narwhals 2.21.2

Extremely lightweight compatibility layer between dataframe libraries

Narwhals

Extremely lightweight and extensible compatibility layer between dataframe libraries!

• Full API support: cuDF, Modin, pandas, Polars, PyArrow.
• Lazy-only support: Daft, Dask, DuckDB, Ibis, PySpark, SQLFrame.

Seamlessly support all, without depending on any!

• ✅ Just use a subset of the Polars API, no need to learn anything new
• ✅ Zero dependencies, Narwhals only uses what the user passes in so your library can stay lightweight
• ✅ Separate lazy and eager APIs, use expressions
• ✅ Support pandas' complicated type system and index, without either getting in the way
• ✅ 100% branch coverage, tested against pandas and Polars nightly builds
• ✅ Negligible overhead, see overhead
• ✅ Let your IDE help you thanks to full static typing, see typing
• ✅ Perfect backwards compatibility policy, see stable api for how to opt-in

Get started!

• Read the documentation
• Chat with us on Discord!
• Join our community call
• Read the contributing guide

Table of contents

• Narwhals
  • Installation
  • Usage
  • Example
  • Scope
  • Roadmap
  • Used by
  • Sponsors and institutional partners
  • Support
  • Appears on
  • Why "Narwhals"?

Installation

• pip (recommended, as it's the most up-to-date)

  pip install narwhals
  

• conda-forge (also fine, but the latest version may take longer to appear)

  conda install -c conda-forge narwhals
  

Usage

There are three steps to writing dataframe-agnostic code using Narwhals:

1. use narwhals.from_native to wrap a pandas/Polars/Modin/cuDF/PyArrow DataFrame/LazyFrame in a Narwhals class

2. use the subset of the Polars API supported by Narwhals

3. use narwhals.to_native to return an object to the user in its original dataframe flavour. For example:

   • if you started with pandas, you'll get pandas back
   • if you started with Polars, you'll get Polars back
   • if you started with Modin, you'll get Modin back (and compute will be distributed)
   • if you started with cuDF, you'll get cuDF back (and compute will happen on GPU)
   • if you started with PyArrow, you'll get PyArrow back

Example

Narwhals allows you to define dataframe-agnostic functions. For example:

import narwhals as nw
from narwhals.typing import IntoFrameT


def agnostic_function(df_native: IntoFrameT) -> IntoFrameT:
    return (
        nw.from_native(df_native)
        .with_columns(
            category=nw.when(nw.col("animal").str.contains("whale"))
            .then(nw.lit("whale"))
            .otherwise(nw.lit("other"))
        )
        .to_native()
    )

You can then pass pandas.DataFrame, polars.DataFrame, polars.LazyFrame, duckdb.DuckDBPyRelation, pyspark.sql.DataFrame, pyarrow.Table, and more, to agnostic_function. In each case, no additional dependencies will be required, and computation will stay native to the input library:

import duckdb
import polars as pl
import pandas as pd

data = {
    "animal": ["blue whale", "orca", "dolphin", "humpback whale", "seal"],
    "length_m": [30.0, 8, 2.5, 17, 2.2],
    "weight_kg": [150000, 4000, 200, 30000, 85],
}

print("Polars result")
df_pl = pl.DataFrame(data)
print(agnostic_function(df_pl))

print("DuckDB result")
print(agnostic_function(duckdb.sql("select * from df_pl")))

print("pandas result")
df_pd = pd.DataFrame(data)
print(agnostic_function(df_pd))

Polars result
shape: (5, 4)
┌────────────────┬──────────┬───────────┬──────────┐
│ animal         ┆ length_m ┆ weight_kg ┆ category │
│ ---            ┆ ---      ┆ ---       ┆ ---      │
│ str            ┆ f64      ┆ i64       ┆ str      │
╞════════════════╪══════════╪═══════════╪══════════╡
│ blue whale     ┆ 30.0     ┆ 150000    ┆ whale    │
│ orca           ┆ 8.0      ┆ 4000      ┆ other    │
│ dolphin        ┆ 2.5      ┆ 200       ┆ other    │
│ humpback whale ┆ 17.0     ┆ 30000     ┆ whale    │
│ seal           ┆ 2.2      ┆ 85        ┆ other    │
└────────────────┴──────────┴───────────┴──────────┘
DuckDB result
┌────────────────┬──────────┬───────────┬──────────┐
│     animal     │ length_m │ weight_kg │ category │
│    varchar     │  double  │   int64   │ varchar  │
├────────────────┼──────────┼───────────┼──────────┤
│ blue whale     │     30.0 │    150000 │ whale    │
│ orca           │      8.0 │      4000 │ other    │
│ dolphin        │      2.5 │       200 │ other    │
│ humpback whale │     17.0 │     30000 │ whale    │
│ seal           │      2.2 │        85 │ other    │
└────────────────┴──────────┴───────────┴──────────┘

pandas result
           animal  length_m  weight_kg category
0      blue whale      30.0     150000    whale
1            orca       8.0       4000    other
2         dolphin       2.5        200    other
3  humpback whale      17.0      30000    whale
4            seal       2.2         85    other

See the tutorial for several examples!

Scope

• Do you maintain a dataframe-consuming library?
• Do you have a specific Polars function in mind that you would like Narwhals to have in order to make your work easier?

If you said yes to both, we'd love to hear from you!

Roadmap

See roadmap discussion on GitHub for an up-to-date plan of future work.

Used by

Join the party!

• altair
• bokeh
• darts
• fairlearn
• formulaic
• gt-extras
• hierarchicalforecast
• marimo
• metalearners
• mosaic
• panel-graphic-walker
• plotly
• pointblank
• pymarginaleffects
• pyreadstat
• py-shiny
• pysummaries
• rio
• scikit-lego
• scikit-playtime
• tabmat
• tea-tasting
• timebasedcv
• tubular
• Validoopsie
• vegafusion
• wimsey

Feel free to add your project to the list if it's missing, and/or chat with us on Discord if you'd like any support.

Sponsors and institutional partners

Narwhals is 100% independent, community-driven, and community-owned. We are extremely grateful to the following organisations for having provided some funding / development time:

• Quansight Labs
• Quansight Futures
• OpenTeams
• POSSEE initiative
• BYU-Idaho

If you contribute to Narwhals on your organization's time, please let us know. We'd be happy to add your employer to this list!

Support

If you'd like to say "thank you", please give us a ⭐ star ⭐.

Please contact [e-mail] if you would like to:

• Receive professional support (e.g., if you're using or would like to use Narwhals at your company).
• Have any Narwhals fixes / features prioritised.
• Commission any Narwhals plugins for new backends.

Appears on

Narwhals has been featured in several talks, podcasts, and blog posts:

• Inspiring Computing Podcast The Rise of Narwhals in Open-Source

• PyCon DE & PyData 2025 How Narwhals is silently bringing pandas, Polars, DuckDB, PyArrow, and more together

• The Python Exchange March 2025 What Can Narwhals Do for You?

• PyData London 2025 How Narwhals brings Polars, DuckDB, PyArrow, & pandas together

• Talk Python to me Podcast Ahoy, Narwhals are bridging the data science APIs

• Python Bytes Podcast Episode 402, topic #2

• Super Data Science: ML & AI Podcast
  Narwhals: For Pandas-to-Polars DataFrame Compatibility

• Sample Space Podcast | probabl
  How Narwhals has many end users ... that never use it directly. - Marco Gorelli

• The Real Python Podcast Narwhals: Expanding DataFrame Compatibility Between Libraries

• Pycon Lithuania 2024
  Marco Gorelli - DataFrame interoperatiblity - what's been achieved, and what comes next?

• Pycon Italy 2024
  How you can write a dataframe-agnostic library - Marco Gorelli

• Polars Blog Post
  Polars has a new lightweight plotting backend

• Quansight Labs blog post (w/ Scikit-Lego)
  How Narwhals and scikit-lego came together to achieve dataframe-agnosticism

Why "Narwhals"?

Coz they are so awesome.

Thanks to Olha Urdeichuk for the illustration!