1 /*
2 * "streamable kanji code filter and converter"
3 * Copyright (c) 1998-2002 HappySize, Inc. All rights reserved.
4 *
5 * LICENSE NOTICES
6 *
7 * This file is part of "streamable kanji code filter and converter",
8 * which is distributed under the terms of GNU Lesser General Public
9 * License (version 2) as published by the Free Software Foundation.
10 *
11 * This software is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with "streamable kanji code filter and converter";
18 * if not, write to the Free Software Foundation, Inc., 59 Temple Place,
19 * Suite 330, Boston, MA 02111-1307 USA
20 *
21 * The author of this file:
22 *
23 */
24 /*
25 * the source code included in this files was separated from mbfilter_cp936.c
26 * by rui hirokawa <hirokawa@php.net> on 11 Aug 2011.
27 *
28 */
29
30 #include "mbfilter.h"
31 #include "mbfilter_gb18030.h"
32
33 #include "unicode_table_cp936.h"
34 #include "unicode_table_gb18030.h"
35
36 static int mbfl_filt_ident_gb18030(int c, mbfl_identify_filter *filter);
37
38 static const char *mbfl_encoding_gb18030_aliases[] = {"gb-18030", "gb-18030-2000", NULL};
39
40 const mbfl_encoding mbfl_encoding_gb18030 = {
41 mbfl_no_encoding_gb18030,
42 "GB18030",
43 "GB18030",
44 (const char *(*)[])&mbfl_encoding_gb18030_aliases,
45 NULL,
46 MBFL_ENCTYPE_MBCS | MBFL_ENCTYPE_GL_UNSAFE,
47 &vtbl_gb18030_wchar,
48 &vtbl_wchar_gb18030
49 };
50
51 const struct mbfl_identify_vtbl vtbl_identify_gb18030 = {
52 mbfl_no_encoding_gb18030,
53 mbfl_filt_ident_common_ctor,
54 mbfl_filt_ident_gb18030
55 };
56
57 const struct mbfl_convert_vtbl vtbl_gb18030_wchar = {
58 mbfl_no_encoding_gb18030,
59 mbfl_no_encoding_wchar,
60 mbfl_filt_conv_common_ctor,
61 NULL,
62 mbfl_filt_conv_gb18030_wchar,
63 mbfl_filt_conv_common_flush,
64 NULL,
65 };
66
67 const struct mbfl_convert_vtbl vtbl_wchar_gb18030 = {
68 mbfl_no_encoding_wchar,
69 mbfl_no_encoding_gb18030,
70 mbfl_filt_conv_common_ctor,
71 NULL,
72 mbfl_filt_conv_wchar_gb18030,
73 mbfl_filt_conv_common_flush,
74 NULL,
75 };
76
77 #define CK(statement) do { if ((statement) < 0) return (-1); } while (0)
78
79
80 int
mbfl_bisec_srch(int w,const unsigned short * tbl,int n)81 mbfl_bisec_srch(int w, const unsigned short *tbl, int n)
82 {
83 int k, k1 = 0, k2 = n-1;
84
85 while (k1 < k2) {
86 k = (k1+k2) >> 1;
87 if (w <= tbl[2*k+1]) {
88 k2 = k;
89 } else if (w >= tbl[2*k+2]) {
90 k1 = k + 1;
91 } else {
92 return -1;
93 }
94 }
95 return k1;
96 }
97
98 int
mbfl_bisec_srch2(int w,const unsigned short tbl[],int n)99 mbfl_bisec_srch2(int w, const unsigned short tbl[], int n)
100 {
101 int k, k1 = 0, k2 = n;
102
103 if (w == tbl[0]) {
104 return 0;
105 }
106
107 while (k2 - k1 > 1) {
108 k = (k1 + k2) >> 1;
109 if (w < tbl[k]) {
110 k2 = k;
111 } else if (w > tbl[k]) {
112 k1 = k;
113 } else {
114 return k;
115 }
116 }
117 return -1;
118 }
119
120 /*
121 * GB18030 => wchar
122 */
123 int
mbfl_filt_conv_gb18030_wchar(int c,mbfl_convert_filter * filter)124 mbfl_filt_conv_gb18030_wchar(int c, mbfl_convert_filter *filter)
125 {
126 int k;
127 int c1, c2, c3, w = -1;
128
129 switch (filter->status) {
130 case 0:
131 if (c >= 0 && c < 0x80) { /* latin */
132 CK((*filter->output_function)(c, filter->data));
133 } else if (c == 0x80) { /* euro sign */
134 CK((*filter->output_function)(0x20ac, filter->data));
135 } else if (c == 0xff) {
136 CK((*filter->output_function)(0x00ff, filter->data));
137 } else if (c > 0x80 && c < 0xff) { /* dbcs/qbcs lead byte */
138 filter->status = 1;
139 filter->cache = c;
140 } else {
141 w = c & MBFL_WCSGROUP_MASK;
142 w |= MBFL_WCSGROUP_THROUGH;
143 CK((*filter->output_function)(w, filter->data));
144 }
145 break;
146
147 case 1: /* dbcs/qbcs second byte */
148 c1 = filter->cache;
149 filter->status = 0;
150
151 if (c1 >= 0x81 && c1 <= 0x84 && c >= 0x30 && c <= 0x39) { /* 4 byte range: Unicode BMP */
152 filter->status = 2;
153 filter->cache = (c1 << 8) | c;
154 return c;
155 } else if (c1 >= 0x90 && c1 <= 0xe3 && c >= 0x30 && c <= 0x39) {
156 /* 4 byte range: Unicode 16 planes */
157 filter->status = 2;
158 filter->cache = (c1 << 8) | c;
159 return c;
160 } else if (((c1 >= 0xaa && c1 <= 0xaf) || (c1 >= 0xf8 && c1 <= 0xfe)) &&
161 (c >= 0xa1 && c <= 0xfe)) { /* UDA part1,2: U+E000-U+E4C5 */
162 w = 94*(c1 >= 0xf8 ? c1 - 0xf2 : c1 - 0xaa) + (c - 0xa1) + 0xe000;
163 CK((*filter->output_function)(w, filter->data));
164 } else if (c1 >= 0xa1 && c1 <= 0xa7 && c >= 0x40 && c < 0xa1 && c != 0x7f) {
165 /* UDA part3 : U+E4C6-U+E765*/
166 w = 96*(c1 - 0xa1) + c - (c >= 0x80 ? 0x41 : 0x40) + 0xe4c6;
167 CK((*filter->output_function)(w, filter->data));
168 }
169
170 c2 = (c1 << 8) | c;
171
172 if (w <= 0 &&
173 ((c2 >= 0xa2ab && c2 <= 0xa9f0 + (0xe80f-0xe801)) ||
174 (c2 >= 0xd7fa && c2 <= 0xd7fa + (0xe814-0xe810)) ||
175 (c2 >= 0xfe50 && c2 <= 0xfe80 + (0xe864-0xe844)))) {
176 for (k = 0; k < mbfl_gb18030_pua_tbl_max; k++) {
177 if (c2 >= mbfl_gb18030_pua_tbl[k][2] &&
178 c2 <= mbfl_gb18030_pua_tbl[k][2] + mbfl_gb18030_pua_tbl[k][1]
179 - mbfl_gb18030_pua_tbl[k][0]) {
180 w = c2 - mbfl_gb18030_pua_tbl[k][2] + mbfl_gb18030_pua_tbl[k][0];
181 CK((*filter->output_function)(w, filter->data));
182 break;
183 }
184 }
185 }
186
187 if (w <= 0) {
188 if ((c1 >= 0xa1 && c1 <= 0xa9 && c >= 0xa1 && c <= 0xfe) ||
189 (c1 >= 0xb0 && c1 <= 0xf7 && c >= 0xa1 && c <= 0xfe) ||
190 (c1 >= 0x81 && c1 <= 0xa0 && c >= 0x40 && c <= 0xfe && c != 0x7f) ||
191 (c1 >= 0xaa && c1 <= 0xfe && c >= 0x40 && c <= 0xa0 && c != 0x7f) ||
192 (c1 >= 0xa8 && c1 <= 0xa9 && c >= 0x40 && c <= 0xa0 && c != 0x7f)) {
193 w = (c1 - 0x81)*192 + (c - 0x40);
194 if (w >= 0 && w < cp936_ucs_table_size) {
195 w = cp936_ucs_table[w];
196 } else {
197 w = 0;
198 }
199 if (w <= 0) {
200 w = (c1 << 8) | c;
201 w &= MBFL_WCSPLANE_MASK;
202 w |= MBFL_WCSPLANE_GB18030;
203 }
204 CK((*filter->output_function)(w, filter->data));
205 } else if ((c >= 0 && c < 0x21) || c == 0x7f) { /* CTLs */
206 CK((*filter->output_function)(c, filter->data));
207 } else {
208 w = (c1 << 8) | c;
209 w &= MBFL_WCSGROUP_MASK;
210 w |= MBFL_WCSGROUP_THROUGH;
211 CK((*filter->output_function)(w, filter->data));
212 }
213 }
214 break;
215 case 2: /* qbcs third byte */
216 c1 = (filter->cache >> 8) & 0xff;
217 c2 = filter->cache & 0xff;
218 filter->status = 0;
219 filter->cache = 0;
220 if (((c1 >= 0x81 && c1 <= 0x84) || (c1 >= 0x90 && c1 <= 0xe3)) &&
221 c2 >= 0x30 && c2 <= 0x39 && c >= 0x81 && c <= 0xfe) {
222 filter->cache = (c1 << 16) | (c2 << 8) | c;
223 filter->status = 3;
224 } else {
225 w = (c1 << 16) | (c2 << 8) | c;
226 w &= MBFL_WCSGROUP_MASK;
227 w |= MBFL_WCSGROUP_THROUGH;
228 CK((*filter->output_function)(w, filter->data));
229 }
230 break;
231
232 case 3: /* qbcs fourth byte */
233 c1 = (filter->cache >> 16) & 0xff;
234 c2 = (filter->cache >> 8) & 0xff;
235 c3 = filter->cache & 0xff;
236 filter->status = 0;
237 filter->cache = 0;
238 if (((c1 >= 0x81 && c1 <= 0x84) || (c1 >= 0x90 && c1 <= 0xe3)) &&
239 c2 >= 0x30 && c2 <= 0x39 && c3 >= 0x81 && c3 <= 0xfe && c >= 0x30 && c <= 0x39) {
240 if (c1 >= 0x90 && c1 <= 0xe3) {
241 w = ((((c1 - 0x90)*10 + (c2 - 0x30))*126 + (c3 - 0x81)))*10 + (c - 0x30) + 0x10000;
242 } else { /* Unicode BMP */
243 w = (((c1 - 0x81)*10 + (c2 - 0x30))*126 + (c3 - 0x81))*10 + (c - 0x30);
244 if (w >= 0 && w <= 39419) {
245 k = mbfl_bisec_srch(w, mbfl_gb2uni_tbl, mbfl_gb_uni_max);
246 if (k<0) {
247 /* error */
248 w = (c1 << 24) | (c2 << 16) | (c3 << 8) | c;
249 w &= MBFL_WCSGROUP_MASK;
250 w |= MBFL_WCSGROUP_THROUGH;
251 CK((*filter->output_function)(w, filter->data));
252 return c;
253 }
254 w += mbfl_gb_uni_ofst[k];
255 } else {
256 w = (c1 << 24) | (c2 << 16) | (c3 << 8) | c;
257 w &= MBFL_WCSGROUP_MASK;
258 w |= MBFL_WCSGROUP_THROUGH;
259 CK((*filter->output_function)(w, filter->data));
260 return c;
261 }
262 }
263 CK((*filter->output_function)(w, filter->data));
264 } else {
265 w = (c1 << 24) | (c2 << 16) | (c3 << 8) | c;
266 w &= MBFL_WCSGROUP_MASK;
267 w |= MBFL_WCSGROUP_THROUGH;
268 CK((*filter->output_function)(w, filter->data));
269 }
270 break;
271
272 default:
273 filter->status = 0;
274 break;
275 }
276
277 return c;
278 }
279
280 /*
281 * wchar => GB18030
282 */
283 int
mbfl_filt_conv_wchar_gb18030(int c,mbfl_convert_filter * filter)284 mbfl_filt_conv_wchar_gb18030(int c, mbfl_convert_filter *filter)
285 {
286 int k, k1, k2;
287 int c1, s = 0, s1 = 0;
288
289 if (c >= ucs_a1_cp936_table_min && c < ucs_a1_cp936_table_max) {
290 s = ucs_a1_cp936_table[c - ucs_a1_cp936_table_min];
291 } else if (c >= ucs_a2_cp936_table_min && c < ucs_a2_cp936_table_max) {
292 s = ucs_a2_cp936_table[c - ucs_a2_cp936_table_min];
293 } else if (c >= ucs_a3_cp936_table_min && c < ucs_a3_cp936_table_max) {
294 s = ucs_a3_cp936_table[c - ucs_a3_cp936_table_min];
295 } else if (c >= ucs_i_cp936_table_min && c < ucs_i_cp936_table_max) {
296 s = ucs_i_cp936_table[c - ucs_i_cp936_table_min];
297 } else if (c >= ucs_ci_cp936_table_min && c < ucs_ci_cp936_table_max) {
298 /* U+F900-FA2F CJK Compatibility Ideographs */
299 if (c == 0xf92c) {
300 s = 0xfd9c;
301 } else if (c == 0xf979) {
302 s = 0xfd9d;
303 } else if (c == 0xf995) {
304 s = 0xfd9e;
305 } else if (c == 0xf9e7) {
306 s = 0xfd9f;
307 } else if (c == 0xf9f1) {
308 s = 0xfda0;
309 } else if (c >= 0xfa0c && c <= 0xfa29) {
310 s = ucs_ci_s_cp936_table[c - 0xfa0c];
311 }
312 } else if (c >= ucs_cf_cp936_table_min && c < ucs_cf_cp936_table_max) {
313 /* FE30h CJK Compatibility Forms */
314 s = ucs_cf_cp936_table[c - ucs_cf_cp936_table_min];
315 } else if (c >= ucs_sfv_cp936_table_min && c < ucs_sfv_cp936_table_max) {
316 /* U+FE50-FE6F Small Form Variants */
317 s = ucs_sfv_cp936_table[c - ucs_sfv_cp936_table_min];
318 } else if (c >= ucs_hff_cp936_table_min && c < ucs_hff_cp936_table_max) {
319 /* U+FF00-FFFF HW/FW Forms */
320 if (c == 0xff04) {
321 s = 0xa1e7;
322 } else if (c == 0xff5e) {
323 s = 0xa1ab;
324 } else if (c >= 0xff01 && c <= 0xff5d) {
325 s = c - 0xff01 + 0xa3a1;
326 } else if (c >= 0xffe0 && c <= 0xffe5) {
327 s = ucs_hff_s_cp936_table[c-0xffe0];
328 }
329 }
330
331 if (c == 0x20ac) { /* euro-sign */
332 s = 0xa2e3;
333 }
334
335 if (s <= 0 && c >= mbfl_gb18030_c_tbl_key[0] &&
336 c <= mbfl_gb18030_c_tbl_key[mbfl_gb18030_c_tbl_max-1]) {
337 k1 = mbfl_bisec_srch2(c, mbfl_gb18030_c_tbl_key, mbfl_gb18030_c_tbl_max);
338 if (k1 >= 0) {
339 s = mbfl_gb18030_c_tbl_val[k1];
340 }
341 }
342
343 if (c >= 0xe000 && c <= 0xe864) { /* PUA */
344 if (c < 0xe766) {
345 if (c < 0xe4c6) {
346 c1 = c - 0xe000;
347 s = (c1 % 94) + 0xa1; c1 /= 94;
348 s |= (c1 < 0x06 ? c1 + 0xaa : c1 + 0xf2) << 8;
349 } else {
350 c1 = c - 0xe4c6;
351 s = ((c1 / 96) + 0xa1) << 8; c1 %= 96;
352 s |= c1 + (c1 >= 0x3f ? 0x41 : 0x40);
353 }
354 } else {
355 /* U+E766..U+E864 */
356 k1 = 0; k2 = mbfl_gb18030_pua_tbl_max;
357 while (k1 < k2) {
358 k = (k1 + k2) >> 1;
359 if (c < mbfl_gb18030_pua_tbl[k][0]) {
360 k2 = k;
361 } else if (c > mbfl_gb18030_pua_tbl[k][1]) {
362 k1 = k + 1;
363 } else {
364 s = c - mbfl_gb18030_pua_tbl[k][0] + mbfl_gb18030_pua_tbl[k][2];
365 break;
366 }
367 }
368 }
369 }
370
371 if (s <= 0 && c >= 0x0080 && c <= 0xffff) { /* BMP */
372 s = mbfl_bisec_srch(c, mbfl_uni2gb_tbl, mbfl_gb_uni_max);
373 if (s >= 0) {
374 c1 = c - mbfl_gb_uni_ofst[s];
375 s = (c1 % 10) + 0x30; c1 /= 10;
376 s |= ((c1 % 126) + 0x81) << 8; c1 /= 126;
377 s |= ((c1 % 10) + 0x30) << 16; c1 /= 10;
378 s1 = c1 + 0x81;
379 }
380 } else if (c >= 0x10000 && c <= 0x10ffff) { /* Code set 3: Unicode U+10000..U+10FFFF */
381 c1 = c - 0x10000;
382 s = (c1 % 10) + 0x30; c1 /= 10;
383 s |= ((c1 % 126) + 0x81) << 8; c1 /= 126;
384 s |= ((c1 % 10) + 0x30) << 16; c1 /= 10;
385 s1 = c1 + 0x90;
386 }
387
388 if (s <= 0) {
389 c1 = c & ~MBFL_WCSPLANE_MASK;
390 if (c1 == MBFL_WCSPLANE_WINCP936) {
391 s = c & MBFL_WCSPLANE_MASK;
392 }
393 if (c == 0) {
394 s = 0;
395 } else if (s <= 0) {
396 s = -1;
397 }
398 }
399 if (s >= 0) {
400 if (s <= 0x80) { /* latin */
401 CK((*filter->output_function)(s, filter->data));
402 } else if (s1 > 0) { /* qbcs */
403 CK((*filter->output_function)(s1 & 0xff, filter->data));
404 CK((*filter->output_function)((s >> 16) & 0xff, filter->data));
405 CK((*filter->output_function)((s >> 8) & 0xff, filter->data));
406 CK((*filter->output_function)(s & 0xff, filter->data));
407 } else { /* dbcs */
408 CK((*filter->output_function)((s >> 8) & 0xff, filter->data));
409 CK((*filter->output_function)(s & 0xff, filter->data));
410 }
411 } else {
412 CK(mbfl_filt_conv_illegal_output(c, filter));
413 }
414
415 return c;
416 }
417
mbfl_filt_ident_gb18030(int c,mbfl_identify_filter * filter)418 static int mbfl_filt_ident_gb18030(int c, mbfl_identify_filter *filter)
419 {
420 int c1;
421
422 c1 = (filter->status >> 8) & 0xff;
423 filter->status &= 0xff;
424
425 if (filter->status == 0) {
426 if (c <= 0x80 || c == 0xff) {
427 filter->status = 0;
428 } else {
429 filter->status = 1;
430 filter->status |= (c << 8);
431 }
432 } else if (filter->status == 1) { /* dbcs/qbcs 2nd byte */
433 if (((c1 >= 0x81 && c1 <= 0x84) || (c1 >= 0x90 && c1 <= 0xe3)) && c >= 0x30 && c <= 0x39) { /* qbcs */
434 filter->status = 2;
435 } else if (((c1 >= 0xaa && c1 <= 0xaf) || (c1 >= 0xf8 && c1 <= 0xfe)) && (c >= 0xa1 && c <= 0xfe)) {
436 filter->status = 0; /* UDA part 1,2 */
437 } else if (c1 >= 0xa1 && c1 <= 0xa7 && c >= 0x40 && c < 0xa1 && c != 0x7f) {
438 filter->status = 0; /* UDA part 3 */
439 } else if ((c1 >= 0xa1 && c1 <= 0xa9 && c >= 0xa1 && c <= 0xfe) ||
440 (c1 >= 0xb0 && c1 <= 0xf7 && c >= 0xa1 && c <= 0xfe) ||
441 (c1 >= 0x81 && c1 <= 0xa0 && c >= 0x40 && c <= 0xfe && c != 0x7f) ||
442 (c1 >= 0xaa && c1 <= 0xfe && c >= 0x40 && c <= 0xa0 && c != 0x7f) ||
443 (c1 >= 0xa8 && c1 <= 0xa9 && c >= 0x40 && c <= 0xa0 && c != 0x7f)) {
444 filter->status = 0; /* DBCS */
445 } else {
446 filter->flag = 1; /* bad */
447 filter->status = 0;
448 }
449 } else if (filter->status == 2) { /* qbcs 3rd byte */
450 if (c > 0x80 && c < 0xff) {
451 filter->status = 3;
452 } else {
453 filter->flag = 1; /* bad */
454 filter->status = 0;
455 }
456 } else if (filter->status == 3) { /* qbcs 4th byte */
457 if (c >= 0x30 && c < 0x40) {
458 filter->status = 0;
459 } else {
460 filter->flag = 1; /* bad */
461 filter->status = 0;
462 }
463 } else { /* bad */
464 filter->flag = 1;
465 }
466
467 return c;
468 }
469
