ENH:Add EA types to read CSV

Closes GH23228

Author: kprestel

pandas/_libs/parsers.pyx

@@ -1207,7 +1207,10 @@ cdef class TextReader:
                 na_count = 0
 
             if result is not None and dtype != 'int64':
-                result = result.astype(dtype)
+                try:
+                    result = result.astype(dtype)
+                except TypeError:
+                    result = result.astype(dtype.numpy_dtype)
 
             return result, na_count
 
@@ -1216,7 +1219,10 @@ cdef class TextReader:
                                            na_filter, na_hashset, na_flist)
 
             if result is not None and dtype != 'float64':
-                result = result.astype(dtype)
+                try:
+                    result = result.astype(dtype)
+                except TypeError:
+                    result = result.astype(dtype.numpy_dtype)
             return result, na_count
 
         elif is_bool_dtype(dtype):

pandas/core/dtypes/common.py

@@ -1963,7 +1963,10 @@ def _get_dtype_type(arr_or_dtype):
     try:
         return arr_or_dtype.dtype.type
     except AttributeError:
-        return type(None)
+        try:
+            return arr_or_dtype.numpy_dtype.type
+        except AttributeError:
+            return type(None)
 
 
 def _get_dtype_from_object(dtype):

pandas/tests/io/parser/common.py

@@ -22,6 +22,7 @@
 
 from pandas.io.common import URLError
 from pandas.io.parsers import TextFileReader, TextParser
+from pandas.core.arrays.integer import Int64Dtype
 
 
 class ParserTests(object):
@@ -1618,3 +1619,24 @@ def test_skip_bad_lines(self):
         val = sys.stderr.getvalue()
         assert 'Skipping line 3' in val
         assert 'Skipping line 5' in val
+
+    def test_buffer_rd_bytes_bad_unicode(self):
+        # Regression test for #22748
+        t = BytesIO(b"\xB0")
+        if PY3:
+            t = TextIOWrapper(t, encoding='ascii', errors='surrogateescape')
+        with pytest.raises(UnicodeError):
+            pd.read_csv(t, encoding='UTF-8')
+
+    def test_EA_types(self):
+        df = pd.DataFrame({'Int': pd.Series([1, 2, 3], dtype='Int64'),
+                           'A': [1, 2, 1]})
+        data = df.to_csv(index=False)
+        result = pd.read_csv(StringIO(data), dtype={'Int': Int64Dtype})
+        assert result is not None
+
+        df = pd.DataFrame({'Int': pd.Series([1, 2, 3], dtype='Int8'),
+                           'A': [1, 2, 1]})
+        data = df.to_csv(index=False)
+        result = pd.read_csv(StringIO(data), dtype={'Int': 'Int8'})
+        assert result is not None