---
id: grizz
version: "0.1.1"
license: BSD-3-Clause
license_treatment: permissive
maintenance: active
---

# grizz — A light library to preprocess data with polars

License: permissive · Maintenance: active · Downloads: 92.2K/mo

## What it is and what it does

grizz is a lightweight library for ingesting and transforming data within polars DataFrames. It uses an object-oriented design where ingestors load data and transformers reshape it, and these building blocks can be combined into pipelines. The library wraps common operations like column casting, filtering, and type conversion into reusable transformer objects that can be chained together.

The package depends on polars for DataFrame operations, plus coola, iden, and objectory for internal utilities. It supports Python 3.9 through 3.13 and is currently in beta—the documentation explicitly warns that the API is not stable and may change significantly before a 1.0 release. Optional dependencies like clickhouse-connect, pyarrow, and tqdm extend functionality for specific use cases.

Use it for:

- Cast multiple DataFrame columns to a target type in a single composable step
- Build reusable data preprocessing pipelines by combining ingestors and transformers
- Extend the library with custom ingestors to load data from domain-specific sources
- Preprocess tabular data for machine learning workflows using polars as the backend

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

grizz provides composable ingestors and transformers to load and reshape data in polars DataFrames using an object-oriented building-block approach.

Yes, if you are already using polars and want a lightweight, composable approach to data transformation. The low install friction, active maintenance, and permissive license make it low-risk. However, be aware that the API is unstable in beta—pin your version and review release notes before upgrading, as breaking changes are expected.

## Install

pip install grizz
uv add grizz
poetry add grizz

## Installing grizz

Before you install: Low install friction with a pure-Python wheel. Actively maintained with recent commits; however, the package is in beta stage and the fact sheet notes that API stability is not guaranteed between releases, so breaking changes are possible.

License in practice: BSD-3-Clause is permissive and allows commercial use, modification, and distribution with minimal restrictions—suitable for most projects.

Quickstart:

pip install grizz

import polars as pl
from grizz.transformer import Cast

transformer = Cast(columns=["col1", "col3"], dtype=pl.Int32)
frame = pl.DataFrame({"col1": [1, 2, 3], "col3": ["1", "2", "3"]})
out = transformer.transform(frame)

Verify before relying:

- Whether custom ingestors and transformers can be easily extended beyond the built-in set
- Performance characteristics when handling large DataFrames or complex transformation chains
- Availability and completeness of documentation for all transformer and ingestor types

## Package facts

- License: BSD-3-Clause (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: active
- Downloads: 92.2K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

polars dataframe transformation, data ingestion pipeline, dataframe preprocessing, polars data casting, composable data transformers, dataframe column operations, polars etl library, polars-ecosystem, data-pipeline, etl

[View on SkillFed](https://skillfed.io/packages/grizz) · [View on PyPI](https://pypi.org/project/grizz/)