gh-109638: Keep space-delimiter matching linear

Author: Punisheroot

Lib/csv.py

@@ -352,8 +352,10 @@ def _detect_doublequote(self, data, delimiter, quotechar):
         escaped_quote = re.escape(quotechar)
         values = {'delim': escaped_delimiter, 'quote': escaped_quote}
         if delimiter:
+            # Spaces after a space delimiter are delimiters, not padding.
+            values['space'] = '' if delimiter == ' ' else ' *+'
             candidate = re.compile(
-                    r"(?:%(delim)s|\r|^) *+%(quote)s"
+                    r"(?:%(delim)s|\r|^)%(space)s%(quote)s"
                     r"[^%(quote)s]*+%(quote)s%(quote)s"
                     r"(?:%(quote)s%(quote)s|[^%(quote)s]++)*+"
                     r"%(quote)s(?:%(delim)s|(?=\r)|$)"

Lib/test/test_csv.py

@@ -1507,6 +1507,19 @@ def test_sniff_regex_backtracking(self):
         sample = '"",' * 100 + '"' * 100 + '0' + '"' * 100 + '0'
         self.assertEqual(sniffer.sniff(sample).delimiter, ',')
 
+    def test_sniff_space_delimiter(self):
+        # This sample used to be quadratic.
+        sniffer = csv.Sniffer()
+        sample = '"a" "b"\n' + ' ' * 100_000
+        dialect = sniffer.sniff(sample)
+        self.assertEqual(dialect.delimiter, ' ')
+        self.assertIs(dialect.doublequote, False)
+
+        # A quoted field can still start after multiple space delimiters.
+        dialect = sniffer.sniff('"a"   "b""c"')
+        self.assertEqual(dialect.delimiter, ' ')
+        self.assertIs(dialect.doublequote, True)
+
     def test_doublequote(self):
         sniffer = csv.Sniffer()
         dialect = sniffer.sniff(self.header1)