Improve unicode table handling (bellard#286)

- Document table and index formats
- Add size statistics
- Fix UBSAN issue in `get_le24()`

Fixes bellard#285

Author: chqrlie

libunicode-table.h

@@ -189,9 +189,13 @@ static const uint8_t unicode_prop_Cased1_table[196] = {
 };
 
 static const uint8_t unicode_prop_Cased1_index[21] = {
-    0xb9, 0x02, 0xe0, 0xc0, 0x1d, 0x20, 0xe5, 0x2c,
-    0x20, 0xb1, 0x07, 0x21, 0xc1, 0xd6, 0x21, 0x4a,
-    0xf1, 0x01, 0x8a, 0xf1, 0x01,
+    0xb9, 0x02, 0xe0,  //  002B9 at 39
+    0xc0, 0x1d, 0x20,  //  01DC0 at 65
+    0xe5, 0x2c, 0x20,  //  02CE5 at 97
+    0xb1, 0x07, 0x21,  //  107B1 at 129
+    0xc1, 0xd6, 0x21,  //  1D6C1 at 161
+    0x4a, 0xf1, 0x01,  //  1F14A at 192
+    0x8a, 0xf1, 0x01,  //  1F18A at 224 (upper bound)
 };
 
 static const uint8_t unicode_prop_Case_Ignorable_table[737] = {
@@ -291,15 +295,29 @@ static const uint8_t unicode_prop_Case_Ignorable_table[737] = {
 };
 
 static const uint8_t unicode_prop_Case_Ignorable_index[69] = {
-    0xbe, 0x05, 0x00, 0xfe, 0x07, 0x00, 0x52, 0x0a,
-    0xa0, 0xc1, 0x0b, 0x00, 0x82, 0x0d, 0x00, 0x3f,
-    0x10, 0x80, 0xd4, 0x17, 0x40, 0xcf, 0x1a, 0x20,
-    0xf5, 0x1c, 0x00, 0x80, 0x20, 0x00, 0x16, 0xa0,
-    0x00, 0xc6, 0xa8, 0x00, 0xc2, 0xaa, 0x60, 0x56,
-    0xfe, 0x20, 0xb1, 0x07, 0x01, 0x75, 0x10, 0x01,
-    0xeb, 0x12, 0x21, 0x41, 0x16, 0x01, 0x5c, 0x1a,
-    0x01, 0x43, 0x1f, 0x01, 0x2e, 0xcf, 0x41, 0x25,
-    0xe0, 0x01, 0xf0, 0x01, 0x0e,
+    0xbe, 0x05, 0x00,  //  005BE at 32
+    0xfe, 0x07, 0x00,  //  007FE at 64
+    0x52, 0x0a, 0xa0,  //  00A52 at 101
+    0xc1, 0x0b, 0x00,  //  00BC1 at 128
+    0x82, 0x0d, 0x00,  //  00D82 at 160
+    0x3f, 0x10, 0x80,  //  0103F at 196
+    0xd4, 0x17, 0x40,  //  017D4 at 226
+    0xcf, 0x1a, 0x20,  //  01ACF at 257
+    0xf5, 0x1c, 0x00,  //  01CF5 at 288
+    0x80, 0x20, 0x00,  //  02080 at 320
+    0x16, 0xa0, 0x00,  //  0A016 at 352
+    0xc6, 0xa8, 0x00,  //  0A8C6 at 384
+    0xc2, 0xaa, 0x60,  //  0AAC2 at 419
+    0x56, 0xfe, 0x20,  //  0FE56 at 449
+    0xb1, 0x07, 0x01,  //  107B1 at 480
+    0x75, 0x10, 0x01,  //  11075 at 512
+    0xeb, 0x12, 0x21,  //  112EB at 545
+    0x41, 0x16, 0x01,  //  11641 at 576
+    0x5c, 0x1a, 0x01,  //  11A5C at 608
+    0x43, 0x1f, 0x01,  //  11F43 at 640
+    0x2e, 0xcf, 0x41,  //  1CF2E at 674
+    0x25, 0xe0, 0x01,  //  1E025 at 704
+    0xf0, 0x01, 0x0e,  //  E01F0 at 736 (upper bound)
 };
 
 static const uint8_t unicode_prop_ID_Start_table[1100] = {
@@ -444,20 +462,41 @@ static const uint8_t unicode_prop_ID_Start_table[1100] = {
 };
 
 static const uint8_t unicode_prop_ID_Start_index[105] = {
-    0xf6, 0x03, 0x20, 0xa6, 0x07, 0x00, 0xa9, 0x09,
-    0x20, 0xb1, 0x0a, 0x00, 0xba, 0x0b, 0x20, 0x3b,
-    0x0d, 0x20, 0xc7, 0x0e, 0x20, 0x49, 0x12, 0x00,
-    0x9b, 0x16, 0x00, 0xac, 0x19, 0x00, 0xc0, 0x1d,
-    0x80, 0x80, 0x20, 0x20, 0x70, 0x2d, 0x00, 0x00,
-    0x32, 0x00, 0xda, 0xa7, 0x00, 0x4c, 0xaa, 0x20,
-    0xc7, 0xd7, 0x20, 0xfc, 0xfd, 0x20, 0x9d, 0x02,
-    0x21, 0x96, 0x05, 0x01, 0xf3, 0x08, 0x01, 0xb3,
-    0x0c, 0x21, 0x73, 0x11, 0x61, 0x34, 0x13, 0x01,
-    0x1b, 0x17, 0x21, 0x8a, 0x1a, 0x01, 0x34, 0x1f,
-    0x21, 0xbf, 0x6a, 0x01, 0x23, 0xb1, 0xa1, 0xad,
-    0xd4, 0x01, 0x6f, 0xd7, 0x01, 0xff, 0xe7, 0x61,
-    0x5e, 0xee, 0x01, 0xe1, 0xeb, 0x22, 0xb0, 0x23,
-    0x03,
+    0xf6, 0x03, 0x20,  //  003F6 at 33
+    0xa6, 0x07, 0x00,  //  007A6 at 64
+    0xa9, 0x09, 0x20,  //  009A9 at 97
+    0xb1, 0x0a, 0x00,  //  00AB1 at 128
+    0xba, 0x0b, 0x20,  //  00BBA at 161
+    0x3b, 0x0d, 0x20,  //  00D3B at 193
+    0xc7, 0x0e, 0x20,  //  00EC7 at 225
+    0x49, 0x12, 0x00,  //  01249 at 256
+    0x9b, 0x16, 0x00,  //  0169B at 288
+    0xac, 0x19, 0x00,  //  019AC at 320
+    0xc0, 0x1d, 0x80,  //  01DC0 at 356
+    0x80, 0x20, 0x20,  //  02080 at 385
+    0x70, 0x2d, 0x00,  //  02D70 at 416
+    0x00, 0x32, 0x00,  //  03200 at 448
+    0xda, 0xa7, 0x00,  //  0A7DA at 480
+    0x4c, 0xaa, 0x20,  //  0AA4C at 513
+    0xc7, 0xd7, 0x20,  //  0D7C7 at 545
+    0xfc, 0xfd, 0x20,  //  0FDFC at 577
+    0x9d, 0x02, 0x21,  //  1029D at 609
+    0x96, 0x05, 0x01,  //  10596 at 640
+    0xf3, 0x08, 0x01,  //  108F3 at 672
+    0xb3, 0x0c, 0x21,  //  10CB3 at 705
+    0x73, 0x11, 0x61,  //  11173 at 739
+    0x34, 0x13, 0x01,  //  11334 at 768
+    0x1b, 0x17, 0x21,  //  1171B at 801
+    0x8a, 0x1a, 0x01,  //  11A8A at 832
+    0x34, 0x1f, 0x21,  //  11F34 at 865
+    0xbf, 0x6a, 0x01,  //  16ABF at 896
+    0x23, 0xb1, 0xa1,  //  1B123 at 933
+    0xad, 0xd4, 0x01,  //  1D4AD at 960
+    0x6f, 0xd7, 0x01,  //  1D76F at 992
+    0xff, 0xe7, 0x61,  //  1E7FF at 1027
+    0x5e, 0xee, 0x01,  //  1EE5E at 1056
+    0xe1, 0xeb, 0x22,  //  2EBE1 at 1089
+    0xb0, 0x23, 0x03,  //  323B0 at 1120 (upper bound)
 };
 
 static const uint8_t unicode_prop_ID_Continue1_table[660] = {
@@ -547,14 +586,27 @@ static const uint8_t unicode_prop_ID_Continue1_table[660] = {
 };
 
 static const uint8_t unicode_prop_ID_Continue1_index[63] = {
-    0xfa, 0x06, 0x00, 0x70, 0x09, 0x00, 0xf0, 0x0a,
-    0x40, 0x57, 0x0c, 0x00, 0xf0, 0x0d, 0x60, 0xc7,
-    0x0f, 0x20, 0xea, 0x17, 0x40, 0x05, 0x1b, 0x00,
-    0x41, 0x20, 0x00, 0x0c, 0xa8, 0x80, 0x37, 0xaa,
-    0x20, 0x50, 0xfe, 0x20, 0x3a, 0x0d, 0x21, 0x74,
-    0x11, 0x01, 0x5a, 0x14, 0x21, 0x44, 0x19, 0x81,
-    0x5a, 0x1d, 0xa1, 0xf5, 0x6a, 0x21, 0x45, 0xd2,
-    0x41, 0xaf, 0xe2, 0x21, 0xf0, 0x01, 0x0e,
+    0xfa, 0x06, 0x00,  //  006FA at 32
+    0x70, 0x09, 0x00,  //  00970 at 64
+    0xf0, 0x0a, 0x40,  //  00AF0 at 98
+    0x57, 0x0c, 0x00,  //  00C57 at 128
+    0xf0, 0x0d, 0x60,  //  00DF0 at 163
+    0xc7, 0x0f, 0x20,  //  00FC7 at 193
+    0xea, 0x17, 0x40,  //  017EA at 226
+    0x05, 0x1b, 0x00,  //  01B05 at 256
+    0x41, 0x20, 0x00,  //  02041 at 288
+    0x0c, 0xa8, 0x80,  //  0A80C at 324
+    0x37, 0xaa, 0x20,  //  0AA37 at 353
+    0x50, 0xfe, 0x20,  //  0FE50 at 385
+    0x3a, 0x0d, 0x21,  //  10D3A at 417
+    0x74, 0x11, 0x01,  //  11174 at 448
+    0x5a, 0x14, 0x21,  //  1145A at 481
+    0x44, 0x19, 0x81,  //  11944 at 516
+    0x5a, 0x1d, 0xa1,  //  11D5A at 549
+    0xf5, 0x6a, 0x21,  //  16AF5 at 577
+    0x45, 0xd2, 0x41,  //  1D245 at 610
+    0xaf, 0xe2, 0x21,  //  1E2AF at 641
+    0xf0, 0x01, 0x0e,  //  E01F0 at 672 (upper bound)
 };
 
 #ifdef CONFIG_ALL_UNICODE
@@ -676,17 +728,35 @@ static const uint8_t unicode_cc_table[899] = {
 };
 
 static const uint8_t unicode_cc_index[87] = {
-    0x4d, 0x03, 0x00, 0x97, 0x05, 0x20, 0xc6, 0x05,
-    0x00, 0xe7, 0x06, 0x00, 0x45, 0x07, 0x00, 0x9c,
-    0x08, 0x00, 0x4d, 0x09, 0x00, 0x3c, 0x0b, 0x00,
-    0x3d, 0x0d, 0x00, 0x36, 0x0f, 0x00, 0x38, 0x10,
-    0x20, 0x3a, 0x19, 0x00, 0xcb, 0x1a, 0x20, 0xd3,
-    0x1c, 0x00, 0xcf, 0x1d, 0x00, 0xe2, 0x20, 0x00,
-    0x2e, 0x30, 0x20, 0x2b, 0xa9, 0x20, 0xed, 0xab,
-    0x00, 0x39, 0x0a, 0x01, 0x51, 0x0f, 0x01, 0x73,
-    0x11, 0x01, 0x75, 0x13, 0x01, 0x2b, 0x17, 0x21,
-    0x3f, 0x1c, 0x21, 0x9e, 0xbc, 0x21, 0x08, 0xe0,
-    0x01, 0x44, 0xe9, 0x01, 0x4b, 0xe9, 0x01,
+    0x4d, 0x03, 0x00,  //  0034D at 32
+    0x97, 0x05, 0x20,  //  00597 at 65
+    0xc6, 0x05, 0x00,  //  005C6 at 96
+    0xe7, 0x06, 0x00,  //  006E7 at 128
+    0x45, 0x07, 0x00,  //  00745 at 160
+    0x9c, 0x08, 0x00,  //  0089C at 192
+    0x4d, 0x09, 0x00,  //  0094D at 224
+    0x3c, 0x0b, 0x00,  //  00B3C at 256
+    0x3d, 0x0d, 0x00,  //  00D3D at 288
+    0x36, 0x0f, 0x00,  //  00F36 at 320
+    0x38, 0x10, 0x20,  //  01038 at 353
+    0x3a, 0x19, 0x00,  //  0193A at 384
+    0xcb, 0x1a, 0x20,  //  01ACB at 417
+    0xd3, 0x1c, 0x00,  //  01CD3 at 448
+    0xcf, 0x1d, 0x00,  //  01DCF at 480
+    0xe2, 0x20, 0x00,  //  020E2 at 512
+    0x2e, 0x30, 0x20,  //  0302E at 545
+    0x2b, 0xa9, 0x20,  //  0A92B at 577
+    0xed, 0xab, 0x00,  //  0ABED at 608
+    0x39, 0x0a, 0x01,  //  10A39 at 640
+    0x51, 0x0f, 0x01,  //  10F51 at 672
+    0x73, 0x11, 0x01,  //  11173 at 704
+    0x75, 0x13, 0x01,  //  11375 at 736
+    0x2b, 0x17, 0x21,  //  1172B at 769
+    0x3f, 0x1c, 0x21,  //  11C3F at 801
+    0x9e, 0xbc, 0x21,  //  1BC9E at 833
+    0x08, 0xe0, 0x01,  //  1E008 at 864
+    0x44, 0xe9, 0x01,  //  1E944 at 896
+    0x4b, 0xe9, 0x01,  //  1E94B at 928 (upper bound)
 };
 
 static const uint32_t unicode_decomp_table1[699] = {
@@ -4484,3 +4554,4 @@ static const uint16_t unicode_prop_len_table[] = {
 };
 
 #endif /* CONFIG_ALL_UNICODE */
+/* 62 tables / 32261 bytes, 5 index / 345 bytes */

libunicode.c

@@ -262,11 +262,7 @@ int lre_canonicalize(uint32_t c, BOOL is_unicode)
 
 static uint32_t get_le24(const uint8_t *ptr)
 {
-#if defined(__x86__) || defined(__x86_64__)
-    return *(uint16_t *)ptr | (ptr[2] << 16);
-#else
     return ptr[0] | (ptr[1] << 8) | (ptr[2] << 16);
-#endif
 }
 
 #define UNICODE_INDEX_BLOCK_LEN 32
@@ -317,6 +313,14 @@ static BOOL lre_is_in_table(uint32_t c, const uint8_t *table,
         return FALSE; /* outside the table */
     p = table + pos;
     bit = 0;
+    /* Compressed run length encoding:
+       00..3F: 2 packed lengths: 3-bit + 3-bit
+       40..5F: 5-bits plus extra byte for length
+       60..7F: 5-bits plus 2 extra bytes for length
+       80..FF: 7-bit length
+       lengths must be incremented to get character count
+       Ranges alternate between false and true return value.
+     */
     for(;;) {
         b = *p++;
         if (b < 64) {
@@ -833,6 +837,13 @@ static int unicode_get_cc(uint32_t c)
     if (pos < 0)
         return 0;
     p = unicode_cc_table + pos;
+    /* Compressed run length encoding:
+       - 2 high order bits are combining class type
+       -         0:0, 1:230, 2:extra byte linear progression, 3:extra byte
+       - 00..2F: range length (add 1)
+       - 30..37: 3-bit range-length + 1 extra byte
+       - 38..3F: 3-bit range-length + 2 extra byte
+     */
     for(;;) {
         b = *p++;
         type = b >> 6;
@@ -1185,6 +1196,15 @@ static int unicode_general_category1(CharRange *cr, uint32_t gc_mask)
     p = unicode_gc_table;
     p_end = unicode_gc_table + countof(unicode_gc_table);
     c = 0;
+    /* Compressed range encoding:
+       initial byte:
+       bits 0..4: category number (special case 31)
+       bits 5..7: range length (add 1)
+       special case bits 5..7 == 7: read an extra byte
+       - 00..7F: range length (add 7 + 1)
+       - 80..BF: 6-bits plus extra byte for range length (add 7 + 128)
+       - C0..FF: 6-bits plus 2 extra bytes for range length (add 7 + 128 + 16384)
+     */
     while (p < p_end) {
         b = *p++;
         n = b >> 5;
@@ -1238,6 +1258,14 @@ static int unicode_prop1(CharRange *cr, int prop_idx)
     p_end = p + unicode_prop_len_table[prop_idx];
     c = 0;
     bit = 0;
+    /* Compressed range encoding:
+       00..3F: 2 packed lengths: 3-bit + 3-bit
+       40..5F: 5-bits plus extra byte for length
+       60..7F: 5-bits plus 2 extra bytes for length
+       80..FF: 7-bit length
+       lengths must be incremented to get character count
+       Ranges alternate between false and true return value.
+     */
     while (p < p_end) {
         c0 = c;
         b = *p++;