parquet

What it is and what it does

parquet is a pure-Python implementation of the Apache Parquet file format with read-only support. It provides both a command-line tool (`parquet` command) for inspecting Parquet files and a programmatic API with DictReader and reader classes similar to Python's csv module. The package is useful for debugging and quick viewing of Parquet data without the overhead of a JVM, and it can read data files from the parquet-compatibility project.

The package depends on thriftpy2 for Thrift serialization and backports.csv for CSV output. However, it has not been actively maintained since 2020, and its test coverage was limited to Python 2.7, 3.6, and 3.7. Not all Parquet format features have been implemented—notably nested data handling and deprecated bitpacking remain incomplete. Performance has not been optimized.

Use it for:

• Quickly inspect Parquet file contents from the command line without installing Java or Spark.
• Convert Parquet data to JSON or TSV format for integration with non-Parquet tools.
• Read Parquet files programmatically in Python for data debugging and exploratory analysis.
• Extract specific columns from Parquet files using the DictReader or reader interface.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Reads Apache Parquet files in pure Python and outputs data as JSON or TSV via a command-line tool or programmatic reader interface.

No. The package is abandoned (last commit 2021-10-26, no release since 2020-04-30) and was only tested on Python 2.7, 3.6, and 3.7—all now end-of-life. Compatibility with modern Python versions is unverified, and critical Parquet features remain unimplemented. For production use, prefer actively maintained alternatives like pyarrow or fastparquet.

Install

parquet on PyPI

pip install parquet

uv add parquet

poetry add parquet

Installing parquet

Before you install

Low install friction with only two runtime dependencies. However, the package is abandoned—last commit was 2021-10-26 and no release since 2020-04-30. It was tested on Python 2.7, 3.6, and 3.7, which are now end-of-life versions; compatibility with modern Python is unverified.

License in practice

Licensed under Apache License 2.0 (permissive), so you can use it freely in commercial and open-source projects without significant legal constraints.

Quickstart

import parquet

with open("test.parquet") as fo:
    for row in parquet.DictReader(fo, columns=['foo', 'bar']):
        print(row)

Requires thriftpy2 and backports.csv as runtime dependencies; compatibility with Python versions newer than 3.7 is unverified.

Verify before relying

• Whether the package works reliably on Python 3.8 and later versions.
• Current state of nested data support and which parquet-format features remain unimplemented.
• Performance characteristics compared to modern alternatives.
• Whether snappy compression support (optional extra) is still functional.

Package facts

Evidence: parquet-1.3.1-py3-none-any.whl

Tags

More Database packages