Skip to main content

Polars io-plugin for reading and writing avro files

Project description

polars-avro

build pypi docs

A polars io plugin for reading and writing Apache Avro files, built on arrow-avro. It provides scan support with predicate pushdown, map type reading, and continued avro support as polars deprecates its built-in implementation.

Python Usage

from polars_avro import scan_avro, read_avro, write_avro

lazy = scan_avro(path)
frame = read_avro(path)
write_avro([frame], path)

Rust Usage

There are two main exports: [Reader] for iterating DataFrames from avro sources, and [Writer] for writing DataFrames to an avro file.

use polars_avro::{Reader, Writer, ReadOptions};

// read
let reader = Reader::try_new(
    [File::open("data.avro")],
    ReadOptions::basic(),
).unwrap();
for batch in reader {
    let frame = batch.unwrap();
}

// write
let mut writer = Writer::try_new(file, frame.schema(), None).unwrap();
writer.write(&frame).unwrap();

ℹ️ Avro supports writing with file compression schemes. In rust these need to be enabled via feature flags: deflate, snappy, bzip2, xz, zstd. Decompression is handled automatically.

Idiosyncrasies

Avro and Arrow don't align fully, and polars only supports a subset of arrow. Some types require casting before writing, and some avro types map to different polars types than you might expect when reading.

Writing

The following polars types error when writing and must be cast first:

Polars Type Cast To
Int8 Int32
Int16 Int32
UInt8 Int32
UInt16 Int32
UInt32 Int64
UInt64 Int64 (lossy for > 2⁶³)
Time Int64
Categorical Int32 or String
Enum Int32 or String

Compression is supported via feature flags: deflate, snappy, bzip2, xz, zstd.

Reading

utf8_view behavior — the utf8_view option (default false) changes how certain types are read:

Type utf8_view=false (default) utf8_view=true
UUID binary (16 bytes) formatted string
nullable strings preserves nulls replaces null with "" (lossy)

Since polars tends to work with string views internally, utf8_view=true is likely faster if you don't mind losing null string distinctions.

Type mappings of note:

Avro Type Polars Type
Enum Categorical (not Enum)
Map List of Struct {key, value}
BigDecimal Binary
Duration unsupported (errors)
Date Date (days since epoch)
TimeMillis, TimeMicros Time (nanoseconds)
TimestampMillis/Micros/Nanos Datetime with matching precision and UTC tz
LocalTimestampMillis/Micros/Nanos Datetime with matching precision and no tz

Constraints: the root avro schema must be a Record, and all files in a multi-file read must share the same schema.

Benchmarks

Python reports median (file reads, in-memory writes). Rust reports mean. native = polars built-in avro. Ratio relative to native; bold = fastest. Complex rows use nested/struct types.

Benchmark native polars-avro jetliner
python read 1K × 2 64 µs (1.00x) 99 µs (1.54x) 180 µs (2.79x)
python read 64K × 2 2.7 ms (1.00x) 2.1 ms (0.78x) 2.8 ms (1.04x)
python read 1K × 8 183 µs (1.00x) 242 µs (1.32x) 337 µs (1.84x)
python read 1M × 8 159 ms (1.00x) 114 ms (0.72x) 145 ms (0.91x)
python read 1M × 128 2.6 s (1.00x) 1.8 s (0.69x) 2.8 s (1.09x)
python read complex 1K × 8 449 µs 592 µs
python read complex 1M × 8 181 ms 260 ms
python read proj 1M × 128 → 8 1.6 s (1.00x) 1.2 s (0.75x) 1.2 s (0.77x)
python read proj 1K × 8 → 2 133 µs (1.00x) 297 µs (2.24x) 264 µs (1.99x)
python write 1K × 2 42 µs (1.00x) 30 µs (0.72x)
python write 64K × 2 1.5 ms (1.00x) 1.1 ms (0.71x)
python write 1K × 8 143 µs (1.00x) 114 µs (0.80x)
python write 1M × 8 87 ms (1.00x) 93 ms (1.07x)
python write 1M × 128 1.5 s (1.00x) 2.2 s (1.48x)
rust read 1K × 2 42 µs (1.00x) 34 µs (0.80x)
rust read 1M × 128 2.8 s (1.00x) 2.0 s (0.69x)
rust read proj 1M × 128 → 8 1.3 s (1.00x) 1.2 s (0.87x)
rust read proj 1K × 8 → 2 109 µs (1.00x) 116 µs (1.06x)
rust write 1K × 2 42 µs (1.00x) 22 µs (0.53x)
rust write 64K × 2 1.5 ms (1.00x) 1.0 ms (0.67x)
rust write 1K × 8 135 µs (1.00x) 93 µs (0.69x)
rust write 1M × 8 97 ms (1.00x) 89 ms (0.92x)
rust write 1M × 128 1.6 s (1.00x) 1.4 s (0.88x)

Development

Rust

Standard cargo commands will build and test the rust library.

Python

The python library is built with uv and maturin. The rust components should build once, ance otherwise allow usage and testing.

You may need to recompile the python bindings with uv run maturin develop.

Testing

cargo fmt --check
cargo clippy --all-features --tests
cargo test
uv run ruff format --check
uv run ruff check
uv run pyright
uv run pytest

Benchmarking

Python benchmarks are disabled by default. To run them:

cargo +nightly bench
uv run pytest --benchmark-only

Releasing

rm -rf dist
uv build --sdist
uv run maturin build -r -o dist --target aarch64-apple-darwin
uv run maturin build -r -o dist --target aarch64-unknown-linux-gnu --zig
uv publish --username __token__

To Do

  • reimplement single column reader?
  • reimplement better workarounds for types that don't exist, e.g. serialize polars cat/enum to arrow enum and vice versa

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

polars_avro-0.8.1.tar.gz (164.8 kB view details)

Uploaded Source

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

polars_avro-0.8.1-cp310-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (12.1 MB view details)

Uploaded CPython 3.10+manylinux: glibc 2.17+ ARM64

polars_avro-0.8.1-cp310-abi3-macosx_11_0_arm64.whl (12.3 MB view details)

Uploaded CPython 3.10+macOS 11.0+ ARM64

File details

Details for the file polars_avro-0.8.1.tar.gz.

File metadata

  • Download URL: polars_avro-0.8.1.tar.gz
  • Upload date:
  • Size: 164.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.6.17

File hashes

Hashes for polars_avro-0.8.1.tar.gz
Algorithm Hash digest
SHA256 d51207e34fd5317e210b1b0ce173a8a73c898cec699fab51716a503601787eb8
MD5 1bf8f4d60b40ba07bd00e542062a6121
BLAKE2b-256 ab6ba92423c3c71238447a10b796e9aa3c535d5288711565990bfeddf7dbd071

See more details on using hashes here.

File details

Details for the file polars_avro-0.8.1-cp310-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl.

File metadata

File hashes

Hashes for polars_avro-0.8.1-cp310-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
Algorithm Hash digest
SHA256 e33fa7a39180de4a379538c901ae15b42c30fce699f57c7299359d38eb5d8fc5
MD5 6f807280b31608e0ae10c9de266a6835
BLAKE2b-256 5dc6e6d2019aeb17edcb3b6b399ba4436719873d5f810a5cdd47240c8b87124c

See more details on using hashes here.

File details

Details for the file polars_avro-0.8.1-cp310-abi3-macosx_11_0_arm64.whl.

File metadata

File hashes

Hashes for polars_avro-0.8.1-cp310-abi3-macosx_11_0_arm64.whl
Algorithm Hash digest
SHA256 6dfd0807b129b5d0c363ab4174ab40837c20c3ec3d4c9454587963788bc726e7
MD5 53e01ae78c461f235a7bb16b07fae0f2
BLAKE2b-256 9c3858f54120741b08ceb899809bde71ea4e621d7d094b00a7bb6456acd4e02e

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page