBool(args.align_inst != .none);
const is_const: u4 = @intFromBool(args.is_const);
const is_comptime: u4 = @intFromBool(args.is_comptime);
const small: u16 = has_type | (has_align << 1) | (is_const << 2) | (is_comptime << 3);
const new_index: Zir.Inst.Index = @enumFromInt(astgen.instructions.len);
astgen.instructions.appendAssumeCapacity(.{
.tag = .extended,
.data = .{ .extended = .{
.opcode = .alloc,
.small = small,
.operand = payload_index,
} },
});
gz.instructions.appendAssumeCapacity(new_index);
return new_index.toRef();
}
fn addAsm(
gz: *GenZir,
args: struct {
tag: Zir.Inst.Extended,
/// Absolute node index. This function does the conversion to offset from Decl.
node: Ast.Node.Index,
asm_source: Zir.NullTerminatedString,
output_type_bits: u32,
is_volatile: bool,
outputs: []const Zir.Inst.Asm.Output,
inputs: []const Zir.Inst.Asm.Input,
clobbers: []const u32,
},
) !Zir.Inst.Ref {
const astgen = gz.astgen;
const gpa = astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
try astgen.instructions.ensureUnusedCapacity(gpa, 1);
try astgen.extra.ensureUnusedCapacity(gpa, @typeInfo(Zir.Inst.Asm).Struct.fields.len +
args.outputs.len * @typeInfo(Zir.Inst.Asm.Output).Struct.fields.len +
args.inputs.len * @typeInfo(Zir.Inst.Asm.Input).Struct.fields.len +
args.clobbers.len);
const payload_index = gz.astgen.addExtraAssumeCapacity(Zir.Inst.Asm{
.src_node = gz.nodeIndexToRelative(args.node),
.asm_source = args.asm_source,
.output_type_bits = args.output_type_bits,
});
for (args.outputs) |output| {
_ = gz.astgen.addExtraAssumeCapacity(output);
}
for (args.inputs) |input| {
_ = gz.astgen.addExtraAssumeCapacity(input);
}
gz.astgen.extra.appendSliceAssumeCapacity(args.clobbers);
// * 0b00000000_000XXXXX - `outputs_len`.
// * 0b000000XX_XXX00000 - `inputs_len`.
// * 0b0XXXXX00_00000000 - `clobbers_len`.
// * 0bX0000000_00000000 - is volatile
const small: u16 = @as(u16, @intCast(args.outputs.len)) |
@as(u16, @intCast(args.inputs.len << 5)) |
@as(u16, @intCast(args.clobbers.len << 10)) |
(@as(u16, @intFromBool(args.is_volatile)) << 15);
const new_index: Zir.Inst.Index = @enumFromInt(astgen.instructions.len);
astgen.instructions.appendAssumeCapacity(.{
.tag = .extended,
.data = .{ .extended = .{
.opcode = args.tag,
.small = small,
.operand = payload_index,
} },
});
gz.instructions.appendAssumeCapacity(new_index);
return new_index.toRef();
}
/// Note that this returns a `Zir.Inst.Index` not a ref.
/// Does *not* append the block instruction to the scope.
/// Leaves the `payload_index` field undefined.
fn makeBlockInst(gz: *GenZir, tag: Zir.Inst.Tag, node: Ast.Node.Index) !Zir.Inst.Index {
const new_index: Zir.Inst.Index = @enumFromInt(gz.astgen.instructions.len);
const gpa = gz.astgen.gpa;
try gz.astgen.instructions.append(gpa, .{
.tag = tag,
.data = .{ .pl_node = .{
.src_node = gz.nodeIndexToRelative(node),
.payload_index = undefined,
} },
});
return new_index;
}
/// Note that this returns a `Zir.Inst.Index` not a ref.
/// Leaves the `payload_index` field undefined.
fn addCondBr(gz: *GenZir, tag: Zir.Inst.Tag, node: Ast.Node.Index) !Zir.Inst.Index {
const gpa = gz.astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
const new_index: Zir.Inst.Index = @enumFromInt(gz.astgen.instructions.len);
try gz.astgen.instructions.append(gpa, .{
.tag = tag,
.data = .{ .pl_node = .{
.src_node = gz.nodeIndexToRelative(node),
.payload_index = undefined,
} },
});
gz.instructions.appendAssumeCapacity(new_index);
return new_index;
}
fn setStruct(gz: *GenZir, inst: Zir.Inst.Index, args: struct {
src_node: Ast.Node.Index,
captures_len: u32,
fields_len: u32,
decls_len: u32,
has_backing_int: bool,
layout: std.builtin.Type.ContainerLayout,
known_non_opv: bool,
known_comptime_only: bool,
is_tuple: bool,
any_comptime_fields: bool,
any_default_inits: bool,
any_aligned_fields: bool,
fields_hash: std.zig.SrcHash,
}) !void {
const astgen = gz.astgen;
const gpa = astgen.gpa;
// Node 0 is valid for the root `struct_decl` of a file!
assert(args.src_node != 0 or gz.parent.tag == .top);
const fields_hash_arr: [4]u32 = @bitCast(args.fields_hash);
try astgen.extra.ensureUnusedCapacity(gpa, @typeInfo(Zir.Inst.StructDecl).Struct.fields.len + 3);
const payload_index = astgen.addExtraAssumeCapacity(Zir.Inst.StructDecl{
.fields_hash_0 = fields_hash_arr[0],
.fields_hash_1 = fields_hash_arr[1],
.fields_hash_2 = fields_hash_arr[2],
.fields_hash_3 = fields_hash_arr[3],
.src_node = gz.nodeIndexToRelative(args.src_node),
});
if (args.captures_len != 0) {
astgen.extra.appendAssumeCapacity(args.captures_len);
}
if (args.fields_len != 0) {
astgen.extra.appendAssumeCapacity(args.fields_len);
}
if (args.decls_len != 0) {
astgen.extra.appendAssumeCapacity(args.decls_len);
}
astgen.instructions.set(@intFromEnum(inst), .{
.tag = .extended,
.data = .{ .extended = .{
.opcode = .struct_decl,
.small = @bitCast(Zir.Inst.StructDecl.Small{
.has_captures_len = args.captures_len != 0,
.has_fields_len = args.fields_len != 0,
.has_decls_len = args.decls_len != 0,
.has_backing_int = args.has_backing_int,
.known_non_opv = args.known_non_opv,
.known_comptime_only = args.known_comptime_only,
.is_tuple = args.is_tuple,
.name_strategy = gz.anon_name_strategy,
.layout = args.layout,
.any_comptime_fields = args.any_comptime_fields,
.any_default_inits = args.any_default_inits,
.any_aligned_fields = args.any_aligned_fields,
}),
.operand = payload_index,
} },
});
}
fn setUnion(gz: *GenZir, inst: Zir.Inst.Index, args: struct {
src_node: Ast.Node.Index,
tag_type: Zir.Inst.Ref,
captures_len: u32,
body_len: u32,
fields_len: u32,
decls_len: u32,
layout: std.builtin.Type.ContainerLayout,
auto_enum_tag: bool,
any_aligned_fields: bool,
fields_hash: std.zig.SrcHash,
}) !void {
const astgen = gz.astgen;
const gpa = astgen.gpa;
assert(args.src_node != 0);
const fields_hash_arr: [4]u32 = @bitCast(args.fields_hash);
try astgen.extra.ensureUnusedCapacity(gpa, @typeInfo(Zir.Inst.UnionDecl).Struct.fields.len + 5);
const payload_index = astgen.addExtraAssumeCapacity(Zir.Inst.UnionDecl{
.fields_hash_0 = fields_hash_arr[0],
.fields_hash_1 = fields_hash_arr[1],
.fields_hash_2 = fields_hash_arr[2],
.fields_hash_3 = fields_hash_arr[3],
.src_node = gz.nodeIndexToRelative(args.src_node),
});
if (args.tag_type != .none) {
astgen.extra.appendAssumeCapacity(@intFromEnum(args.tag_type));
}
if (args.captures_len != 0) {
astgen.extra.appendAssumeCapacity(args.captures_len);
}
if (args.body_len != 0) {
astgen.extra.appendAssumeCapacity(args.body_len);
}
if (args.fields_len != 0) {
astgen.extra.appendAssumeCapacity(args.fields_len);
}
if (args.decls_len != 0) {
astgen.extra.appendAssumeCapacity(args.decls_len);
}
astgen.instructions.set(@intFromEnum(inst), .{
.tag = .extended,
.data = .{ .extended = .{
.opcode = .union_decl,
.small = @bitCast(Zir.Inst.UnionDecl.Small{
.has_tag_type = args.tag_type != .none,
.has_captures_len = args.captures_len != 0,
.has_body_len = args.body_len != 0,
.has_fields_len = args.fields_len != 0,
.has_decls_len = args.decls_len != 0,
.name_strategy = gz.anon_name_strategy,
.layout = args.layout,
.auto_enum_tag = args.auto_enum_tag,
.any_aligned_fields = args.any_aligned_fields,
}),
.operand = payload_index,
} },
});
}
fn setEnum(gz: *GenZir, inst: Zir.Inst.Index, args: struct {
src_node: Ast.Node.Index,
tag_type: Zir.Inst.Ref,
captures_len: u32,
body_len: u32,
fields_len: u32,
decls_len: u32,
nonexhaustive: bool,
fields_hash: std.zig.SrcHash,
}) !void {
const astgen = gz.astgen;
const gpa = astgen.gpa;
assert(args.src_node != 0);
const fields_hash_arr: [4]u32 = @bitCast(args.fields_hash);
try astgen.extra.ensureUnusedCapacity(gpa, @typeInfo(Zir.Inst.EnumDecl).Struct.fields.len + 5);
const payload_index = astgen.addExtraAssumeCapacity(Zir.Inst.EnumDecl{
.fields_hash_0 = fields_hash_arr[0],
.fields_hash_1 = fields_hash_arr[1],
.fields_hash_2 = fields_hash_arr[2],
.fields_hash_3 = fields_hash_arr[3],
.src_node = gz.nodeIndexToRelative(args.src_node),
});
if (args.tag_type != .none) {
astgen.extra.appendAssumeCapacity(@intFromEnum(args.tag_type));
}
if (args.captures_len != 0) {
astgen.extra.appendAssumeCapacity(args.captures_len);
}
if (args.body_len != 0) {
astgen.extra.appendAssumeCapacity(args.body_len);
}
if (args.fields_len != 0) {
astgen.extra.appendAssumeCapacity(args.fields_len);
}
if (args.decls_len != 0) {
astgen.extra.appendAssumeCapacity(args.decls_len);
}
astgen.instructions.set(@intFromEnum(inst), .{
.tag = .extended,
.data = .{ .extended = .{
.opcode = .enum_decl,
.small = @bitCast(Zir.Inst.EnumDecl.Small{
.has_tag_type = args.tag_type != .none,
.has_captures_len = args.captures_len != 0,
.has_body_len = args.body_len != 0,
.has_fields_len = args.fields_len != 0,
.has_decls_len = args.decls_len != 0,
.name_strategy = gz.anon_name_strategy,
.nonexhaustive = args.nonexhaustive,
}),
.operand = payload_index,
} },
});
}
fn setOpaque(gz: *GenZir, inst: Zir.Inst.Index, args: struct {
src_node: Ast.Node.Index,
captures_len: u32,
decls_len: u32,
}) !void {
const astgen = gz.astgen;
const gpa = astgen.gpa;
assert(args.src_node != 0);
try astgen.extra.ensureUnusedCapacity(gpa, @typeInfo(Zir.Inst.OpaqueDecl).Struct.fields.len + 2);
const payload_index = astgen.addExtraAssumeCapacity(Zir.Inst.OpaqueDecl{
.src_node = gz.nodeIndexToRelative(args.src_node),
});
if (args.captures_len != 0) {
astgen.extra.appendAssumeCapacity(args.captures_len);
}
if (args.decls_len != 0) {
astgen.extra.appendAssumeCapacity(args.decls_len);
}
astgen.instructions.set(@intFromEnum(inst), .{
.tag = .extended,
.data = .{ .extended = .{
.opcode = .opaque_decl,
.small = @bitCast(Zir.Inst.OpaqueDecl.Small{
.has_captures_len = args.captures_len != 0,
.has_decls_len = args.decls_len != 0,
.name_strategy = gz.anon_name_strategy,
}),
.operand = payload_index,
} },
});
}
fn add(gz: *GenZir, inst: Zir.Inst) !Zir.Inst.Ref {
return (try gz.addAsIndex(inst)).toRef();
}
fn addAsIndex(gz: *GenZir, inst: Zir.Inst) !Zir.Inst.Index {
const gpa = gz.astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
try gz.astgen.instructions.ensureUnusedCapacity(gpa, 1);
const new_index: Zir.Inst.Index = @enumFromInt(gz.astgen.instructions.len);
gz.astgen.instructions.appendAssumeCapacity(inst);
gz.instructions.appendAssumeCapacity(new_index);
return new_index;
}
fn reserveInstructionIndex(gz: *GenZir) !Zir.Inst.Index {
const gpa = gz.astgen.gpa;
try gz.instructions.ensureUnusedCapacity(gpa, 1);
try gz.astgen.instructions.ensureUnusedCapacity(gpa, 1);
const new_index: Zir.Inst.Index = @enumFromInt(gz.astgen.instructions.len);
gz.astgen.instructions.len += 1;
gz.instructions.appendAssumeCapacity(new_index);
return new_index;
}
fn addRet(gz: *GenZir, ri: ResultInfo, operand: Zir.Inst.Ref, node: Ast.Node.Index) !void {
switch (ri.rl) {
.ptr => |ptr_res| _ = try gz.addUnNode(.ret_load, ptr_res.inst, node),
.coerced_ty => _ = try gz.addUnNode(.ret_node, operand, node),
else => unreachable,
}
}
fn addDbgVar(gz: *GenZir, tag: Zir.Inst.Tag, name: Zir.NullTerminatedString, inst: Zir.Inst.Ref) !void {
if (gz.is_comptime) return;
_ = try gz.add(.{ .tag = tag, .data = .{
.str_op = .{
.str = name,
.operand = inst,
},
} });
}
};
/// This can only be for short-lived references; the memory becomes invalidated
/// when another string is added.
fn nullTerminatedString(astgen: AstGen, index: Zir.NullTerminatedString) [*:0]const u8 {
return @ptrCast(astgen.string_bytes.items[@intFromEnum(index)..]);
}
/// Local variables shadowing detection, including function parameters.
fn detectLocalShadowing(
astgen: *AstGen,
scope: *Scope,
ident_name: Zir.NullTerminatedString,
name_token: Ast.TokenIndex,
token_bytes: []const u8,
id_cat: Scope.IdCat,
) !void {
const gpa = astgen.gpa;
if (token_bytes[0] != '@' and isPrimitive(token_bytes)) {
return astgen.failTokNotes(name_token, "name shadows primitive '{s}'", .{
token_bytes,
}, &[_]u32{
try astgen.errNoteTok(name_token, "consider using @\"{s}\" to disambiguate", .{
token_bytes,
}),
});
}
var s = scope;
var outer_scope = false;
while (true) switch (s.tag) {
.local_val => {
const local_val = s.cast(Scope.LocalVal).?;
if (local_val.name == ident_name) {
const name_slice = mem.span(astgen.nullTerminatedString(ident_name));
const name = try gpa.dupe(u8, name_slice);
defer gpa.free(name);
if (outer_scope) {
return astgen.failTokNotes(name_token, "{s} '{s}' shadows {s} from outer scope", .{
@tagName(id_cat), name, @tagName(local_val.id_cat),
}, &[_]u32{
try astgen.errNoteTok(
local_val.token_src,
"previous declaration here",
.{},
),
});
}
return astgen.failTokNotes(name_token, "redeclaration of {s} '{s}'", .{
@tagName(local_val.id_cat), name,
}, &[_]u32{
try astgen.errNoteTok(
local_val.token_src,
"previous declaration here",
.{},
),
});
}
s = local_val.parent;
},
.local_ptr => {
const local_ptr = s.cast(Scope.LocalPtr).?;
if (local_ptr.name == ident_name) {
const name_slice = mem.span(astgen.nullTerminatedString(ident_name));
const name = try gpa.dupe(u8, name_slice);
defer gpa.free(name);
if (outer_scope) {
return astgen.failTokNotes(name_token, "{s} '{s}' shadows {s} from outer scope", .{
@tagName(id_cat), name, @tagName(local_ptr.id_cat),
}, &[_]u32{
try astgen.errNoteTok(
local_ptr.token_src,
"previous declaration here",
.{},
),
});
}
return astgen.failTokNotes(name_token, "redeclaration of {s} '{s}'", .{
@tagName(local_ptr.id_cat), name,
}, &[_]u32{
try astgen.errNoteTok(
local_ptr.token_src,
"previous declaration here",
.{},
),
});
}
s = local_ptr.parent;
},
.namespace => {
outer_scope = true;
const ns = s.cast(Scope.Namespace).?;
const decl_node = ns.decls.get(ident_name) orelse {
s = ns.parent;
continue;
};
const name_slice = mem.span(astgen.nullTerminatedString(ident_name));
const name = try gpa.dupe(u8, name_slice);
defer gpa.free(name);
return astgen.failTokNotes(name_token, "{s} shadows declaration of '{s}'", .{
@tagName(id_cat), name,
}, &[_]u32{
try astgen.errNoteNode(decl_node, "declared here", .{}),
});
},
.gen_zir => {
s = s.cast(GenZir).?.parent;
outer_scope = true;
},
.defer_normal, .defer_error => s = s.cast(Scope.Defer).?.parent,
.top => break,
};
}
const LineColumn = struct { u32, u32 };
/// Advances the source cursor to the main token of `node` if not in comptime scope.
/// Usually paired with `emitDbgStmt`.
fn maybeAdvanceSourceCursorToMainToken(gz: *GenZir, node: Ast.Node.Index) LineColumn {
if (gz.is_comptime) return .{ gz.astgen.source_line - gz.decl_line, gz.astgen.source_column };
const tree = gz.astgen.tree;
const token_starts = tree.tokens.items(.start);
const main_tokens = tree.nodes.items(.main_token);
const node_start = token_starts[main_tokens[node]];
gz.astgen.advanceSourceCursor(node_start);
return .{ gz.astgen.source_line - gz.decl_line, gz.astgen.source_column };
}
/// Advances the source cursor to the beginning of `node`.
fn advanceSourceCursorToNode(astgen: *AstGen, node: Ast.Node.Index) void {
const tree = astgen.tree;
const token_starts = tree.tokens.items(.start);
const node_start = token_starts[tree.firstToken(node)];
astgen.advanceSourceCursor(node_start);
}
/// Advances the source cursor to an absolute byte offset `end` in the file.
fn advanceSourceCursor(astgen: *AstGen, end: usize) void {
const source = astgen.tree.source;
var i = astgen.source_offset;
var line = astgen.source_line;
var column = astgen.source_column;
assert(i <= end);
while (i < end) : (i += 1) {
if (source[i] == '\n') {
line += 1;
column = 0;
} else {
column += 1;
}
}
astgen.source_offset = i;
astgen.source_line = line;
astgen.source_column = column;
}
fn scanDecls(astgen: *AstGen, namespace: *Scope.Namespace, members: []const Ast.Node.Index) !u32 {
const gpa = astgen.gpa;
const tree = astgen.tree;
const node_tags = tree.nodes.items(.tag);
const main_tokens = tree.nodes.items(.main_token);
const token_tags = tree.tokens.items(.tag);
// We don't have shadowing for test names, so we just track those for duplicate reporting locally.
var named_tests: std.AutoHashMapUnmanaged(Zir.NullTerminatedString, Ast.Node.Index) = .{};
var decltests: std.AutoHashMapUnmanaged(Zir.NullTerminatedString, Ast.Node.Index) = .{};
defer {
named_tests.deinit(gpa);
decltests.deinit(gpa);
}
var decl_count: u32 = 0;
for (members) |member_node| {
const name_token = switch (node_tags[member_node]) {
.global_var_decl,
.local_var_decl,
.simple_var_decl,
.aligned_var_decl,
=> blk: {
decl_count += 1;
break :blk main_tokens[member_node] + 1;
},
.fn_proto_simple,
.fn_proto_multi,
.fn_proto_one,
.fn_proto,
.fn_decl,
=> blk: {
decl_count += 1;
const ident = main_tokens[member_node] + 1;
if (token_tags[ident] != .identifier) {
switch (astgen.failNode(member_node, "missing function name", .{})) {
error.AnalysisFail => continue,
error.OutOfMemory => return error.OutOfMemory,
}
}
break :blk ident;
},
.@"comptime", .@"usingnamespace" => {
decl_count += 1;
continue;
},
.test_decl => {
decl_count += 1;
// We don't want shadowing detection here, and test names work a bit differently, so
// we must do the redeclaration detection ourselves.
const test_name_token = main_tokens[member_node] + 1;
switch (token_tags[test_name_token]) {
else => {}, // unnamed test
.string_literal => {
const name = try astgen.strLitAsString(test_name_token);
const gop = try named_tests.getOrPut(gpa, name.index);
if (gop.found_existing) {
const name_slice = astgen.string_bytes.items[@intFromEnum(name.index)..][0..name.len];
const name_duped = try gpa.dupe(u8, name_slice);
defer gpa.free(name_duped);
try astgen.appendErrorNodeNotes(member_node, "duplicate test name '{s}'", .{name_duped}, &.{
try astgen.errNoteNode(gop.value_ptr.*, "other test here", .{}),
});
} else {
gop.value_ptr.* = member_node;
}
},
.identifier => {
const name = try astgen.identAsString(test_name_token);
const gop = try decltests.getOrPut(gpa, name);
if (gop.found_existing) {
const name_slice = mem.span(astgen.nullTerminatedString(name));
const name_duped = try gpa.dupe(u8, name_slice);
defer gpa.free(name_duped);
try astgen.appendErrorNodeNotes(member_node, "duplicate decltest '{s}'", .{name_duped}, &.{
try astgen.errNoteNode(gop.value_ptr.*, "other decltest here", .{}),
});
} else {
gop.value_ptr.* = member_node;
}
},
}
continue;
},
else => continue,
};
const token_bytes = astgen.tree.tokenSlice(name_token);
if (token_bytes[0] != '@' and isPrimitive(token_bytes)) {
switch (astgen.failTokNotes(name_token, "name shadows primitive '{s}'", .{
token_bytes,
}, &[_]u32{
try astgen.errNoteTok(name_token, "consider using @\"{s}\" to disambiguate", .{
token_bytes,
}),
})) {
error.AnalysisFail => continue,
error.OutOfMemory => return error.OutOfMemory,
}
}
const name_str_index = try astgen.identAsString(name_token);
const gop = try namespace.decls.getOrPut(gpa, name_str_index);
if (gop.found_existing) {
const name = try gpa.dupe(u8, mem.span(astgen.nullTerminatedString(name_str_index)));
defer gpa.free(name);
switch (astgen.failNodeNotes(member_node, "redeclaration of '{s}'", .{
name,
}, &[_]u32{
try astgen.errNoteNode(gop.value_ptr.*, "other declaration here", .{}),
})) {
error.AnalysisFail => continue,
error.OutOfMemory => return error.OutOfMemory,
}
}
var s = namespace.parent;
while (true) switch (s.tag) {
.local_val => {
const local_val = s.cast(Scope.LocalVal).?;
if (local_val.name == name_str_index) {
return astgen.failTokNotes(name_token, "declaration '{s}' shadows {s} from outer scope", .{
token_bytes, @tagName(local_val.id_cat),
}, &[_]u32{
try astgen.errNoteTok(
local_val.token_src,
"previous declaration here",
.{},
),
});
}
s = local_val.parent;
},
.local_ptr => {
const local_ptr = s.cast(Scope.LocalPtr).?;
if (local_ptr.name == name_str_index) {
return astgen.failTokNotes(name_token, "declaration '{s}' shadows {s} from outer scope", .{
token_bytes, @tagName(local_ptr.id_cat),
}, &[_]u32{
try astgen.errNoteTok(
local_ptr.token_src,
"previous declaration here",
.{},
),
});
}
s = local_ptr.parent;
},
.namespace => s = s.cast(Scope.Namespace).?.parent,
.gen_zir => s = s.cast(GenZir).?.parent,
.defer_normal, .defer_error => s = s.cast(Scope.Defer).?.parent,
.top => break,
};
gop.value_ptr.* = member_node;
}
return decl_count;
}
fn isInferred(astgen: *AstGen, ref: Zir.Inst.Ref) bool {
const inst = ref.toIndex() orelse return false;
const zir_tags = astgen.instructions.items(.tag);
return switch (zir_tags[@intFromEnum(inst)]) {
.alloc_inferred,
.alloc_inferred_mut,
.alloc_inferred_comptime,
.alloc_inferred_comptime_mut,
=> true,
.extended => {
const zir_data = astgen.instructions.items(.data);
if (zir_data[@intFromEnum(inst)].extended.opcode != .alloc) return false;
const small: Zir.Inst.AllocExtended.Small = @bitCast(zir_data[@intFromEnum(inst)].extended.small);
return !small.has_type;
},
else => false,
};
}
/// Assumes capacity for body has already been added. Needed capacity taking into
/// account fixups can be found with `countBodyLenAfterFixups`.
fn appendBodyWithFixups(astgen: *AstGen, body: []const Zir.Inst.Index) void {
return appendBodyWithFixupsArrayList(astgen, &astgen.extra, body);
}
fn appendBodyWithFixupsArrayList(
astgen: *AstGen,
list: *std.ArrayListUnmanaged(u32),
body: []const Zir.Inst.Index,
) void {
for (body) |body_inst| {
appendPossiblyRefdBodyInst(astgen, list, body_inst);
}
}
fn appendPossiblyRefdBodyInst(
astgen: *AstGen,
list: *std.ArrayListUnmanaged(u32),
body_inst: Zir.Inst.Index,
) void {
list.appendAssumeCapacity(@intFromEnum(body_inst));
const kv = astgen.ref_table.fetchRemove(body_inst) orelse return;
const ref_inst = kv.value;
return appendPossiblyRefdBodyInst(astgen, list, ref_inst);
}
fn countBodyLenAfterFixups(astgen: *AstGen, body: []const Zir.Inst.Index) u32 {
var count = body.len;
for (body) |body_inst| {
var check_inst = body_inst;
while (astgen.ref_table.get(check_inst)) |ref_inst| {
count += 1;
check_inst = ref_inst;
}
}
return @intCast(count);
}
fn emitDbgStmt(gz: *GenZir, lc: LineColumn) !void {
if (gz.is_comptime) return;
if (gz.instructions.items.len > gz.instructions_top) {
const astgen = gz.astgen;
const last = gz.instructions.items[gz.instructions.items.len - 1];
if (astgen.instructions.items(.tag)[@intFromEnum(last)] == .dbg_stmt) {
astgen.instructions.items(.data)[@intFromEnum(last)].dbg_stmt = .{
.line = lc[0],
.column = lc[1],
};
return;
}
}
_ = try gz.add(.{ .tag = .dbg_stmt, .data = .{
.dbg_stmt = .{
.line = lc[0],
.column = lc[1],
},
} });
}
/// In some cases, Sema expects us to generate a `dbg_stmt` at the instruction
/// *index* directly preceding the next instruction (e.g. if a call is %10, it
/// expects a dbg_stmt at %9). TODO: this logic may allow redundant dbg_stmt
/// instructions; fix up Sema so we don't need it!
fn emitDbgStmtForceCurrentIndex(gz: *GenZir, lc: LineColumn) !void {
const astgen = gz.astgen;
if (gz.instructions.items.len > gz.instructions_top and
@intFromEnum(gz.instructions.items[gz.instructions.items.len - 1]) == astgen.instructions.len - 1)
{
const last = astgen.instructions.len - 1;
if (astgen.instructions.items(.tag)[last] == .dbg_stmt) {
astgen.instructions.items(.data)[last].dbg_stmt = .{
.line = lc[0],
.column = lc[1],
};
return;
}
}
_ = try gz.add(.{ .tag = .dbg_stmt, .data = .{
.dbg_stmt = .{
.line = lc[0],
.column = lc[1],
},
} });
}
fn lowerAstErrors(astgen: *AstGen) !void {
const tree = astgen.tree;
assert(tree.errors.len > 0);
const gpa = astgen.gpa;
const parse_err = tree.errors[0];
var msg: std.ArrayListUnmanaged(u8) = .{};
defer msg.deinit(gpa);
const token_starts = tree.tokens.items(.start);
const token_tags = tree.tokens.items(.tag);
var notes: std.ArrayListUnmanaged(u32) = .{};
defer notes.deinit(gpa);
if (token_tags[parse_err.token + @intFromBool(parse_err.token_is_prev)] == .invalid) {
const tok = parse_err.token + @intFromBool(parse_err.token_is_prev);
const bad_off: u32 = @intCast(tree.tokenSlice(parse_err.token + @intFromBool(parse_err.token_is_prev)).len);
const byte_abs = token_starts[parse_err.token + @intFromBool(parse_err.token_is_prev)] + bad_off;
try notes.append(gpa, try astgen.errNoteTokOff(tok, bad_off, "invalid byte: '{'}'", .{
std.zig.fmtEscapes(tree.source[byte_abs..][0..1]),
}));
}
for (tree.errors[1..]) |note| {
if (!note.is_note) break;
msg.clearRetainingCapacity();
try tree.renderError(note, msg.writer(gpa));
try notes.append(gpa, try astgen.errNoteTok(note.token, "{s}", .{msg.items}));
}
const extra_offset = tree.errorOffset(parse_err);
msg.clearRetainingCapacity();
try tree.renderError(parse_err, msg.writer(gpa));
try astgen.appendErrorTokNotesOff(parse_err.token, extra_offset, "{s}", .{msg.items}, notes.items);
}
const DeclarationName = union(enum) {
named: Ast.TokenIndex,
named_test: Ast.TokenIndex,
unnamed_test,
decltest: Zir.NullTerminatedString,
@"comptime",
@"usingnamespace",
};
/// Sets all extra data for a `declaration` instruction.
/// Unstacks `value_gz`, `align_gz`, `linksection_gz`, and `addrspace_gz`.
fn setDeclaration(
decl_inst: Zir.Inst.Index,
src_hash: std.zig.SrcHash,
name: DeclarationName,
line_offset: u32,
is_pub: bool,
is_export: bool,
doc_comment: Zir.NullTerminatedString,
value_gz: *GenZir,
/// May be `null` if all these blocks would be empty.
/// If `null`, then `value_gz` must have nothing stacked on it.
extra_gzs: ?struct {
/// Must be stacked on `value_gz`.
align_gz: *GenZir,
/// Must be stacked on `align_gz`.
linksection_gz: *GenZir,
/// Must be stacked on `linksection_gz`, and have nothing stacked on it.
addrspace_gz: *GenZir,
},
) !void {
const astgen = value_gz.astgen;
const gpa = astgen.gpa;
const empty_body: []Zir.Inst.Index = &.{};
const value_body, const align_body, const linksection_body, const addrspace_body = if (extra_gzs) |e| .{
value_gz.instructionsSliceUpto(e.align_gz),
e.align_gz.instructionsSliceUpto(e.linksection_gz),
e.linksection_gz.instructionsSliceUpto(e.addrspace_gz),
e.addrspace_gz.instructionsSlice(),
} else .{ value_gz.instructionsSlice(), empty_body, empty_body, empty_body };
const value_len = astgen.countBodyLenAfterFixups(value_body);
const align_len = astgen.countBodyLenAfterFixups(align_body);
const linksection_len = astgen.countBodyLenAfterFixups(linksection_body);
const addrspace_len = astgen.countBodyLenAfterFixups(addrspace_body);
const true_doc_comment: Zir.NullTerminatedString = switch (name) {
.decltest => |test_name| test_name,
else => doc_comment,
};
const src_hash_arr: [4]u32 = @bitCast(src_hash);
const extra: Zir.Inst.Declaration = .{
.src_hash_0 = src_hash_arr[0],
.src_hash_1 = src_hash_arr[1],
.src_hash_2 = src_hash_arr[2],
.src_hash_3 = src_hash_arr[3],
.name = switch (name) {
.named => |tok| @enumFromInt(@intFromEnum(try astgen.identAsString(tok))),
.named_test => |tok| @enumFromInt(@intFromEnum(try astgen.testNameString(tok))),
.unnamed_test => .unnamed_test,
.decltest => .decltest,
.@"comptime" => .@"comptime",
.@"usingnamespace" => .@"usingnamespace",
},
.line_offset = line_offset,
.flags = .{
.value_body_len = @intCast(value_len),
.is_pub = is_pub,
.is_export = is_export,
.has_doc_comment = true_doc_comment != .empty,
.has_align_linksection_addrspace = align_len != 0 or linksection_len != 0 or addrspace_len != 0,
},
};
astgen.instructions.items(.data)[@intFromEnum(decl_inst)].pl_node.payload_index = try astgen.addExtra(extra);
if (extra.flags.has_doc_comment) {
try astgen.extra.append(gpa, @intFromEnum(true_doc_comment));
}
if (extra.flags.has_align_linksection_addrspace) {
try astgen.extra.appendSlice(gpa, &.{
align_len,
linksection_len,
addrspace_len,
});
}
try astgen.extra.ensureUnusedCapacity(gpa, value_len + align_len + linksection_len + addrspace_len);
astgen.appendBodyWithFixups(value_body);
if (extra.flags.has_align_linksection_addrspace) {
astgen.appendBodyWithFixups(align_body);
astgen.appendBodyWithFixups(linksection_body);
astgen.appendBodyWithFixups(addrspace_body);
}
if (extra_gzs) |e| {
e.addrspace_gz.unstack();
e.linksection_gz.unstack();
e.align_gz.unstack();
}
value_gz.unstack();
}