[3.15] gh-109638: Fix excessive backtracking in csv.Sniffer doublequote detection

Lib/csv.py

@@ -330,8 +330,14 @@ def _guess_quote_and_delimiter(self, data, delimiters):
 
         # if we see an extra quote between delimiters, we've got a
         # double quoted format
+        #
+        # The first character class must exclude the quote character.  If it
+        # can cross a quote, the engine has to try every way of splitting a
+        # run of quotes between the two classes, which is pathological on
+        # quote-heavy input (gh-109638).  The second class must keep admitting
+        # the quote, since that is what lets `"a""b"` match.
         dq_regexp = re.compile(
-                               r"((%(delim)s)|^)\W*%(quote)s[^%(delim)s\n]*%(quote)s[^%(delim)s\n]*%(quote)s\W*((%(delim)s)|$)" % \
+                               r"((%(delim)s)|^)\W*%(quote)s[^%(delim)s%(quote)s\n]*%(quote)s[^%(delim)s\n]*%(quote)s\W*((%(delim)s)|$)" % \
                                {'delim':re.escape(delim), 'quote':quotechar}, re.MULTILINE)
 
 

Lib/test/test_csv.py

@@ -1501,6 +1501,12 @@ def test_delimiters(self):
         self.assertEqual(dialect.delimiter, ',')
         self.assertEqual(dialect.quotechar, '"')
 
+    def test_sniff_regex_backtracking(self):
+        # gh-109638: this artificial sample used to take about ten seconds.
+        sniffer = csv.Sniffer()
+        sample = '"",' * 60 + '"' * 60 + '0' + '"' * 60 + '0'
+        self.assertEqual(sniffer.sniff(sample).delimiter, ',')
+
     def test_doublequote(self):
         sniffer = csv.Sniffer()
         dialect = sniffer.sniff(self.header1)

Misc/NEWS.d/next/Library/2026-07-29-12-33-49.gh-issue-109638.Rq7bXm.rst

@@ -0,0 +1,4 @@
+Fix excessive regular expression backtracking in :class:`csv.Sniffer` on
+quote-heavy samples.  Detecting a doubled-quote dialect could try every way of
+splitting a run of quotes between two character classes, making
+:meth:`~csv.Sniffer.sniff` take seconds on samples of a few hundred characters.