BUG: fix df concat containing mix of localized and non-localized Timestamps

[tsdlovell]

BUG: fix issue with concat of localized timestamps
TST: test concat of dataframes with non-None timezone columns

[jreback]

add the issue number here as well

[tsdlovell]

Created #12467 and added it in test comment

[jreback]

xref to #12396 which is closed by #12403

[jreback]

ping!

[tsdlovell]

Hey Jeff, I've looked into the other failures a little but don't fully understand when/why the expected dims change. It looks like merge (the tests that fail) always wants a 1d array and is failing because we are using atleast_2d on one of our inputs to make our (2-d, 1-d) input combo compatible. Perhaps we should instead be making the 2-d into a 1-d? Is this just playing wack-a-mole?

[jreback]

a bit of a hack, but assure 2d-ness only when we are axis=1 concat.

diff --git a/pandas/tseries/common.py b/pandas/tseries/common.py
index f32e9b9..11a5fdc 100644
--- a/pandas/tseries/common.py
+++ b/pandas/tseries/common.py
@@ -260,8 +260,7 @@ def _concat_compat(to_concat, axis=0, typs=None):
         # if dtype is of datetimetz or timezone
         if x.dtype.kind == _NS_DTYPE.kind:
             if getattr(x, 'tz', None) is not None:
-                x = x.asobject
-                x = np.atleast_2d(x)
+                x = x.asobject.values
             else:
                 shape = x.shape
                 x = tslib.ints_to_pydatetime(x.view(np.int64).ravel())
@@ -272,6 +271,8 @@ def _concat_compat(to_concat, axis=0, typs=None):
             x = tslib.ints_to_pytimedelta(x.view(np.int64).ravel())
             x = x.reshape(shape)

+        if axis == 1:
+            x = np.atleast_2d(x)
         return x

     if typs is None:

[tsdlovell]

That passes the tests that were failing. Unfortunately, can't push from work. Will make that change this evening.

[jreback]

@tsdlovell np.

[tsdlovell]

travis is failing on OSX and py3.5 at the same points

ERROR: test_range_slice_day (pandas.tseries.tests.test_period.TestPeriodIndex)

Traceback (most recent call last):
File "/Users/travis/build/pydata/pandas/pandas/tseries/tests/test_period.py", line 2619, in test_range_slice_day
idx[v:]
File "/Users/travis/build/pydata/pandas/pandas/tseries/base.py", line 176, in getitem
result = getitem(key)

IndexError: failed to coerce slice entry of type str to integer

ERROR: test_range_slice_seconds (pandas.tseries.tests.test_period.TestPeriodIndex)

Traceback (most recent call last):
File "/Users/travis/build/pydata/pandas/pandas/tseries/tests/test_period.py", line 2676, in test_range_slice_seconds
idx[v:]
File "/Users/travis/build/pydata/pandas/pandas/tseries/base.py", line 176, in getitem
result = getitem(key)
IndexError: failed to coerce slice entry of type str to integer

Was working in the env I created a bit ago, but fails in an env I just created, so maybe something to do with change in dependencies. Will look into it.

[jreback]

numpy 1.11 was just updated fully in conda
I am about to push the fixes for this

[jreback]

rebase and should be good to go

[jreback]

@tsdlovell thank you sir!