arrow-zig

Apache Arrow columnar format in Zig.

Requires Zig 0.16.0+.

Docs: https://notfilippo.github.io/arrow-zig/

Getting Started

From a Zig project:

zig fetch --save=arrow git+https://github.com/notfilippo/arrow-zig.git

Then add the dependency module to the executable or library in build.zig:

const std = @import("std");

pub fn build(b: *std.Build) void {
    const target = b.standardTargetOptions(.{});
    const optimize = b.standardOptimizeOption(.{});
    const arrow_dep = b.dependency("arrow", .{
        .target = target,
        .optimize = optimize,
    });

    const exe = b.addExecutable(.{
        .name = "my_app",
        .root_module = b.createModule(.{
            .root_source_file = b.path("src/main.zig"),
            .target = target,
            .optimize = optimize,
        }),
    });
    exe.root_module.addImport("arrow", arrow_dep.module("arrow"));

    b.installArtifact(exe);
}

Pass .single_threaded = true to b.dependency("arrow", .{ ... }) to disable atomic buffer refcounts for single-threaded use.

Example

const arrow = @import("arrow");
const std = @import("std");

pub fn main() !void {
    var gpa: std.heap.DebugAllocator(.{}) = .init;
    defer _ = gpa.deinit();
    const allocator = gpa.allocator();

    var b = arrow.builder.NumericBuilder(i32).init(allocator);
    defer b.deinit();

    try b.append(10);
    try b.appendNull();
    try b.append(30);

    const data = try b.finish();
    defer data.deinit();

    const values = try arrow.array.NumericArray(i32).fromData(data);
    std.debug.print("{}\n", .{values.value(2)});
}

C++ Parity Matrix

Area	Status	Notes
Data types	Complete	Core C++ type metadata, including dictionary, extension, union, run-end encoded, views, decimals, intervals
Array storage	Strong	ArrayData, slicing, ref counts, quick validate, full validateFull, logical null counts
Typed arrays	Strong	Thin views over storage for all implemented layouts, no generic dynamic Array facade
Builders	Good	Primitive, binary, nested, dictionary with existing dictionaries, no generic MakeBuilder
Schema and batches	Good	Schemas, fields, record batches, struct batch conversion
C Data Interface	Strong	Schema, array, record batch, and stream import/export
Compute, IPC, files	Missing	No compute kernels, IPC, Parquet, dataset, table, or chunked array layer yet

Contributing

Run the full local check before sending changes:

zig build ci

This runs license checks, generated docs, regular tests, single threaded tests, and nanoarrow interop tests in both thread modes.

Format and ABI details follow the Arrow Columnar Format and Arrow C Data Interface.