dartarrow

A PyArrow-style API for Apache Arrow in pure Dart.

If you know pyarrow, you already know this: pa.array, pa.schema, pa.table, ipc.openStream. Same object model, Dart spelling. No native library, no runtime dependencies beyond a pure Dart compression package.

Guide · API reference · PyArrow parity · Changelog

Install

dependencies:
  dartarrow:
    git:
      url: https://github.com/richardrh/dartarrow.git
      ref: main # pin a full commit SHA for reproducible builds

Not yet published to pub.dev.

Usage

The API looks like PyArrow but in Dart.

import 'package:dartarrow/arrow.dart' as pa;
import 'package:dartarrow/ipc.dart' as ipc;

void main() {
  final values = pa.array([10, null, 30], type: pa.int32());
  print(values[1].isValid);            // false: a typed int32 null Scalar
  print(values.slice(1, 2).toDartList()); // [null, 30]

  final table = pa.table({
    'id': pa.chunkedArray([values.slice(0, 1), values.slice(1)]),
  });

  // Arrow IPC round trip (format V5, in memory)
  final bytes = (ipc.newStream(table.schema)..writeTable(table)).close();
  print(ipc.openStream(bytes).readAll().numRows); // 3
}

More in example/.

Features

Not included: compute kernels, datasets, Parquet/CSV.

Verified against PyArrow 26.0.0 fixtures, including IPC interop in both directions.

Dart-specific behaviour

• Wide integers use BigInt; decimals keep an unscaled BigInt and scale, never a double.
• Array does not implement Iterable (its cast<R>() clashes with Arrow's cast(type)). Use values.scalars or values.toDartList().
• Array.fromBuffers borrows memory. Keep buffers alive and unmutated while arrays or slices exist.

See CONTRIBUTING.md for the PyArrow conformance workflow.

License

Apache-2.0