Maintain row order after cross join (#463)

Fixes the failing polars tests.

By default, `polars` gives no guarantees on the resulting row order of a
join (see [here](pola-rs/polars#20725)),
meaning that our tests used to pass just by luck. This has changed since
`polars==0.19.0`, which apparently included changes that affect the row
order of our test dataframes. The PR fixes these tests by ignoring the
row order during the equality check.

CHANGELOG.md

@@ -36,6 +36,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   compatible with `strtobool`
 - All arguments to `MetaRecommender.select_recommender` are now optional
 - `MetaRecommender`s can now be composed of other `MetaRecommender`s
+- For performance reasons, search space manipulation using `polars` is no longer
+  guaranteed to produce the same row order as the corresponding `pandas` operations
 
 ### Fixed
 - Rare bug arising from degenerate `SubstanceParameter.comp_df` rows that caused

docs/userguide/envvars.md

@@ -85,6 +85,12 @@ changing the Python environment. To do so, you can set the environment variable
 `BAYBE_DEACTIVATE_POLARS` to any truthy value accepted by
 [`strtobool`](baybe.utils.boolean.strtobool).
 
+```{admonition} Row Order
+:class: caution
+
+For performance reasons, search space manipulation using `polars` is not
+guaranteed to produce the same row order as the corresponding `pandas` operations.
+```
 
 ## Disk Caching
 For some components, such as the

tests/constraints/test_constraints_polars.py

@@ -189,7 +189,7 @@ def test_polars_product(constraints, parameters):
     # Do Pandas product
     df_pd = parameter_cartesian_prod_pandas(parameters)
 
-    # Assert equality of lengths before filtering
+    # Assert equality before filtering
     assert_frame_equal(df_pl.to_pandas(), df_pd)
 
     # Apply constraints
@@ -198,5 +198,9 @@ def test_polars_product(constraints, parameters):
         _apply_constraint_filter_polars(ldf, constraints)[0].collect().to_pandas()
     )
 
-    # Assert strict equality of two dataframes
-    assert_frame_equal(df_pl_filtered, df_pd_filtered)
+    # Assert order-agnostic equality of the two dataframes
+    cols = df_pd_filtered.columns.tolist()
+    assert_frame_equal(
+        df_pd_filtered.sort_values(cols).reset_index(drop=True),
+        df_pl_filtered.sort_values(cols).reset_index(drop=True),
+    )