feat: Add having to group_by context

[borchero]

Fixes #23290.

A few comments:

• This implementation is still missing a specialized implementation for the streaming engine as I had a little bit of a hard time figuring out how to achieve this.
• I only added one test case so far as I first wanted to get feedback on the current implementation

[codecov]

Codecov Report

Attention: Patch coverage is 44.92754% with 38 lines in your changes missing coverage. Please review.

Project coverage is 81.28%. Comparing base (e99abdc) to head (6378db6).

Files with missing lines	Patch %	Lines
py-polars/polars/dataframe/group_by.py	50.00%	14 Missing and 5 partials ⚠️
crates/polars-python/src/lazygroupby.rs	0.00%	8 Missing ⚠️
crates/polars-lazy/src/frame/mod.rs	61.53%	5 Missing ⚠️
.../polars-plan/src/plans/conversion/dsl_to_ir/mod.rs	0.00%	3 Missing ⚠️
py-polars/polars/lazyframe/group_by.py	25.00%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #23550      +/-   ##
==========================================
- Coverage   81.28%   81.28%   -0.01%     
==========================================
  Files        1644     1644              
  Lines      223249   223312      +63     
  Branches     2841     2847       +6     
==========================================
+ Hits       181479   181511      +32     
- Misses      41071    41097      +26     
- Partials      699      704       +5     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ritchie46]

I am not sure we should implement this as opposed this lowering to a group-by with a post filter. I think it should be the latter.

[borchero]

I am not sure we should implement this as opposed this lowering to a group-by with a post filter. I think it should be the latter.

Fair, the only problem I see is that I now need to choose column names for the predicates which may not clash with each other and any keys/aggregations. How do you typically solve such an issue here?

[ritchie46]

You can use _POLARS_HAVING_ as a temporary name or as a prefix.

[borchero]

One more question @ritchie46: would you want to do the

group_by(key, predicate, agg)
-> group_by(key, predicate | agg) + filter(predicate) + drop(predicate)

translation when building the DslPlan (i.e. from LazyGroupBy.agg) or when going from DslPlan to IR (i.e. in to_alp_impl) or some place I haven't identified yet (😄)?

[ritchie46]

Unless absolutely trivial, I think it should be when converting from DslPlan -> IR.

[borchero]

I removed the predicate field from the IR and do the translation in to_alp_impl now. This generally seems much more ergonomic 😄

[borchero]

@ritchie46 do you have any thoughts on the updated implementation? :)

[etiennebacher]

Isn't it "that contain more than one element" instead? ("filter" is equivalent to "keep" for me)

Or "Keep groups that contain more than one element" ?