charclass.c revision 20d2c4d2 
1/* $XTermId: charclass.c,v 1.22 2009/11/05 23:46:15 tom Exp $ */
2
3/*
4 * Compact and efficient reimplementation of the
5 * xterm character class mechanism for large character sets
6 *
7 * Markus Kuhn -- mkuhn@acm.org -- 2000-07-03
8 *
9 * Xterm allows users to select entire words with a double-click on the left
10 * mouse button.  Opinions might differ on what type of characters are part of
11 * separate words, therefore xterm allows users to configure a class code for
12 * each 8-bit character.  Words are maximum length sequences of neighboring
13 * characters with identical class code.  Extending this mechanism to Unicode
14 * naively would create an at least 2^16 entries (128 kB) long class code
15 * table.
16 *
17 * Instead, we transform the character class table into a list of intervals,
18 * that will be accessed via a linear search.  Changes made to the table by the
19 * user will be appended.  A special class code IDENT (default) marks
20 * characters who have their code number as the class code.
21 *
22 * We could alternatively use a sorted table of non-overlapping intervals that
23 * can be accessed via binary search, but merging in new intervals is
24 * significantly more hassle and not worth the effort here.
25 */
26
27#include <xterm.h>
28#include <charclass.h>
29
30#if OPT_WIDE_CHARS
31
32static struct classentry {
33    int cclass;
34    int first;
35    int last;
36} *classtab;
37
38/*
39 * Special convention for classtab[0]:
40 * - classtab[0].cclass is the allocated number of entries in classtab
41 * - classtab[0].first = 1 (first used entry in classtab)
42 * - classtab[0].last is the last used entry in classtab
43 */
44
45int
46SetCharacterClassRange(int low, int high, int value)
47{
48    if (high < low)
49	return -1;		/* nothing to do */
50
51    /* make sure we have at least one free entry left at table end */
52    if (classtab[0].last > classtab[0].cclass - 2) {
53	classtab[0].cclass += 5 + classtab[0].cclass / 4;
54	classtab = TypeRealloc(struct classentry,
55			         (unsigned) classtab[0].cclass, classtab);
56	if (!classtab)
57	    abort();
58    }
59
60    /* simply append new interval to end of interval array */
61    classtab[0].last++;
62    classtab[classtab[0].last].first = low;
63    classtab[classtab[0].last].last = high;
64    classtab[classtab[0].last].cclass = value;
65
66    return 0;
67}
68
69typedef enum {
70    IDENT = -1,
71    ALNUM = 48,
72    CNTRL = 1,
73    BLANK = 32
74} Classes;
75
76void
77init_classtab(void)
78{
79    const int size = 50;
80
81    classtab = TypeMallocN(struct classentry, (unsigned) size);
82    if (!classtab)
83	abort();
84    classtab[0].cclass = size;
85    classtab[0].first = 1;
86    classtab[0].last = 0;
87
88    /* old xterm default classes */
89    SetCharacterClassRange(0, 0, BLANK);
90    SetCharacterClassRange(1, 31, CNTRL);
91    SetCharacterClassRange('\t', '\t', BLANK);
92    SetCharacterClassRange('0', '9', ALNUM);
93    SetCharacterClassRange('A', 'Z', ALNUM);
94    SetCharacterClassRange('_', '_', ALNUM);
95    SetCharacterClassRange('a', 'z', ALNUM);
96    SetCharacterClassRange(127, 159, CNTRL);
97    SetCharacterClassRange(160, 191, IDENT);
98    SetCharacterClassRange(192, 255, ALNUM);
99    SetCharacterClassRange(215, 215, IDENT);
100    SetCharacterClassRange(247, 247, IDENT);
101
102    /* added Unicode classes */
103    SetCharacterClassRange(0x0100, 0xffdf, ALNUM);	/* mostly characters */
104    SetCharacterClassRange(0x037e, 0x037e, IDENT);	/* Greek question mark */
105    SetCharacterClassRange(0x0387, 0x0387, IDENT);	/* Greek ano teleia */
106    SetCharacterClassRange(0x055a, 0x055f, IDENT);	/* Armenian punctuation */
107    SetCharacterClassRange(0x0589, 0x0589, IDENT);	/* Armenian full stop */
108    SetCharacterClassRange(0x0700, 0x070d, IDENT);	/* Syriac punctuation */
109    SetCharacterClassRange(0x104a, 0x104f, IDENT);	/* Myanmar punctuation */
110    SetCharacterClassRange(0x10fb, 0x10fb, IDENT);	/* Georgian punctuation */
111    SetCharacterClassRange(0x1361, 0x1368, IDENT);	/* Ethiopic punctuation */
112    SetCharacterClassRange(0x166d, 0x166e, IDENT);	/* Canadian Syl. punctuation */
113    SetCharacterClassRange(0x17d4, 0x17dc, IDENT);	/* Khmer punctuation */
114    SetCharacterClassRange(0x1800, 0x180a, IDENT);	/* Mongolian punctuation */
115    SetCharacterClassRange(0x2000, 0x200a, BLANK);	/* spaces */
116    SetCharacterClassRange(0x200b, 0x27ff, IDENT);	/* punctuation and symbols */
117    SetCharacterClassRange(0x2070, 0x207f, 0x2070);	/* superscript */
118    SetCharacterClassRange(0x2080, 0x208f, 0x2080);	/* subscript */
119    SetCharacterClassRange(0x3000, 0x3000, BLANK);	/* ideographic space */
120    SetCharacterClassRange(0x3001, 0x3020, IDENT);	/* ideographic punctuation */
121    SetCharacterClassRange(0x3040, 0x309f, 0x3040);	/* Hiragana */
122    SetCharacterClassRange(0x30a0, 0x30ff, 0x30a0);	/* Katakana */
123    SetCharacterClassRange(0x3300, 0x9fff, 0x4e00);	/* CJK Ideographs */
124    SetCharacterClassRange(0xac00, 0xd7a3, 0xac00);	/* Hangul Syllables */
125    SetCharacterClassRange(0xf900, 0xfaff, 0x4e00);	/* CJK Ideographs */
126    SetCharacterClassRange(0xfe30, 0xfe6b, IDENT);	/* punctuation forms */
127    SetCharacterClassRange(0xff00, 0xff0f, IDENT);	/* half/fullwidth ASCII */
128    SetCharacterClassRange(0xff1a, 0xff20, IDENT);	/* half/fullwidth ASCII */
129    SetCharacterClassRange(0xff3b, 0xff40, IDENT);	/* half/fullwidth ASCII */
130    SetCharacterClassRange(0xff5b, 0xff64, IDENT);	/* half/fullwidth ASCII */
131
132    return;
133}
134
135int
136CharacterClass(int c)
137{
138    int i, cclass = IDENT;
139
140    for (i = classtab[0].first; i <= classtab[0].last; i++)
141	if (classtab[i].first <= c && classtab[i].last >= c)
142	    cclass = classtab[i].cclass;
143
144    if (cclass < 0)
145	cclass = c;
146
147    return cclass;
148}
149
150#ifdef NO_LEAKS
151void
152noleaks_CharacterClass(void)
153{
154    if (classtab != 0) {
155	free(classtab);
156	classtab = 0;
157    }
158}
159#endif
160
161#endif /* OPT_WIDE_CHARS */
162