Zig Coding

Zig Coding Conventions

This skill ensures all Zig code follows the conventions from Zig: Zero to Hero guide.

When to Use

• Writing new Zig code
• Reviewing Zig code changes
• Refactoring existing Zig code
• Designing new modules or types

Naming Conventions

Error Handling Patterns

Use Error Unions (!T) for Failures

fn loadConfig(path: []const u8) !Config {
    const file = try std.fs.openFileAbsolute(path, .{});
    defer file.close();
    // ...
}

Use Optionals (?T) for Valid Absence

fn findService(name: []const u8) ?*Service {
    return services.get(name);  // null is valid - service doesn't exist
}

Error Propagation

// Propagate with try
const config = try loadConfig(path);

// Provide fallback with catch
const port = parsePort(input) catch 7946;

// Handle optionals with orelse
const service = findService(name) orelse return error.ServiceNotFound;

Resource Management

Immediate Defer Pattern

fn processFile(path: []const u8) !void {
    const file = try std.fs.openFileAbsolute(path, .{});
    defer file.close();  // Immediately after acquisition

    const buffer = try allocator.alloc(u8, 4096);
    defer allocator.free(buffer);

    // Work with file and buffer...
}

Partial Failure Rollback with errdefer

fn initializeNode(allocator: Allocator) !*Node {
    const node = try allocator.create(Node);
    errdefer allocator.destroy(node);  // Cleanup if subsequent operations fail

    node.config = try loadConfig();
    errdefer node.config.deinit();

    node.network = try initNetwork();
    return node;
}

Nested Blocks in Loops

// WRONG: defers accumulate
for (items) |item| {
    const resource = try acquire(item);
    defer release(resource);  // All defers run at function end!
}

// CORRECT: nested block scopes the defer
for (items) |item| {
    {
        const resource = try acquire(item);
        defer release(resource);  // Runs at block end
        process(resource);
    }
}

Module Organization

Visibility Control

// Public API - exported
pub fn startNode(config: Config) !*Node { ... }

// Internal implementation - not exported
fn validateConfig(config: Config) !void { ... }

Explicit Re-exports

// In orchestrator/orchestrator.zig
pub const Scheduler = @import("scheduler.zig").Scheduler;
pub const HealthMonitor = @import("health.zig").HealthMonitor;
pub const StateManager = @import("state.zig").StateManager;

Interface Pattern (for Fizz simulation testing)

pub const Time = struct {
    nowFn: *const fn () i64,

    pub fn now(self: Time) i64 {
        return self.nowFn();
    }
};

// Production implementation
pub fn realTime() Time {
    return .{ .nowFn = struct {
        fn now() i64 {
            return std.time.timestamp();
        }
    }.now };
}

Anti-Patterns to Avoid

1. Optionals for errors - Use !T when absence indicates failure
2. Defer in loops - Use nested blocks to scope cleanup
3. Unmarked comptime - Always mark type parameters with comptime
4. Direct syscalls in core logic - Use injected interfaces for testability

Fizz-Specific Conventions

Allocator Naming

• gpa - General Purpose Allocator (explicit deinit required)
• arena - Arena Allocator (bulk-free at end)
• page_allocator - Direct page allocation

Interface Injection

All core logic must accept interfaces rather than using system calls directly:

pub const Orchestrator = struct {
    time: Time,
    network: Network,
    runtime: Runtime,

    pub fn init(time: Time, network: Network, runtime: Runtime) Orchestrator {
        return .{ .time = time, .network = network, .runtime = runtime };
    }
};

Review Checklist

• Naming follows conventions (PascalCase types, camelCase functions, snake_case variables)
• Error unions used for failures, optionals for valid absence
• Resources have immediate defer cleanup
• errdefer used for partial failure rollback
• No defer inside loops without nested blocks
• Core logic uses injected interfaces (no direct syscalls)
• Public API marked with pub
• File names use snake_case