bpo-40328: Add tool for generating cjk mapping headers

Misc/NEWS.d/next/Core and Builtins/2020-04-19-22-23-32.bpo-40328.gWJ53f.rst

@@ -0,0 +1 @@
+Add tools for generating mappings headers for CJKCodecs.

Modules/cjkcodecs/README

@@ -1,8 +1,6 @@
 To generate or modify mapping headers
 -------------------------------------
-Mapping headers are imported from CJKCodecs as pre-generated form.
-If you need to tweak or add something on it, please look at tools/
-subdirectory of CJKCodecs' distribution.
+Mapping headers are generated from Tools/unicode/genmap_*.py
 
 
 

Modules/cjkcodecs/mappings_cn.h

@@ -1,3 +1,4 @@
+// AUTO-GENERATED FILE FROM genmap_schinese.py: DO NOT EDIT
 static const ucs2_t __gb2312_decmap[7482] = {
 12288,12289,12290,12539,713,711,168,12291,12293,8213,65374,8214,8230,8216,
 8217,8220,8221,12308,12309,12296,12297,12298,12299,12300,12301,12302,12303,

Modules/cjkcodecs/mappings_jisx0213_pair.h

@@ -1,3 +1,4 @@
+// AUTO-GENERATED FILE FROM genmap_japanese.py: DO NOT EDIT
 #define JISX0213_ENCPAIRS 46
 #ifdef EXTERN_JISX0213_PAIR
 static const struct widedbcs_index *jisx0213_pair_decmap;

Modules/cjkcodecs/mappings_jp.h

@@ -1,3 +1,4 @@
+// AUTO-GENERATED FILE FROM genmap_japanese.py: DO NOT EDIT
 static const ucs2_t __jisx0208_decmap[6956] = {
 12288,12289,12290,65292,65294,12539,65306,65307,65311,65281,12443,12444,180,
 65344,168,65342,65507,65343,12541,12542,12445,12446,12291,20189,12293,12294,

Modules/cjkcodecs/mappings_kr.h

@@ -1,3 +1,4 @@
+// AUTO-GENERATED FILE FROM genmap_korean.py: DO NOT EDIT
 static const ucs2_t __ksx1001_decmap[8264] = {
 12288,12289,12290,183,8229,8230,168,12291,173,8213,8741,65340,8764,8216,8217,
 8220,8221,12308,12309,12296,12297,12298,12299,12300,12301,12302,12303,12304,
@@ -3249,3 +3250,4 @@ __cp949_encmap+31959,0,255},{__cp949_encmap+32215,0,255},{__cp949_encmap+32471
 __cp949_encmap+32891,0,11},{0,0,0},{0,0,0},{0,0,0},{0,0,0},{__cp949_encmap+
 32903,1,230},
 };
+

Tools/unicode/genmap_japanese.py

@@ -0,0 +1,251 @@
+#
+# genmap_ja_codecs.py: Japanese Codecs Map Generator
+#
+# Original Author:  Hye-Shik Chang <[email protected]>
+# Modified Author:  Dong-hee Na <[email protected]>
+#
+import os
+
+from genmap_support import *
+
+JISX0208_C1 = (0x21, 0x74)
+JISX0208_C2 = (0x21, 0x7e)
+JISX0212_C1 = (0x22, 0x6d)
+JISX0212_C2 = (0x21, 0x7e)
+JISX0213_C1 = (0x21, 0x7e)
+JISX0213_C2 = (0x21, 0x7e)
+CP932P0_C1  = (0x81, 0x81) # patches between shift-jis and cp932
+CP932P0_C2  = (0x5f, 0xca)
+CP932P1_C1  = (0x87, 0x87) # CP932 P1
+CP932P1_C2  = (0x40, 0x9c)
+CP932P2_C1  = (0xed, 0xfc) # CP932 P2
+CP932P2_C2  = (0x40, 0xfc)
+
+MAPPINGS_JIS0208 = 'http://www.unicode.org/Public/MAPPINGS/OBSOLETE/EASTASIA/JIS/JIS0208.TXT'
+MAPPINGS_JIS0212 = 'http://www.unicode.org/Public/MAPPINGS/OBSOLETE/EASTASIA/JIS/JIS0212.TXT'
+MAPPINGS_CP932 = 'http://www.unicode.org/Public/MAPPINGS/VENDORS/MICSFT/WINDOWS/CP932.TXT'
+MAPPINGS_JISX0213_2004 = 'http://wakaba-web.hp.infoseek.co.jp/table/jisx0213-2004-std.txt'
+
+
+def loadmap_jisx0213(fo):
+    decmap3, decmap4 = {}, {} # maps to BMP for level 3 and 4
+    decmap3_2, decmap4_2 = {}, {} # maps to U+2xxxx for level 3 and 4
+    decmap3_pair = {} # maps to BMP-pair for level 3
+    for line in fo:
+        line = line.split('#', 1)[0].strip()
+        if not line or len(line.split()) < 2:
+            continue
+
+        row = line.split()
+        loc = eval('0x' + row[0][2:])
+        level = eval(row[0][0])
+        m = None
+        if len(row[1].split('+')) == 2: # single unicode
+            uni = eval('0x' + row[1][2:])
+            if level == 3:
+                if uni < 0x10000:
+                    m = decmap3
+                elif 0x20000 <= uni < 0x30000:
+                    uni -= 0x20000
+                    m = decmap3_2
+            elif level == 4:
+                if uni < 0x10000:
+                    m = decmap4
+                elif 0x20000 <= uni < 0x30000:
+                    uni -= 0x20000
+                    m = decmap4_2
+            m.setdefault((loc >> 8), {})
+            m[(loc >> 8)][(loc & 0xff)] = uni
+        else: # pair
+            uniprefix = eval('0x' + row[1][2:6]) # body
+            uni = eval('0x' + row[1][7:11]) # modifier
+            if level != 3:
+                raise ValueError("invalid map")
+            decmap3_pair.setdefault(uniprefix, {})
+            m = decmap3_pair[uniprefix]
+
+        if m is None:
+            raise ValueError("invalid map")
+        m.setdefault((loc >> 8), {})
+        m[(loc >> 8)][(loc & 0xff)] = uni
+
+    return decmap3, decmap4, decmap3_2, decmap4_2, decmap3_pair
+
+
+def main():
+    jisx0208file = open_mapping_file('python-mappings/JIS0208.TXT', MAPPINGS_JIS0208)
+    jisx0212file = open_mapping_file('python-mappings/JIS0212.TXT', MAPPINGS_JIS0212)
+    cp932file = open_mapping_file('python-mappings/CP932.TXT', MAPPINGS_CP932)
+    jisx0213file = open_mapping_file('python-mappings/jisx0213-2004-std.txt', MAPPINGS_JISX0213_2004)
+
+    print("Loading Mapping File...")
+
+    sjisdecmap = loadmap(jisx0208file, natcol=0, unicol=2)
+    jisx0208decmap = loadmap(jisx0208file, natcol=1, unicol=2)
+    jisx0212decmap = loadmap(jisx0212file)
+    cp932decmap = loadmap(cp932file)
+    jis3decmap, jis4decmap, jis3_2_decmap, jis4_2_decmap, jis3_pairdecmap = loadmap_jisx0213(jisx0213file)
+
+    if jis3decmap[0x21][0x24] != 0xff0c:
+        raise SystemExit('Please adjust your JIS X 0213 map using jisx0213-2000-std.txt.diff')
+
+    sjisencmap, cp932encmap = {}, {}
+    jisx0208_0212encmap = {}
+    for c1, m in sjisdecmap.items():
+        for c2, code in m.items():
+            sjisencmap.setdefault(code >> 8, {})
+            sjisencmap[code >> 8][code & 0xff] = c1 << 8 | c2
+    for c1, m in cp932decmap.items():
+        for c2, code in m.items():
+            cp932encmap.setdefault(code >> 8, {})
+            if (code & 0xff) not in cp932encmap[code >> 8]:
+                cp932encmap[code >> 8][code & 0xff] = c1 << 8 | c2
+    for c1, m in cp932encmap.copy().items():
+        for c2, code in m.copy().items():
+            if c1 in sjisencmap and c2 in sjisencmap[c1] and sjisencmap[c1][c2] == code:
+                del cp932encmap[c1][c2]
+                if not cp932encmap[c1]:
+                    del cp932encmap[c1]
+
+    jisx0213pairdecmap = {}
+    jisx0213pairencmap = []
+    for unibody, m1 in jis3_pairdecmap.items():
+        for c1, m2 in m1.items():
+            for c2, modifier in m2.items():
+                jisx0213pairencmap.append((unibody, modifier, c1 << 8 | c2))
+                jisx0213pairdecmap.setdefault(c1, {})
+                jisx0213pairdecmap[c1][c2] = unibody << 16 | modifier
+
+    # Twinmap for both of JIS X 0208 (MSB unset) and JIS X 0212 (MSB set)
+    for c1, m in jisx0208decmap.items():
+        for c2, code in m.items():
+            jisx0208_0212encmap.setdefault(code >> 8, {})
+            jisx0208_0212encmap[code >> 8][code & 0xff] = c1 << 8 | c2
+
+    for c1, m in jisx0212decmap.items():
+        for c2, code in m.items():
+            jisx0208_0212encmap.setdefault(code >> 8, {})
+            if (code & 0xff) in jisx0208_0212encmap[code >> 8]:
+                print("OOPS!!!", (code))
+            jisx0208_0212encmap[code >> 8][code & 0xff] = 0x8000 | c1 << 8 | c2
+
+    jisx0213bmpencmap = {}
+    for c1, m in jis3decmap.copy().items():
+        for c2, code in m.copy().items():
+            if c1 in jisx0208decmap and c2 in jisx0208decmap[c1]:
+                if code in jis3_pairdecmap:
+                    jisx0213bmpencmap[code >> 8][code & 0xff] = (0,) # pair
+                    jisx0213pairencmap.append((code, 0, c1 << 8 | c2))
+                elif jisx0208decmap[c1][c2] == code:
+                    del jis3decmap[c1][c2]
+                    if not jis3decmap[c1]:
+                        del jis3decmap[c1]
+                else:
+                    raise ValueError("Difference between JIS X 0208 and JIS X 0213 Plane 1 is found.")
+            else:
+                jisx0213bmpencmap.setdefault(code >> 8, {})
+                if code not in jis3_pairdecmap:
+                    jisx0213bmpencmap[code >> 8][code & 0xff] = c1 << 8 | c2
+                else:
+                    jisx0213bmpencmap[code >> 8][code & 0xff] = (0,) # pair
+                    jisx0213pairencmap.append((code, 0, c1 << 8 | c2))
+
+    for c1, m in jis4decmap.items():
+        for c2, code in m.items():
+            jisx0213bmpencmap.setdefault(code >> 8, {})
+            jisx0213bmpencmap[code >> 8][code & 0xff] = 0x8000 | c1 << 8 | c2
+
+    jisx0213empencmap = {}
+    for c1, m in jis3_2_decmap.items():
+        for c2, code in m.items():
+            jisx0213empencmap.setdefault(code >> 8, {})
+            jisx0213empencmap[code >> 8][code & 0xff] = c1 << 8 | c2
+    for c1, m in jis4_2_decmap.items():
+        for c2, code in m.items():
+            jisx0213empencmap.setdefault(code >> 8, {})
+            jisx0213empencmap[code >> 8][code & 0xff] = 0x8000 | c1 << 8 | c2
+
+    with open("mappings_jp.h", "w") as fp:
+        print_autogen(fp, os.path.basename(__file__))
+        print("Generating JIS X 0208 decode map...")
+        writer = DecodeMapWriter(fp, "jisx0208", jisx0208decmap)
+        writer.update_decode_map(JISX0208_C1, JISX0208_C2)
+        writer.generate()
+
+        print("Generating JIS X 0212 decode map...")
+        writer = DecodeMapWriter(fp, "jisx0212", jisx0212decmap)
+        writer.update_decode_map(JISX0212_C1, JISX0212_C2)
+        writer.generate()
+
+        print("Generating JIS X 0208 && JIS X 0212 encode map...")
+        writer = EncodeMapWriter(fp, "jisxcommon", jisx0208_0212encmap)
+        writer.generate()
+
+        print("Generating CP932 Extension decode map...")
+        writer = DecodeMapWriter(fp, "cp932ext", cp932decmap)
+        writer.update_decode_map(CP932P0_C1, CP932P0_C2)
+        writer.update_decode_map(CP932P1_C1, CP932P1_C2)
+        writer.update_decode_map(CP932P2_C1, CP932P2_C2)
+        writer.generate()
+
+        print("Generating CP932 Extension encode map...")
+        writer = EncodeMapWriter(fp, "cp932ext", cp932encmap)
+        writer.generate()
+
+        print("Generating JIS X 0213 Plane 1 BMP decode map...")
+        writer = DecodeMapWriter(fp, "jisx0213_1_bmp", jis3decmap)
+        writer.update_decode_map(JISX0213_C1, JISX0213_C2)
+        writer.generate()
+
+        print("Generating JIS X 0213 Plane 2 BMP decode map...")
+        writer = DecodeMapWriter(fp, "jisx0213_2_bmp", jis4decmap)
+        writer.update_decode_map(JISX0213_C1, JISX0213_C2)
+        writer.generate()
+
+        print("Generating JIS X 0213 BMP encode map...")
+        writer = EncodeMapWriter(fp, "jisx0213_bmp", jisx0213bmpencmap)
+        writer.generate()
+
+        print("Generating JIS X 0213 Plane 1 EMP decode map...")
+        writer = DecodeMapWriter(fp, "jisx0213_1_emp", jis3_2_decmap)
+        writer.update_decode_map(JISX0213_C1, JISX0213_C2)
+        writer.generate()
+
+        print("Generating JIS X 0213 Plane 2 EMP decode map...")
+        writer = DecodeMapWriter(fp, "jisx0213_2_emp", jis4_2_decmap)
+        writer.update_decode_map(JISX0213_C1, JISX0213_C2)
+        writer.generate()
+
+        print("Generating JIS X 0213 EMP encode map...")
+        writer = EncodeMapWriter(fp, "jisx0213_emp", jisx0213empencmap)
+        writer.generate()
+
+    with open('mappings_jisx0213_pair.h', 'w') as fp:
+        print_autogen(fp, os.path.basename(__file__))
+        fp.write(f"#define JISX0213_ENCPAIRS {len(jisx0213pairencmap)}\n")
+        fp.write("""\
+#ifdef EXTERN_JISX0213_PAIR
+static const struct widedbcs_index *jisx0213_pair_decmap;
+static const struct pair_encodemap *jisx0213_pair_encmap;
+#else
+""")
+
+        print("Generating JIS X 0213 unicode-pair decode map...")
+        writer = DecodeMapWriter(fp, "jisx0213_pair", jisx0213pairdecmap)
+        writer.update_decode_map(JISX0213_C1, JISX0213_C2)
+        writer.generate(wide=True)
+
+        print("Generating JIS X 0213 unicode-pair encode map...")
+        jisx0213pairencmap.sort()
+        fp.write("static const struct pair_encodemap jisx0213_pair_encmap[JISX0213_ENCPAIRS] = {\n")
+        filler = BufferedFiller()
+        for body, modifier, jis in jisx0213pairencmap:
+            filler.write('{', '0x%04x%04x,' % (body, modifier), '0x%04x' % jis, '},')
+        filler.printout(fp)
+        fp.write("};\n")
+        fp.write("#endif\n")
+
+    print("Done!")
+
+if __name__ == '__main__':
+    main()

Tools/unicode/genmap_korean.py

@@ -0,0 +1,62 @@
+#
+# genmap_korean.py: Korean Codecs Map Generator
+#
+# Original Author:  Hye-Shik Chang <[email protected]>
+# Modified Author:  Dong-hee Na <[email protected]>
+#
+import os
+
+from genmap_support import *
+
+
+KSX1001_C1 = (0x21, 0x7e)
+KSX1001_C2 = (0x21, 0x7e)
+UHCL1_C1 = (0x81, 0xa0)
+UHCL1_C2 = (0x41, 0xfe)
+UHCL2_C1 = (0xa1, 0xfe)
+UHCL2_C2 = (0x41, 0xa0)
+MAPPINGS_CP949 = 'http://www.unicode.org/Public/MAPPINGS/VENDORS/MICSFT/WINDOWS/CP949.TXT'
+
+
+def main():
+    mapfile = open_mapping_file('python-mappings/CP949.TXT', MAPPINGS_CP949)
+    print("Loading Mapping File...")
+    decmap = loadmap(mapfile)
+    uhcdecmap, ksx1001decmap, cp949encmap = {}, {}, {}
+    for c1, c2map in decmap.items():
+        for c2, code in c2map.items():
+            if c1 >= 0xa1 and c2 >= 0xa1:
+                ksx1001decmap.setdefault(c1 & 0x7f, {})
+                ksx1001decmap[c1 & 0x7f][c2 & 0x7f] = c2map[c2]
+                cp949encmap.setdefault(code >> 8, {})
+                cp949encmap[code >> 8][code & 0xFF] = (c1 << 8 | c2) & 0x7f7f
+            else:
+                # uhc
+                uhcdecmap.setdefault(c1, {})
+                uhcdecmap[c1][c2] = c2map[c2]
+                cp949encmap.setdefault(code >> 8, {})  # MSB set
+                cp949encmap[code >> 8][code & 0xFF] = (c1 << 8 | c2)
+
+    with open('mappings_kr.h', 'w') as fp:
+        print_autogen(fp, os.path.basename(__file__))
+
+        print("Generating KS X 1001 decode map...")
+        writer = DecodeMapWriter(fp, "ksx1001", ksx1001decmap)
+        writer.update_decode_map(KSX1001_C1, KSX1001_C2)
+        writer.generate()
+
+        print("Generating UHC decode map...")
+        writer = DecodeMapWriter(fp, "cp949ext", uhcdecmap)
+        writer.update_decode_map(UHCL1_C1, UHCL1_C2)
+        writer.update_decode_map(UHCL2_C1, UHCL2_C2)
+        writer.generate()
+
+        print("Generating CP949 (includes KS X 1001) encode map...")
+        writer = EncodeMapWriter(fp, "cp949", cp949encmap)
+        writer.generate()
+
+    print("Done!")
+
+
+if __name__ == '__main__':
+    main()