That's what the parameter method="table" is for. It is however only implemented when given engine="numba" in the method, so is a bit of a hit & miss for custom functions given to apply.

So, this is currently more of an limitation of the current API. and changing it,so method="table" also works for non-numba functions, could be worth an investigation but would be an enhancement proposal, not a bug fix IMO.

Is rolling.apply part of your work on fixing groupby.apply, @rhshadrach?

Hi topper-123, thank you so much for your reply.

The info for reference of numba is limited. I suppose this code should work, replacing the last row of code in the question.
df.rolling(window=2, min_periods=1, method="table").apply(lambda x: x, raw=True, engine='numba'), which returns an error. ValueError: cannot broadcast source array for assignment

I am not familiar with the restrictions on when numba works with rolling.apply, but yeah I tried that too and it didn't work, I haven't checked why...

I looked through the issue list, and there is already #40374, which started out as a enhancement proposal, but ended up. as a doc issue, but is still open, though the related doc issue has been resolved. So IDK, can you rephrase this to be an enhancement proposal, requesting to method="table" in rolling work outside of numba also? Then I'll close the other as completed.

Is rolling.apply part of your work on fixing groupby.apply, @rhshadrach?

Yes, it is!

Is rolling.apply part of your work on fixing groupby.apply, @rhshadrach?

Yes, it is!

Nice. Would we then be able to roll over a DataFrame like requested here?

Yes - I am aiming to add user-control as to whether the columns or the entire frame is passed to the UDF across the board, so in particular in rolling.

I am not familiar with the restrictions on when numba works with rolling.apply, but yeah I tried that too and it didn't work, I haven't checked why...

For context, this was done because the current window aggregations (including apply) in cython were only written to accepts 1D inputs. At the time, it wasn't in my scope to rewrite them to 2D

Hello @mb-hz and @mroeschke,

My project team is looking for Pandas enhancement features for our grad school semester long project. We saw this task and would like to contribute if possible!

Hello @mroeschke

Has this issue been fixed on your end with your groupby.apply fix? I am currently working on trying to modify the Rolling class _apply function to do this, but don't want to double work.

To go with the above comment, @rhshadrach where you mention aiming to add user control on this, would or does this fix this issue?

@iangainey - adding user control on what is given to the passed callable (column-by-column or the entire frame at once) would close this issue. Currently there is no concrete proposal for this and it is something I plan to take up eventually - but would be happy if someone else takes it up. That being said, I would like such a proposal to encompass agg, apply, transform, filter (in some instances), and map across Series, DataFrame, groupby, rolling, and resample rather than going piece-meal creating a (more) inconsistent API.

@rhshadrach Hi Richard, thanks for the response! I currently have a working fix in the scope of this issue for rolling.apply by adding user control with a multi_column = true parameter. Is it a requirement that the parameter to be all encompassing, or would you accept a solution that would only close this issue that could be incorporated into a bigger multi_column fix later on.