---

name: zig-expert description: Write idiomatic Zig code following the Zen of Zig philosophy. Use for systems programming with Zig (.zig files), covering manual memory management with allocators, error unions and explicit error handling, compile-time programming (comptime), data-oriented design, and the new async/Io model. Applies functional programming parallels (Result types, ADTs, explicit effects) for C-free systems development.

Idiomatic Zig Programming

Expert guidance for writing idiomatic Zig code that embodies the Zen of Zig: explicit intent, no hidden control flow, and compile-time over runtime.

Zen of Zig (Core Philosophy)

These principles govern all idiomatic Zig code:

Principle	Implication
Communicate intent precisely	Explicit code; APIs make requirements obvious
Edge cases matter	No undefined behaviors glossed over
Favor reading over writing	Optimize for clarity and maintainability
One obvious way	Avoid multiple complex features for same task
Runtime crashes > bugs	Fail fast and loudly, never corrupt state silently
Compile errors > runtime crashes	Catch issues at compile-time when possible
Resource deallocation must succeed	Design APIs with allocation failure in mind
Memory is a resource	Manage memory as consciously as any other resource
No hidden control flow	No exceptions, no GC, no implicit allocations

FP Conceptual Parallels

Zig shares key concepts with functional programming:

FP Concept	Zig Equivalent
Result/Either type	Error union !T (either error or value)
Option/Maybe	Optional ?T (nullable type)
ADTs / Sum types	Tagged unions with union(enum)
Pattern matching	switch with exhaustive handling
Explicit effects	Allocator/Io parameters (dependency injection)
Immutability preference	const by default, var only when needed
Pure functions	Functions without hidden state or allocations

Workflow Decision Tree

1. Declaring a binding? → Use const unless mutation required
2. Function needs memory? → Accept Allocator parameter, never global alloc
3. Function can fail? → Return error union !T, use try to propagate
4. Handling an error? → Use catch with explicit handler or try to propagate
5. Need cleanup on exit? → Use defer immediately after acquisition
6. Cleanup only on error? → Use errdefer for conditional cleanup
7. Need generic code? → Use comptime type parameters
8. Compile-time known value? → Use comptime to evaluate at build time
9. Calling C code? → Use @cImport for seamless FFI
10. Need async I/O? → Pass Io interface, use io.async() and future.await()
11. Optimizing hot path? → Consider data-oriented design (SoA vs AoS)

Essential Patterns

Error Unions (Result Type Equivalent)

const FileError = error{ NotFound, PermissionDenied, InvalidPath };

fn readConfig(path: []const u8) FileError!Config {
    const file = std.fs.cwd().openFile(path, .{}) catch |err| {
        return switch (err) {
            error.FileNotFound => error.NotFound,
            error.AccessDenied => error.PermissionDenied,
            else => error.InvalidPath,
        };
    };
    defer file.close();
    // ... parse config
    return config;
}

// Propagate with try (like Rust's ?)
pub fn main() !void {
    const config = try readConfig("app.conf");
    // ...
}

// Handle explicitly with catch
pub fn mainSafe() void {
    const config = readConfig("app.conf") catch |err| {
        std.debug.print("Failed: {}\n", .{err});
        return;
    };
    // ...
}

Allocator Pattern (Explicit Effects)

const std = @import("std");

// Function signature communicates: "I need to allocate"
fn processData(allocator: std.mem.Allocator, input: []const u8) ![]u8 {
    var result = try allocator.alloc(u8, input.len * 2);
    errdefer allocator.free(result); // cleanup only on error path
    
    // ... process into result
    
    return result; // caller owns this memory
}

pub fn main() !void {
    var gpa = std.heap.GeneralPurposeAllocator(.{}){};
    defer _ = gpa.deinit();
    const allocator = gpa.allocator();
    
    const data = try processData(allocator, "input");
    defer allocator.free(data); // caller responsible for cleanup
}

Tagged Unions (ADTs / Sum Types)

const PaymentState = union(enum) {
    pending: void,
    processing: struct { transaction_id: []const u8 },
    completed: Receipt,
    failed: PaymentError,
    
    // Methods on the union
    pub fn describe(self: PaymentState) []const u8 {
        return switch (self) {
            .pending => "Waiting for payment",
            .processing => |p| p.transaction_id,
            .completed => |r| r.summary,
            .failed => |e| e.message,
        };
    }
};

// Exhaustive switch (compiler enforces all cases)
fn handlePayment(state: PaymentState) void {
    switch (state) {
        .pending => startProcessing(),
        .processing => |p| pollStatus(p.transaction_id),
        .completed => |receipt| sendConfirmation(receipt),
        .failed => |err| notifyFailure(err),
    }
}

Compile-Time Programming

// comptime function for generics
fn max(comptime T: type, a: T, b: T) T {
    return if (a > b) a else b;
}

// Compile-time computed constants
const LOOKUP_TABLE = blk: {
    var table: [256]u8 = undefined;
    for (&table, 0..) |*entry, i| {
        entry.* = @intCast((i * 7) % 256);
    }
    break :blk table;
};

// Generic container (like TypeScript generics)
fn ArrayList(comptime T: type) type {
    return struct {
        items: []T,
        allocator: std.mem.Allocator,
        
        const Self = @This();
        
        pub fn init(allocator: std.mem.Allocator) Self {
            return .{ .items = &[_]T{}, .allocator = allocator };
        }
        
        pub fn append(self: *Self, item: T) !void {
            // ...
        }
    };
}

Resource Management with defer

fn processFile(allocator: std.mem.Allocator, path: []const u8) !void {
    // Open file
    const file = try std.fs.cwd().openFile(path, .{});
    defer file.close(); // ALWAYS runs on scope exit
    
    // Allocate buffer
    const buffer = try allocator.alloc(u8, 4096);
    defer allocator.free(buffer); // cleanup guaranteed
    
    // errdefer for conditional cleanup
    var result = try allocator.alloc(u8, 1024);
    errdefer allocator.free(result); // only on error
    
    // If we reach here successfully, caller owns result
    // ...
}

Quick Reference

// Imports
const std = @import("std");

// Variables
const immutable: u32 = 42;        // prefer const
var mutable: u32 = 0;             // only when needed

// Optionals (?T) - like Option/Maybe
var maybe_value: ?u32 = null;
const unwrapped = maybe_value orelse 0;        // default value
const ptr = maybe_value orelse return error.Missing;  // early return

// Error unions (!T) - like Result/Either
fn canFail() !u32 { return error.SomeError; }
const value = try canFail();                   // propagate error
const safe = canFail() catch |err| handleError(err);  // catch error

// Slices (pointer + length, not null-terminated)
const slice: []const u8 = "hello";             // string literal is []const u8
const arr: [5]u8 = .{ 1, 2, 3, 4, 5 };
const sub = arr[1..3];                         // slice of array

// Iteration
for (slice, 0..) |byte, index| { }             // value and index
for (slice) |byte| { }                         // value only

// Switch (exhaustive, can capture)
switch (tagged_union) {
    .variant => |captured| doSomething(captured),
    else => {},  // or handle all cases
}

// Comptime
const SIZE = comptime blk: { break :blk 64; };
fn generic(comptime T: type, val: T) T { return val; }

Detailed References

• references/idioms.md - Data-oriented design, memory patterns, testing
• references/async-io.md - New async/Io model, futures, cancellation
• references/c-interop.md - C FFI, @cImport, ABI compatibility

Forbidden Patterns

❌ Never	✅ Instead
Global allocator / hidden malloc	Pass Allocator explicitly
Exceptions / panic for errors	Return error union !T
Null pointers without type	Use optional ?*T
Preprocessor macros	Use comptime and inline functions
C-style strings in Zig code	Use slices []const u8
Ignoring errors silently	Handle with catch or propagate with try
var when const works	Default to const, mutate only when necessary
Hidden control flow	Make all branches explicit
OOP inheritance hierarchies	Use composition and tagged unions