root / lib / filters / wordsep-filter.c @ 2d9cf61c
History | View | Annotate | Download (8.9 kB)
| 1 | /* SylFilter - a message filter
|
|---|---|
| 2 | * |
| 3 | * Copyright (C) 2011 Hiroyuki Yamamoto |
| 4 | * Copyright (C) 2011 Sylpheed Development Team |
| 5 | */ |
| 6 | |
| 7 | #include <glib.h> |
| 8 | #include <string.h> |
| 9 | |
| 10 | #include "filter.h" |
| 11 | #include "wordsep-filter.h" |
| 12 | |
| 13 | typedef enum |
| 14 | {
|
| 15 | UC_OTHER, |
| 16 | UC_HIRA, |
| 17 | UC_KATA, |
| 18 | UC_HAN |
| 19 | } UCType; |
| 20 | |
| 21 | #define MAX_TOKEN_LEN 42 |
| 22 | #define NGRAM_LEN 4 /* 4-gram */ |
| 23 | |
| 24 | #define URANGE(c, start, end) ((c) >= start && (c) <= end)
|
| 25 | |
| 26 | #define u_is_hira(c) (URANGE(c, 0x3041, 0x3096) || \ |
| 27 | URANGE(c, 0x309d, 0x309f)) |
| 28 | #define u_is_kata(c) (URANGE(c, 0x30a1, 0x30fa) || \ |
| 29 | URANGE(c, 0x30fd, 0x30ff) || \ |
| 30 | URANGE(c, 0x31f0, 0x31ff) || \ |
| 31 | URANGE(c, 0xff66, 0xff6f) || \ |
| 32 | URANGE(c, 0xff71, 0xff9d)) |
| 33 | #define u_is_han(c) (URANGE(c, 0x2e80, 0x2fdf) || \ |
| 34 | (c) == 0x3005 || \
|
| 35 | (c) == 0x3007 || \
|
| 36 | URANGE(c, 0x3021, 0x3029) || \ |
| 37 | URANGE(c, 0x3038, 0x303b) || \ |
| 38 | URANGE(c, 0x3400, 0x4dbf) || \ |
| 39 | URANGE(c, 0x4e00, 0x9fff) || \ |
| 40 | URANGE(c, 0xf900, 0xfaff) || \ |
| 41 | URANGE(c, 0x20000, 0x2ffff)) |
| 42 | |
| 43 | #define APPEND_STR(s, w) \
|
| 44 | { \
|
| 45 | if (s->len > 0) \ |
| 46 | g_string_append_c(s, ' '); \
|
| 47 | if (is_url) \
|
| 48 | g_string_append(s, "Url*"); \
|
| 49 | g_string_append(s, w); \ |
| 50 | } |
| 51 | |
| 52 | |
| 53 | static void append_ngram_str(GString *str, const char *word, gboolean is_url) |
| 54 | {
|
| 55 | const char *p = word; |
| 56 | const char *bp; |
| 57 | int len;
|
| 58 | |
| 59 | xfilter_debug_print("append_ngram_str: %s\n", word);
|
| 60 | |
| 61 | while (*p != '\0') { |
| 62 | len = 0;
|
| 63 | bp = p; |
| 64 | while (len < NGRAM_LEN) {
|
| 65 | p = g_utf8_next_char(p); |
| 66 | len++; |
| 67 | if (*p == '\0') |
| 68 | break;
|
| 69 | } |
| 70 | if (len < NGRAM_LEN)
|
| 71 | break;
|
| 72 | |
| 73 | if (str->len > 0) |
| 74 | g_string_append_c(str, ' ');
|
| 75 | if (is_url)
|
| 76 | g_string_append(str, "Url*");
|
| 77 | g_string_append_len(str, bp, p - bp); |
| 78 | xfilter_debug_print("n-gram: %.*s\n", p - bp, bp);
|
| 79 | |
| 80 | p = g_utf8_next_char(bp); |
| 81 | } |
| 82 | } |
| 83 | |
| 84 | static char *do_wordsep(const char *content, gboolean drop_punct, |
| 85 | gboolean drop_single_letter, gboolean drop_two_hira, |
| 86 | gboolean normalize, gboolean mb_ngram, |
| 87 | gboolean is_url, gboolean is_header) |
| 88 | {
|
| 89 | GString *out_str; |
| 90 | const char *p = content; |
| 91 | const char *bp = content; |
| 92 | UCType prev_uctype = UC_OTHER; |
| 93 | UCType prev2_uctype = UC_OTHER; |
| 94 | GUnicodeType prev_utype = G_UNICODE_OTHER_LETTER; |
| 95 | GUnicodeBreakType prev_btype; |
| 96 | gboolean prev_is_wide = FALSE; |
| 97 | gboolean prev_is_space = FALSE; |
| 98 | gboolean prev_is_punct = FALSE; |
| 99 | gboolean prev_is_digit = FALSE; |
| 100 | gint word_len = 0;
|
| 101 | |
| 102 | out_str = g_string_new("");
|
| 103 | |
| 104 | while (*p != '\0') { |
| 105 | gunichar wc; |
| 106 | UCType uctype = UC_OTHER; |
| 107 | GUnicodeType utype = G_UNICODE_OTHER_LETTER; |
| 108 | GUnicodeBreakType btype; |
| 109 | gboolean is_wide, is_space, is_punct, is_digit; |
| 110 | gboolean do_break = FALSE; |
| 111 | gboolean is_drop = FALSE; |
| 112 | gboolean do_ngram = FALSE; |
| 113 | |
| 114 | if (!is_header && !is_url && p == bp && *p == 'h') { |
| 115 | if (!strncmp(p, "http://", 7)) |
| 116 | p += 7;
|
| 117 | else if (!strncmp(p, "https://", 8)) |
| 118 | p += 8;
|
| 119 | if (p > bp) {
|
| 120 | if (!strncmp(p, "www.", 4)) |
| 121 | p += 4;
|
| 122 | } |
| 123 | |
| 124 | if (p > bp) {
|
| 125 | const char *ep = p; |
| 126 | char *url_str, *sep_str;
|
| 127 | |
| 128 | while (*ep != '\0' && g_ascii_isgraph(*ep)) |
| 129 | ep++; |
| 130 | url_str = g_strndup(p, ep - p); |
| 131 | sep_str = do_wordsep(url_str, drop_punct, drop_single_letter, drop_two_hira, normalize, mb_ngram, TRUE, is_header); |
| 132 | if (out_str->len > 0) |
| 133 | g_string_append_c(out_str, ' ');
|
| 134 | g_string_append(out_str, sep_str); |
| 135 | g_free(url_str); |
| 136 | word_len = 0;
|
| 137 | bp = p = ep; |
| 138 | continue;
|
| 139 | } |
| 140 | } |
| 141 | |
| 142 | wc = g_utf8_get_char(p); |
| 143 | utype = g_unichar_type(wc); |
| 144 | btype = g_unichar_break_type(wc); |
| 145 | is_wide = g_unichar_iswide(wc) || wc == 0x25cb ||
|
| 146 | URANGE(wc, 0x25ce, 0x25d1); |
| 147 | is_space = g_unichar_isspace(wc); |
| 148 | is_punct = (*p != '!' && *p != '$' && *p != '\'' && *p != '-') |
| 149 | && g_unichar_ispunct(wc); |
| 150 | is_digit = g_unichar_isdigit(wc); |
| 151 | if (is_wide) {
|
| 152 | if (u_is_hira(wc))
|
| 153 | uctype = UC_HIRA; |
| 154 | else if (u_is_kata(wc)) |
| 155 | uctype = UC_KATA; |
| 156 | else if (u_is_han(wc)) |
| 157 | uctype = UC_HAN; |
| 158 | } |
| 159 | #if 0
|
| 160 | if (utype == G_UNICODE_DASH_PUNCTUATION || |
| 161 | utype == G_UNICODE_OTHER_SYMBOL) |
| 162 | is_punct = FALSE; |
| 163 | #endif |
| 164 | /* handle 1,234,567 / 192.168.0.1 */
|
| 165 | if (*p == '.' || *p == ',') { |
| 166 | if (prev_is_digit) {
|
| 167 | const char *np; |
| 168 | gunichar nwc; |
| 169 | np = g_utf8_next_char(p); |
| 170 | nwc = g_utf8_get_char(np); |
| 171 | if (g_unichar_isdigit(nwc))
|
| 172 | is_punct = FALSE; |
| 173 | } |
| 174 | } |
| 175 | |
| 176 | #if 0
|
| 177 | {
|
| 178 | gchar s[7] = {0};
|
| 179 | g_unichar_to_utf8(wc, s); |
| 180 | xfilter_debug_print("%s: utype=%d word_len=%d\n", s, utype, word_len);
|
| 181 | } |
| 182 | #endif |
| 183 | |
| 184 | if (p > bp) {
|
| 185 | if (prev_is_wide != is_wide)
|
| 186 | do_break = TRUE; |
| 187 | else if (prev_is_punct != is_punct) |
| 188 | do_break = TRUE; |
| 189 | else if (is_space || is_punct) |
| 190 | do_break = TRUE; |
| 191 | else if (prev_uctype != UC_HAN && uctype == UC_HAN) |
| 192 | do_break = TRUE; |
| 193 | else if (prev_uctype == UC_HAN && uctype != UC_HAN) |
| 194 | do_break = TRUE; |
| 195 | else if (prev_uctype == UC_KATA && |
| 196 | uctype != UC_KATA && |
| 197 | utype != G_UNICODE_MODIFIER_LETTER && |
| 198 | utype != G_UNICODE_OTHER_SYMBOL) |
| 199 | do_break = TRUE; |
| 200 | else if (prev_uctype == UC_HIRA && |
| 201 | uctype != UC_HIRA && |
| 202 | utype != G_UNICODE_MODIFIER_LETTER && |
| 203 | utype != G_UNICODE_OTHER_SYMBOL) |
| 204 | do_break = TRUE; |
| 205 | else if (prev_uctype == UC_OTHER && |
| 206 | uctype != UC_OTHER && |
| 207 | prev_utype != G_UNICODE_MODIFIER_LETTER) |
| 208 | do_break = TRUE; |
| 209 | else if (prev_utype == G_UNICODE_MODIFIER_LETTER && |
| 210 | prev2_uctype != uctype) |
| 211 | do_break = TRUE; |
| 212 | } |
| 213 | |
| 214 | if (drop_punct) {
|
| 215 | is_drop = prev_is_punct; |
| 216 | } |
| 217 | if (drop_single_letter) {
|
| 218 | if (word_len == 1 && do_break && prev_uctype != UC_HAN) |
| 219 | is_drop = TRUE; |
| 220 | } |
| 221 | if (drop_two_hira) {
|
| 222 | if (word_len == 2 && do_break && prev2_uctype == UC_HIRA && prev_uctype == UC_HIRA) |
| 223 | is_drop = TRUE; |
| 224 | } |
| 225 | |
| 226 | if (word_len > MAX_TOKEN_LEN && do_break)
|
| 227 | is_drop = TRUE; |
| 228 | |
| 229 | if (mb_ngram && !is_drop && do_break &&
|
| 230 | prev_uctype == UC_HAN && word_len > NGRAM_LEN) |
| 231 | do_ngram = TRUE; |
| 232 | |
| 233 | prev2_uctype = prev_uctype; |
| 234 | prev_uctype = uctype; |
| 235 | prev_utype = utype; |
| 236 | prev_btype = btype; |
| 237 | prev_is_wide = is_wide; |
| 238 | prev_is_space = is_space; |
| 239 | prev_is_punct = is_punct; |
| 240 | prev_is_digit = is_digit; |
| 241 | |
| 242 | if (do_break) {
|
| 243 | gchar *word; |
| 244 | |
| 245 | word_len = 0;
|
| 246 | |
| 247 | if (!is_drop) {
|
| 248 | if (normalize)
|
| 249 | word = g_ascii_strdown(bp, p - bp); |
| 250 | else
|
| 251 | word = g_strndup(bp, p - bp); |
| 252 | if (do_ngram)
|
| 253 | append_ngram_str(out_str, word, is_url); |
| 254 | else
|
| 255 | APPEND_STR(out_str, word); |
| 256 | g_free(word); |
| 257 | } |
| 258 | |
| 259 | if (is_space) {
|
| 260 | do {
|
| 261 | p = g_utf8_next_char(p); |
| 262 | wc = g_utf8_get_char(p); |
| 263 | } while (g_unichar_isspace(wc));
|
| 264 | bp = p; |
| 265 | continue;
|
| 266 | } |
| 267 | bp = p; |
| 268 | } else {
|
| 269 | if (is_space) {
|
| 270 | word_len = 0;
|
| 271 | do {
|
| 272 | p = g_utf8_next_char(p); |
| 273 | wc = g_utf8_get_char(p); |
| 274 | } while (g_unichar_isspace(wc));
|
| 275 | bp = p; |
| 276 | continue;
|
| 277 | } |
| 278 | } |
| 279 | |
| 280 | p = g_utf8_next_char(p); |
| 281 | word_len++; |
| 282 | } |
| 283 | |
| 284 | if (p > bp) {
|
| 285 | gchar *word; |
| 286 | gboolean is_drop = FALSE; |
| 287 | gboolean do_ngram = FALSE; |
| 288 | |
| 289 | if (drop_punct) {
|
| 290 | is_drop = prev_is_punct; |
| 291 | } |
| 292 | if (drop_single_letter) {
|
| 293 | if (word_len == 1 && prev_uctype != UC_HAN) |
| 294 | is_drop = TRUE; |
| 295 | } |
| 296 | if (drop_two_hira) {
|
| 297 | if (word_len == 2 && prev2_uctype == UC_HIRA && prev_uctype == UC_HIRA) |
| 298 | is_drop = TRUE; |
| 299 | } |
| 300 | |
| 301 | if (word_len > MAX_TOKEN_LEN)
|
| 302 | is_drop = TRUE; |
| 303 | |
| 304 | if (mb_ngram && prev_uctype == UC_HAN && word_len > NGRAM_LEN)
|
| 305 | do_ngram = TRUE; |
| 306 | |
| 307 | if (!is_drop) {
|
| 308 | if (normalize)
|
| 309 | word = g_ascii_strdown(bp, -1);
|
| 310 | else
|
| 311 | word = g_strdup(bp); |
| 312 | if (do_ngram)
|
| 313 | append_ngram_str(out_str, word, is_url); |
| 314 | else
|
| 315 | APPEND_STR(out_str, word); |
| 316 | g_free(word); |
| 317 | } |
| 318 | } |
| 319 | |
| 320 | return g_string_free(out_str, FALSE);
|
| 321 | } |
| 322 | |
| 323 | static XFilterStatus xfilter_wordsep_func(XFilter *filter, const XMessageData *data, XFilterResult *result) |
| 324 | {
|
| 325 | const char *mime_type; |
| 326 | const char *content = NULL; |
| 327 | char *processed_content;
|
| 328 | XMessageData *msgdata; |
| 329 | |
| 330 | g_return_val_if_fail(result != NULL, XF_ERROR);
|
| 331 | |
| 332 | mime_type = xfilter_message_data_get_mime_type(data); |
| 333 | if (!mime_type) {
|
| 334 | xfilter_result_set_status(result, XF_UNSUPPORTED_TYPE); |
| 335 | return XF_UNSUPPORTED_TYPE;
|
| 336 | } |
| 337 | |
| 338 | if (!g_strncasecmp(mime_type, "text/", 5)) |
| 339 | content = xfilter_message_data_get_content(data); |
| 340 | else {
|
| 341 | xfilter_result_set_status(result, XF_UNSUPPORTED_TYPE); |
| 342 | return XF_UNSUPPORTED_TYPE;
|
| 343 | } |
| 344 | |
| 345 | processed_content = do_wordsep(content, TRUE, TRUE, TRUE, FALSE, TRUE, FALSE, FALSE); |
| 346 | |
| 347 | msgdata = xfilter_message_data_new(NULL, mime_type);
|
| 348 | xfilter_message_data_set_content(msgdata, processed_content); |
| 349 | |
| 350 | #define WORDSEP_HEADER(attr) \
|
| 351 | content = xfilter_message_data_get_attribute(data, attr); \ |
| 352 | if (content) { \
|
| 353 | xfilter_debug_print("header: %s\n", content); \
|
| 354 | processed_content = \ |
| 355 | do_wordsep(content, TRUE, TRUE, TRUE, FALSE, TRUE, FALSE, TRUE); \ |
| 356 | xfilter_message_data_set_attribute \ |
| 357 | (msgdata, attr, processed_content, FALSE); \ |
| 358 | g_free(processed_content); \ |
| 359 | } |
| 360 | |
| 361 | WORDSEP_HEADER(XM_FROM); |
| 362 | WORDSEP_HEADER(XM_TO); |
| 363 | WORDSEP_HEADER(XM_CC); |
| 364 | WORDSEP_HEADER(XM_SUBJECT); |
| 365 | WORDSEP_HEADER(XM_RECEIVED); |
| 366 | |
| 367 | #undef WORDSEP_HEADER
|
| 368 | |
| 369 | xfilter_result_set_message_data(result, msgdata); |
| 370 | |
| 371 | xfilter_result_set_status(result, XF_REWRITTEN); |
| 372 | |
| 373 | return XF_REWRITTEN;
|
| 374 | } |
| 375 | |
| 376 | XFilter *xfilter_wordsep_new(void)
|
| 377 | {
|
| 378 | XFilter *filter; |
| 379 | |
| 380 | filter = xfilter_new(XF_CONTENT, "wordsep");
|
| 381 | xfilter_set_content_filter_func(X_CONTENT_FILTER(filter), |
| 382 | xfilter_wordsep_func); |
| 383 | |
| 384 | return filter;
|
| 385 | } |