Strale

A memory efficient compact, copy-on-write (COW) string type for Zig.
The type occupies 16 bytes on 64-bit platforms.

Features:

• Small String Optimization (SSO)
• Reference counting
• Zero-copy substring views
• Copy-on-write mutation
• Optional UTF-8 support
• Optional thread-safety support
• Support global allocator to further reduce heap memory usage

Installation

Add to your build.zig.zon:

    .dependencies = .{
        .strale = .{ 
            .url = "https://github.com/Arcelyth/strale/archive/refs/heads/main.tar.gz", 
            .hash = "..." 
        },
    },

Add to your build.zig:

    const strale = b.dependency("strale", .{
        .target = target,
        .optimize = optimize,
    });
    exe_module.addImport("strale", strale.module("strale"));

Example

const std = @import("std");
const strale = @import("strale");
const StraleBytes = strale.StraleBytes;
const StraleUtf8Atomic = strale.StraleUtf8Atomic;

pub fn main() !void {
    const alloc = std.heap.page_allocator;

    var a = try StraleBytes.initSlice(alloc, "hello");
    defer a.deinit();

    var b = a.clone();
    try b.push(alloc, '!');
    std.debug.print("a = {s}\n", .{a.slice()});
    std.debug.print("b = {s}\n", .{b.slice()});

    var long = try StraleBytes.initSlice(
        alloc,
        "abcdefghijklmnopqrstuvwxyz",
    );
    defer long.deinit();

    var sub = long.substr(10, 10);
    defer sub.deinit();

    std.debug.print("sub = {s}\n", .{sub.slice()});

    var utf8 = try StraleUtf8Atomic.initSlice(alloc, "你好,世界");
    defer utf8.deinit();

    defer utf8.deinit();

    var iter = utf8.split(",");

    while (iter.next()) |c| {
        std.debug.print("{s}\n", .{c});
    }
}

Benchmarks

Benchmark environment:

• OS: macOS 26.5.2 (Darwin 25.5.0)
• Architecture: arm64
• Zig: 0.16.0

Compact string operations

Operation	Bytes	Strale	Heap-Allocated []u8
Creation	0	2.77 ns	2.78 ns
Creation	11	3.31 ns	11.97 ns
Creation	12	3.57 ns	12.84 ns
Creation	22	16.38 ns	15.48 ns
Creation	23	17.99 ns	16.15 ns
Creation	24	15.64 ns	14.86 ns
Creation	25	14.12 ns	14.78 ns
Creation	50	15.55 ns	14.88 ns
Clone	0	1.27 ns	2.77 ns
Clone	11	1.33 ns	12.63 ns
Clone	12	1.33 ns	12.64 ns
Clone	22	1.95 ns	15.97 ns
Clone	23	1.93 ns	15.54 ns
Clone	24	1.92 ns	14.81 ns
Clone	25	1.96 ns	14.52 ns
Clone	50	1.97 ns	15.01 ns
Access	0	1.02 ns	0.80 ns
Access	11	1.01 ns	0.76 ns
Access	12	1.01 ns	0.77 ns
Access	22	1.10 ns	0.76 ns
Access	23	1.09 ns	0.76 ns
Access	24	1.10 ns	0.76 ns
Access	25	1.10 ns	0.76 ns
Access	50	1.09 ns	0.76 ns

Word indexing

The idea and input for this benchmark come from html5ever Tendril's benchmark.

Input	Size	Heap-Allocated []u8	Strale
English 1	64 KiB	235.41 us	161.17 us
English 1	1 MiB	3606.88 us	2365.52 us
English 2	64 KiB	238.81 us	163.89 us
English 2	1 MiB	3660.98 us	2329.87 us
Korean	64 KiB	115.07 us	76.25 us
Korean	1 MiB	1746.92 us	1104.08 us
Korean HTML	64 KiB	98.25 us	64.47 us
Korean HTML	1 MiB	1497.99 us	934.74 us

Run benchmarks by yourself:

zig build bench -- [iterations]

LICENSE

MIT License