BUG: Fix bugs in stata

Fix incorrect skipping in strl writer
Fix incorrect byteorder when exporting bigendian
Fix incorrect byteorder parsing when importing bigendian
Improve test coverage for errors

Author: bashtage

pandas/io/stata.py

@@ -1050,7 +1050,7 @@ def _read_new_header(self, first_char):
         if self.format_version not in [117, 118]:
             raise ValueError(_version_error)
         self.path_or_buf.read(21)  # </release><byteorder>
-        self.byteorder = self.path_or_buf.read(3) == "MSF" and '>' or '<'
+        self.byteorder = self.path_or_buf.read(3) == b'MSF' and '>' or '<'
         self.path_or_buf.read(15)  # </byteorder><K>
         self.nvar = struct.unpack(self.byteorder + 'H',
                                   self.path_or_buf.read(2))[0]
@@ -1824,9 +1824,7 @@ def _dtype_to_stata_type(dtype, column):
     type inserted.
     """
     # TODO: expand to handle datetime to integer conversion
-    if dtype.type == np.string_:
-        return dtype.itemsize
-    elif dtype.type == np.object_:  # try to coerce it to the biggest string
+    if dtype.type == np.object_:  # try to coerce it to the biggest string
         # not memory efficient, what else could we
         # do?
         itemsize = max_len_string_array(_ensure_object(column.values))
@@ -2347,25 +2345,30 @@ def _prepare_data(self):
         data = self._convert_strls(data)
 
         # 3. Convert bad string data to '' and pad to correct length
-        dtype = []
+        dtypes = []
         data_cols = []
         has_strings = False
+        native_byteorder = self._byteorder == _set_endianness(sys.byteorder)
         for i, col in enumerate(data):
             typ = typlist[i]
             if typ <= self._max_string_length:
                 has_strings = True
                 data[col] = data[col].fillna('').apply(_pad_bytes, args=(typ,))
                 stype = 'S%d' % typ
-                dtype.append(('c' + str(i), stype))
+                dtypes.append(('c' + str(i), stype))
                 string = data[col].str.encode(self._encoding)
                 data_cols.append(string.values.astype(stype))
             else:
-                dtype.append(('c' + str(i), data[col].dtype))
-                data_cols.append(data[col].values)
-        dtype = np.dtype(dtype)
-
-        if has_strings:
-            self.data = np.fromiter(zip(*data_cols), dtype=dtype)
+                values = data[col].values
+                dtype = data[col].dtype
+                if not native_byteorder:
+                    dtype = dtype.newbyteorder(self._byteorder)
+                dtypes.append(('c' + str(i), dtype))
+                data_cols.append(values)
+        dtypes = np.dtype(dtypes)
+
+        if has_strings or not native_byteorder:
+            self.data = np.fromiter(zip(*data_cols), dtype=dtypes)
         else:
             self.data = data.to_records(index=False)
 
@@ -2403,9 +2406,7 @@ def _dtype_to_stata_type_117(dtype, column, force_strl):
     # TODO: expand to handle datetime to integer conversion
     if force_strl:
         return 32768
-    if dtype.type == np.string_:
-        return chr(dtype.itemsize)
-    elif dtype.type == np.object_:  # try to coerce it to the biggest string
+    if dtype.type == np.object_:  # try to coerce it to the biggest string
         # not memory efficient, what else could we
         # do?
         itemsize = max_len_string_array(_ensure_object(column.values))
@@ -2599,7 +2600,7 @@ def generate_blob(self, table):
         o_type = self._byteorder + self._gso_o_type
         len_type = self._byteorder + 'I'
         for strl, vo in table.items():
-            if vo == 0:
+            if vo == (0, 0):
                 continue
             v, o = vo
             # GSO

pandas/tests/io/test_stata.py

@@ -499,6 +499,15 @@ def test_timestamp_and_label(self, version):
                 assert reader.time_stamp == '29 Feb 2000 14:21'
                 assert reader.data_label == data_label
 
+    @pytest.mark.parametrize('version', [114, 117])
+    def test_invalid_timestamp(self, version):
+        original = DataFrame([(1,)], columns=['variable'])
+        time_stamp = '01 Jan 2000, 00:00:00'
+        with tm.ensure_clean() as path:
+            with pytest.raises(ValueError):
+                original.to_stata(path, time_stamp=time_stamp,
+                                  version=version)
+
     def test_numeric_column_names(self):
         original = DataFrame(np.reshape(np.arange(25.0), (5, 5)))
         original.index.name = 'index'
@@ -639,7 +648,8 @@ def test_write_missing_strings(self):
                                   expected)
 
     @pytest.mark.parametrize('version', [114, 117])
-    def test_bool_uint(self, version):
+    @pytest.mark.parametrize('byteorder', ['>', '<'])
+    def test_bool_uint(self, byteorder, version):
         s0 = Series([0, 1, True], dtype=np.bool)
         s1 = Series([0, 1, 100], dtype=np.uint8)
         s2 = Series([0, 1, 255], dtype=np.uint8)
@@ -658,7 +668,7 @@ def test_bool_uint(self, version):
             expected[c] = expected[c].astype(t)
 
         with tm.ensure_clean() as path:
-            original.to_stata(path, version=version)
+            original.to_stata(path, byteorder=byteorder, version=version)
             written_and_read_again = self.read_dta(path)
             written_and_read_again = written_and_read_again.set_index('index')
             tm.assert_frame_equal(written_and_read_again, expected)
@@ -1173,6 +1183,29 @@ def test_write_variable_labels(self, version):
                 read_labels = sr.variable_labels()
             assert read_labels == variable_labels
 
+    @pytest.mark.parametrize('version', [114, 117])
+    def test_invalid_variable_labels(self, version):
+        original = pd.DataFrame({'a': [1, 2, 3, 4],
+                                 'b': [1.0, 3.0, 27.0, 81.0],
+                                 'c': ['Atlanta', 'Birmingham',
+                                       'Cincinnati', 'Detroit']})
+        original.index.name = 'index'
+        variable_labels = {'a': 'very long' * 10,
+                           'b': 'City Exponent',
+                           'c': 'City'}
+        with tm.ensure_clean() as path:
+            with pytest.raises(ValueError):
+                original.to_stata(path,
+                                  variable_labels=variable_labels,
+                                  version=version)
+
+        variable_labels['a'] = u'invalid character Œ'
+        with tm.ensure_clean() as path:
+            with pytest.raises(ValueError):
+                original.to_stata(path,
+                                  variable_labels=variable_labels,
+                                  version=version)
+
     def test_write_variable_label_errors(self):
         original = pd.DataFrame({'a': [1, 2, 3, 4],
                                  'b': [1.0, 3.0, 27.0, 81.0],
@@ -1220,6 +1253,12 @@ def test_default_date_conversion(self):
             direct = read_stata(path, convert_dates=True)
             tm.assert_frame_equal(reread, direct)
 
+            original.to_stata(path,
+                              write_index=False,
+                              convert_dates={2: 'tc'})
+            direct = read_stata(path, convert_dates=True)
+            tm.assert_frame_equal(reread, direct)
+
     def test_unsupported_type(self):
         original = pd.DataFrame({'a': [1 + 2j, 2 + 4j]})
 
@@ -1419,3 +1458,17 @@ def test_convert_strl_name_swap(self):
                 reread.columns = original.columns
                 tm.assert_frame_equal(reread, original,
                                       check_index_type=False)
+
+    def test_invalid_date_conversion(self):
+        # GH 12259
+        dates = [dt.datetime(1999, 12, 31, 12, 12, 12, 12000),
+                 dt.datetime(2012, 12, 21, 12, 21, 12, 21000),
+                 dt.datetime(1776, 7, 4, 7, 4, 7, 4000)]
+        original = pd.DataFrame({'nums': [1.0, 2.0, 3.0],
+                                 'strs': ['apple', 'banana', 'cherry'],
+                                 'dates': dates})
+
+        with tm.ensure_clean() as path:
+            with pytest.raises(ValueError):
+                original.to_stata(path,
+                                  convert_dates={'wrong_name': 'tc'})