Detect a doubled quote at a record boundary

Normalize the record separators in the sample, so that ^ and $ match
at every record boundary: a doubled quote in the last field of a CRLF
record was not detected, and a match could cross a CR separated record.

Skip the detection if there is no delimiter, because an empty one
matches at every position and makes the search quadratic.

Author: serhiy-storchaka

Lib/csv.py

@@ -243,6 +243,8 @@ def sniff(self, sample, delimiters=None):
         Returns a dialect (or None) corresponding to the sample
         """
 
+        sample = sample.replace('\r\n', '\n').replace('\r', '\n')
+
         quotechar, doublequote, delimiter, skipinitialspace = \
                    self._guess_quote_and_delimiter(sample, delimiters)
         if not delimiter:
@@ -331,14 +333,17 @@ def _guess_quote_and_delimiter(self, data, delimiters):
         # A doubled quote character inside a quoted field means
         # a double quoted format.  Match whole fields, so that a match
         # cannot slide across field boundaries.
-        dq_regexp = re.compile(
-                r"(?:(?<=%(delim)s)|^) *+%(quote)s"           # ,"
-                r"((?:%(quote)s%(quote)s|[^%(quote)s]++)*+)"  # the body
-                r"%(quote)s(?:%(delim)s|$)"                   # ",
-                % {'delim': re.escape(delim), 'quote': quotechar},
-                re.MULTILINE)
-        dquotechar = quotechar * 2
-        doublequote = any(dquotechar in m[1] for m in dq_regexp.finditer(data))
+        doublequote = False
+        if delim:
+            dq_regexp = re.compile(
+                    r"(?:(?<=%(delim)s)|^) *+%(quote)s"           # ,"
+                    r"((?:%(quote)s%(quote)s|[^%(quote)s]++)*+)"  # the body
+                    r"%(quote)s(?:%(delim)s|$)"                   # ",
+                    % {'delim': re.escape(delim), 'quote': quotechar},
+                    re.MULTILINE)
+            dquotechar = quotechar * 2
+            doublequote = any(dquotechar in m[1]
+                              for m in dq_regexp.finditer(data))
 
         return (quotechar, doublequote, delim, skipinitialspace)
 

Lib/test/test_csv.py

@@ -1582,6 +1582,23 @@ def test_sniff_doublequote_across_fields(self):
         self.assertEqual(next(csv.reader(StringIO(sample), dialect)),
                          [',', '', ','])
 
+    def test_sniff_doublequote_record_separators(self):
+        # The record separator ends a field as a delimiter does.
+        sniffer = csv.Sniffer()
+        for sep in '\n', '\r\n', '\r':
+            with self.subTest(sep=sep):
+                sample = ('x,"a""b"' + sep + 'y,"c"' + sep) * 2
+                self.assertIs(sniffer.sniff(sample).doublequote, True)
+                sample = ('"",","' + sep) * 4
+                self.assertIs(sniffer.sniff(sample).doublequote, False)
+
+    def test_sniff_single_column(self):
+        # This sample used to be quadratic.
+        sniffer = csv.Sniffer()
+        sample = '"a"\n' + ' ' * 100000
+        with self.assertRaisesRegex(csv.Error, "Could not determine delimiter"):
+            sniffer.sniff(sample, delimiters=',;')
+
 
 class NUL:
     def write(s, *args):