Merge pull request pandas-dev#5101 from cpcloud/fix-period-datetime-join-stack-overflow

BUG: fix PeriodIndex join with DatetimeIndex stack overflow

Author: cpcloud

doc/source/release.rst

@@ -553,6 +553,8 @@ Bug Fixes
     passed ``index_col=0`` (:issue:`5066`).
   - Fixed a bug where :func:`~pandas.read_html` was incorrectly infering the
     type of headers (:issue:`5048`).
+  - Fixed a bug where ``DatetimeIndex`` joins with ``PeriodIndex`` caused a
+    stack overflow (:issue:`3899`).
 
 
 pandas 0.12.0

pandas/tseries/period.py

@@ -12,8 +12,8 @@
 import pandas.tseries.frequencies as _freq_mod
 
 import pandas.core.common as com
-from pandas.core.common import (isnull, _NS_DTYPE, _INT64_DTYPE,
-                                _maybe_box, _values_from_object)
+from pandas.core.common import (isnull, _INT64_DTYPE, _maybe_box,
+                                _values_from_object)
 from pandas import compat
 from pandas.lib import Timestamp
 import pandas.lib as lib
@@ -712,13 +712,10 @@ def _array_values(self):
     def astype(self, dtype):
         dtype = np.dtype(dtype)
         if dtype == np.object_:
-            result = np.empty(len(self), dtype=dtype)
-            result[:] = [x for x in self]
-            return result
+            return Index(np.array(list(self), dtype), dtype)
         elif dtype == _INT64_DTYPE:
-            return self.values.copy()
-        else:  # pragma: no cover
-            raise ValueError('Cannot cast PeriodIndex to dtype %s' % dtype)
+            return Index(self.values, dtype)
+        raise ValueError('Cannot cast PeriodIndex to dtype %s' % dtype)
 
     def __iter__(self):
         for val in self.values:

pandas/tseries/tests/test_period.py

@@ -22,8 +22,8 @@
 import pandas.core.datetools as datetools
 import pandas as pd
 import numpy as np
-from pandas.compat import range, lrange, lmap, map, zip
-randn = np.random.randn
+from numpy.random import randn
+from pandas.compat import range, lrange, lmap, zip
 
 from pandas import Series, TimeSeries, DataFrame
 from pandas.util.testing import(assert_series_equal, assert_almost_equal,
@@ -1207,7 +1207,6 @@ def test_is_(self):
         self.assertFalse(index.is_(index - 2))
         self.assertFalse(index.is_(index - 0))
 
-
     def test_comp_period(self):
         idx = period_range('2007-01', periods=20, freq='M')
 
@@ -1913,6 +1912,17 @@ def test_join_self(self):
             res = index.join(index, how=kind)
             self.assert_(index is res)
 
+    def test_join_does_not_recur(self):
+        df = tm.makeCustomDataframe(3, 2, data_gen_f=lambda *args:
+                                    np.random.randint(2), c_idx_type='p',
+                                    r_idx_type='dt')
+        s = df.iloc[:2, 0]
+
+        res = s.index.join(df.columns, how='outer')
+        expected = Index([s.index[0], s.index[1],
+                          df.columns[0], df.columns[1]], object)
+        tm.assert_index_equal(res, expected)
+
     def test_align_series(self):
         rng = period_range('1/1/2000', '1/1/2010', freq='A')
         ts = Series(np.random.randn(len(rng)), index=rng)
@@ -2185,15 +2195,15 @@ def test_minutely(self):
 
     def test_secondly(self):
         self._check_freq('S', '1970-01-01')
-        
+
     def test_millisecondly(self):
         self._check_freq('L', '1970-01-01')
 
     def test_microsecondly(self):
         self._check_freq('U', '1970-01-01')
-        
+
     def test_nanosecondly(self):
-        self._check_freq('N', '1970-01-01')        
+        self._check_freq('N', '1970-01-01')
 
     def _check_freq(self, freq, base_date):
         rng = PeriodIndex(start=base_date, periods=10, freq=freq)

pandas/tseries/tests/test_timeseries.py

@@ -31,16 +31,12 @@
 
 import pandas.index as _index
 
-from pandas.compat import(
-    range, long, StringIO, lrange, lmap, map, zip, cPickle as pickle, product
-)
-from pandas import read_pickle
+from pandas.compat import range, long, StringIO, lrange, lmap, zip, product
 import pandas.core.datetools as dt
 from numpy.random import rand
 from numpy.testing import assert_array_equal
 from pandas.util.testing import assert_frame_equal
 import pandas.compat as compat
-from pandas.core.datetools import BDay
 import pandas.core.common as com
 from pandas import concat
 from pandas import _np_version_under1p7
@@ -2064,6 +2060,18 @@ def test_ns_index(self):
         new_index = pd.DatetimeIndex(start=index[0], end=index[-1], freq=index.freq)
         self.assert_index_parameters(new_index)
 
+    def test_join_with_period_index(self):
+        df = tm.makeCustomDataframe(10, 10, data_gen_f=lambda *args:
+                                    np.random.randint(2), c_idx_type='p',
+                                    r_idx_type='dt')
+        s = df.iloc[:5, 0]
+        joins = 'left', 'right', 'inner', 'outer'
+
+        for join in joins:
+            with tm.assertRaisesRegexp(ValueError, 'can only call with other '
+                                       'PeriodIndex-ed objects'):
+                df.columns.join(s.index, how=join)
+
 
 class TestDatetime64(unittest.TestCase):
     """