ENH: raise a more helpful error message when merging frames with duplicate columns. close #2649

Author: wesm

RELEASE.rst

@@ -101,6 +101,8 @@ pandas 0.10.1
   - Fix setitem on a Series with a boolean key and a non-scalar as value (GH2686_)
   - Box datetime64 values in Series.apply/map (GH2627_, GH2689_)
   - Upconvert datetime + datetime64 values when concatenating frames (GH2624_)
+  - Raise a more helpful error message in merge operations when one DataFrame
+    has duplicate columns (GH2649_)
 
 **API Changes**
 
@@ -124,10 +126,11 @@ pandas 0.10.1
 .. _GH2624: https://github.com/pydata/pandas/issues/2624
 .. _GH2625: https://github.com/pydata/pandas/issues/2625
 .. _GH2627: https://github.com/pydata/pandas/issues/2627
-.. _GH2643: https://github.com/pydata/pandas/issues/2643
 .. _GH2631: https://github.com/pydata/pandas/issues/2631
 .. _GH2633: https://github.com/pydata/pandas/issues/2633
 .. _GH2637: https://github.com/pydata/pandas/issues/2637
+.. _GH2643: https://github.com/pydata/pandas/issues/2643
+.. _GH2649: https://github.com/pydata/pandas/issues/2649
 .. _GH2668: https://github.com/pydata/pandas/issues/2668
 .. _GH2689: https://github.com/pydata/pandas/issues/2689
 .. _GH2690: https://github.com/pydata/pandas/issues/2690

pandas/tools/merge.py

@@ -378,6 +378,14 @@ def _validate_specification(self):
                 if self.left_on is None:
                     raise MergeError('Must pass left_on or left_index=True')
             else:
+                if not self.left.columns.is_unique:
+                    raise MergeError("Left data columns not unique: %s"
+                                     % repr(self.left.columns))
+
+                if not self.right.columns.is_unique:
+                    raise MergeError("Right data columns not unique: %s"
+                                     % repr(self.right.columns))
+
                 # use the common columns
                 common_cols = self.left.columns.intersection(
                     self.right.columns)

pandas/tools/tests/test_merge.py

@@ -713,6 +713,19 @@ def test_merge_nosort(self):
 
         self.assert_((df.var3.unique() == result.var3.unique()).all())
 
+    def test_overlapping_columns_error_message(self):
+        # #2649
+        df = DataFrame({'key': [1, 2, 3],
+                        'v1': [4, 5, 6],
+                        'v2': [7, 8, 9]})
+        df2 = DataFrame({'key': [1, 2, 3],
+                         'v1': [4, 5, 6],
+                         'v2': [7, 8, 9]})
+
+        df.columns = ['key', 'foo', 'foo']
+        df2.columns = ['key', 'bar', 'bar']
+
+        self.assertRaises(Exception, merge, df, df2)
 
 def _check_merge(x, y):
     for how in ['inner', 'left', 'outer']: