Added int nanometer example in README

README.md

@@ -1,268 +1,233 @@
-# zig_units
+# dimal — Dimensional Analysis for Zig
 
-`zig_units` lets you attach physical units to numeric values so that dimension mismatches (like adding distance to time) become **compile errors** rather than silent bugs. 
+A comptime-first dimensional analysis module for Zig. If you try to add meters to seconds, **it won't compile**. That's the point.
 
-At runtime, a `Quantity` is just its underlying numeric value — **zero memory overhead.**
+Born from a space simulation where `i128` positions were needed to avoid float imprecision far from the origin, this module grew into a full physical-unit type system with zero runtime overhead.
 
-```zig
+> **Source:** [git.bouvais.lu/adrien/zig-dimal](https://git.bouvais.lu/adrien/zig-dimal)  
-const velocity = distance.divBy(time);  // Result type: L¹T⁻¹  ✓
+> **Minimum Zig version:** `0.16.0`
-const error    = mass.add(velocity);    // COMPILE ERROR: M¹ != L¹T⁻¹
-```
 
-**Requirements:** Zig `0.16.0`
+---
+
+## Features
+
+- **100% comptime** — all dimension and unit tracking happens at compile time. No added memory, *almost* native performance.
+- **Compile-time dimension errors** — adding `Meter` to `Second` is a compile error, not a runtime panic.
+- **Automatic unit conversion** — use `.to()` to convert between compatible units (e.g. `km/h` → `m/s`). Scale factors are resolved at comptime.
+- **Full SI prefix support** — `pico`, `nano`, `micro`, `milli`, `centi`, `deci`, `kilo`, `mega`, `giga`, `tera`, `peta`, and more.
+- **Time scale support** — `min`, `hour`, `year` built in.
+- **Scalar and Vector types** — operate on individual values or fixed-size arrays with the same dimensional safety.
+- **Built-in physical quantities** — `dma.Base` provides ready-made types for `Velocity`, `Acceleration`, `Force`, `Energy`, `Pressure`, `ElectricCharge`, `ThermalConductivity`, and many more.
+- **Rich formatting** — values print with their unit automatically: `9.81m.s⁻²`, `42m.kg.s⁻¹`, `0.172km`.
+- **`i128` support** — the whole reason this exists. Use large integers for high-precision fixed-point positions without manual conversion.
+- **Tests and benchmarks included** — run them and see how it performs on your machine (results welcome!).
+
+---
+
+## The 7 SI Base Dimensions
+
+| Symbol | Dimension            | SI Unit  |
+|--------|----------------------|----------|
+| `L`    | Length               | `m`      |
+| `M`    | Mass                 | `g`      |
+| `T`    | Time                 | `s`      |
+| `I`    | Electric Current     | `A`      |
+| `Tp`   | Temperature          | `K`      |
+| `N`    | Amount of Substance  | `mol`    |
+| `J`    | Luminous Intensity   | `cd`     |
 
 ---
 
 ## Installation
 
-### 1. Add as a Zig dependency
+### 1. Fetch the dependency
-```bash
+
-zig fetch --save https://github.com/YOUR_USERNAME/zig_units/archive/refs/heads/main.tar.gz
+```sh
+zig fetch --save git+https://git.bouvais.lu/adrien/zig-dimal#b9647e04266e3f395cfd26b41622b0c119a1e5be
 ```
 
-### 2. Configure `build.zig`
+This will add the following to your `build.zig.zon` automatically:
+
 ```zig
-const zig_units = b.dependency("zig_units", .{
+.dependencies = .{
-    .target = target,
+    .dimal = .{
-    .optimize = optimize,
+        .url = "git+https://git.bouvais.lu/adrien/zig-dimal#b9647e04266e3f395cfd26b41622b0c119a1e5be",
-});
+        .hash = "dimal-0.1.0-WNhSHvomAQAX1ISvq9ZBal-Gam6078y8hE67aC82l63V",
-// Add to your module or executable
+    },
-exe.root_module.addImport("units", zig_units.module("zig_units"));
+},
 ```
 
----
+### 2. Wire it up in `build.zig`
-
-## Quick Start: Using Predefined Quantities
-
-`units.Base` provides a clean way to instantiate common physical types without manually defining dimensions.
 
 ```zig
 const std = @import("std");
-const units = @import("units");
 
-pub fn main() !void {
+pub fn build(b: *std.Build) void {
-    // Instantiate types for f32 backing
+    const target = b.standardTargetOptions(.{});
-    const Meter  = units.Base.Meter.Of(f32);
+    const dimal = b.dependency("dimal", .{}).module("dimal");
-    const Second = units.Base.Second.Of(f32);
-    
-    const dist = Meter{ .value = 10.0 };
-    const time = Second{ .value = 2.0 };
 
-    // Arithmetic is type-safe and creates the correct resulting dimension
+    const exe = b.addExecutable(.{
-    const vel = dist.divBy(time); // Type is Velocity (L/T)
+        .name = "my_project",
-    
+        .root_module = b.createModule(.{
-    std.debug.print("Speed: {f}\n", .{vel}); // Output: 5m.s⁻¹
+            .root_source_file = b.path("src/main.zig"),
+            .target = target,
+            .imports = &.{.{
+                .name = "dimal",
+                .module = dimal,
+            }},
+        }),
+    });
+    b.installArtifact(exe);
 }
 ```
 
----
+### 3. Import in your code
-
-### Core Arithmetic Operations
-
-Dimensional analysis is handled entirely at compile-time. If the math doesn't make physical sense, it won't compile.
 
 ```zig
-const M = units.Base.Meter.Of(f32);
+const dma = @import("dimal");
-const S = units.Base.Second.Of(f32);
+const Scalar     = dma.Scalar;
-
+const Dimensions = dma.Dimensions;
-const dist = M{ .value = 100.0 };
+const Scales     = dma.Scales;
-const time = S{ .value = 5.0 };
-
-// 1. Addition & Subtraction (Must have same dimensions)
-const two_dist = dist.add(dist); // 200.0m
-const zero     = dist.sub(dist); // 0.0m
-
-// 2. Division (Subtracts dimension exponents)
-// Result: Velocity (L¹ T⁻¹)
-const vel = dist.divBy(time); // 20.0 m.s⁻¹
-
-// 3. Multiplication (Adds dimension exponents)
-// Result: Area (L²)
-const area = dist.mulBy(dist); // 100.0 m²
-
-// 4. Chained Operations
-// Result: Acceleration (L¹ T⁻²)
-const accel = dist.divBy(time).divBy(time); // 4.0 m.s⁻²
 ```
 
 ---
 
-## Defining Custom Quantities
+## Quick Start
 
-You aren't limited to the built-in library. You can define any physical quantity by specifying its **Dimensions**
+### Defining unit types
-(powers of base units) and its **Scale** (SI prefixes).
 
-### 1. Create a custom dimension
+A `Scalar` type is parameterized by three things: the numeric type (`f64`, `i128`, …), the dimensions (which physical quantities, and their exponents), and the scales (SI prefixes or custom time units).
-Dimensions are defined by 7 base SI units: `L` (Length), `M` (Mass), `T` (Time), `I` (Current), `Tp` (Temp), `N` (Substance), `J` (Intensity).
 
 ```zig
-const Dims = units.Dimensions;
+const Meter     = Scalar(f64, .init(.{ .L = 1 }),            .init(.{}));
-const Scales = units.Scales;
+const NanoMeter = Scalar(i64, .init(.{ .L = 1 }),            .init(.{ .L = .n }));
-
+const KiloMeter = Scalar(f64, .init(.{ .L = 1 }),            .init(.{ .L = .k }));
-// Frequency is T⁻¹
+const Second    = Scalar(f64, .init(.{ .T = 1 }),            .init(.{}));
-const FreqDims = Dims.init(.{ .T = -1 });
+const Velocity  = Scalar(f64, .init(.{ .L = 1, .T = -1 }),  .init(.{}));
-
+const Kmh       = Scalar(f64, .init(.{ .L = 1, .T = -1 }),  .init(.{ .L = .k, .T = .hour }));
-// Force is M¹ L¹ T⁻²
-const ForceDims = Dims.init(.{ .M = 1, .L = 1, .T = -2 });
 ```
 
-### 2. Create a custom Type
+Or use the pre-built helpers from `dma.Base`:
-Combine a numeric type, the dimensions, and a scale.
 
 ```zig
-const Hertz = units.Quantity(f32, FreqDims, Scales.init(.{}));
+const Acceleration = dma.Base.Acceleration.Of(f64);
-
+const KmhSpeed     = dma.Base.Speed.Scaled(f64, Scales.init(.{ .L = .k, .T = .hour }));
-// A specialized scale: Millimeters per Second Squared
-const MmPerSecSq = units.Quantity(f32, 
-    Dims.init(.{ .L = 1, .T = -2 }), 
-    Scales.init(.{ .L = .m }) // .m = milli
-);
 ```
 
+### Kinematics example
+
+```zig
+const v0    = Velocity{ .value = 10.0 };  // 10 m/s
+const accel = Acceleration{ .value = 9.81 }; // 9.81 m/s²
+const time  = Second{ .value = 5.0 };     // 5 s
+
+// d = v₀t + ½at²
+const d1 = v0.mulBy(time);                     // → Meter
+const d2 = accel.mulBy(time.mulBy(time)).scale(0.5); // → Meter
+const dist = d1.add(d2);
+
+const v_final = v0.add(accel.mulBy(time));
+
+std.debug.print("Distance: {d} | {d}\n", .{ dist, dist.to(KiloMeter) });
+// Distance: 172.625m | 0.172625km
+
+std.debug.print("Final speed: {d:.2}\n", .{v_final});
+// Final speed: 59.05m.s⁻¹
+```
+
+### Unit conversion
+
+`.to()` converts between compatible units at comptime. Mixing incompatible dimensions is a **compile error**.
+
+```zig
+const speed_kmh = Kmh{ .value = 120.0 };
+const speed_ms  = speed_kmh.to(Velocity); // 33.333... m/s — comptime ratio
+
+// This would NOT compile:
+// const bad = speed_kmh.to(Second); // "Dimension mismatch in to: L1T-1 vs T1"
+```
+
+### Working with Vectors
+
+Every `Scalar` type exposes a `.Vec3` and a generic `.Vec(n)`:
+
+```zig
+const Vec3Meter = Meter.Vec3; // or: Vector(3, Meter)
+
+const pos = Vec3Meter{ .data = .{ 100, 200, 300 } };
+const t   = Second{ .value = 10 };
+
+const vel = pos.divByScalar(t); // → Vec3 of Velocity (m/s)
+
+std.debug.print("{d}\n", .{vel}); // (10, 20, 30)m.s⁻¹
+```
+
+Vectors support: `add`, `sub`, `mulBy`, `divBy`, `mulByScalar`, `divByScalar`, `negate`, `to`, `length`, `lengthSqr`.
+
 ---
 
-## Unit Conversions
+## API Reference
 
-The library handles SI prefixes (`k`, `m`, `u`, `n`, etc.) and time aliases (`.min`, `.hour`) automatically.
+### `Scalar(T, dims, scales)`
-When performing arithmetic between different scales, the **finer (smaller) scale wins** to preserve precision.
 
-```zig
+| Method | Description |
-const KM = units.Base.Meter.Scaled(f32, Scales.init(.{ .L = .k })); // Kilometers
+|---|---|
-const M  = units.Base.Meter.Of(f32);                               // Meters
+| `.add(rhs)` | Add two quantities of the same dimension. Auto-converts scales. |
+| `.sub(rhs)` | Subtract. Auto-converts scales. |
+| `.mulBy(rhs)` | Multiply — dimensions are **summed**. `m * s⁻¹` → `m·s⁻¹`. |
+| `.divBy(rhs)` | Divide — dimensions are **subtracted**. `m / s` → `m·s⁻¹`. |
+| `.to(DestType)` | Convert to another unit of the same dimension. Compile error on mismatch. |
+| `.vec3()` | Wrap the value in a `Vec3` of the same type. |
+| `.Vec(n)` | Get the `Vector(n, Self)` type. |
 
-const d1 = KM{ .value = 1.2 };  // 1.2 km
+### `dma.Base` — Pre-built quantities
-const d2 = M{ .value = 300.0 }; // 300 m
 
-const total = d1.add(d2);      // Result is 1500.0 (Meters)
+A selection of what's available (call `.Of(T)` for base units, `.Scaled(T, scales)` for custom scales):
-const final = total.to(KM);    // Explicitly convert back to KM -> 1.5
+
-```
+`Meter`, `Second`, `Gramm`, `Kelvin`, `ElectricCurrent`, `Speed`, `Acceleration`, `Inertia`, `Force`, `Pressure`, `Energy`, `Power`, `Area`, `Volume`, `Density`, `Frequency`, `Viscosity`, `ElectricCharge`, `ElectricPotential`, `ElectricResistance`, `MagneticFlux`, `ThermalCapacity`, `ThermalConductivity`, and more.
+
+### `Scales` — SI prefixes
+
+| Tag | Factor |
+|---|---|
+| `.P` | 10¹⁵ |
+| `.T` | 10¹² |
+| `.G` | 10⁹ |
+| `.M` | 10⁶ |
+| `.k` | 10³ |
+| `.none` | 1 |
+| `.c` | 10⁻² |
+| `.m` | 10⁻³ |
+| `.u` | 10⁻⁶ |
+| `.n` | 10⁻⁹ |
+| `.p` | 10⁻¹² |
+| `.f` | 10⁻¹⁵ |
+| `.min` | 60 |
+| `.hour` | 3600 |
+| `.year` | 31 536 000 |
 
 ---
 
-## Physical Vectors (Vec3)
+## Running Tests and Benchmarks
 
-Physical quantities often come in 3D vectors (Position, Velocity, Force). Every `Quantity` type has a `.Vec3` alias built-in.
+```sh
-
-```zig
-const Vec3M = units.Base.Meter.Of(f32).Vec3;
-
-const gravity = Vec3M{ .data = .{ 0, -9.81, 0 } };
-const pos     = Vec3M.initDefault(0); // [0, 0, 0]
-
-// Vectors support standard operations
-const length = gravity.length(); // Returns f32: 9.81
-const double = gravity.scale(2.0);
-```
-
-You can also create a Vector of any length.
-Vec3 found in a Quantity is just a convenience.
-
-```zig
-const M  = units.Base.Meter.Of(f32);
-const Vec10M = units.QuantityVec(10, Meter);
-
-const gravity = Vec10M.initDefault(1);
-const length = gravity.length(); // Returns f32: 1.0
-```
-
----
-
-## High Precision & Integer Backing
-
-While most libraries default to `f32` or `f64`, `zig_units` is mainly designed to support **large-bit integers (`i128`, `i256`)**. 
-
-This is critical for applications like **space simulations**, where floating-point numbers suffer from "jitter" or "flickering"
-once you travel far from the origin. By using an `i128` with a millimeter scale, you can represent the diameter
-of the observable universe with millimeter precision—something impossible with `f64`.
-
-### Avoiding Floating-Point Jitter
-```zig
-// Millimeter precision using 128-bit integers
-const MM  = units.Base.Meter.Scaled(i128, units.Scales.init(.{ .L = .m }));
-const KM  = units.Base.Meter.Scaled(i128, units.Scales.init(.{ .L = .k }));
-
-const solar_system_dist = KM{ .value = 150_000_000 }; // 150 million km
-const ship_nudge        = MM{ .value = 5 };           // 5 mm
-
-// The library performs exact integer math for conversions.
-// Resulting type is MM (the finer scale), maintaining perfect precision.
-const new_pos = solar_system_dist.add(ship_nudge); 
-```
-
-### Integer-Specific Features
- *   **Exact Conversions:** When converting between integer scales (e.g., `km` to `m`), the library uses fast-path native multiplication.
- *   **Safe Vector Lengths:** `QuantityVec.length()` includes a custom integer square root implementation, allowing you to calculate distances between coordinates without ever casting to a float.
- *   **Zero Drift:** Unlike floats, repeated additions and subtractions of integers never accumulate "epsilon" drift, ensuring your simulation remains deterministic.
- *   **Precision-First Scaling:** When operating on two different scales (e.g., adding `km` and `mm`), the result automatically adopts the finer scale (`mm`). This ensures **zero implicit data loss** during calculation. You only lose precision if you *explicitly* choose to convert back to a coarser scale using `.to()`.
-
----
-
-## SI Scales Reference
-
-| Prefix | Enum | Factor |
-| :--- | :--- | :--- |
-| **Kilo** | `.k` | 10³ |
-| **Mega** | `.M` | 10⁶ |
-| **Giga** | `.G` | 10⁹ |
-| **Milli** | `.m` | 10⁻³ |
-| **Micro** | `.u` | 10⁻⁶ |
-| **Minute**| `.min` | 60 |
-| **Hour**  | `.hour`| 3,600 |
-
----
-
-## API Summary
-
-### `Quantity(T, dims, scales)`
-- `.add(rhs)` / `.sub(rhs)`: Automatic scaling, requires same dimensions.
-- `.mulBy(rhs)` / `.divBy(rhs)`: Composes dimensions (e.g., $L \times L = L^2$).
-- `.scale(scalar)`: Multiply by a raw number (preserves dimensions).
-- `.to(OtherType)`: Safely convert between scales of the same dimension.
-- `.vec3()`: Create a 3D vector from a scalar.
-
-### `Dimensions`
-- `L`: Length (m)
-- `M`: Mass (g)
-- `T`: Time (s)
-- `I`: Current (A)
-- `Tp`: Temperature (K)
-- `N`: Amount (mol)
-- `J`: Intensity (cd)
-
----
-
-## Testing & Benchmarks
-
-`zig_units` comes with a comprehensive test suite that verifies dimensional correctness, SI prefix scaling,
-and vector math accuracy across all numeric types.
-
-### Running Tests
-To run the full suite of unit tests and performance benchmarks:
-```bash
 zig build test
+zig build benchmark
 ```
 
-### Benchmarks
+Benchmark results are very welcome — feel free to share yours!
-When you run the tests, the library also executes a performance benchmark.
-This measures the cost of operations (in nanoseconds per operation) across different backing types (`i32` to `f128`) and vector lengths.
 
-Because all dimensional logic is resolved at **compile-time**, you will see that `Quantity` operations perform at the same speed as raw primitive math.
+---
 
-**Example Benchmark Output:**
+## Roadmap / Known Limitations
-```text
- Quantity<T> benchmark — 100,000 iterations, 10 samples/cell
 
-┌───────────────────┬──────┬─────────────────────┬─────────────────────┐
+- More operations beyond `add`, `sub`, `mulBy`, `divBy` (e.g. `pow`, `sqrt`).
-│ Operation         │ Type │ ns / op (± delta)   │ Throughput (ops/s)  │
+- SIMD acceleration for `Vector` operations.
-├───────────────────┼──────┼─────────────────────┼─────────────────────┤
+- Some paths may still fall back to runtime computation — optimization ongoing.
-│ add               │ i32  │     0.18 ns ±0.02   │          5555555556 │
+- More test coverage.
-│ mulBy             │ f64  │     0.22 ns ±0.01   │          4545454545 │
-└───────────────────┴──────┴─────────────────────┴─────────────────────┘
-```
 
-### Verification Examples
+---
-The test suite ensures that:
+
-- **Dimension Safety:** Chained operations like `dist.divBy(time).divBy(time)` correctly result in an Acceleration type.
+## License
-- **Scale Accuracy:** Adding `1km + 1mm` results exactly in `1000001mm` without truncation.
+
-- **Formatting:** Quantities print correctly with Unicode superscripts (e.g., `9.81m.s⁻²`).
+See the repository for license details.
-- **Vector Math:** Euclidean lengths for both floats and integers are verified against known constants.

build.zig

@@ -2,40 +2,38 @@ const std = @import("std");
 
 pub fn build(b: *std.Build) void {
     const target = b.standardTargetOptions(.{});
-    const optimize = b.standardOptimizeOption(.{ .preferred_optimize_mode = .ReleaseSmall });
+    const optimize = b.standardOptimizeOption(.{});
-    const exe = b.addExecutable(.{
+
-        .name = "Zig_Units",
+    // 1. Define the module so other projects can import it
+    _ = b.addModule("dimal", .{
+        .root_source_file = b.path("src/main.zig"),
+    });
+
+    const exe_tests = b.addTest(.{
         .root_module = b.createModule(.{
             .root_source_file = b.path("src/main.zig"),
             .target = target,
             .optimize = optimize,
+        }),
+        .test_runner = .{ .path = b.path("test_runner.zig"), .mode = .simple },
+    });
+
+    const run_exe_tests = b.addRunArtifact(exe_tests);
+    const test_step = b.step("test", "Run tests");
+    test_step.dependOn(&run_exe_tests.step);
+
+    const bench_exe = b.addExecutable(.{
+        .name = "benchmark",
+        .root_module = b.createModule(.{
+            .root_source_file = b.path("src/benchmark.zig"),
+            .target = target,
             .imports = &.{},
         }),
     });
 
-    b.installArtifact(exe);
+    b.installArtifact(bench_exe);
-    const run_step = b.step("run", "Run the app");
 
-    const run_cmd = b.addRunArtifact(exe);
+    const run_bench = b.addRunArtifact(bench_exe);
-    run_step.dependOn(&run_cmd.step);
+    const bench_step = b.step("benchmark", "Run the benchmark suite");
-
+    bench_step.dependOn(&run_bench.step);
-    run_cmd.step.dependOn(b.getInstallStep());
-
-    if (b.args) |args| {
-        run_cmd.addArgs(args);
-    }
-
-    const exe_tests = b.addTest(.{
-        .root_module = exe.root_module,
-        .test_runner = .{
-            .path = .{ .cwd_relative = "test_runner.zig" },
-            .mode = .simple,
-        },
-    });
-
-    // A run step that will run the second test executable.
-    const run_exe_tests = b.addRunArtifact(exe_tests);
-
-    const test_step = b.step("test", "Run tests");
-    test_step.dependOn(&run_exe_tests.step);
 }

build.zig.zon

@@ -1,81 +1,13 @@
 .{
-    // This is the default name used by packages depending on this one. For
+    .name = .dimal,
-    // example, when a user runs `zig fetch --save <url>`, this field is used
+    .version = "0.1.0",
-    // as the key in the `dependencies` table. Although the user can choose a
+    .fingerprint = 0x9453b1ff1e52d858,
-    // different name, most users will stick with this provided value.
-    //
-    // It is redundant to include "zig" in this name because it is already
-    // within the Zig package namespace.
-    .name = .Zig_Units,
-    // This is a [Semantic Version](https://semver.org/).
-    // In a future version of Zig it will be used for package deduplication.
-    .version = "0.0.0",
-    // Together with name, this represents a globally unique package
-    // identifier. This field is generated by the Zig toolchain when the
-    // package is first created, and then *never changes*. This allows
-    // unambiguous detection of one package being an updated version of
-    // another.
-    //
-    // When forking a Zig project, this id should be regenerated (delete the
-    // field and run `zig build`) if the upstream project is still maintained.
-    // Otherwise, the fork is *hostile*, attempting to take control over the
-    // original project's identity. Thus it is recommended to leave the comment
-    // on the following line intact, so that it shows up in code reviews that
-    // modify the field.
-    .fingerprint = 0x934b00cf88f69b22, // Changing this has security and trust implications.
-    // Tracks the earliest Zig version that the package considers to be a
-    // supported use case.
     .minimum_zig_version = "0.16.0",
-    // This field is optional.
+    .dependencies = .{},
-    // Each dependency must either provide a `url` and `hash`, or a `path`.
-    // `zig build --fetch` can be used to fetch all dependencies of a package, recursively.
-    // Once all dependencies are fetched, `zig build` no longer requires
-    // internet connectivity.
-    .dependencies = .{
-        // See `zig fetch --save <url>` for a command-line interface for adding dependencies.
-        //.example = .{
-        //    // When updating this field to a new URL, be sure to delete the corresponding
-        //    // `hash`, otherwise you are communicating that you expect to find the old hash at
-        //    // the new URL. If the contents of a URL change this will result in a hash mismatch
-        //    // which will prevent zig from using it.
-        //    .url = "https://example.com/foo.tar.gz",
-        //
-        //    // This is computed from the file contents of the directory of files that is
-        //    // obtained after fetching `url` and applying the inclusion rules given by
-        //    // `paths`.
-        //    //
-        //    // This field is the source of truth; packages do not come from a `url`; they
-        //    // come from a `hash`. `url` is just one of many possible mirrors for how to
-        //    // obtain a package matching this `hash`.
-        //    //
-        //    // Uses the [multihash](https://multiformats.io/multihash/) format.
-        //    .hash = "...",
-        //
-        //    // When this is provided, the package is found in a directory relative to the
-        //    // build root. In this case the package's hash is irrelevant and therefore not
-        //    // computed. This field and `url` are mutually exclusive.
-        //    .path = "foo",
-        //
-        //    // When this is set to `true`, a package is declared to be lazily
-        //    // fetched. This makes the dependency only get fetched if it is
-        //    // actually used.
-        //    .lazy = false,
-        //},
-    },
-    // Specifies the set of files and directories that are included in this package.
-    // Only files and directories listed here are included in the `hash` that
-    // is computed for this package. Only files listed here will remain on disk
-    // when using the zig package manager. As a rule of thumb, one should list
-    // files required for compilation plus any license(s).
-    // Paths are relative to the build root. Use the empty string (`""`) to refer to
-    // the build root itself.
-    // A directory listed here means that all files within, recursively, are included.
     .paths = .{
         "build.zig",
         "build.zig.zon",
         "src",
-        // For example...
+        "test_runner.zig",
-        //"LICENSE",
-        //"README.md",
     },
 }

src/Base.zig

@@ -25,6 +25,8 @@ pub fn BaseScalar(comptime d: anytype) type {
     };
 }
 
+pub const Dimless = BaseScalar(.{});
+
 // ==========================================
 // Base Quantities
 // ==========================================

src/Scalar.zig

@@ -8,7 +8,7 @@ const Dimensions = @import("Dimensions.zig");
 const Dimension = Dimensions.Dimension;
 
 pub fn Scalar(comptime T: type, comptime d: Dimensions, comptime s: Scales) type {
-    @setEvalBranchQuota(100_000);
+    @setEvalBranchQuota(10_000_000);
     return struct {
         value: T,
 
@@ -86,10 +86,6 @@ pub fn Scalar(comptime T: type, comptime d: Dimensions, comptime s: Scales) type
             }
         }
 
-        pub inline fn scale(self: Self, sc: T) Self {
-            return .{ .value = self.value * sc };
-        }
-
         pub inline fn to(self: Self, comptime Dest: type) Dest {
             if (comptime !dims.eql(Dest.dims))
                 @compileError("Dimension mismatch in to: " ++ dims.str() ++ " vs " ++ Dest.dims.str());
@@ -140,11 +136,21 @@ pub fn Scalar(comptime T: type, comptime d: Dimensions, comptime s: Scales) type
             return Vec3.initDefault(self.value);
         }
 
-        pub fn format(
+        pub fn formatNumber(
             self: Self,
             writer: *std.Io.Writer,
+            options: std.fmt.Number,
         ) !void {
-            try writer.print("{d}", .{self.value});
+            switch (@typeInfo(T)) {
+                .float, .comptime_float => try writer.printFloat(self.value, options),
+                .int, .comptime_int => try writer.printInt(self.value, 10, .lower, .{
+                    .width = options.width,
+                    .alignment = options.alignment,
+                    .fill = options.fill,
+                    .precision = options.precision,
+                }),
+                else => unreachable,
+            }
             var first = true;
             inline for (std.enums.values(Dimension)) |bu| {
                 const v = dims.get(bu);
@@ -187,48 +193,39 @@ test "Add" {
     const added = distance.add(distance2);
     try std.testing.expectEqual(30, added.value);
     try std.testing.expectEqual(1, @TypeOf(added).dims.get(.L));
-    std.debug.print("KiloMeter {f} + {f} = {f} OK\n", .{ distance, distance2, added });
 
     const KiloMeter = Scalar(i128, Dimensions.init(.{ .L = 1 }), Scales.init(.{ .L = .k }));
     const distance3 = KiloMeter{ .value = 2 };
     const added2 = distance.add(distance3);
     try std.testing.expectEqual(2010, added2.value);
     try std.testing.expectEqual(1, @TypeOf(added2).dims.get(.L));
-    std.debug.print("KiloMeter {f} + {f} = {f} OK\n", .{ distance, distance3, added2 });
 
     const added3 = distance3.add(distance).to(KiloMeter);
     try std.testing.expectEqual(2, added3.value);
     try std.testing.expectEqual(1, @TypeOf(added3).dims.get(.L));
-    std.debug.print("KiloMeter {f} + {f} = {f} OK\n", .{ distance3, distance, added3 });
 
     const KiloMeter_f = Scalar(f64, Dimensions.init(.{ .L = 1 }), Scales.init(.{ .L = .k }));
     const distance4 = KiloMeter_f{ .value = 2 };
     const added4 = distance4.add(distance).to(KiloMeter_f);
     try std.testing.expectApproxEqAbs(2.01, added4.value, 0.000001);
     try std.testing.expectEqual(1, @TypeOf(added4).dims.get(.L));
-    std.debug.print("KiloMeter_f {f} + {f} = {f} OK\n", .{ distance4, distance, added4 });
 }
 
 test "Sub" {
     const Meter = Scalar(i128, Dimensions.init(.{ .L = 1 }), Scales.init(.{}));
-    const KiloMeter = Scalar(i128, Dimensions.init(.{ .L = 1 }), Scales.init(.{ .L = .k }));
     const KiloMeter_f = Scalar(f64, Dimensions.init(.{ .L = 1 }), Scales.init(.{ .L = .k }));
 
     const a = Meter{ .value = 500 };
     const b = Meter{ .value = 200 };
     const diff = a.sub(b);
     try std.testing.expectEqual(300, diff.value);
-    std.debug.print("Sub: {f} - {f} = {f} OK\n", .{ a, b, diff });
+    const diff2 = b.sub(a);
-
+    try std.testing.expectEqual(-300, diff2.value);
-    const km = KiloMeter{ .value = 1 };
-    const diff2 = a.sub(km);
-    std.debug.print("Sub cross-scale: {f} - {f} = {f}\n", .{ a, km, diff2 });
 
     const km_f = KiloMeter_f{ .value = 2.5 };
     const m_f = Meter{ .value = 500 };
     const diff3 = km_f.sub(m_f);
     try std.testing.expectApproxEqAbs(2000, diff3.value, 1e-4);
-    std.debug.print("Sub float cross-scale: {f} - {f} = {f} OK\n", .{ km_f, m_f, diff3 });
 }
 
 test "MulBy" {
@@ -242,14 +239,12 @@ test "MulBy" {
     try std.testing.expectEqual(12, area_time.value);
     try std.testing.expectEqual(1, @TypeOf(area_time).dims.get(.L));
     try std.testing.expectEqual(1, @TypeOf(area_time).dims.get(.T));
-    std.debug.print("MulBy: {f} * {f} = {f} OK\n", .{ d, t, area_time });
 
     const d2 = Meter{ .value = 5.0 };
     const area = d.mulBy(d2);
     try std.testing.expectEqual(15, area.value);
     try std.testing.expectEqual(2, @TypeOf(area).dims.get(.L));
     try std.testing.expectEqual(0, @TypeOf(area).dims.get(.T));
-    std.debug.print("MulBy: {f} * {f} = {f} OK\n", .{ d, d2, area });
 }
 
 test "MulBy with scale" {
@@ -261,7 +256,6 @@ test "MulBy with scale" {
     const prod = dist.mulBy(mass);
     try std.testing.expectEqual(1, @TypeOf(prod).dims.get(.L));
     try std.testing.expectEqual(1, @TypeOf(prod).dims.get(.M));
-    std.debug.print("MulBy scaled: {f} * {f} = {f} OK\n", .{ dist, mass, prod });
 }
 
 test "MulBy with type change" {
@@ -279,7 +273,6 @@ test "MulBy with type change" {
     try std.testing.expectApproxEqAbs(12, area_time_f.value, 0.0001);
     try std.testing.expectEqual(1, @TypeOf(area_time).dims.get(.L));
     try std.testing.expectEqual(1, @TypeOf(area_time).dims.get(.T));
-    std.debug.print("MulBy: {f} * {f} = {f} OK\n", .{ d, t, area_time });
 }
 
 test "MulBy small" {
@@ -293,23 +286,16 @@ test "MulBy small" {
     try std.testing.expectEqual(12, area_time.value);
     try std.testing.expectEqual(1, @TypeOf(area_time).dims.get(.L));
     try std.testing.expectEqual(1, @TypeOf(area_time).dims.get(.T));
-    std.debug.print("MulBy: {f} * {f} = {f} OK\n", .{ d, t, area_time });
 }
 
-test "Scale" {
+test "MulBy dimensionless" {
+    const DimLess = Scalar(i128, Dimensions.init(.{}), Scales.init(.{}));
     const Meter = Scalar(i128, Dimensions.init(.{ .L = 1 }), Scales.init(.{}));
-    const Second = Scalar(f32, Dimensions.init(.{ .T = 1 }), Scales.init(.{}));
 
     const d = Meter{ .value = 7 };
-    const scaled = d.scale(3);
+    const scaled = d.mulBy(DimLess{ .value = 3 });
     try std.testing.expectEqual(21, scaled.value);
     try std.testing.expectEqual(1, @TypeOf(scaled).dims.get(.L));
-    std.debug.print("Scale int: {f} * 3 = {f} OK\n", .{ d, scaled });
-
-    const t = Second{ .value = 1.5 };
-    const scaled_f = t.scale(4.0);
-    try std.testing.expectApproxEqAbs(@as(f32, 6.0), scaled_f.value, 1e-4);
-    std.debug.print("Scale float: {f} * 4 = {f} OK\n", .{ t, scaled_f });
 }
 
 test "Chained: velocity and acceleration" {
@@ -328,8 +314,6 @@ test "Chained: velocity and acceleration" {
     try std.testing.expectEqual(5, accel.value);
     try std.testing.expectEqual(1, @TypeOf(accel).dims.get(.L));
     try std.testing.expectEqual(-2, @TypeOf(accel).dims.get(.T));
-
-    std.debug.print("Velocity: {f}, Acceleration: {f} OK\n", .{ velocity, accel });
 }
 
 test "DivBy integer exact" {
@@ -343,7 +327,6 @@ test "DivBy integer exact" {
     try std.testing.expectEqual(30, vel.value);
     try std.testing.expectEqual(1, @TypeOf(vel).dims.get(.L));
     try std.testing.expectEqual(-1, @TypeOf(vel).dims.get(.T));
-    std.debug.print("DivBy int: {f} / {f} = {f} OK\n", .{ dist, time, vel });
 }
 
 test "Conversion chain: km -> m -> cm" {
@@ -357,7 +340,6 @@ test "Conversion chain: km -> m -> cm" {
 
     try std.testing.expectEqual(15_000, m.value);
     try std.testing.expectEqual(1_500_000, cm.value);
-    std.debug.print("Chain: {f} -> {f} -> {f} OK\n", .{ km, m, cm });
 }
 
 test "Conversion: hours -> minutes -> seconds" {
@@ -371,7 +353,6 @@ test "Conversion: hours -> minutes -> seconds" {
 
     try std.testing.expectEqual(60, min.value);
     try std.testing.expectEqual(3600, sec.value);
-    std.debug.print("Time chain: {f} -> {f} -> {f} OK\n", .{ h, min, sec });
 }
 
 test "Negative values" {
@@ -381,10 +362,9 @@ test "Negative values" {
     const b = Meter{ .value = 20 };
     const diff = a.sub(b);
     try std.testing.expectEqual(-15, diff.value);
-    std.debug.print("Negative sub: {f} - {f} = {f} OK\n", .{ a, b, diff });
 }
 
-test "Format Quantity" {
+test "Format Scalar" {
     const MeterPerSecondSq = Scalar(
         f32,
         Dimensions.init(.{ .L = 1, .T = -2 }),
@@ -395,396 +375,22 @@ test "Format Quantity" {
         Dimensions.init(.{ .M = 1, .L = 1, .T = -1 }),
         Scales.init(.{ .M = .k }),
     );
+    const Meter = Scalar(f32, Dimensions.init(.{ .L = 1 }), Scales.init(.{}));
 
+    const m = Meter{ .value = 1.23456 };
     const accel = MeterPerSecondSq{ .value = 9.81 };
     const momentum = KgMeterPerSecond{ .value = 42.0 };
 
-    std.debug.print("Acceleration: {f}\n", .{accel});
+    var buf: [64]u8 = undefined;
-    std.debug.print("Momentum: {f}\n", .{momentum});
+    var res = try std.fmt.bufPrint(&buf, "{d:.2}", .{m});
-}
+    try std.testing.expectEqualStrings("1.23m", res);
 
-test "Benchmark" {
+    res = try std.fmt.bufPrint(&buf, "{d}", .{accel});
-    const Io = std.Io;
+    try std.testing.expectEqualStrings("9.81m.ns⁻²", res);
-    const ITERS: usize = 100_000;
-    const SAMPLES: usize = 10; // Number of samples for stats
 
-    var gsink: f64 = 0;
+    res = try std.fmt.bufPrint(&buf, "{d}", .{momentum});
-    const io = std.testing.io;
+    try std.testing.expectEqualStrings("42m.kg.s⁻¹", res);
 
-    // Standard Zig 0.16 timestamp retrieval
+    res = try std.fmt.bufPrint(&buf, "{d:_>10.1}", .{m});
-    const getTime = struct {
+    try std.testing.expectEqualStrings("_______1.2m", res);
-        fn f(i: Io) Io.Timestamp {
-            return Io.Clock.awake.now(i);
-        }
-    }.f;
-
-    const fold = struct {
-        fn f(comptime TT: type, s: *f64, v: TT) void {
-            s.* += if (comptime @typeInfo(TT) == .float)
-                @as(f64, @floatCast(v))
-            else
-                @as(f64, @floatFromInt(v));
-        }
-    }.f;
-
-    const getVal = struct {
-        fn f(comptime TT: type, i: usize, comptime mask: u7) TT {
-            const v: u8 = @as(u8, @truncate(i & @as(usize, mask))) + 1;
-            return if (comptime @typeInfo(TT) == .float) @floatFromInt(v) else @intCast(v);
-        }
-    }.f;
-
-    const Stats = struct {
-        median: f64,
-        delta: f64,
-        ops_per_sec: f64,
-    };
-
-    const computeStats = struct {
-        fn f(samples: []f64, iters: usize) Stats {
-            std.mem.sort(f64, samples, {}, std.sort.asc(f64));
-            const mid = samples.len / 2;
-            const median_ns = if (samples.len % 2 == 0) (samples[mid - 1] + samples[mid]) / 2.0 else samples[mid];
-
-            const low = samples[0];
-            const high = samples[samples.len - 1];
-            const delta_ns = (high - low) / 2.0;
-
-            const ns_per_op = median_ns / @as(f64, @floatFromInt(iters));
-            return .{
-                .median = ns_per_op,
-                .delta = (delta_ns / @as(f64, @floatFromInt(iters))),
-                .ops_per_sec = 1_000_000_000.0 / ns_per_op,
-            };
-        }
-    }.f;
-
-    std.debug.print(
-        \\
-        \\ Quantity<T> benchmark — {d} iterations, {d} samples/cell
-        \\
-        \\┌───────────────────┬──────┬─────────────────────┬─────────────────────┐
-        \\│ Operation         │ Type │ ns / op (± delta)   │ Throughput (ops/s)  │
-        \\├───────────────────┼──────┼─────────────────────┼─────────────────────┤
-        \\
-    , .{ ITERS, SAMPLES });
-
-    const Types = .{ i16, i32, i64, i128, i256, f32, f64, f128 };
-    const TNames = .{ "i16", "i32", "i64", "i128", "i256", "f32", "f64", "f128" };
-    const Ops = .{ "add", "sub", "mulBy", "divBy", "scale", "to" };
-
-    var results_matrix: [Ops.len][Types.len]f64 = undefined;
-
-    comptime var tidx: usize = 0;
-    inline for (Types, TNames) |T, tname| {
-        const M = Scalar(T, Dimensions.init(.{ .L = 1 }), Scales.init(.{}));
-        const KM = Scalar(T, Dimensions.init(.{ .L = 1 }), Scales.init(.{ .L = .k }));
-        const S = Scalar(T, Dimensions.init(.{ .T = 1 }), Scales.init(.{}));
-
-        inline for (Ops, 0..) |op_name, oidx| {
-            var samples: [SAMPLES]f64 = undefined;
-
-            for (0..SAMPLES) |s_idx| {
-                var sink: T = 0;
-                const t_start = getTime(io);
-
-                for (0..ITERS) |i| {
-                    const r = if (comptime std.mem.eql(u8, op_name, "add"))
-                        (M{ .value = getVal(T, i, 63) }).add(M{ .value = getVal(T, i +% 7, 63) })
-                    else if (comptime std.mem.eql(u8, op_name, "sub"))
-                        (M{ .value = getVal(T, i +% 10, 63) }).sub(M{ .value = getVal(T, i, 63) })
-                    else if (comptime std.mem.eql(u8, op_name, "mulBy"))
-                        (M{ .value = getVal(T, i, 63) }).mulBy(M{ .value = getVal(T, i +% 1, 63) })
-                    else if (comptime std.mem.eql(u8, op_name, "divBy"))
-                        (M{ .value = getVal(T, i +% 10, 63) }).divBy(S{ .value = getVal(T, i, 63) })
-                    else if (comptime std.mem.eql(u8, op_name, "scale"))
-                        (M{ .value = getVal(T, i, 63) }).scale(getVal(T, i +% 2, 63))
-                    else
-                        (KM{ .value = getVal(T, i, 15) }).to(M);
-
-                    if (comptime @typeInfo(T) == .float) sink += r.value else sink ^= r.value;
-                }
-
-                const t_end = getTime(io);
-                samples[s_idx] = @as(f64, @floatFromInt(t_start.durationTo(t_end).toNanoseconds()));
-                fold(T, &gsink, sink);
-            }
-
-            const stats = computeStats(&samples, ITERS);
-            results_matrix[oidx][tidx] = stats.median;
-
-            std.debug.print("│ {s:<17} │ {s:<4} │ {d:>8.2} ns ±{d:<6.2} │ {d:>19.0} │\n", .{ op_name, tname, stats.median, stats.delta, stats.ops_per_sec });
-        }
-
-        if (comptime tidx < Types.len - 1) {
-            std.debug.print("├───────────────────┼──────┼─────────────────────┼─────────────────────┤\n", .{});
-        }
-        tidx += 1;
-    }
-
-    // Median Summary Table
-    std.debug.print("└───────────────────┴──────┴─────────────────────┴─────────────────────┘\n\n", .{});
-    std.debug.print("Median Summary (ns/op):\n", .{});
-
-    std.debug.print("┌──────────────┬───────┬───────┬───────┬───────┬───────┬───────┬───────┬───────┐\n", .{});
-    std.debug.print("│  Operation   │  i16  │  i32  │  i64  │  i128 │  i256 │  f32  │  f64  │  f128 │\n", .{});
-    std.debug.print("├──────────────┼───────┼───────┼───────┼───────┼───────┼───────┼───────┼───────┤\n", .{});
-
-    inline for (Ops, 0..) |op_name, oidx| {
-        std.debug.print("│  {s:<11} │", .{op_name});
-        var i: usize = 0;
-        while (i < Types.len) : (i += 1) {
-            std.debug.print("{d:>6.1} │", .{results_matrix[oidx][i]});
-        }
-        std.debug.print("\n", .{});
-    }
-
-    std.debug.print("└──────────────┴───────┴───────┴───────┴───────┴───────┴───────┴───────┴───────┘\n", .{});
-
-    std.debug.print("\nAnti-optimisation sink: {d:.4}\n", .{gsink});
-    try std.testing.expect(gsink != 0);
-}
-
-test "Overhead Analysis: Quantity vs Native" {
-    const Io = std.Io;
-    const ITERS: usize = 100_000;
-    const SAMPLES: usize = 5;
-    const io = std.testing.io;
-
-    const getTime = struct {
-        fn f(i: Io) Io.Timestamp {
-            return Io.Clock.awake.now(i);
-        }
-    }.f;
-
-    const fold = struct {
-        fn f(comptime TT: type, s: *f64, v: TT) void {
-            s.* += if (comptime @typeInfo(TT) == .float)
-                @as(f64, @floatCast(v))
-            else
-                @as(f64, @floatFromInt(v));
-        }
-    }.f;
-
-    // Helper to safely get a value of type T from a loop index
-    const getValT = struct {
-        fn f(comptime TT: type, i: usize) TT {
-            const v = (i % 100) + 1;
-            return if (comptime @typeInfo(TT) == .float) @floatFromInt(v) else @intCast(v);
-        }
-    }.f;
-
-    const Types = .{ i32, i64, i128, f32, f64 };
-    const TNames = .{ "i32", "i64", "i128", "f32", "f64" };
-    const Ops = .{ "add", "mulBy", "divBy" };
-
-    var gsink: f64 = 0;
-
-    std.debug.print(
-        \\
-        \\ Quantity vs Native Overhead Analysis
-        \\
-        \\┌───────────┬──────┬───────────┬───────────┬───────────┐
-        \\│ Operation │ Type │ Native    │ Quantity  │ Slowdown  │
-        \\├───────────┼──────┼───────────┼───────────┼───────────┤
-        \\
-    , .{});
-
-    inline for (Ops, 0..) |op_name, j| {
-        inline for (Types, 0..) |T, tidx| {
-            var native_total_ns: f64 = 0;
-            var quantity_total_ns: f64 = 0;
-
-            const M = Scalar(T, Dimensions.init(.{ .L = 1 }), Scales.init(.{}));
-            const S = Scalar(T, Dimensions.init(.{ .T = 1 }), Scales.init(.{}));
-
-            for (0..SAMPLES) |_| {
-                // --- 1. Benchmark Native ---
-                var n_sink: T = 0;
-                const n_start = getTime(io);
-                for (0..ITERS) |i| {
-                    const a = getValT(T, i);
-                    const b = getValT(T, 2);
-                    const r = if (comptime std.mem.eql(u8, op_name, "add"))
-                        a + b
-                    else if (comptime std.mem.eql(u8, op_name, "mulBy"))
-                        a * b
-                    else if (comptime @typeInfo(T) == .int) @divTrunc(a, b) else a / b;
-
-                    if (comptime @typeInfo(T) == .float) n_sink += r else n_sink ^= r;
-                }
-                const n_end = getTime(io);
-                native_total_ns += @as(f64, @floatFromInt(n_start.durationTo(n_end).toNanoseconds()));
-                fold(T, &gsink, n_sink);
-
-                // --- 2. Benchmark Quantity ---
-                var q_sink: T = 0;
-                const q_start = getTime(io);
-                for (0..ITERS) |i| {
-                    const qa = M{ .value = getValT(T, i) };
-                    const qb = if (comptime std.mem.eql(u8, op_name, "divBy")) S{ .value = getValT(T, 2) } else M{ .value = getValT(T, 2) };
-
-                    const r = if (comptime std.mem.eql(u8, op_name, "add"))
-                        qa.add(qb)
-                    else if (comptime std.mem.eql(u8, op_name, "mulBy"))
-                        qa.mulBy(qb)
-                    else
-                        qa.divBy(qb);
-
-                    if (comptime @typeInfo(T) == .float) q_sink += r.value else q_sink ^= r.value;
-                }
-                const q_end = getTime(io);
-                quantity_total_ns += @as(f64, @floatFromInt(q_start.durationTo(q_end).toNanoseconds()));
-                fold(T, &gsink, q_sink);
-            }
-
-            const avg_n = (native_total_ns / SAMPLES) / @as(f64, @floatFromInt(ITERS));
-            const avg_q = (quantity_total_ns / SAMPLES) / @as(f64, @floatFromInt(ITERS));
-            const slowdown = avg_q / avg_n;
-
-            std.debug.print("│ {s:<9} │ {s:<4} │ {d:>7.2}ns │ {d:>7.2}ns │ {d:>8.2}x │\n", .{
-                op_name, TNames[tidx], avg_n, avg_q, slowdown,
-            });
-        }
-        if (j != Ops.len - 1) std.debug.print("├───────────┼──────┼───────────┼───────────┼───────────┤\n", .{});
-    }
-
-    std.debug.print("└───────────┴──────┴───────────┴───────────┴───────────┘\n", .{});
-    try std.testing.expect(gsink != 0);
-}
-
-test "Cross-Type Overhead Analysis: Quantity vs Native" {
-    const Io = std.Io;
-    const ITERS: usize = 100_000;
-    const SAMPLES: usize = 5;
-    const io = std.testing.io;
-
-    const getTime = struct {
-        fn f(i: Io) Io.Timestamp {
-            return Io.Clock.awake.now(i);
-        }
-    }.f;
-
-    const fold = struct {
-        fn f(comptime TT: type, s: *f64, v: TT) void {
-            s.* += if (comptime @typeInfo(TT) == .float)
-                @as(f64, @floatCast(v))
-            else
-                @as(f64, @floatFromInt(v));
-        }
-    }.f;
-
-    const getValT = struct {
-        fn f(comptime TT: type, i: usize) TT {
-            // Keep values safe and non-zero to avoid division by zero or overflows during cross-casting
-            const v = (i % 50) + 1;
-            return if (comptime @typeInfo(TT) == .float) @floatFromInt(v) else @intCast(v);
-        }
-    }.f;
-
-    // Helper for the Native baseline: explicitly casting T2 to T1 before the operation
-    const castTo = struct {
-        fn f(comptime DestT: type, comptime SrcT: type, val: SrcT) DestT {
-            if (comptime DestT == SrcT) return val;
-            const src_info = @typeInfo(SrcT);
-            const dest_info = @typeInfo(DestT);
-
-            if (dest_info == .int and src_info == .int) return @intCast(val);
-            if (dest_info == .float and src_info == .int) return @floatFromInt(val);
-            if (dest_info == .int and src_info == .float) return @intFromFloat(val);
-            if (dest_info == .float and src_info == .float) return @floatCast(val);
-            unreachable;
-        }
-    }.f;
-
-    const Types = .{ i16, i64, i128, f32, f64 };
-    const TNames = .{ "i16", "i64", "i128", "f32", "f64" };
-    const Ops = .{ "add", "mulBy", "divBy" };
-
-    var gsink: f64 = 0;
-
-    std.debug.print(
-        \\
-        \\ Cross-Type Overhead Analysis: Quantity vs Native
-        \\
-        \\┌─────────┬──────┬──────┬───────────┬───────────┬───────────┐
-        \\│ Op      │ T1   │ T2   │ Native    │ Quantity  │ Slowdown  │
-        \\├─────────┼──────┼──────┼───────────┼───────────┼───────────┤
-        \\
-    , .{});
-
-    inline for (Ops, 0..) |op_name, j| {
-        inline for (Types, 0..) |T1, t1_idx| {
-            inline for (Types, 0..) |T2, t2_idx| {
-                var native_total_ns: f64 = 0;
-                var quantity_total_ns: f64 = 0;
-
-                const M1 = Scalar(T1, Dimensions.init(.{ .L = 1 }), Scales.init(.{}));
-                const M2 = Scalar(T2, Dimensions.init(.{ .L = 1 }), Scales.init(.{}));
-                const S2 = Scalar(T2, Dimensions.init(.{ .T = 1 }), Scales.init(.{}));
-
-                for (0..SAMPLES) |_| {
-                    // --- 1. Benchmark Native (Cast T2 to T1, then math) ---
-                    var n_sink: T1 = 0;
-                    const n_start = getTime(io);
-                    for (0..ITERS) |i| {
-                        const a = getValT(T1, i);
-                        const b_raw = getValT(T2, 2);
-                        const b = castTo(T1, T2, b_raw);
-
-                        const r = if (comptime std.mem.eql(u8, op_name, "add"))
-                            a + b
-                        else if (comptime std.mem.eql(u8, op_name, "mulBy"))
-                            a * b
-                        else if (comptime @typeInfo(T1) == .int)
-                            @divTrunc(a, b)
-                        else
-                            a / b;
-
-                        if (comptime @typeInfo(T1) == .float) n_sink += r else n_sink ^= r;
-                    }
-                    const n_end = getTime(io);
-                    native_total_ns += @as(f64, @floatFromInt(n_start.durationTo(n_end).toNanoseconds()));
-                    fold(T1, &gsink, n_sink);
-
-                    // --- 2. Benchmark Quantity ---
-                    var q_sink: T1 = 0;
-                    const q_start = getTime(io);
-                    for (0..ITERS) |i| {
-                        const qa = M1{ .value = getValT(T1, i) };
-                        const qb = if (comptime std.mem.eql(u8, op_name, "divBy"))
-                            S2{ .value = getValT(T2, 2) }
-                        else
-                            M2{ .value = getValT(T2, 2) };
-
-                        const r = if (comptime std.mem.eql(u8, op_name, "add"))
-                            qa.add(qb)
-                        else if (comptime std.mem.eql(u8, op_name, "mulBy"))
-                            qa.mulBy(qb)
-                        else
-                            qa.divBy(qb);
-
-                        if (comptime @typeInfo(T1) == .float) q_sink += r.value else q_sink ^= r.value;
-                    }
-                    const q_end = getTime(io);
-                    quantity_total_ns += @as(f64, @floatFromInt(q_start.durationTo(q_end).toNanoseconds()));
-                    fold(T1, &gsink, q_sink);
-                }
-
-                const avg_n = (native_total_ns / SAMPLES) / @as(f64, @floatFromInt(ITERS));
-                const avg_q = (quantity_total_ns / SAMPLES) / @as(f64, @floatFromInt(ITERS));
-                const slowdown = avg_q / avg_n;
-
-                std.debug.print("│ {s:<7} │ {s:<4} │ {s:<4} │ {d:>7.2}ns │ {d:>7.2}ns │ {d:>8.2}x │\n", .{
-                    op_name, TNames[t1_idx], TNames[t2_idx], avg_n, avg_q, slowdown,
-                });
-            }
-        }
-        if (j != Ops.len - 1) {
-            std.debug.print("├─────────┼──────┼──────┼───────────┼───────────┼───────────┤\n", .{});
-        }
-    }
-
-    std.debug.print("└─────────┴──────┴──────┴───────────┴───────────┴───────────┘\n", .{});
-    try std.testing.expect(gsink != 0);
 }

src/Vector.zig

@@ -26,74 +26,74 @@ pub fn Vector(comptime len: usize, comptime Q: type) type {
 
         pub fn initDefault(v: T) Self {
             var data: [len]T = undefined;
-            for (&data) |*item| item.* = v;
+            inline for (&data) |*item| item.* = v;
             return .{ .data = data };
         }
 
-        pub fn add(self: Self, rhs: anytype) Vector(len, Scalar(T, d, s.min(@TypeOf(rhs).scales))) {
+        pub inline fn add(self: Self, rhs: anytype) Vector(len, Scalar(T, d, s.min(@TypeOf(rhs).scales))) {
             const Tr = @TypeOf(rhs);
             var res: Vector(len, Scalar(T, d, s.min(Tr.scales))) = undefined;
-            for (self.data, 0..) |v, i| {
+            inline for (self.data, 0..) |v, i| {
                 const q = (Q{ .value = v }).add(Tr.ScalarType{ .value = rhs.data[i] });
                 res.data[i] = q.value;
             }
             return res;
         }
 
-        pub fn sub(self: Self, rhs: anytype) Vector(len, Scalar(T, d, s.min(@TypeOf(rhs).scales))) {
+        pub inline fn sub(self: Self, rhs: anytype) Vector(len, Scalar(T, d, s.min(@TypeOf(rhs).scales))) {
             const Tr = @TypeOf(rhs);
             var res: Vector(len, Scalar(T, d, s.min(Tr.scales))) = undefined;
-            for (self.data, 0..) |v, i| {
+            inline for (self.data, 0..) |v, i| {
                 const q = (Q{ .value = v }).sub(Tr.ScalarType{ .value = rhs.data[i] });
                 res.data[i] = q.value;
             }
             return res;
         }
 
-        pub fn divBy(
+        pub inline fn divBy(
             self: Self,
             rhs: anytype,
         ) Vector(len, Scalar(T, d.sub(@TypeOf(rhs).dims), s.min(@TypeOf(rhs).scales))) {
             const Tr = @TypeOf(rhs);
             var res: Vector(len, Scalar(T, d.sub(Tr.dims), s.min(Tr.scales))) = undefined;
-            for (self.data, 0..) |v, i| {
+            inline for (self.data, 0..) |v, i| {
                 const q = (Q{ .value = v }).divBy(Tr.ScalarType{ .value = rhs.data[i] });
                 res.data[i] = q.value;
             }
             return res;
         }
 
-        pub fn mulBy(
+        pub inline fn mulBy(
             self: Self,
             rhs: anytype,
         ) Vector(len, Scalar(T, d.add(@TypeOf(rhs).dims), s.min(@TypeOf(rhs).scales))) {
             const Tr = @TypeOf(rhs);
             var res: Vector(len, Scalar(T, d.add(Tr.dims), s.min(Tr.scales))) = undefined;
-            for (self.data, 0..) |v, i| {
+            inline for (self.data, 0..) |v, i| {
                 const q = (Q{ .value = v }).mulBy(Tr.ScalarType{ .value = rhs.data[i] });
                 res.data[i] = q.value;
             }
             return res;
         }
 
-        pub fn divByScalar(
+        pub inline fn divByScalar(
             self: Self,
             scalar: anytype,
         ) Vector(len, Scalar(T, d.sub(@TypeOf(scalar).dims), s.min(@TypeOf(scalar).scales))) {
             var res: Vector(len, Scalar(T, d.sub(@TypeOf(scalar).dims), s.min(@TypeOf(scalar).scales))) = undefined;
-            for (self.data, 0..) |v, i| {
+            inline for (self.data, 0..) |v, i| {
                 const q = Q{ .value = v };
                 res.data[i] = q.divBy(scalar).value;
             }
             return res;
         }
 
-        pub fn mulByScalar(
+        pub inline fn mulByScalar(
             self: Self,
             scalar: anytype,
         ) Vector(len, Scalar(T, d.add(@TypeOf(scalar).dims), s.min(@TypeOf(scalar).scales))) {
             var res: Vector(len, Scalar(T, d.add(@TypeOf(scalar).dims), s.min(@TypeOf(scalar).scales))) = undefined;
-            for (self.data, 0..) |v, i| {
+            inline for (self.data, 0..) |v, i| {
                 const q = Q{ .value = v };
                 res.data[i] = q.mulBy(scalar).value;
             }
@@ -102,37 +102,29 @@ pub fn Vector(comptime len: usize, comptime Q: type) type {
 
         pub fn negate(self: Self) Self {
             var res: Self = undefined;
-            for (self.data, 0..) |v, i| {
+            inline for (self.data, 0..) |v, i| {
                 res.data[i] = -v;
             }
             return res;
         }
 
-        pub fn scale(self: Self, rhs: T) Self {
+        pub inline fn to(self: Self, comptime DestQ: type) Vector(len, DestQ) {
-            var res: Self = undefined;
-            for (self.data, 0..) |v, i| {
-                res.data[i] = (Q{ .value = v }).scale(rhs).value;
-            }
-            return res;
-        }
-
-        pub fn to(self: Self, comptime DestQ: type) Vector(len, DestQ) {
             var res: Vector(len, DestQ) = undefined;
-            for (self.data, 0..) |v, i| {
+            inline for (self.data, 0..) |v, i| {
                 res.data[i] = (Q{ .value = v }).to(DestQ).value;
             }
             return res;
         }
 
-        pub fn lengthSqr(self: Self) T {
+        pub inline fn lengthSqr(self: Self) T {
             var sum: T = 0;
-            for (self.data) |v| {
+            inline for (self.data) |v| {
                 sum += v * v;
             }
             return sum;
         }
 
-        pub fn length(self: Self) T {
+        pub inline fn length(self: Self) T {
             const len_sq = self.lengthSqr();
 
             if (comptime @typeInfo(T) == .int) {
@@ -147,21 +139,43 @@ pub fn Vector(comptime len: usize, comptime Q: type) type {
             }
         }
 
-        pub fn format(self: Self, writer: *std.Io.Writer) !void {
+        pub fn formatNumber(
+            self: Self,
+            writer: *std.Io.Writer,
+            options: std.fmt.Number,
+        ) !void {
             try writer.writeAll("(");
             for (self.data, 0..) |v, i| {
                 if (i > 0) try writer.writeAll(", ");
-                try writer.print("{d:.2}", .{v});
+                switch (@typeInfo(T)) {
+                    .float, .comptime_float => try writer.printFloat(v, options),
+                    .int, .comptime_int => try writer.printInt(v, 10, .lower, .{
+                        .width = options.width,
+                        .alignment = options.alignment,
+                        .fill = options.fill,
+                        .precision = options.precision,
+                    }),
+                    else => unreachable,
+                }
             }
             try writer.writeAll(")");
             var first = true;
             inline for (std.enums.values(Dimension)) |bu| {
                 const v = dims.get(bu);
-                if (v == 0) continue;
+                if (comptime v == 0) continue;
-                if (!first) try writer.writeAll(".");
+                if (!first)
+                    try writer.writeAll(".");
+
                 first = false;
-                try writer.print("{s}{s}", .{ scales.get(bu).str(), bu.unit() });
+
-                if (v != 1) try hlp.printSuperscript(writer, v);
+                const uscale = scales.get(bu);
+                if (bu == .T and (uscale == .min or uscale == .hour or uscale == .year))
+                    try writer.print("{s}", .{uscale.str()})
+                else
+                    try writer.print("{s}{s}", .{ uscale.str(), bu.unit() });
+
+                if (v != 1)
+                    try hlp.printSuperscript(writer, v);
             }
         }
     };
@@ -182,8 +196,12 @@ test "Format VectorX" {
     const accel = MeterPerSecondSq.Vec3.initDefault(9.81);
     const momentum = KgMeterPerSecond.Vec3{ .data = .{ 43, 0, 11 } };
 
-    std.debug.print("Acceleration: {f}\n", .{accel});
+    var buf: [64]u8 = undefined;
-    std.debug.print("Momentum: {f}\n", .{momentum});
+    var res = try std.fmt.bufPrint(&buf, "{d}", .{accel});
+    try std.testing.expectEqualStrings("(9.81, 9.81, 9.81)m.ns⁻²", res);
+
+    res = try std.fmt.bufPrint(&buf, "{d:.2}", .{momentum});
+    try std.testing.expectEqualStrings("(43.00, 0.00, 11.00)m.kg.s⁻¹", res);
 }
 
 test "VecX Init and Basic Arithmetic" {
@@ -265,12 +283,6 @@ test "VecX Element-wise Math and Scaling" {
     try std.testing.expectEqual(4, div.data[1]);
     try std.testing.expectEqual(3, div.data[2]);
     try std.testing.expectEqual(0, @TypeOf(div).dims.get(.L)); // M / M = Dimensionless
-
-    // Scale by primitive
-    const scaled = v1.scale(2);
-    try std.testing.expectEqual(20, scaled.data[0]);
-    try std.testing.expectEqual(40, scaled.data[1]);
-    try std.testing.expectEqual(60, scaled.data[2]);
 }
 
 test "VecX Conversions" {
@@ -304,124 +316,3 @@ test "VecX Length" {
     try std.testing.expectApproxEqAbs(@as(f32, 25.0), v_float.lengthSqr(), 1e-4);
     try std.testing.expectApproxEqAbs(@as(f32, 5.0), v_float.length(), 1e-4);
 }
-
-test "Benchmark Vector ops" {
-    const Io = std.Io;
-    const ITERS: usize = 10_000;
-    const SAMPLES: usize = 10;
-    var gsink: f64 = 0;
-
-    // In Zig 0.14+, we use the testing IO for clock access in tests
-    const io = std.testing.io;
-
-    const getTime = struct {
-        fn f(i: Io) Io.Timestamp {
-            return Io.Clock.awake.now(i);
-        }
-    }.f;
-
-    const getVal = struct {
-        fn f(comptime TT: type, i: usize, comptime mask: u7) TT {
-            const v: u8 = @as(u8, @truncate(i & @as(usize, mask))) + 1;
-            return if (comptime @typeInfo(TT) == .float) @floatFromInt(v) else @intCast(v);
-        }
-    }.f;
-
-    const fold = struct {
-        fn f(comptime TT: type, s: *f64, v: TT) void {
-            s.* += if (comptime @typeInfo(TT) == .float)
-                @as(f64, @floatCast(v))
-            else
-                @as(f64, @floatFromInt(v));
-        }
-    }.f;
-
-    const computeStats = struct {
-        fn f(samples: []f64, iters: usize) f64 {
-            std.mem.sort(f64, samples, {}, std.sort.asc(f64));
-            const mid = samples.len / 2;
-            const median_ns = if (samples.len % 2 == 0)
-                (samples[mid - 1] + samples[mid]) / 2.0
-            else
-                samples[mid];
-            return median_ns / @as(f64, @floatFromInt(iters));
-        }
-    }.f;
-
-    std.debug.print(
-        \\
-        \\ Vector<N, T> benchmark — {d} iterations, {d} samples/cell
-        \\ (Results in ns/op)
-        \\
-        \\┌─────────────┬──────┬─────────┬─────────┬─────────┐
-        \\│ Operation   │ Type │   Len=3 │   Len=4 │  Len=16 │
-        \\├─────────────┼──────┼─────────┼─────────┼─────────┤
-        \\
-    , .{ ITERS, SAMPLES });
-
-    const Types = .{ i32, i64, i128, f32, f64 };
-    const TNames = .{ "i32", "i64", "i128", "f32", "f64" };
-    const Lengths = .{ 3, 4, 16 };
-    const Ops = .{ "add", "scale", "mulByScalar", "length" };
-
-    inline for (Ops, 0..) |op_name, o_idx| {
-        inline for (Types, TNames) |T, tname| {
-            std.debug.print("│ {s:<11} │ {s:<4} │", .{ op_name, tname });
-
-            inline for (Lengths) |len| {
-                const Q_base = Scalar(T, Dimensions.init(.{ .L = 1 }), Scales.init(.{}));
-                const Q_time = Scalar(T, Dimensions.init(.{ .T = 1 }), Scales.init(.{}));
-                const V = Vector(len, Q_base);
-
-                var samples: [SAMPLES]f64 = undefined;
-
-                for (0..SAMPLES) |s_idx| {
-                    var sink: T = 0;
-
-                    const t_start = getTime(io);
-                    for (0..ITERS) |i| {
-                        const v1 = V.initDefault(getVal(T, i, 63));
-
-                        if (comptime std.mem.eql(u8, op_name, "add")) {
-                            const v2 = V.initDefault(getVal(T, i +% 7, 63));
-                            const res = v1.add(v2);
-                            for (res.data) |val| {
-                                if (comptime @typeInfo(T) == .float) sink += val else sink ^= val;
-                            }
-                        } else if (comptime std.mem.eql(u8, op_name, "scale")) {
-                            const sc = getVal(T, i +% 2, 63);
-                            const res = v1.scale(sc);
-                            for (res.data) |val| {
-                                if (comptime @typeInfo(T) == .float) sink += val else sink ^= val;
-                            }
-                        } else if (comptime std.mem.eql(u8, op_name, "mulByScalar")) {
-                            const s_val = Q_time{ .value = getVal(T, i +% 2, 63) };
-                            const res = v1.mulByScalar(s_val);
-                            for (res.data) |val| {
-                                if (comptime @typeInfo(T) == .float) sink += val else sink ^= val;
-                            }
-                        } else if (comptime std.mem.eql(u8, op_name, "length")) {
-                            const r = v1.length();
-                            if (comptime @typeInfo(T) == .float) sink += r else sink ^= r;
-                        }
-                    }
-                    const t_end = getTime(io);
-
-                    samples[s_idx] = @as(f64, @floatFromInt(t_start.durationTo(t_end).toNanoseconds()));
-                    fold(T, &gsink, sink);
-                }
-
-                const median_ns_per_op = computeStats(&samples, ITERS);
-                std.debug.print(" {d:>7.1} │", .{median_ns_per_op});
-            }
-            std.debug.print("\n", .{});
-        }
-
-        if (o_idx < Ops.len - 1) {
-            std.debug.print("├─────────────┼──────┼─────────┼─────────┼─────────┤\n", .{});
-        }
-    }
-    std.debug.print("└─────────────┴──────┴─────────┴─────────┴─────────┘\n", .{});
-    std.debug.print("\nAnti-optimisation sink: {d:.4}\n", .{gsink});
-    try std.testing.expect(gsink != 0);
-}

src/benchmark.zig

@@ -0,0 +1,429 @@
+const std = @import("std");
+const Io = std.Io;
+const Scalar = @import("Scalar.zig").Scalar;
+const Vector = @import("Vector.zig").Vector;
+
+var io: Io = undefined;
+pub fn main(init: std.process.Init) !void {
+    var stdout_buf: [4096]u8 = undefined;
+    var stdout_writer: std.Io.File.Writer = .init(.stdout(), init.io, &stdout_buf);
+    try stdout_writer.interface.print("Starting Benchmarks...", .{});
+
+    io = init.io;
+
+    try bench_Scalar(&stdout_writer.interface);
+    try stdout_writer.flush();
+    try bench_vsNative(&stdout_writer.interface);
+    try stdout_writer.flush();
+    try bench_crossTypeVsNative(&stdout_writer.interface);
+    try stdout_writer.flush();
+    try bench_Vector(&stdout_writer.interface);
+    try stdout_writer.flush();
+}
+
+fn getTime() Io.Timestamp {
+    return Io.Clock.awake.now(io);
+}
+
+fn fold(comptime TT: type, s: *f64, v: TT) void {
+    s.* += if (comptime @typeInfo(TT) == .float)
+        @as(f64, @floatCast(v))
+    else
+        @as(f64, @floatFromInt(v));
+}
+
+fn bench_Scalar(writer: *std.Io.Writer) !void {
+    const ITERS: usize = 100_000;
+    const SAMPLES: usize = 10;
+
+    const getVal = struct {
+        fn f(comptime TT: type, i: usize, comptime mask: u7) TT {
+            const v: u8 = @as(u8, @truncate(i & @as(usize, mask))) + 1;
+            return if (comptime @typeInfo(TT) == .float) @floatFromInt(v) else @intCast(v);
+        }
+    }.f;
+
+    const Stats = struct {
+        median: f64,
+        delta: f64,
+        ops_per_sec: f64,
+    };
+
+    const computeStats = struct {
+        fn f(samples: []f64, iters: usize) Stats {
+            std.mem.sort(f64, samples, {}, std.sort.asc(f64));
+            const mid = samples.len / 2;
+            const median_ns = if (samples.len % 2 == 0) (samples[mid - 1] + samples[mid]) / 2.0 else samples[mid];
+
+            const low = samples[0];
+            const high = samples[samples.len - 1];
+            const delta_ns = (high - low) / 2.0;
+
+            const ns_per_op = median_ns / @as(f64, @floatFromInt(iters));
+            return .{
+                .median = ns_per_op,
+                .delta = (delta_ns / @as(f64, @floatFromInt(iters))),
+                .ops_per_sec = 1_000_000_000.0 / ns_per_op,
+            };
+        }
+    }.f;
+
+    try writer.print(
+        \\
+        \\ Scalar<T> benchmark — {d} iterations, {d} samples/cell
+        \\
+        \\┌───────────────────┬──────┬─────────────────────┬─────────────────────┐
+        \\│ Operation         │ Type │ ns / op (± delta)   │ Throughput (ops/s)  │
+        \\├───────────────────┼──────┼─────────────────────┼─────────────────────┤
+        \\
+    , .{ ITERS, SAMPLES });
+
+    const Types = .{ i16, i32, i64, i128, i256, f32, f64, f128 };
+    const TNames = .{ "i16", "i32", "i64", "i128", "i256", "f32", "f64", "f128" };
+    const Ops = .{ "add", "sub", "mulBy", "divBy", "to" };
+
+    var results_matrix: [Ops.len][Types.len]f64 = undefined;
+
+    comptime var tidx: usize = 0;
+    inline for (Types, TNames) |T, tname| {
+        const M = Scalar(T, .init(.{ .L = 1 }), .init(.{}));
+        const KM = Scalar(T, .init(.{ .L = 1 }), .init(.{ .L = .k }));
+        const S = Scalar(T, .init(.{ .T = 1 }), .init(.{}));
+
+        inline for (Ops, 0..) |op_name, oidx| {
+            var samples: [SAMPLES]f64 = undefined;
+
+            for (0..SAMPLES) |s_idx| {
+                const t_start = getTime();
+
+                for (0..ITERS) |i| {
+                    std.mem.doNotOptimizeAway(
+                        {
+                            _ = if (comptime std.mem.eql(u8, op_name, "add"))
+                                (M{ .value = getVal(T, i, 63) }).add(M{ .value = getVal(T, i +% 7, 63) })
+                            else if (comptime std.mem.eql(u8, op_name, "sub"))
+                                (M{ .value = getVal(T, i +% 10, 63) }).sub(M{ .value = getVal(T, i, 63) })
+                            else if (comptime std.mem.eql(u8, op_name, "mulBy"))
+                                (M{ .value = getVal(T, i, 63) }).mulBy(M{ .value = getVal(T, i +% 1, 63) })
+                            else if (comptime std.mem.eql(u8, op_name, "divBy"))
+                                (M{ .value = getVal(T, i +% 10, 63) }).divBy(S{ .value = getVal(T, i, 63) })
+                            else
+                                (KM{ .value = getVal(T, i, 15) }).to(M);
+                        },
+                    );
+                }
+
+                const t_end = getTime();
+                samples[s_idx] = @as(f64, @floatFromInt(t_start.durationTo(t_end).toNanoseconds()));
+            }
+
+            const stats = computeStats(&samples, ITERS);
+            results_matrix[oidx][tidx] = stats.median;
+
+            try writer.print("│ {s:<17} │ {s:<4} │ {d:>8.2} ns ±{d:<6.2} │ {d:>19.0} │\n", .{ op_name, tname, stats.median, stats.delta, stats.ops_per_sec });
+        }
+
+        if (comptime tidx < Types.len - 1) {
+            try writer.print("├───────────────────┼──────┼─────────────────────┼─────────────────────┤\n", .{});
+        }
+        tidx += 1;
+    }
+
+    // Median Summary Table
+    try writer.print("└───────────────────┴──────┴─────────────────────┴─────────────────────┘\n\n", .{});
+    try writer.print("Median Summary (ns/op):\n", .{});
+
+    try writer.print("┌──────────────┬───────┬───────┬───────┬───────┬───────┬───────┬───────┬───────┐\n", .{});
+    try writer.print("│  Operation   │  i16  │  i32  │  i64  │  i128 │  i256 │  f32  │  f64  │  f128 │\n", .{});
+    try writer.print("├──────────────┼───────┼───────┼───────┼───────┼───────┼───────┼───────┼───────┤\n", .{});
+
+    inline for (Ops, 0..) |op_name, oidx| {
+        try writer.print("│  {s:<11} │", .{op_name});
+        var i: usize = 0;
+        while (i < Types.len) : (i += 1)
+            try writer.print("{d:>6.1} │", .{results_matrix[oidx][i]});
+
+        try writer.print("\n", .{});
+    }
+
+    try writer.print("└──────────────┴───────┴───────┴───────┴───────┴───────┴───────┴───────┴───────┘\n", .{});
+}
+
+fn bench_vsNative(writer: *std.Io.Writer) !void {
+    const ITERS: usize = 100_000;
+    const SAMPLES: usize = 5;
+
+    // Helper to safely get a value of type T from a loop index
+    const getValT = struct {
+        fn f(comptime TT: type, i: usize) TT {
+            const v = (i % 100) + 1;
+            return if (comptime @typeInfo(TT) == .float) @floatFromInt(v) else @intCast(v);
+        }
+    }.f;
+
+    const Types = .{ i32, i64, i128, f32, f64 };
+    const TNames = .{ "i32", "i64", "i128", "f32", "f64" };
+    const Ops = .{ "add", "mulBy", "divBy" };
+
+    try writer.print(
+        \\
+        \\ Scalar vs Native Overhead Analysis
+        \\
+        \\┌───────────┬──────┬───────────┬───────────┬───────────┐
+        \\│ Operation │ Type │ Native    │ Scalar    │ Slowdown  │
+        \\├───────────┼──────┼───────────┼───────────┼───────────┤
+        \\
+    , .{});
+
+    inline for (Ops, 0..) |op_name, j| {
+        inline for (Types, 0..) |T, tidx| {
+            var native_total_ns: f64 = 0;
+            var quantity_total_ns: f64 = 0;
+
+            const M = Scalar(T, .init(.{ .L = 1 }), .init(.{}));
+            const S = Scalar(T, .init(.{ .T = 1 }), .init(.{}));
+
+            std.mem.doNotOptimizeAway({
+                for (0..SAMPLES) |_| {
+                    // --- 1. Benchmark Native ---
+                    const n_start = getTime();
+                    for (0..ITERS) |i| {
+                        const a = getValT(T, i);
+                        const b = getValT(T, 2);
+                        _ = if (comptime std.mem.eql(u8, op_name, "add"))
+                            a + b
+                        else if (comptime std.mem.eql(u8, op_name, "mulBy"))
+                            a * b
+                        else if (comptime @typeInfo(T) == .int) @divTrunc(a, b) else a / b;
+                    }
+                    const n_end = getTime();
+                    native_total_ns += @as(f64, @floatFromInt(n_start.durationTo(n_end).toNanoseconds()));
+
+                    // --- 2. Benchmark Scalar ---
+                    const q_start = getTime();
+                    for (0..ITERS) |i| {
+                        const qa = M{ .value = getValT(T, i) };
+                        const qb = if (comptime std.mem.eql(u8, op_name, "divBy")) S{ .value = getValT(T, 2) } else M{ .value = getValT(T, 2) };
+
+                        _ = if (comptime std.mem.eql(u8, op_name, "add"))
+                            qa.add(qb)
+                        else if (comptime std.mem.eql(u8, op_name, "mulBy"))
+                            qa.mulBy(qb)
+                        else
+                            qa.divBy(qb);
+                    }
+                    const q_end = getTime();
+                    quantity_total_ns += @as(f64, @floatFromInt(q_start.durationTo(q_end).toNanoseconds()));
+                }
+            });
+
+            const avg_n = (native_total_ns / SAMPLES) / @as(f64, @floatFromInt(ITERS));
+            const avg_q = (quantity_total_ns / SAMPLES) / @as(f64, @floatFromInt(ITERS));
+            const slowdown = avg_q / avg_n;
+
+            try writer.print("│ {s:<9} │ {s:<4} │ {d:>7.2}ns │ {d:>7.2}ns │ {d:>8.2}x │\n", .{
+                op_name, TNames[tidx], avg_n, avg_q, slowdown,
+            });
+        }
+        if (j != Ops.len - 1) try writer.print("├───────────┼──────┼───────────┼───────────┼───────────┤\n", .{});
+    }
+
+    try writer.print("└───────────┴──────┴───────────┴───────────┴───────────┘\n", .{});
+}
+
+fn bench_crossTypeVsNative(writer: *std.Io.Writer) !void {
+    const ITERS: usize = 100_000;
+    const SAMPLES: usize = 5;
+
+    const getValT = struct {
+        fn f(comptime TT: type, i: usize) TT {
+            // Keep values safe and non-zero to avoid division by zero or overflows during cross-casting
+            const v = (i % 50) + 1;
+            return if (comptime @typeInfo(TT) == .float) @floatFromInt(v) else @intCast(v);
+        }
+    }.f;
+
+    // Helper for the Native baseline: explicitly casting T2 to T1 before the operation
+    const castTo = struct {
+        fn f(comptime DestT: type, comptime SrcT: type, val: SrcT) DestT {
+            if (comptime DestT == SrcT) return val;
+            const src_info = @typeInfo(SrcT);
+            const dest_info = @typeInfo(DestT);
+
+            if (dest_info == .int and src_info == .int) return @intCast(val);
+            if (dest_info == .float and src_info == .int) return @floatFromInt(val);
+            if (dest_info == .int and src_info == .float) return @intFromFloat(val);
+            if (dest_info == .float and src_info == .float) return @floatCast(val);
+            unreachable;
+        }
+    }.f;
+
+    const Types = .{ i16, i64, i128, f32, f64 };
+    const TNames = .{ "i16", "i64", "i128", "f32", "f64" };
+    const Ops = .{ "add", "mulBy", "divBy" };
+
+    try writer.print(
+        \\
+        \\ Cross-Type Overhead Analysis: Scalar vs Native
+        \\
+        \\┌─────────┬──────┬──────┬───────────┬───────────┬───────────┐
+        \\│ Op      │ T1   │ T2   │ Native    │ Scalar    │ Slowdown  │
+        \\├─────────┼──────┼──────┼───────────┼───────────┼───────────┤
+        \\
+    , .{});
+
+    inline for (Ops, 0..) |op_name, j| {
+        inline for (Types, 0..) |T1, t1_idx| {
+            inline for (Types, 0..) |T2, t2_idx| {
+                var native_total_ns: f64 = 0;
+                var quantity_total_ns: f64 = 0;
+
+                const M1 = Scalar(T1, .init(.{ .L = 1 }), .init(.{}));
+                const M2 = Scalar(T2, .init(.{ .L = 1 }), .init(.{}));
+                const S2 = Scalar(T2, .init(.{ .T = 1 }), .init(.{}));
+
+                std.mem.doNotOptimizeAway({
+                    for (0..SAMPLES) |_| {
+                        // --- 1. Benchmark Native (Cast T2 to T1, then math) ---
+                        const n_start = getTime();
+                        for (0..ITERS) |i| {
+                            const a = getValT(T1, i);
+                            const b_raw = getValT(T2, 2);
+                            const b = castTo(T1, T2, b_raw);
+
+                            _ = if (comptime std.mem.eql(u8, op_name, "add"))
+                                a + b
+                            else if (comptime std.mem.eql(u8, op_name, "mulBy"))
+                                a * b
+                            else if (comptime @typeInfo(T1) == .int)
+                                @divTrunc(a, b)
+                            else
+                                a / b;
+                        }
+                        const n_end = getTime();
+                        native_total_ns += @as(f64, @floatFromInt(n_start.durationTo(n_end).toNanoseconds()));
+
+                        // --- 2. Benchmark Scalar ---
+                        const q_start = getTime();
+                        for (0..ITERS) |i| {
+                            const qa = M1{ .value = getValT(T1, i) };
+                            const qb = if (comptime std.mem.eql(u8, op_name, "divBy"))
+                                S2{ .value = getValT(T2, 2) }
+                            else
+                                M2{ .value = getValT(T2, 2) };
+
+                            _ = if (comptime std.mem.eql(u8, op_name, "add"))
+                                qa.add(qb)
+                            else if (comptime std.mem.eql(u8, op_name, "mulBy"))
+                                qa.mulBy(qb)
+                            else
+                                qa.divBy(qb);
+                        }
+                        const q_end = getTime();
+                        quantity_total_ns += @as(f64, @floatFromInt(q_start.durationTo(q_end).toNanoseconds()));
+                    }
+
+                    const avg_n = (native_total_ns / SAMPLES) / @as(f64, @floatFromInt(ITERS));
+                    const avg_q = (quantity_total_ns / SAMPLES) / @as(f64, @floatFromInt(ITERS));
+                    const slowdown = avg_q / avg_n;
+
+                    try writer.print("│ {s:<7} │ {s:<4} │ {s:<4} │ {d:>7.2}ns │ {d:>7.2}ns │ {d:>8.2}x │\n", .{
+                        op_name, TNames[t1_idx], TNames[t2_idx], avg_n, avg_q, slowdown,
+                    });
+                });
+            }
+        }
+        if (j != Ops.len - 1) {
+            try writer.print("├─────────┼──────┼──────┼───────────┼───────────┼───────────┤\n", .{});
+        }
+    }
+
+    try writer.print("└─────────┴──────┴──────┴───────────┴───────────┴───────────┘\n", .{});
+}
+
+fn bench_Vector(writer: *std.Io.Writer) !void {
+    const ITERS: usize = 10_000;
+    const SAMPLES: usize = 10;
+
+    const getVal = struct {
+        fn f(comptime TT: type, i: usize, comptime mask: u7) TT {
+            const v: u8 = @as(u8, @truncate(i & @as(usize, mask))) + 1;
+            return if (comptime @typeInfo(TT) == .float) @floatFromInt(v) else @intCast(v);
+        }
+    }.f;
+
+    const computeStats = struct {
+        fn f(samples: []f64, iters: usize) f64 {
+            std.mem.sort(f64, samples, {}, std.sort.asc(f64));
+            const mid = samples.len / 2;
+            const median_ns = if (samples.len % 2 == 0)
+                (samples[mid - 1] + samples[mid]) / 2.0
+            else
+                samples[mid];
+            return median_ns / @as(f64, @floatFromInt(iters));
+        }
+    }.f;
+
+    try writer.print(
+        \\
+        \\ Vector<N, T> benchmark — {d} iterations, {d} samples/cell
+        \\ (Results in ns/op)
+        \\
+        \\┌─────────────┬──────┬─────────┬─────────┬─────────┐
+        \\│ Operation   │ Type │   Len=3 │   Len=4 │  Len=16 │
+        \\├─────────────┼──────┼─────────┼─────────┼─────────┤
+        \\
+    , .{ ITERS, SAMPLES });
+
+    const Types = .{ i32, i64, i128, f32, f64 };
+    const TNames = .{ "i32", "i64", "i128", "f32", "f64" };
+    const Lengths = .{ 3, 4, 16 };
+    const Ops = .{ "add", "divBy", "mulByScalar", "length" };
+
+    inline for (Ops, 0..) |op_name, o_idx| {
+        inline for (Types, TNames) |T, tname| {
+            try writer.print("│ {s:<11} │ {s:<4} │", .{ op_name, tname });
+
+            inline for (Lengths) |len| {
+                const Q_base = Scalar(T, .init(.{ .L = 1 }), .init(.{}));
+                const Q_time = Scalar(T, .init(.{ .T = 1 }), .init(.{}));
+                const V = Vector(len, Q_base);
+
+                var samples: [SAMPLES]f64 = undefined;
+
+                std.mem.doNotOptimizeAway({
+                    for (0..SAMPLES) |s_idx| {
+                        const t_start = getTime();
+                        for (0..ITERS) |i| {
+                            const v1 = V.initDefault(getVal(T, i, 63));
+
+                            if (comptime std.mem.eql(u8, op_name, "add")) {
+                                const v2 = V.initDefault(getVal(T, i +% 7, 63));
+                                _ = v1.add(v2);
+                            } else if (comptime std.mem.eql(u8, op_name, "divBy")) {
+                                _ = v1.divBy(V.initDefault(getVal(T, i +% 2, 63)));
+                            } else if (comptime std.mem.eql(u8, op_name, "mulByScalar")) {
+                                const s_val = Q_time{ .value = getVal(T, i +% 2, 63) };
+                                _ = v1.mulByScalar(s_val);
+                            } else if (comptime std.mem.eql(u8, op_name, "length")) {
+                                _ = v1.length();
+                            }
+                        }
+                        const t_end = getTime();
+
+                        samples[s_idx] = @as(f64, @floatFromInt(t_start.durationTo(t_end).toNanoseconds()));
+                    }
+
+                    const median_ns_per_op = computeStats(&samples, ITERS);
+                    try writer.print(" {d:>7.1} │", .{median_ns_per_op});
+                });
+            }
+            try writer.print("\n", .{});
+        }
+
+        if (o_idx < Ops.len - 1) {
+            try writer.print("├─────────────┼──────┼─────────┼─────────┼─────────┤\n", .{});
+        }
+    }
+    try writer.print("└─────────────┴──────┴─────────┴─────────┴─────────┘\n", .{});
+}