gh-126505: Fix bugs in compiling case-insensitive character classes (GH-126557)

* upper-case non-BMP character was ignored
* the ASCII flag was ignored when matching a character range whose
  upper bound is beyond the BMP region

diff --git a/Lib/re/_compiler.py b/Lib/re/_compiler.py
index 29109f8812ee7be84520d7ac51bda5d10485344c..20dd561d1c1520fe0ece07bcd92e7e24a297f779 100644 (file)
--- a/Lib/re/_compiler.py
+++ b/Lib/re/_compiler.py
@@ -255,11 +255,11 @@ def _optimize_charset(charset, iscased=None, fixup=None, fixes=None):
         while True:
             try:
                 if op is LITERAL:
-                    if fixup:
-                        lo = fixup(av)
-                        charmap[lo] = 1
-                        if fixes and lo in fixes:
-                            for k in fixes[lo]:
+                    if fixup: # IGNORECASE and not LOCALE
+                        av = fixup(av)
+                        charmap[av] = 1
+                        if fixes and av in fixes:
+                            for k in fixes[av]:
                                 charmap[k] = 1
                         if not hascased and iscased(av):
                             hascased = True
@@ -267,7 +267,7 @@ def _optimize_charset(charset, iscased=None, fixup=None, fixes=None):
                         charmap[av] = 1
                 elif op is RANGE:
                     r = range(av[0], av[1]+1)
-                    if fixup:
+                    if fixup: # IGNORECASE and not LOCALE
                         if fixes:
                             for i in map(fixup, r):
                                 charmap[i] = 1
@@ -298,8 +298,7 @@ def _optimize_charset(charset, iscased=None, fixup=None, fixes=None):
                 # Character set contains non-BMP character codes.
                 # For range, all BMP characters in the range are already
                 # proceeded.
-                if fixup:
-                    hascased = True
+                if fixup: # IGNORECASE and not LOCALE
                     # For now, IN_UNI_IGNORE+LITERAL and
                     # IN_UNI_IGNORE+RANGE_UNI_IGNORE work for all non-BMP
                     # characters, because two characters (at least one of
@@ -310,7 +309,13 @@ def _optimize_charset(charset, iscased=None, fixup=None, fixes=None):
                     # Also, both c.lower() and c.lower().upper() are single
                     # characters for every non-BMP character.
                     if op is RANGE:
-                        op = RANGE_UNI_IGNORE
+                        if fixes: # not ASCII
+                            op = RANGE_UNI_IGNORE
+                        hascased = True
+                    else:
+                        assert op is LITERAL
+                        if not hascased and iscased(av):
+                            hascased = True
                 tail.append((op, av))
             break
 

diff --git a/Lib/test/test_re.py b/Lib/test/test_re.py
index ff95f54026e172063b9ba6c3d2c060f17b482fd3..7bc702ec89a4a7b1dc40c5a165dd88631e331164 100644 (file)
--- a/Lib/test/test_re.py
+++ b/Lib/test/test_re.py
@@ -1136,6 +1136,39 @@ class ReTests(unittest.TestCase):
         self.assertTrue(re.match(br'[19a]', b'a', re.I))
         self.assertTrue(re.match(br'[19a]', b'A', re.I))
         self.assertTrue(re.match(br'[19A]', b'a', re.I))
+        self.assertTrue(re.match(r'[19\xc7]', '\xc7', re.I))
+        self.assertTrue(re.match(r'[19\xc7]', '\xe7', re.I))
+        self.assertTrue(re.match(r'[19\xe7]', '\xc7', re.I))
+        self.assertTrue(re.match(r'[19\xe7]', '\xe7', re.I))
+        self.assertTrue(re.match(r'[19\u0400]', '\u0400', re.I))
+        self.assertTrue(re.match(r'[19\u0400]', '\u0450', re.I))
+        self.assertTrue(re.match(r'[19\u0450]', '\u0400', re.I))
+        self.assertTrue(re.match(r'[19\u0450]', '\u0450', re.I))
+        self.assertTrue(re.match(r'[19\U00010400]', '\U00010400', re.I))
+        self.assertTrue(re.match(r'[19\U00010400]', '\U00010428', re.I))
+        self.assertTrue(re.match(r'[19\U00010428]', '\U00010400', re.I))
+        self.assertTrue(re.match(r'[19\U00010428]', '\U00010428', re.I))
+
+        self.assertTrue(re.match(br'[19A]', b'A', re.I))
+        self.assertTrue(re.match(br'[19a]', b'a', re.I))
+        self.assertTrue(re.match(br'[19a]', b'A', re.I))
+        self.assertTrue(re.match(br'[19A]', b'a', re.I))
+        self.assertTrue(re.match(r'[19A]', 'A', re.I|re.A))
+        self.assertTrue(re.match(r'[19a]', 'a', re.I|re.A))
+        self.assertTrue(re.match(r'[19a]', 'A', re.I|re.A))
+        self.assertTrue(re.match(r'[19A]', 'a', re.I|re.A))
+        self.assertTrue(re.match(r'[19\xc7]', '\xc7', re.I|re.A))
+        self.assertIsNone(re.match(r'[19\xc7]', '\xe7', re.I|re.A))
+        self.assertIsNone(re.match(r'[19\xe7]', '\xc7', re.I|re.A))
+        self.assertTrue(re.match(r'[19\xe7]', '\xe7', re.I|re.A))
+        self.assertTrue(re.match(r'[19\u0400]', '\u0400', re.I|re.A))
+        self.assertIsNone(re.match(r'[19\u0400]', '\u0450', re.I|re.A))
+        self.assertIsNone(re.match(r'[19\u0450]', '\u0400', re.I|re.A))
+        self.assertTrue(re.match(r'[19\u0450]', '\u0450', re.I|re.A))
+        self.assertTrue(re.match(r'[19\U00010400]', '\U00010400', re.I|re.A))
+        self.assertIsNone(re.match(r'[19\U00010400]', '\U00010428', re.I|re.A))
+        self.assertIsNone(re.match(r'[19\U00010428]', '\U00010400', re.I|re.A))
+        self.assertTrue(re.match(r'[19\U00010428]', '\U00010428', re.I|re.A))
 
         # Two different characters have the same lowercase.
         assert 'K'.lower() == '\u212a'.lower() == 'k' # 'K'
@@ -1172,8 +1205,10 @@ class ReTests(unittest.TestCase):
         self.assertTrue(re.match(br'[9-a]', b'_', re.I))
         self.assertIsNone(re.match(br'[9-A]', b'_', re.I))
         self.assertTrue(re.match(r'[\xc0-\xde]', '\xd7', re.I))
+        self.assertTrue(re.match(r'[\xc0-\xde]', '\xe7', re.I))
         self.assertIsNone(re.match(r'[\xc0-\xde]', '\xf7', re.I))
         self.assertTrue(re.match(r'[\xe0-\xfe]', '\xf7', re.I))
+        self.assertTrue(re.match(r'[\xe0-\xfe]', '\xc7', re.I))
         self.assertIsNone(re.match(r'[\xe0-\xfe]', '\xd7', re.I))
         self.assertTrue(re.match(r'[\u0430-\u045f]', '\u0450', re.I))
         self.assertTrue(re.match(r'[\u0430-\u045f]', '\u0400', re.I))
@@ -1184,6 +1219,26 @@ class ReTests(unittest.TestCase):
         self.assertTrue(re.match(r'[\U00010400-\U00010427]', '\U00010428', re.I))
         self.assertTrue(re.match(r'[\U00010400-\U00010427]', '\U00010400', re.I))
 
+        self.assertTrue(re.match(r'[\xc0-\xde]', '\xd7', re.I|re.A))
+        self.assertIsNone(re.match(r'[\xc0-\xde]', '\xe7', re.I|re.A))
+        self.assertTrue(re.match(r'[\xe0-\xfe]', '\xf7', re.I|re.A))
+        self.assertIsNone(re.match(r'[\xe0-\xfe]', '\xc7', re.I|re.A))
+        self.assertTrue(re.match(r'[\u0430-\u045f]', '\u0450', re.I|re.A))
+        self.assertIsNone(re.match(r'[\u0430-\u045f]', '\u0400', re.I|re.A))
+        self.assertIsNone(re.match(r'[\u0400-\u042f]', '\u0450', re.I|re.A))
+        self.assertTrue(re.match(r'[\u0400-\u042f]', '\u0400', re.I|re.A))
+        self.assertTrue(re.match(r'[\U00010428-\U0001044f]', '\U00010428', re.I|re.A))
+        self.assertIsNone(re.match(r'[\U00010428-\U0001044f]', '\U00010400', re.I|re.A))
+        self.assertIsNone(re.match(r'[\U00010400-\U00010427]', '\U00010428', re.I|re.A))
+        self.assertTrue(re.match(r'[\U00010400-\U00010427]', '\U00010400', re.I|re.A))
+
+        self.assertTrue(re.match(r'[N-\x7f]', 'A', re.I|re.A))
+        self.assertTrue(re.match(r'[n-\x7f]', 'Z', re.I|re.A))
+        self.assertTrue(re.match(r'[N-\uffff]', 'A', re.I|re.A))
+        self.assertTrue(re.match(r'[n-\uffff]', 'Z', re.I|re.A))
+        self.assertTrue(re.match(r'[N-\U00010000]', 'A', re.I|re.A))
+        self.assertTrue(re.match(r'[n-\U00010000]', 'Z', re.I|re.A))
+
         # Two different characters have the same lowercase.
         assert 'K'.lower() == '\u212a'.lower() == 'k' # 'K'
         self.assertTrue(re.match(r'[J-M]', '\u212a', re.I))

diff --git a/Misc/NEWS.d/next/Library/2024-11-07-22-41-47.gh-issue-126505.iztYE1.rst b/Misc/NEWS.d/next/Library/2024-11-07-22-41-47.gh-issue-126505.iztYE1.rst
new file mode 100644 (file)
index 0000000..0a0f893
--- /dev/null
+++ b/Misc/NEWS.d/next/Library/2024-11-07-22-41-47.gh-issue-126505.iztYE1.rst
@@ -0,0 +1,4 @@
+Fix bugs in compiling case-insensitive :mod:`regular expressions <re>` with
+character classes containing non-BMP characters: upper-case non-BMP
+character did was ignored and the ASCII flag was ignored when
+matching a character range whose upper bound is beyond the BMP region.