narwhals-dev/narwhals

Health v2 · maintenance only

Adoption not part of the score

1722 stars · 226 forks observed · 2026-09-22

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Narwhals is a lightweight, zero-dependency Python compatibility layer that lets code written against a subset of the Polars API run on pandas, Polars, PyArrow, cuDF, Modin, Dask, DuckDB, Ibis, PySpark, and more. It is aimed at library authors who want to make their dataframe-consuming code dataframe-agnostic without hard dependencies on any specific backend.

Use cases

• write dataframe-agnostic library code that works with pandas and polars
• support multiple dataframe backends without adding dependencies
• migrate code between pandas and polars with minimal changes
• run the same dataframe logic on spark, duckdb, or dask
• build a tool that accepts any dataframe library as input
• use polars expressions api on top of pandas dataframes

When to choose

• you are building a library or service that consumes dataframes and want backend flexibility
• you want zero-dependency interoperability across dataframe libraries
• you prefer the polars api but need to support pandas users
• you need lazy and eager dataframe support with full static typing

When to avoid

• you only ever use a single dataframe library and need no interoperability
• you need dataframe features outside narwhals' supported api subset
• you want a full dataframe implementation rather than a compatibility layer

Facets

library · maturity active

data-science developer-tools data-science developer-tools python dataframes interoperability compatibility-layer polars-api pandas zero-dependency data-engineering

Member repositories

For agents

markdown · JSON · MCP: product_card(name="narwhals-dev/narwhals")