charclass.c revision 2eaa94a1 
1/* $XTermId: charclass.c,v 1.21 2008/12/30 17:35:09 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, (unsigned) classtab[0].cclass, classtab);
55	if (!classtab)
56	    abort();
57    }
58
59    /* simply append new interval to end of interval array */
60    classtab[0].last++;
61    classtab[classtab[0].last].first = low;
62    classtab[classtab[0].last].last = high;
63    classtab[classtab[0].last].cclass = value;
64
65    return 0;
66}
67
68typedef enum {
69    IDENT = -1,
70    ALNUM = 48,
71    CNTRL = 1,
72    BLANK = 32
73} Classes;
74
75void
76init_classtab(void)
77{
78    const int size = 50;
79
80    classtab = TypeMallocN(struct classentry, size);
81    if (!classtab)
82	abort();
83    classtab[0].cclass = size;
84    classtab[0].first = 1;
85    classtab[0].last = 0;
86
87    /* old xterm default classes */
88    SetCharacterClassRange(0, 0, BLANK);
89    SetCharacterClassRange(1, 31, CNTRL);
90    SetCharacterClassRange('\t', '\t', BLANK);
91    SetCharacterClassRange('0', '9', ALNUM);
92    SetCharacterClassRange('A', 'Z', ALNUM);
93    SetCharacterClassRange('_', '_', ALNUM);
94    SetCharacterClassRange('a', 'z', ALNUM);
95    SetCharacterClassRange(127, 159, CNTRL);
96    SetCharacterClassRange(160, 191, IDENT);
97    SetCharacterClassRange(192, 255, ALNUM);
98    SetCharacterClassRange(215, 215, IDENT);
99    SetCharacterClassRange(247, 247, IDENT);
100
101    /* added Unicode classes */
102    SetCharacterClassRange(0x0100, 0xffdf, ALNUM);	/* mostly characters */
103    SetCharacterClassRange(0x037e, 0x037e, IDENT);	/* Greek question mark */
104    SetCharacterClassRange(0x0387, 0x0387, IDENT);	/* Greek ano teleia */
105    SetCharacterClassRange(0x055a, 0x055f, IDENT);	/* Armenian punctuation */
106    SetCharacterClassRange(0x0589, 0x0589, IDENT);	/* Armenian full stop */
107    SetCharacterClassRange(0x0700, 0x070d, IDENT);	/* Syriac punctuation */
108    SetCharacterClassRange(0x104a, 0x104f, IDENT);	/* Myanmar punctuation */
109    SetCharacterClassRange(0x10fb, 0x10fb, IDENT);	/* Georgian punctuation */
110    SetCharacterClassRange(0x1361, 0x1368, IDENT);	/* Ethiopic punctuation */
111    SetCharacterClassRange(0x166d, 0x166e, IDENT);	/* Canadian Syl. punctuation */
112    SetCharacterClassRange(0x17d4, 0x17dc, IDENT);	/* Khmer punctuation */
113    SetCharacterClassRange(0x1800, 0x180a, IDENT);	/* Mongolian punctuation */
114    SetCharacterClassRange(0x2000, 0x200a, BLANK);	/* spaces */
115    SetCharacterClassRange(0x200b, 0x27ff, IDENT);	/* punctuation and symbols */
116    SetCharacterClassRange(0x2070, 0x207f, 0x2070);	/* superscript */
117    SetCharacterClassRange(0x2080, 0x208f, 0x2080);	/* subscript */
118    SetCharacterClassRange(0x3000, 0x3000, BLANK);	/* ideographic space */
119    SetCharacterClassRange(0x3001, 0x3020, IDENT);	/* ideographic punctuation */
120    SetCharacterClassRange(0x3040, 0x309f, 0x3040);	/* Hiragana */
121    SetCharacterClassRange(0x30a0, 0x30ff, 0x30a0);	/* Katakana */
122    SetCharacterClassRange(0x3300, 0x9fff, 0x4e00);	/* CJK Ideographs */
123    SetCharacterClassRange(0xac00, 0xd7a3, 0xac00);	/* Hangul Syllables */
124    SetCharacterClassRange(0xf900, 0xfaff, 0x4e00);	/* CJK Ideographs */
125    SetCharacterClassRange(0xfe30, 0xfe6b, IDENT);	/* punctuation forms */
126    SetCharacterClassRange(0xff00, 0xff0f, IDENT);	/* half/fullwidth ASCII */
127    SetCharacterClassRange(0xff1a, 0xff20, IDENT);	/* half/fullwidth ASCII */
128    SetCharacterClassRange(0xff3b, 0xff40, IDENT);	/* half/fullwidth ASCII */
129    SetCharacterClassRange(0xff5b, 0xff64, IDENT);	/* half/fullwidth ASCII */
130
131    return;
132}
133
134int
135CharacterClass(int c)
136{
137    int i, cclass = IDENT;
138
139    for (i = classtab[0].first; i <= classtab[0].last; i++)
140	if (classtab[i].first <= c && classtab[i].last >= c)
141	    cclass = classtab[i].cclass;
142
143    if (cclass < 0)
144	cclass = c;
145
146    return cclass;
147}
148
149#ifdef NO_LEAKS
150void
151noleaks_CharacterClass(void)
152{
153    if (classtab != 0) {
154	free(classtab);
155	classtab = 0;
156    }
157}
158#endif
159
160#endif /* OPT_WIDE_CHARS */
161