InterpolationJoiner - polars

[zbenmo]

Describe the bug

Tried a simple join as follows:

joiner = InterpolationJoiner(
data_store["depth_0"][0],
key=["case_id"],
suffix="_depth_0",
).fit(data_store["df_base"])
join = joiner.transform(data_store["df_base"])
join.head()

--

data_store["depth_0"][0] - polars Dataframe
data_store["df_base"] - polars Dataframe

--

Steps/Code to Reproduce

joiner = InterpolationJoiner(
data_store["depth_0"][0],
key=["case_id"],
suffix="_depth_0",
).fit(data_store["df_base"])
join = joiner.transform(data_store["df_base"])
join.head()

Expected Results

Wanted to see the join, as in: https://skrub-data.org/stable/auto_examples/09_interpolation_join.html

Actual Results

---

KeyError Traceback (most recent call last)
File /opt/conda/lib/python3.10/site-packages/polars/_utils/deprecation.py:95, in deprecate_parameter_as_positional..decorate..wrapper(*args, **kwargs)
94 try:
---> 95 param_args = kwargs.pop(old_name)
96 except KeyError:

KeyError: 'columns'

During handling of the above exception, another exception occurred:

TypeError Traceback (most recent call last)
Cell In[8], line 5
1 joiner = InterpolationJoiner(
2 data_store["depth_0"][0],
3 key=["case_id"],
4 suffix="_depth_0",
----> 5 ).fit(data_store["df_base"])
6 join = joiner.transform(data_store["df_base"])
7 join.head()

File /opt/conda/lib/python3.10/site-packages/skrub/_interpolation_joiner.py:225, in InterpolationJoiner.fit(failed resolving arguments)
223 _join_utils.check_missing_columns(X, self.main_key, "'X' (the main table)")
224 key_values = self.vectorizer.fit_transform(self.aux_table[self._aux_key])
--> 225 estimators = self._get_estimator_assignments()
226 fit_results = joblib.Parallel(self.n_jobs)(
227 joblib.delayed(_fit)(
228 key_values,
(...)
233 for assignment in estimators
234 )
235 fit_results = self._check_fit_results(fit_results)

File /opt/conda/lib/python3.10/site-packages/skrub/_interpolation_joiner.py:356, in InterpolationJoiner._get_estimator_assignments(self)
339 def _get_estimator_assignments(self):
340 """Identify column groups to be predicted together and assign them an estimator.
341
342 In many cases, a single estimator cannot handle all the target columns.
(...)
354 separately to each column.
355 """
--> 356 aux_table = self.aux_table.drop(self._aux_key, axis=1)
357 assignments = []
358 regression_table = aux_table.select_dtypes("number")

File /opt/conda/lib/python3.10/site-packages/polars/_utils/deprecation.py:97, in deprecate_parameter_as_positional..decorate..wrapper(*args, **kwargs)
95 param_args = kwargs.pop(old_name)
96 except KeyError:
---> 97 return function(*args, **kwargs)
99 issue_deprecation_warning(
100 f"named {old_name} param is deprecated; use positional *args instead.",
101 version=version,
102 )
103 if not isinstance(param_args, Sequence) or isinstance(param_args, str):

TypeError: DataFrame.drop() got an unexpected keyword argument 'axis'

Versions

'0.1.0'

[jeromedockes]

thanks for reporting this bug. Indeed, InterpolationJoiner does not yet have support for polars, although that should be added soon. in the meanwhile it should be documented and provide a better error message

• document the fact that interpolationjoiner is missing polars support ATM
• add polars support