put date parsing after individual column inferencing so integers can be parsed better

Author: changhiskhan

pandas/io/parsers.py

@@ -807,12 +807,12 @@ def get_chunk(self, rows=None):
                 col = self.columns[col]
             data[col] = lib.map_infer(data[col], f)
 
+        data = _convert_to_ndarrays(data, self.na_values, self.verbose)
+
         columns = list(self.columns)
         if self.parse_dates is not None:
             data, columns = self._process_date_conversion(data)
 
-        data = _convert_to_ndarrays(data, self.na_values, self.verbose)
-
         df = DataFrame(data=data, columns=columns, index=index)
         if self._has_complex_date_col and self.index_col is not None:
             if not self._name_processed:
@@ -1129,7 +1129,9 @@ def _concat_date_cols(date_cols):
         return date_cols[0]
 
     # stripped = [map(str.strip, x) for x in date_cols]
-    return np.array([' '.join(x) for x in zip(*date_cols)], dtype=object)
+    rs = np.array([' '.join([str(y) for y in x])
+                   for x in zip(*date_cols)], dtype=object)
+    return rs
 
 
 class FixedWidthReader(object):

pandas/io/tests/test_date_converters.py

@@ -57,7 +57,7 @@ def test_parse_date_fields(self):
         expected = np.array([datetime(2007, 1, 3), datetime(2008, 2, 4)])
         self.assert_((result == expected).all())
 
-        data = "year, month, day, a\n 2001, 01, 10, 10.\n 2001, 02, 1, 11."
+        data = "year, month, day, a\n 2001 , 01 , 10 , 10.\n 2001 , 02 , 1 , 11."
         datecols = {'ymd': [0, 1, 2]}
         df = read_table(StringIO(data), sep=',', header=0,
                         parse_dates=datecols,