fix: dask group by with kwargs #1676
Conversation
| def var( | ||
| ddof: int = 1, | ||
| ) -> Callable[ | ||
| [pd.core.groupby.generic.SeriesGroupBy], pd.core.groupby.generic.SeriesGroupBy |
There was a problem hiding this comment.
I need to double check the return type
|
|
||
| import dask_expr as dx | ||
|
|
||
| return partial(dx._groupby.GroupBy.var, ddof=ddof) |
There was a problem hiding this comment.
The minimum version we support for dask is "dask[dataframe]==2024.7", which installs dask-expr==1.1.8, and dx._groupby.GroupBy is already available.
FBruzzesi
added
fix
high priority
MarcoGorelli
left a comment
MarcoGorelli
left a comment
There was a problem hiding this comment.
awesome, thanks a lot @FBruzzesi !
if we can use a public variant for dask expr's groupby, then that'd probably be better, else i think this is ok (we test against their nightlies anyway)
|
Thanks for reviewing @MarcoGorelli 👌
It's definitly not exposed in dask-expr. As mentioned in the discourse thread we could use |
|
Just caught up with Patrick from Dask, and he's said that they plan to merge dask-expr into dask and then they'll expose some public apis, and that we can merge as-is for now |
What type of PR is this? (check all applicable)
Related issues
group_bycontext ignores expr arguments #1606 by implementing var/std with custom ddof for dask as explained in the dask discourse threadChecklist
If you have comments or can explain your changes, please do so below