MAINT: Nicer error msg for NULL byte in read_csv

Provides a nicer error message for the Python engine
in read_csv when the data contains a NULL byte.

Closes pandas-devgh-2741.

Author: gfyoung

pandas/io/parsers.py

@@ -2190,7 +2190,17 @@ def _next_line(self):
                 next(self.data)
 
             while True:
-                orig_line = next(self.data)
+                try:
+                    orig_line = next(self.data)
+                except csv.Error as e:
+                    if 'NULL byte' in str(e):
+                        raise csv.Error(
+                            'NULL byte detected. This byte '
+                            'cannot be processed in Python\'s '
+                            'native csv library at the moment, '
+                            'so please pass in engine=\'c\' instead.')
+                    else:
+                        raise
                 line = self._check_comments([orig_line])[0]
                 self.pos += 1
                 if (not self.skip_blank_lines and

pandas/io/tests/parser/common.py

@@ -1503,3 +1503,19 @@ def test_memory_map(self):
 
         out = self.read_csv(mmap_file, memory_map=True)
         tm.assert_frame_equal(out, expected)
+
+    def test_null_byte_char(self):
+        # see gh-2741
+        data = '\x00,foo'
+        cols = ['a', 'b']
+
+        expected = DataFrame([[np.nan, 'foo']],
+                             columns=cols)
+
+        if self.engine == 'c':
+            out = self.read_csv(StringIO(data), names=cols)
+            tm.assert_frame_equal(out, expected)
+        else:
+            msg = "NULL byte detected"
+            with tm.assertRaisesRegexp(csv.Error, msg):
+                self.read_csv(StringIO(data), names=cols)