Address merge comments

Author: kprestel

pandas/_libs/parsers.pyx

@@ -50,7 +50,8 @@ from pandas.core.dtypes.common import (
     is_integer_dtype, is_float_dtype,
     is_bool_dtype, is_object_dtype,
     is_datetime64_dtype,
-    pandas_dtype)
+    pandas_dtype, is_extension_array_dtype,
+)
 from pandas.core.arrays import Categorical
 from pandas.core.dtypes.concat import union_categoricals
 import pandas.io.common as icom
@@ -1195,10 +1196,10 @@ cdef class TextReader:
                 na_count = 0
 
             if result is not None and dtype != 'int64':
-                try:
-                    result = result.astype(dtype)
-                except TypeError:
+                if is_extension_array_dtype(dtype):
                     result = result.astype(dtype.numpy_dtype)
+                else:
+                    result = result.astype(dtype)
 
             return result, na_count
 
@@ -1207,10 +1208,10 @@ cdef class TextReader:
                                            na_filter, na_hashset, na_flist)
 
             if result is not None and dtype != 'float64':
-                try:
-                    result = result.astype(dtype)
-                except TypeError:
+                if is_extension_array_dtype(dtype):
                     result = result.astype(dtype.numpy_dtype)
+                else:
+                    result = result.astype(dtype)
             return result, na_count
 
         elif is_bool_dtype(dtype):