Fix left join turning into outer join

doc/source/whatsnew/v0.23.0.txt

@@ -772,7 +772,7 @@ Reshaping
 - Bug in timezone comparisons, manifesting as a conversion of the index to UTC in ``.concat()`` (:issue:`18523`)
 - Bug in :func:`concat` when concatting sparse and dense series it returns only a ``SparseDataFrame``. Should be a ``DataFrame``. (:issue:`18914`, :issue:`18686`, and :issue:`16874`)
 - Improved error message for :func:`DataFrame.merge` when there is no common merge key (:issue:`19427`)
--
+- Bug in :func:`DataFrame.join` which does an *outer* instead of a *left* join when being called with multiple DataFrames and some have non-unique indices (:issue:`19624`)
 
 Other
 ^^^^^

pandas/core/frame.py

@@ -5328,18 +5328,17 @@ def _join_compat(self, other, on=None, how='left', lsuffix='', rsuffix='',
                 raise ValueError('Joining multiple DataFrames only supported'
                                  ' for joining on index')
 
-            # join indexes only using concat
-            if how == 'left':
-                how = 'outer'
-                join_axes = [self.index]
-            else:
-                join_axes = None
-
             frames = [self] + list(other)
 
             can_concat = all(df.index.is_unique for df in frames)
 
+            # join indexes only using concat
             if can_concat:
+                if how == 'left':
+                    how = 'outer'
+                    join_axes = [self.index]
+                else:
+                    join_axes = None
                 return concat(frames, axis=1, join=how, join_axes=join_axes,
                               verify_integrity=True)
 

pandas/tests/frame/test_join.py

@@ -165,3 +165,20 @@ def test_join_period_index(frame_with_period_index):
         index=frame_with_period_index.index)
 
     tm.assert_frame_equal(joined, expected)
+
+
+def test_join_left_sequence_non_unique_index():
+    # https://github.com/pandas-dev/pandas/issues/19607
+    df1 = DataFrame({'a': [0, 10, 20]}, index=[1, 2, 3])
+    df2 = DataFrame({'b': [100, 200, 300]}, index=[4, 3, 2])
+    df3 = DataFrame({'c': [400, 500, 600]}, index=[2, 2, 4])
+
+    joined = df1.join([df2, df3], how='left')
+
+    expected = DataFrame({
+        'a': [0, 10, 10, 20],
+        'b': [np.nan, 300, 300, 200],
+        'c': [np.nan, 400, 500, np.nan]
+    }, index=[1, 2, 2, 3])
+
+    tm.assert_frame_equal(joined, expected)