[Enh]: Support `DataFrame|LazyFrame.rolling`

[rwhitten577]

We would like to learn about your use case. For example, if this feature is needed to adopt Narwhals in an open source project, could you please enter the link to it below?

Calculate temporal rolling aggregations where the windows are not constant, e.g. "count of clicks over past 3 days by user"

Please describe the purpose of the new feature or describe the problem to solve.

Polars has .rolling on DataFrame/LazyFrames: https://docs.pola.rs/api/python/stable/reference/dataframe/api/polars.DataFrame.rolling.html

Would be great to support this in Narwhals for easy temporal aggregations. This is the equivalent of a range between interval window function and should be supported by many backends (duckdb, spark, ibis, etc).

Suggest a solution if possible.

No response

If you have tried alternatives, please describe them below.

No response

Additional information that may help us understand your needs.

No response

[MarcoGorelli]

Thanks for the request - I think the equivalent would be rolling_*_by, which should be in scope

[MarcoGorelli]

Note that rolling_mean_by is still marked as unstable in Polars

So, OK probably to add this, so long as it stays out of the stable api

[rwhitten577]

Yeah looks like all the rolling_*_by are unstable while df.rolling is not. I'd be ok with either option assuming the rolling_*_by can be chained with .over. df.rolling also supports offset while the exprs do not. Adding offset_by may be needed to get the same type of lookback functionality with the expr?

[MarcoGorelli]

rolling will probably get rewritten so we may want to avoid that one for now

assuming the rolling_*_by can be chained with .over

yup 👍 #2533 should lay the groundworks for that