eval-toolkit 0.27.1__tar.gz → 0.27.2__tar.gz

{eval_toolkit-0.27.1 → eval_toolkit-0.27.2}/CHANGELOG.md

@@ -7,6 +7,40 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.27.2] — 2026-05-15 — fix base-install pandas import
+
+Base install of `eval-toolkit` (no extras) was broken in 0.27.1: every
+attempt to `from eval_toolkit import evaluate` raised
+`ModuleNotFoundError: No module named 'pandas'` because four modules
+imported pandas at top level despite pandas being declared in the
+`[dataframe]` optional extra. This patch restores the documented
+contract: base install gives pure-numpy primitives; pandas is only
+needed for the `[dataframe]` / `[parquet]` capabilities.
+
+### Fixed
+
+- `src/eval_toolkit/harness.py`: moved `import pandas as pd` under
+  `if TYPE_CHECKING:` (annotation-only use at `EvalSlice.df`).
+- `src/eval_toolkit/loaders.py`: moved top-level pandas import under
+  `if TYPE_CHECKING:` for the `DataFrameLoader.df` annotation; added a
+  function-local `import pandas as pd` inside
+  `ParquetGlobLoader.load_splits` where `pd.read_parquet` / `pd.concat`
+  are actually called. Calling the parquet path still requires pandas
+  (via the `[parquet]` extra) — that's documented behavior, not a
+  regression.
+- `src/eval_toolkit/leakage.py`: removed dead top-level pandas import.
+  Doctest examples already `>>> import pandas as pd` themselves.
+- `src/eval_toolkit/splits.py`: removed dead top-level pandas import
+  and unused `Sequence` import. Doctest example imports pandas itself.
+
+### Added
+
+- `.github/workflows/ci.yml`: new `test-base-install` job. Creates a
+  fresh Py3.13 venv, `pip install .` with no extras, then exercises
+  every public top-level import. Sanity check fails if pandas leaks
+  into the base venv. Regression guard against this entire class of
+  bug. Closes #13.
+
 ## [0.27.1] — 2026-05-15 — first PyPI release (Trusted Publishing)
 
 Functionally identical to v0.27.0; bumped to 0.27.1 because the `v0.27.0`

{eval_toolkit-0.27.1 → eval_toolkit-0.27.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: eval-toolkit
-Version: 0.27.1
+Version: 0.27.2
 Summary: Reusable evaluation contracts for binary classification: metrics, bootstrap CIs, calibration, artifacts, and evidence gates.
 Project-URL: Homepage, https://github.com/brandon-behring/eval-toolkit
 Project-URL: Repository, https://github.com/brandon-behring/eval-toolkit.git

{eval_toolkit-0.27.1 → eval_toolkit-0.27.2}/src/eval_toolkit/_version.py

@@ -2,4 +2,4 @@
 
 __all__ = ["__version__"]
 
-__version__ = "0.27.1"
+__version__ = "0.27.2"

{eval_toolkit-0.27.1 → eval_toolkit-0.27.2}/src/eval_toolkit/harness.py

@@ -40,7 +40,6 @@ from pathlib import Path
 from typing import TYPE_CHECKING, Final, Literal, cast
 
 import numpy as np
-import pandas as pd
 
 from eval_toolkit.artifacts import (
     error_metric,
@@ -66,6 +65,8 @@ from eval_toolkit.protocols import Scorer, SliceAwareScorer
 from eval_toolkit.thresholds import TargetFPRSelector
 
 if TYPE_CHECKING:
+    import pandas as pd
+
     from eval_toolkit.leakage import LeakageCheck
     from eval_toolkit.splits import Splitter
 

{eval_toolkit-0.27.1 → eval_toolkit-0.27.2}/src/eval_toolkit/leakage.py

@@ -42,7 +42,6 @@ from dataclasses import dataclass, field
 from typing import Literal, Protocol, runtime_checkable
 
 import numpy as np
-import pandas as pd
 
 from eval_toolkit.harness import EvalSlice
 from eval_toolkit.text_dedup import (
@@ -1048,5 +1047,6 @@ def run_leakage_checks(
     return LeakageReport(findings=findings)
 
 
-# Suppress unused-import warnings: pd / np are referenced in Examples blocks.
-_ = (pd, np)
+# Suppress unused-import warning: np is referenced in Examples blocks.
+# pandas is imported lazily inside the doctest itself (see Examples above).
+_ = (np,)

{eval_toolkit-0.27.1 → eval_toolkit-0.27.2}/src/eval_toolkit/loaders.py

@@ -32,13 +32,14 @@ import glob as _glob
 from collections.abc import Mapping, Sequence
 from dataclasses import dataclass
 from pathlib import Path
-from typing import Any, Protocol, cast, runtime_checkable
-
-import pandas as pd
+from typing import TYPE_CHECKING, Any, Protocol, cast, runtime_checkable
 
 from eval_toolkit.harness import EvalSlice
 from eval_toolkit.provenance import file_sha256
 
+if TYPE_CHECKING:
+    import pandas as pd
+
 __all__ = [
     "DataFrameLoader",
     "DatasetLoader",
@@ -269,6 +270,8 @@ class ParquetGlobLoader:
             If any split's loaded DataFrame is missing ``feature_col`` or
             ``label_col``.
         """
+        import pandas as pd  # function-local: pandas is in the [parquet]/[dataframe] extra
+
         files_by_split = self._resolve_files()
         out: dict[str, EvalSlice] = {}
         for split_name, files in files_by_split.items():

{eval_toolkit-0.27.1 → eval_toolkit-0.27.2}/src/eval_toolkit/splits.py

@@ -18,12 +18,11 @@ can be fed into the harness directly.
 
 from __future__ import annotations
 
-from collections.abc import Iterator, Sequence
+from collections.abc import Iterator
 from dataclasses import dataclass
 from typing import Protocol, runtime_checkable
 
 import numpy as np
-import pandas as pd
 from sklearn.model_selection import (
     GroupKFold,
     StratifiedKFold,
@@ -514,7 +513,3 @@ def iter_folds_with_pool(
         # PoolBuilder's keys (train, val, possibly more) take precedence;
         # test is reattached from the Splitter.
         yield {**built, "test": test}
-
-
-# Suppress unused-import warnings: pd / Sequence are referenced indirectly.
-_ = (pd, Sequence)