GH-96172 fix unicodedata.east_asian_width being wrong on unassigned code points (#96207)

Lib/test/test_unicodedata.py

@@ -71,7 +71,7 @@ class UnicodeFunctionsTest(UnicodeDatabaseTest):
 
     # Update this if the database changes. Make sure to do a full rebuild
     # (e.g. 'make distclean && make') to get the correct checksum.
-    expectedchecksum = '98d602e1f69d5c5bb8a5910c40bbbad4e18e8370'
+    expectedchecksum = '4975f3ec0acd4a62465d18c9bf8519b1964181f6'
 
     @requires_resource('cpu')
     def test_function_checksum(self):
@@ -90,6 +90,7 @@ class UnicodeFunctionsTest(UnicodeDatabaseTest):
                 self.db.decomposition(char),
                 str(self.db.mirrored(char)),
                 str(self.db.combining(char)),
+                unicodedata.east_asian_width(char),
             ]
             h.update(''.join(data).encode("ascii"))
         result = h.hexdigest()
@@ -220,6 +221,23 @@ class UnicodeFunctionsTest(UnicodeDatabaseTest):
         self.assertEqual(eaw('\u2010'), 'A')
         self.assertEqual(eaw('\U00020000'), 'W')
 
+    def test_east_asian_width_unassigned(self):
+        eaw = self.db.east_asian_width
+        # unassigned
+        for char in '\u0530\u0ece\u10c6\u20fc\uaaca\U000107bd\U000115f2':
+            self.assertEqual(eaw(char), 'N')
+            self.assertIs(self.db.name(char, None), None)
+
+        # unassigned but reserved for CJK
+        for char in '\uFA6E\uFADA\U0002A6E0\U0002FA20\U0003134B\U0003FFFD':
+            self.assertEqual(eaw(char), 'W')
+            self.assertIs(self.db.name(char, None), None)
+
+        # private use areas
+        for char in '\uE000\uF800\U000F0000\U000FFFEE\U00100000\U0010FFF0':
+            self.assertEqual(eaw(char), 'A')
+            self.assertIs(self.db.name(char, None), None)
+
     def test_east_asian_width_9_0_changes(self):
         self.assertEqual(self.db.ucd_3_2_0.east_asian_width('\u231a'), 'N')
         self.assertEqual(self.db.east_asian_width('\u231a'), 'W')

Misc/NEWS.d/next/Library/2022-08-23-13-30-30.gh-issue-96172.7WTHer.rst

@@ -0,0 +1,3 @@
+Fix a bug in ``unicodedata``: ``east_asian_width`` used to return the wrong
+value for unassigned characters; and for yet unassigned, but reserved
+characters.

Tools/unicode/makeunicodedata.py

@@ -77,7 +77,8 @@ BIDIRECTIONAL_NAMES = [ "", "L", "LRE", "LRO", "R", "AL", "RLE", "RLO",
     "PDF", "EN", "ES", "ET", "AN", "CS", "NSM", "BN", "B", "S", "WS",
     "ON", "LRI", "RLI", "FSI", "PDI" ]
 
-EASTASIANWIDTH_NAMES = [ "F", "H", "W", "Na", "A", "N" ]
+# "N" needs to be the first entry, see the comment in makeunicodedata
+EASTASIANWIDTH_NAMES = [ "N", "H", "W", "Na", "A", "F" ]
 
 MANDATORY_LINE_BREAKS = [ "BK", "CR", "LF", "NL" ]
 
@@ -135,6 +136,14 @@ def maketables(trace=0):
 
 def makeunicodedata(unicode, trace):
 
+    # the default value of east_asian_width is "N", for unassigned code points
+    # not mentioned in EastAsianWidth.txt
+    # in addition there are some reserved but unassigned code points in CJK
+    # ranges that are classified as "W". code points in private use areas
+    # have a width of "A". both of these have entries in
+    # EastAsianWidth.txt
+    # see https://unicode.org/reports/tr11/#Unassigned
+    assert EASTASIANWIDTH_NAMES[0] == "N"
     dummy = (0, 0, 0, 0, 0, 0)
     table = [dummy]
     cache = {0: dummy}
@@ -160,12 +169,20 @@ def makeunicodedata(unicode, trace):
                 category, combining, bidirectional, mirrored, eastasianwidth,
                 normalizationquickcheck
                 )
-            # add entry to index and item tables
+        elif unicode.widths[char] is not None:
-            i = cache.get(item)
+            # an unassigned but reserved character, with a known
-            if i is None:
+            # east_asian_width
-                cache[item] = i = len(table)
+            eastasianwidth = EASTASIANWIDTH_NAMES.index(unicode.widths[char])
-                table.append(item)
+            item = (0, 0, 0, 0, eastasianwidth, 0)
-            index[char] = i
+        else:
+            continue
+
+        # add entry to index and item tables
+        i = cache.get(item)
+        if i is None:
+            cache[item] = i = len(table)
+            table.append(item)
+        index[char] = i
 
     # 2) decomposition data
 
@@ -1085,6 +1102,7 @@ class UnicodeData:
         for i in range(0, 0x110000):
             if table[i] is not None:
                 table[i].east_asian_width = widths[i]
+        self.widths = widths
 
         for char, (p,) in UcdFile(DERIVED_CORE_PROPERTIES, version).expanded():
             if table[char]: