BUG: edge case when reading from postgresl with read_sql_query and datetime with timezone types and a chunksize, #11216

- When we don't specifiy a chunksize we get an object dtype
  which is ok
- We create a propery datetime64[ns, tz] type, but its a
  pytz.FixedOffset(....), which ATM is not really a
  useful/palatable type and is mostly confusing for now.
  In the future could attempt to coerce this to a nice tz,
  e.g. US/Eastern, not sure if this is possible
- Note that this is w/o parse_dates specified

Author: jreback

pandas/io/sql.py

@@ -113,6 +113,14 @@ def _parse_date_columns(data_frame, parse_dates):
             fmt = None
         data_frame[col_name] = _handle_date_column(df_col, format=fmt)
 
+
+    # we want to coerce datetime64_tz dtypes for now
+    # we could in theory do a 'nice' conversion from a FixedOffset tz
+    # GH11216
+    for col_name, df_col in data_frame.iteritems():
+        if com.is_datetime64tz_dtype(df_col):
+            data_frame[col_name] = _handle_date_column(df_col)
+
     return data_frame
 
 
@@ -366,7 +374,7 @@ def read_sql_query(sql, con, index_col=None, coerce_float=True, params=None,
     ----------
     sql : string SQL query or SQLAlchemy Selectable (select or text object)
         to be executed.
-    con : SQLAlchemy connectable(engine/connection) or database string URI 
+    con : SQLAlchemy connectable(engine/connection) or database string URI
         or sqlite3 DBAPI2 connection
         Using SQLAlchemy makes it possible to use any DB supported by that
         library.

pandas/io/tests/test_sql.py

@@ -26,13 +26,15 @@
 import nose
 import warnings
 import numpy as np
+import pandas as pd
 
 from datetime import datetime, date, time
 
 from pandas import DataFrame, Series, Index, MultiIndex, isnull, concat
 from pandas import date_range, to_datetime, to_timedelta, Timestamp
 import pandas.compat as compat
 from pandas.compat import StringIO, range, lrange, string_types
+from pandas.core import common as com
 from pandas.core.datetools import format as date_format
 
 import pandas.io.sql as sql
@@ -1248,6 +1250,39 @@ def test_default_date_load(self):
         self.assertTrue(issubclass(df.DateCol.dtype.type, np.datetime64),
                         "DateCol loaded with incorrect type")
 
+    def test_datetime_with_timezone(self):
+        # edge case that converts postgresql datetime with time zone types
+        # to datetime64[ns,psycopg2.tz.FixedOffsetTimezone..], which is ok
+        # but should be more natural, so coerce to datetime64[ns] for now
+
+        # GH11216
+        df = pd.read_sql_query("select * from types_test_data", self.conn)
+        if not hasattr(df,'DateColWithTz'):
+            raise nose.SkipTest("no column with datetime with time zone")
+
+        # this is parsed on Travis (linux), but not on macosx for some reason
+        # even with the same versions of psycopg2 & sqlalchemy, possibly a Postgrsql server
+        # version difference
+        dtype = df.DateColWithTz.dtype
+        self.assertTrue(com.is_object_dtype(dtype) or com.is_datetime64_dtype(dtype),
+                        "DateCol loaded with incorrect type -> {0}".format(dtype))
+
+        df = pd.read_sql_query("select * from types_test_data", self.conn, parse_dates=['DateColWithTz'])
+        if not hasattr(df,'DateColWithTz'):
+            raise nose.SkipTest("no column with datetime with time zone")
+
+        dtype = df.DateColWithTz.dtype
+        self.assertTrue(com.is_datetime64_dtype(dtype),
+                        "DateCol loaded with incorrect type -> {0}".format(dtype))
+
+        df = pd.concat(list(pd.read_sql_query("select * from types_test_data",
+                                              self.conn,chunksize=1)),ignore_index=True)
+        dtype = df.DateColWithTz.dtype
+        self.assertTrue(com.is_datetime64_dtype(dtype),
+                        "DateCol loaded with incorrect type -> {0}".format(dtype))
+        expected = sql.read_sql_table("types_test_data", self.conn)
+        tm.assert_series_equal(df.DateColWithTz, expected.DateColWithTz)
+
     def test_date_parsing(self):
         # No Parsing
         df = sql.read_sql_table("types_test_data", self.conn)