zig
zig
Zig ecosystem for systems programming without hidden control flow.
Atomic Skills
| Skill | Commands | Domain |
|---|---|---|
| zig build | build system | Compile, link, cross-compile |
| zig test | testing | Run test blocks |
| zig fmt | formatter | Canonical formatting |
| zls | LSP | Autocomplete, diagnostics |
Quick Start
# New project
mkdir myproject && cd myproject
zig init
# Build and run
zig build run
# Test
zig build test
# Format
zig fmt src/
# Cross-compile to WASM
zig build -Dtarget=wasm32-freestanding
build.zig (0.15.2)
const std = @import("std");
pub fn build(b: *std.Build) void {
const target = b.standardTargetOptions(.{});
const optimize = b.standardOptimizeOption(.{});
const exe = b.addExecutable(.{
.name = "myapp",
.root_source_file = b.path("src/main.zig"),
.target = target,
.optimize = optimize,
});
b.installArtifact(exe);
const run_cmd = b.addRunArtifact(exe);
run_cmd.step.dependOn(b.getInstallStep());
const run_step = b.step("run", "Run the application");
run_step.dependOn(&run_cmd.step);
const tests = b.addTest(.{
.root_source_file = b.path("src/main.zig"),
.target = target,
.optimize = optimize,
});
const test_step = b.step("test", "Run unit tests");
test_step.dependOn(&b.addRunArtifact(tests).step);
}
build.zig Module Registration (0.15.2)
When registering many modules in build.zig, Zig 0.15.2 enforces unused-const rules:
// If a module has downstream imports, keep the named const:
const message_frame_mod = b.addModule("message_frame", .{
.root_source_file = b.path("src/message_frame.zig"),
});
// If a module has NO downstream imports, discard the return value directly:
_ = b.addModule("terminal", .{
.root_source_file = b.path("src/terminal.zig"),
});
// WRONG — "pointless discard of local constant" error:
// const terminal_mod = b.addModule(...);
// _ = terminal_mod;
// Dependency chaining:
const qrtp_frame_mod = b.addModule("qrtp_frame", .{
.root_source_file = b.path("src/qrtp_frame.zig"),
.imports = &.{ .{ .name = "fountain", .module = fountain_mod } },
});
Core Patterns
Explicit Allocators
const std = @import("std");
pub fn main() !void {
var gpa = std.heap.GeneralPurposeAllocator(.{}){};
defer _ = gpa.deinit();
const allocator = gpa.allocator();
var list = std.ArrayList(u8).init(allocator);
defer list.deinit();
try list.appendSlice("hello");
}
Error Handling
fn readFile(path: []const u8) ![]u8 {
const file = try std.fs.cwd().openFile(path, .{});
defer file.close();
return file.readToEndAlloc(allocator, 1024 * 1024);
}
// Usage with catch
const data = readFile("config.txt") catch |err| {
std.log.err("Failed: {}", .{err});
return err;
};
Comptime Metaprogramming
fn Vec(comptime T: type, comptime N: usize) type {
return struct {
data: [N]T,
const Self = @This();
pub fn dot(self: Self, other: Self) T {
var sum: T = 0;
inline for (0..N) |i| {
sum += self.data[i] * other.data[i];
}
return sum;
}
};
}
const Vec3 = Vec(f32, 3);
Defer/Errdefer
fn process() !void {
const resource = try acquire();
defer release(resource); // Always runs
const temp = try allocate();
errdefer free(temp); // Only on error
try doWork(resource, temp);
// temp ownership transferred, no errdefer needed
}
C Interop
const c = @cImport({
@cInclude("stdio.h");
@cInclude("mylib.h");
});
pub fn main() void {
_ = c.printf("Hello from C\n");
}
Emerging Patterns (zig-syrup)
Pattern 1: SplitMix64 Bijection (Gay.jl Integration)
Comptime modular multiplicative inverse via Newton's method. Enables invertible hashing
with Strong Parallelism Invariance (SPI): same (seed, index) gives the same result
regardless of call order or parallelism.
pub const SplitMix64 = struct {
pub const GOLDEN: u64 = 0x9e3779b97f4a7c15;
pub const MIX1: u64 = 0xbf58476d1ce4e5b9;
pub const MIX2: u64 = 0x94d049bb133111eb;
pub const MIX1_INV: u64 = modInverse64(MIX1);
pub const MIX2_INV: u64 = modInverse64(MIX2);
state: u64,
/// Forward bijection: deterministic, invertible.
pub fn mix(x: u64) u64 {
var z = x +% GOLDEN;
z = (z ^ (z >> 30)) *% MIX1;
z = (z ^ (z >> 27)) *% MIX2;
return z ^ (z >> 31);
}
/// Inverse: unmix(mix(x)) == x for all x.
pub fn unmix(z: u64) u64 {
var x = z;
x ^= x >> 31; x ^= x >> 62;
x *%= MIX2_INV;
x ^= x >> 27; x ^= x >> 54;
x *%= MIX1_INV;
x ^= x >> 30; x ^= x >> 60;
x -%= GOLDEN;
return x;
}
/// O(1) random access. SPI-compatible.
pub fn colorAt(seed: u64, index: u64) u64 {
return mix(seed ^ index);
}
/// Comptime modular inverse mod 2^64 via Newton's method.
fn modInverse64(a: u64) u64 {
@setEvalBranchQuota(10000);
var x: u64 = a;
x *%= 2 -% a *% x; // doubling correct bits each step
x *%= 2 -% a *% x;
x *%= 2 -% a *% x;
x *%= 2 -% a *% x;
x *%= 2 -% a *% x; // 64 bits converged
return x;
}
};
Key insight: modInverse64 runs entirely at comptime — MIX1_INV and MIX2_INV are compile-time constants. The @setEvalBranchQuota is necessary for the comptime evaluator.
Pattern 2: Three-Mode Tagged Union PRNG
Tagged union for GF(3)-aligned PRNG modes. Each mode has a distinct role:
pub const PrngMode = enum {
splitmix, // -1 (MINUS): bijective, invertible, SPI. Default.
xoshiro, // 0 (ERGODIC): fast, non-cryptographic.
chacha, // +1 (PLUS): CSPRNG for identity proofs.
};
pub const Prng = union(PrngMode) {
splitmix: SplitMix64,
xoshiro: std.Random.Xoshiro256,
chacha: std.Random.ChaCha,
pub fn init(prng_mode: PrngMode, seed: u64) Prng {
return switch (prng_mode) {
.splitmix => .{ .splitmix = SplitMix64.init(seed) },
.xoshiro => .{ .xoshiro = std.Random.Xoshiro256.init(seed) },
.chacha => initChaCha(seed),
};
}
pub fn next(self: *Prng) u64 {
return switch (self.*) {
.splitmix => |*s| s.next(),
.xoshiro => |*x| x.next(),
.chacha => |*c| c.random().int(u64),
};
}
// IMPORTANT: Don't name a method the same as the active tag field.
// Use `activeMode` instead of `mode` to avoid shadowing.
pub fn activeMode(self: *const Prng) PrngMode {
return self.*;
}
};
Gotcha: pub fn mode(self) would shadow the union's internal mode tag — renamed to activeMode.
Pattern 3: C ABI Callback Abstraction
Platform-agnostic callbacks for hardware interaction (QRTP QR code rendering):
/// Platform renders QR code from raw bytes. Returns 0 on success.
pub const RenderQRFn = *const fn (
data: [*]const u8,
data_len: usize,
context: ?*anyopaque,
) callconv(.c) c_int;
/// Platform scans QR code from camera. Returns decoded length, 0 if none.
pub const ScanQRFn = *const fn (
buf: [*]u8,
buf_len: usize,
context: ?*anyopaque,
) callconv(.c) usize;
pub const TransportConfig = struct {
render_qr: RenderQRFn,
scan_qr: ScanQRFn,
delay: DelayFn,
context: ?*anyopaque = null,
frame_delay_ms: u32 = 100,
};
Gotcha (0.15.2): Use callconv(.c) (lowercase). callconv(.C) is a compile error.
Pattern 4: SIMD XOR Block Combining
Fixed-size blocks enable SIMD vectorization without allocations:
pub fn xorBlocks(dst: []u8, src: []const u8) void {
const len = @min(dst.len, src.len);
const vec_len = 16;
const full_vecs = len / vec_len;
var i: usize = 0;
while (i < full_vecs * vec_len) : (i += vec_len) {
const d: @Vector(vec_len, u8) = dst[i..][0..vec_len].*;
const s: @Vector(vec_len, u8) = src[i..][0..vec_len].*;
dst[i..][0..vec_len].* = d ^ s;
}
// Scalar tail
while (i < len) : (i += 1) {
dst[i] ^= src[i];
}
}
Pattern 5: Sheaf-Theoretic Decoder (Bumpus StructuredDecompositions.jl)
Fountain decoder maps to adhesion_filter from Bumpus's tree decompositions:
- Source blocks = bags in the tree decomposition
- Encoded blocks = adhesion spans between bags
- XOR = pullback projection on the adhesion
- Belief propagation = sheaf consistency filtering
/// Fountain decoder with adhesion_filter alias.
pub const Decoder = struct {
// ... source blocks, state, pending buffer ...
/// Public alias: StructuredDecompositions.jl naming convention.
pub fn adhesionFilter(self: *Decoder) bool {
return self.propagate();
}
/// Sheaf consistency propagation:
/// When a bag (source block) is solved, check all adhesion spans
/// (pending encoded blocks) for newly degree-1 solvable entries.
fn propagate(self: *Decoder) bool {
var progress = true;
while (progress) {
progress = false;
// XOR out known blocks from pending, solve degree-1 entries
// ... (see fountain.zig for full implementation)
}
return progress;
}
};
Pattern 6: Compact Binary Framing (QRTP)
Fixed-layout binary serialization without allocators, fitting QR code capacity:
pub fn encodeFrame(block: *const fountain.EncodedBlock) QrtpFrame {
var frame = QrtpFrame{};
var pos: usize = 0;
@memcpy(frame.data[pos..][0..4], "qrtp"); // 4-byte tag
pos += 4;
frame.data[pos] = PROTOCOL_VERSION; // 1-byte version
pos += 1;
writeU64BE(frame.data[pos..], block.seed); // 8-byte big-endian
pos += 8;
// ... indices, payload ...
frame.len = pos;
return frame;
}
Zig 0.15.2 Gotchas
| Issue | Wrong | Right |
|---|---|---|
| C calling convention | callconv(.C) |
callconv(.c) |
| Unused module const | const m = b.addModule(...); _ = m; |
_ = b.addModule(...) |
| Method name = tag field | pub fn mode(self) on union(PrngMode) |
pub fn activeMode(self) |
| Hex literal | 0xGAY, 0xPASS |
Only [0-9a-fA-F] valid |
| Wrapping arithmetic | a + b (overflow trap) |
a +% b (wrapping) |
Version Detection
// Feature detection over version checks
const has_new_api = @hasDecl(std, "Build");
const T = if (has_new_api) std.Build else std.build.Builder;
Debug
std.debug.print("value: {any}\n", .{x});
std.log.info("structured: {}", .{data});
@breakpoint(); // Debugger trap
Cross-Compile Targets
# List all targets
zig targets | jq '.native'
# Common targets
zig build -Dtarget=x86_64-linux-gnu
zig build -Dtarget=aarch64-macos
zig build -Dtarget=wasm32-wasi
zig build -Dtarget=thumb-none-eabi # Embedded
Reference: zig-syrup Module Map
| Module | LOC | Trit | Role |
|---|---|---|---|
fountain.zig |
907 | +1 | Luby Transform encoder/decoder, 3-mode PRNG |
qrtp_frame.zig |
427 | 0 | Binary framing for QR/TCP transport |
qrtp_transport.zig |
403 | -1 | Send/recv via C ABI QR callbacks |
message_frame.zig |
— | 0 | TCP length-prefixed framing |
tcp_transport.zig |
— | -1 | TCP OCapN transport |
propagator.zig |
— | 0 | Scoped propagators (lattice cells) |
color.zig |
— | +1 | Gay.jl Okhsl color space |
Related Skills
| Skill | Trit | Role |
|---|---|---|
| zig-programming | -1 | 223 recipes, full docs |
| rama-gay-zig | -1 | Rama + Gay.jl + Zig interleave |
| structured-decomp | 0 | Bumpus tree decompositions |
| zig | -1 | Ecosystem wrapper + emerging patterns |
GF(3) Triads
zig(-1) ⊗ zls-integration(0) ⊗ c-interop(+1) = 0 ✓
zig(-1) ⊗ acsets(0) ⊗ gay-mcp(+1) = 0 ✓ [Schema coloring]
zig(-1) ⊗ babashka(0) ⊗ duckdb-ies(+1) = 0 ✓ [Build analytics]
zig(-1) ⊗ structured-decomp(0) ⊗ fountain(+1) = 0 ✓ [Sheaf decoder]
zig(-1) ⊗ qrtp-frame(0) ⊗ qrtp-transport(+1) = 0 ✓ [Air-gapped transport]
SDF Interleaving
This skill connects to Software Design for Flexibility (Hanson & Sussman, 2021):
Primary Chapter: 2. Domain-Specific Languages
Concepts: DSL, wrapper, pattern-directed, embedding
GF(3) Balanced Triad
zig (−) + SDF.Ch2 (−) + [balancer] (−) = 0
Skill Trit: -1 (MINUS - verification)
Secondary Chapters
- Ch4: Pattern Matching
- Ch6: Layering
- Ch7: Propagators (fountain decoder = belief propagation)
Connection Pattern
DSLs embed domain knowledge. This skill defines domain-specific operations.
Cat# Integration
This skill maps to Cat# = Comod(P) as a bicomodule:
Trit: -1 (MINUS/Validator)
Home: Prof
Poly Op: ⊗
Kan Role: Ran (right Kan extension)
Color: #3B82F6 (blue)
Why -1 (MINUS)?
Zig validates and constrains:
- No hidden allocations
- No hidden control flow
- No exceptions
- Explicit error handling
- Compile-time safety checks
The language itself is a validator — it refuses to compile unsafe patterns.
zig-syrup Cat# Morphisms
fountain.Encoder (+1, Lan_K) →[qrtp_frame (0, Adj)]→ qrtp_transport (-1, Ran_K)
Presheaves Prof Span
SplitMix64.mix (−1, Ran) ↔ SplitMix64.unmix (−1, Ran)
Bijection = isomorphism in Cat#, self-dual in Span
Philosophy
"Zig is not designed to make fancy high-level things. It's designed to make it easy to write correct low-level code."
- Explicit over implicit
- Compile-time over runtime
- No hidden control flow
- Allocator-aware by design
- C interop without FFI overhead
- SPI: same seed → same output regardless of execution order