potential use cases

[MarcoGorelli]

I'm just collecting some projects I notice which may benefit from Narwhals, or who have shown interest. Some of these might not actually end up being places where Narwhals is a good fit, I tend to jot things down if look like they might be from a first glance and then revisit in more detail later

• Make packages (pandas) optional PRQL/pyprql#374
• Refactor dataframe support to use narwhals generic dataframes ApeWorX/ape#2013
• pandas warning with two-way clustering py-econometrics/pyfixest#422 - they mentioned this in a PR. Using pandas via Narwhals avoids the setting-with-copy warning altogether, even for old versions, so it may be of interest
• [Core] Add support for polars Nixtla/tsfeatures#29
• Feature Request: support for Polars ydataai/ydata-profiling#1129
• Switch from pandas to polars for all parts depictio/depictio#59
• Replacing pandas with polars pvlib/pvlib-python#1926
• [ENH] polars support - as sktime datatype sktime/sktime#5423 [ENH] array or DataFrame API support sktime/sktime#5370
• Idea: support narwhals for easy support of multiple df libraries rstudio/pins-python#293
• [ENH] umbrella issue - support for data container types sktime/sktime#6091
• Investigate using Ibis for the common interface library to any DF backend projectmesa/mesa-frames#89

---

Some new ones:

• bokeh
• prophet
• itables
• formulaic

if anyone happens to stumble upon this issue, feel free to let us know and we can add your project to the list

[FBruzzesi]

I will leave it here even though I imagine that it is quite challenging: skrub would be very suitable use case

[MarcoGorelli]

Thanks! Yeah I spoke to them a bit about it, and for now they're hand-rolling their own dispatch mechanism. I think it'll quickly get complex, especially around different pandas/Polars versions, and pandas' complicated type system - I think if we focus on getting Narwhals as easy to use as possible, and fully documented, and robust, then if they decide they'd like to use it, we can be ready 😎

[Excidion]

I am not sure if I can use this issue as a wish-list? If not please delete. In any case I suggested narwhals to featuretools who are looking for ways to speed up their feature matrix calculation: alteryx/featuretools#1967

[dangotbanned]

Not affiliated with, but https://github.com/posit-dev/great-tables

Related

Issues

• support any frames with __dataframe__ posit-dev/great-tables#229
• epic: Remove numpy as a dependency posit-dev/great-tables#296
  • https://github.com/machow/databackend
• Test use of polars selectors throughout library posit-dev/great-tables#35

Blog

• https://posit-dev.github.io/great-tables/blog/polars-styling/

  As it turns out, polars expressions make styling tables very straightforward. The same polars code that you would use to select or filter combines with Great Tables to highlight, circle, or bolden text.
  In this post, I’ll show how Great Tables uses polars expressions to make delightful tables, like the one below.