Migrate ArbitraryOutlierCapper to narwhals, add polars support - #1034
Open
solegalli wants to merge 2 commits into
Open
Migrate ArbitraryOutlierCapper to narwhals, add polars support#1034solegalli wants to merge 2 commits into
solegalli wants to merge 2 commits into
Conversation
Shared base for all outlier transformers (ArbitraryOutlierCapper extends BaseOutlier directly; Winsoriser/OutlierTrimmer extend WinsorizerBase): column reorder + NA/Inf checks in _check_transform_input_and_state(), the fold-limit estimation in WinsorizerBase.fit() (gaussian/iqr/mad/ quantiles), and the capping step in BaseOutlier._transform() are now dataframe-agnostic. Capping (np.clip against per-column bounds) was benchmarked three ways at 10k/50k/100k rows x 1/2/10 columns: pandas-native .clip() loop vs. a single narwhals with_columns(nw.col(v).clip(lo, hi) for v in ...) vs. grouping columns by which bound(s) apply and running up to 3 vectorized numpy calls (np.clip/minimum/maximum) via to_numpy()/new_series(), mirroring ReciprocalTransformer's numpy-acceleration pattern. narwhals-generic alone was already close to parity (0.95-1.49x pandas-native - minimal loss, mergeable per the imputation-base precedent), but the numpy-grouped version was faster still: 0.16-0.82x of pandas-native on the homogeneous case (single tail, all columns share the same bound - the common Winsoriser/ OutlierTrimmer case) and 0.42-1.52x on mixed-coverage dicts (the ArbitraryOutlierCapper case, up to 3 groups). Adopted the numpy-grouped version as the single merged code path for both backends. A first numpy attempt used a blanket -inf/inf sentinel for the missing side per column (like RelativeFeatures-style bound arrays) - that's a correctness bug, not just a style choice: mixing an int64 numpy array with a float -inf/inf bound upcasts the whole column to float64 even when the real, present bound is an int (e.g. ArbitraryOutlierCapper's own docstring example, `max_capping_dict=dict(x1=8)`, expects int64 out). Grouping columns into "both bounds" / "right only" / "left only" buckets and calling np.clip/minimum/maximum with only the bounds that actually exist avoids ever introducing an inf, so dtype promotion matches pandas .clip() exactly - verified byte-for-byte against the old pandas-only implementation across all 4 capping methods x 3 tails, plus the int-dtype and mixed-dict-coverage cases. Also found and fixed a real bug introduced while migrating fit(): plain np.mean/np.std/np.quantile/np.median propagate NaN, unlike pandas' mean/std/quantile/median which skip NaN by default. With missing_values="ignore" and NaN present, this silently produced NaN caps instead of the caps computed from non-null data. Fixed by using the nan-aware numpy variants (np.nanmean/nanstd/nanquantile/nanmedian). Caught by tests/test_outliers/test_winsorizer.py::test_transformer_ignores_na_in_df, which predates this migration but exercises exactly this path. variables/feature names can be int or str; passing a plain list to narwhals' .select() only works for string columns, so every .select() call here uses nw.col(*variables) instead - .select(list_of_ints) raises InvalidIntoExprError. Verified: tests/test_outliers full suite - 83 passed, 3 pre-existing failures in test_check_estimator_outliers.py (sklearn's check_estimator feeds raw numpy arrays, which check_X() has always rejected per the narwhals migration's dataframe-only contract; identical failure set before and after this change). flake8 and mypy clean on the file. Module imports and runs fit/_transform end-to-end on polars with pandas import fully blocked. sphinx -W build clean (only the pre-existing unrelated linkcode_resolve warning). All 4 capping-method x tail combinations and the Winsoriser/OutlierTrimmer/ArbitraryOutlierCapper docstring examples produce byte-identical output to the pre-migration code (checked exact numeric values and dtypes). Not migrated here (belongs to the 3 follow-on transformer branches): ArbitraryOutlierCapper.fit()/transform(), Winsoriser's add_indicators branch (pd.concat), and OutlierTrimmer.transform() (its own .le/.ge/.loc row-filtering, which doesn't go through BaseOutlier._transform at all) all still import pandas directly. Existing tests in tests/test_outliers were left pandas-only rather than parametrized over polars, since they exercise those still-pandas-only subclasses, not BaseOutlier/ WinsorizerBase directly - parametrizing them now would fail on reasons unrelated to this file. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
fit() only builds dicts from user input and validates variables/dtypes via check_numerical_variables (already narwhals-generic) - no numeric computation, so nothing to branch on there. The only pandas-specific lines were the feature_names_in_ assignment (X.columns.to_list(), a pandas-Index method), replaced with the same is_pandas-guarded pattern WinsorizerBase.fit() already uses (list(X.columns) for pandas, nw.from_native(X).columns - already list[str] - otherwise). transform() was already dataframe-agnostic via BaseOutlier._transform(); only its type hints changed (pd.DataFrame -> IntoDataFrame). Benchmarked fit+transform end-to-end at 10k/50k/100k rows x 1/2/10 columns: pandas-native (pre-migration) vs the migrated code on pandas were within noise of each other (~0.9-1.1x), and polars ran 2-4x faster than pandas on both. No pandas/polars branch needed - merged single path, consistent with the is_pandas-only-for-.columns precedent already set in WinsorizerBase. Confirmed the module needs zero pandas: reloaded artbitrary.py in isolation with sys.modules["pandas"] = None (simulating an uninstalled pandas) and ran fit/transform end-to-end on a polars frame - works, and int64 stays int64 for a same-dtype capping dict (the class docstring's own x1 example). Found, while doing so, a real dtype-preservation bug in the already- merged BaseOutlier._transform() (base_outlier.py, commit 71bf7cf on this branch's base) that predates this migration and is not introduced here: when a capping-dict spans columns of different dtypes that land in the same bound-group (e.g. max_capping_dict={"age": 50, "fare": 200} with age int64 and fare float64 - both "right_only"), the group's columns are stacked into one 2D array via to_numpy() before np.clip, which forces a common dtype and upcasts age to float64. The pre- narwhals code (verified against 71bf7cf^) clipped each column independently (X[feature] = X[feature].clip(...)), so int columns never picked up a neighboring float column's dtype. Confirmed this reproduces identically on both pandas and polars (same merged code path) and is untouched by this commit - it lives in base_outlier.py, shared with Winsoriser/OutlierTrimmer, out of this file's scope. Flagged separately rather than fixed here. Rewrote test_arbitrary_capper.py to one parametrized test per behavior over pd.DataFrame/pl.DataFrame (previously pandas-only), using nw.from_native(...).to_dict(as_series=False) for backend-agnostic assertions in place of pd.testing.assert_frame_equal, following the same pattern used for ReciprocalTransformer/ArcsinTransformer. Added a verified "With polars" section to the docs (float dtypes throughout, to sidestep the dtype-upcast issue above rather than put an unexplained surprise in a user-facing example); left the pre-existing pandas Titanic walkthrough untouched - no network access in this environment to re-verify the fetch_openml/CSV-backed output. Verified: tests/test_outliers full suite - 88 passed (up from 83, all 5 new instances are the added polars parametrizations), same 3 pre-existing check_estimator failures as the pre-migration baseline (numpy-array input, unrelated to this change). flake8 and mypy clean. sphinx -W build clean (only the pre-existing linkcode_resolve warning). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Migrates
ArbitraryOutlierCapperto narwhals with polars support.fit()only builds dicts from user input and validates viacheck_numerical_variables(already narwhals-generic) — no numeric computation. The only pandas-specific line wasfeature_names_in_ = X.columns.to_list(), replaced with the sameis_pandas-guarded patternWinsorizerBase.fit()already uses.transform()was already dataframe-agnostic viaBaseOutlier._transform(); only type hints changed (pd.DataFrame→IntoDataFrame).Merge vs split: benchmarked fit+transform end-to-end at 10k/50k/100k rows × 1/2/10 cols. pandas-native (pre-migration) vs migrated-on-pandas were within noise (~0.9–1.1x); polars ran 2–4x faster on both. No split — single merged path. Confirmed the module needs zero pandas (reloaded with
sys.modules["pandas"] = None, ran fit/transform on polars; int64 stays int64 for a same-dtype capping dict).Known pre-existing bug (flagged, not fixed here): in the already-merged
BaseOutlier._transform(), when a capping dict spans columns of different dtypes that land in the same bound-group (e.g.max_capping_dict={"age": 50, "fare": 200}, age int64 / fare float64, both "right only"), the group's columns are stacked into one 2D array viato_numpy()beforenp.clip, forcing a common dtype and upcasting age to float64. The pre-narwhals code clipped each column independently. Reproduces identically on both backends; lives inbase_outlier.py, shared with Winsoriser/OutlierTrimmer, out of this file's scope.Tests rewritten to one parametrized test per behaviour over
pd.DataFrame/pl.DataFrame, usingnw.from_native(...).to_dict(as_series=False)for backend-agnostic assertions. Docs "With polars" section added (float dtypes throughout, to sidestep the dtype-upcast issue above); pandas Titanic walkthrough untouched (no network in sandbox).Verified:
tests/test_outliers— 88 passed (up from 83), same 3 pre-existingcheck_estimatorfailures. flake8 / mypy clean, sphinx -W clean.Stacked on
narwhals-outliers-base(its own PR). Until that merges this PR's diff also contains the sharedBaseOutlier/WinsorizerBasecommit; review that one first.