From 49863eb95544a89ffa7f40b411944a845877455d Mon Sep 17 00:00:00 2001
From: ChrisW09 <50968720+ChrisW09@users.noreply.github.com>
Date: Mon, 27 Jul 2026 22:45:58 +0200
Subject: [PATCH] fix(preprocessor): raise a clear error on duplicate column
 names

Fitting on a DataFrame with a repeated column label failed with

    AttributeError: 'DataFrame' object has no attribute 'dtype'

from inside ``_detect_column_types``, with nothing pointing at the cause:
``X[col]`` returns a DataFrame rather than a Series for a duplicated label, so
the dtype inspection has nothing to read. (``nunique()`` on the line above
happens to work on a frame, which is why it surfaced one line later.)

Check for duplicates up front and raise ``PretabDataError`` naming them. The
ColumnTransformer this builds keys its transformers by column name, so
duplicates could not be routed unambiguously even if detection coped with them
-- rejecting them early is the honest outcome.

Closes #37

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---
 pretab/preprocessor.py   | 13 +++++++++++++
 tests/test_exceptions.py | 33 +++++++++++++++++++++++++++++++++
 2 files changed, 46 insertions(+)

diff --git a/pretab/preprocessor.py b/pretab/preprocessor.py
index fcf228a..f11bbc0 100644
--- a/pretab/preprocessor.py
+++ b/pretab/preprocessor.py
@@ -9,6 +9,7 @@
 
 from .core.exceptions import (
     IncompatibleParamsError,
+    PretabDataError,
     invalid_param_error,
 )
 from .core.logging import configure_logging, get_logger
@@ -285,6 +286,18 @@ def _detect_column_types(self, X):
         elif isinstance(X, np.ndarray):
             X = pd.DataFrame(X, columns=[f"feature_{i}" for i in range(X.shape[1])])
 
+        # ``X[col]`` returns a DataFrame rather than a Series for a duplicated
+        # label, so the dtype inspection below fails with an opaque
+        # ``AttributeError``. The ColumnTransformer this builds also keys its
+        # transformers by column name, so duplicates could not be routed even if
+        # detection coped with them.
+        duplicated = X.columns[X.columns.duplicated()].unique().tolist()
+        if duplicated:
+            raise PretabDataError(
+                f"Duplicate column names are not supported: {duplicated}.\n"
+                "Fix: rename the columns so every name is unique."
+            )
+
         for col in X.columns:
             num_unique_values = X[col].nunique()
             total_samples = len(X[col])
diff --git a/tests/test_exceptions.py b/tests/test_exceptions.py
index 909456a..a9930bb 100644
--- a/tests/test_exceptions.py
+++ b/tests/test_exceptions.py
@@ -249,3 +249,36 @@ def test_preprocessor_unknown_categorical_method():
     with pytest.raises(InvalidParamError) as exc:
         Preprocessor(categorical_method="bogus").fit(df, y)
     assert isinstance(exc.value, ValueError)
+
+
+# --------------------------------------------------------------------------- #
+# Duplicate column names must fail with a clear, actionable error.
+#
+# ``X[col]`` returns a DataFrame rather than a Series for a duplicated label, so
+# detection died on ``.dtype`` with an opaque AttributeError from pandas.
+# --------------------------------------------------------------------------- #
+def test_duplicate_column_names_raise_a_clear_error():
+    rng = np.random.default_rng(0)
+    frame = pd.DataFrame(np.column_stack([rng.normal(size=50)] * 2), columns=pd.Index(["a", "a"]))
+
+    with pytest.raises(PretabDataError, match=r"Duplicate column names are not supported: \['a'\]"):
+        Preprocessor(numerical_method="minmax").fit(frame, rng.normal(size=50))
+
+
+def test_duplicate_column_names_lists_every_offender():
+    rng = np.random.default_rng(0)
+    frame = pd.DataFrame(rng.normal(size=(50, 4)), columns=pd.Index(["a", "a", "b", "b"]))
+
+    with pytest.raises(PretabDataError) as excinfo:
+        Preprocessor(numerical_method="minmax").fit(frame, rng.normal(size=50))
+
+    assert "'a'" in str(excinfo.value) and "'b'" in str(excinfo.value)
+
+
+def test_unique_column_names_are_unaffected():
+    rng = np.random.default_rng(0)
+    frame = pd.DataFrame(rng.normal(size=(50, 2)), columns=pd.Index(["a", "b"]))
+
+    pre = Preprocessor(numerical_method="minmax").fit(frame, rng.normal(size=50))
+
+    assert sorted(pre.output_dims_) == ["a", "b"]