Annotation of mandoc/mansearch.c, Revision 1.38
1.38 ! schwarze 1: /* $Id: mansearch.c,v 1.37 2014/06/20 02:24:40 schwarze Exp $ */
1.1 kristaps 2: /*
3: * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.13 schwarze 4: * Copyright (c) 2013, 2014 Ingo Schwarze <schwarze@openbsd.org>
1.1 kristaps 5: *
6: * Permission to use, copy, modify, and distribute this software for any
7: * purpose with or without fee is hereby granted, provided that the above
8: * copyright notice and this permission notice appear in all copies.
9: *
10: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
13: * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14: * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15: * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16: * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17: */
18: #ifdef HAVE_CONFIG_H
19: #include "config.h"
20: #endif
21:
1.28 schwarze 22: #include <sys/mman.h>
1.1 kristaps 23: #include <assert.h>
24: #include <fcntl.h>
25: #include <getopt.h>
1.6 schwarze 26: #include <limits.h>
1.8 schwarze 27: #include <regex.h>
1.1 kristaps 28: #include <stdio.h>
29: #include <stdint.h>
30: #include <stddef.h>
31: #include <stdlib.h>
32: #include <string.h>
33: #include <unistd.h>
34:
1.4 kristaps 35: #ifdef HAVE_OHASH
1.1 kristaps 36: #include <ohash.h>
1.4 kristaps 37: #else
38: #include "compat_ohash.h"
39: #endif
1.1 kristaps 40: #include <sqlite3.h>
41:
42: #include "mandoc.h"
1.23 schwarze 43: #include "mandoc_aux.h"
1.1 kristaps 44: #include "manpath.h"
45: #include "mansearch.h"
46:
1.20 schwarze 47: extern int mansearch_keymax;
48: extern const char *const mansearch_keynames[];
49:
1.3 kristaps 50: #define SQL_BIND_TEXT(_db, _s, _i, _v) \
1.8 schwarze 51: do { if (SQLITE_OK != sqlite3_bind_text \
1.2 kristaps 52: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
1.8 schwarze 53: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
54: } while (0)
1.3 kristaps 55: #define SQL_BIND_INT64(_db, _s, _i, _v) \
1.8 schwarze 56: do { if (SQLITE_OK != sqlite3_bind_int64 \
1.2 kristaps 57: ((_s), (_i)++, (_v))) \
1.8 schwarze 58: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
59: } while (0)
60: #define SQL_BIND_BLOB(_db, _s, _i, _v) \
61: do { if (SQLITE_OK != sqlite3_bind_blob \
62: ((_s), (_i)++, (&_v), sizeof(_v), SQLITE_STATIC)) \
63: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
64: } while (0)
1.2 kristaps 65:
1.1 kristaps 66: struct expr {
1.38 ! schwarze 67: regex_t regexp; /* compiled regexp, if applicable */
! 68: const char *substr; /* to search for, if applicable */
! 69: struct expr *next; /* next in sequence */
1.34 schwarze 70: uint64_t bits; /* type-mask */
1.38 ! schwarze 71: int equal; /* equality, not subsring match */
1.13 schwarze 72: int open; /* opening parentheses before */
73: int and; /* logical AND before */
74: int close; /* closing parentheses after */
1.1 kristaps 75: };
76:
77: struct match {
1.32 schwarze 78: uint64_t pageid; /* identifier in database */
1.26 schwarze 79: char *desc; /* manual page description */
1.1 kristaps 80: int form; /* 0 == catpage */
81: };
82:
1.17 schwarze 83: static void buildnames(struct manpage *, sqlite3 *,
1.19 schwarze 84: sqlite3_stmt *, uint64_t,
85: const char *, int form);
1.12 schwarze 86: static char *buildoutput(sqlite3 *, sqlite3_stmt *,
87: uint64_t, uint64_t);
1.1 kristaps 88: static void *hash_alloc(size_t, void *);
1.37 schwarze 89: static void hash_free(void *, void *);
90: static void *hash_calloc(size_t, size_t, void *);
1.34 schwarze 91: static struct expr *exprcomp(const struct mansearch *,
1.5 kristaps 92: int, char *[]);
1.1 kristaps 93: static void exprfree(struct expr *);
1.15 schwarze 94: static struct expr *exprspec(struct expr *, uint64_t,
95: const char *, const char *);
1.8 schwarze 96: static struct expr *exprterm(const struct mansearch *, char *, int);
1.13 schwarze 97: static void sql_append(char **sql, size_t *sz,
98: const char *newstr, int count);
1.7 schwarze 99: static void sql_match(sqlite3_context *context,
100: int argc, sqlite3_value **argv);
1.8 schwarze 101: static void sql_regexp(sqlite3_context *context,
102: int argc, sqlite3_value **argv);
1.15 schwarze 103: static char *sql_statement(const struct expr *);
1.28 schwarze 104:
1.34 schwarze 105:
1.28 schwarze 106: int
107: mansearch_setup(int start)
108: {
109: static void *pagecache;
110: int c;
111:
112: #define PC_PAGESIZE 1280
113: #define PC_NUMPAGES 256
114:
115: if (start) {
116: if (NULL != pagecache) {
117: fprintf(stderr, "pagecache already enabled\n");
118: return((int)MANDOCLEVEL_BADARG);
119: }
120:
121: pagecache = mmap(NULL, PC_PAGESIZE * PC_NUMPAGES,
122: PROT_READ | PROT_WRITE, MAP_ANON, -1, 0);
123:
124: if (MAP_FAILED == pagecache) {
125: perror("mmap");
126: pagecache = NULL;
127: return((int)MANDOCLEVEL_SYSERR);
128: }
129:
130: c = sqlite3_config(SQLITE_CONFIG_PAGECACHE,
131: pagecache, PC_PAGESIZE, PC_NUMPAGES);
132:
133: if (SQLITE_OK == c)
134: return((int)MANDOCLEVEL_OK);
135:
136: fprintf(stderr, "pagecache: %s\n", sqlite3_errstr(c));
137:
138: } else if (NULL == pagecache) {
139: fprintf(stderr, "pagecache missing\n");
140: return((int)MANDOCLEVEL_BADARG);
141: }
142:
143: if (-1 == munmap(pagecache, PC_PAGESIZE * PC_NUMPAGES)) {
144: perror("munmap");
145: pagecache = NULL;
146: return((int)MANDOCLEVEL_SYSERR);
147: }
148:
149: pagecache = NULL;
150: return((int)MANDOCLEVEL_OK);
151: }
1.1 kristaps 152:
153: int
1.5 kristaps 154: mansearch(const struct mansearch *search,
1.12 schwarze 155: const struct manpaths *paths,
156: int argc, char *argv[],
157: const char *outkey,
1.1 kristaps 158: struct manpage **res, size_t *sz)
159: {
1.20 schwarze 160: int fd, rc, c, indexbit;
1.32 schwarze 161: int64_t pageid;
1.20 schwarze 162: uint64_t outbit, iterbit;
1.6 schwarze 163: char buf[PATH_MAX];
1.11 schwarze 164: char *sql;
1.10 schwarze 165: struct manpage *mpage;
1.1 kristaps 166: struct expr *e, *ep;
167: sqlite3 *db;
1.12 schwarze 168: sqlite3_stmt *s, *s2;
1.1 kristaps 169: struct match *mp;
170: struct ohash_info info;
171: struct ohash htab;
172: unsigned int idx;
173: size_t i, j, cur, maxres;
174:
1.37 schwarze 175: info.calloc = hash_calloc;
1.1 kristaps 176: info.alloc = hash_alloc;
1.37 schwarze 177: info.free = hash_free;
1.32 schwarze 178: info.key_offset = offsetof(struct match, pageid);
1.1 kristaps 179:
1.2 kristaps 180: *sz = cur = maxres = 0;
1.1 kristaps 181: sql = NULL;
182: *res = NULL;
183: fd = -1;
184: e = NULL;
1.2 kristaps 185: rc = 0;
1.1 kristaps 186:
187: if (0 == argc)
188: goto out;
1.5 kristaps 189: if (NULL == (e = exprcomp(search, argc, argv)))
1.1 kristaps 190: goto out;
191:
1.12 schwarze 192: outbit = 0;
193: if (NULL != outkey) {
1.20 schwarze 194: for (indexbit = 0, iterbit = 1;
195: indexbit < mansearch_keymax;
196: indexbit++, iterbit <<= 1) {
197: if (0 == strcasecmp(outkey,
198: mansearch_keynames[indexbit])) {
199: outbit = iterbit;
1.12 schwarze 200: break;
201: }
202: }
203: }
204:
1.1 kristaps 205: /*
206: * Save a descriptor to the current working directory.
207: * Since pathnames in the "paths" variable might be relative,
208: * and we'll be chdir()ing into them, we need to keep a handle
209: * on our current directory from which to start the chdir().
210: */
211:
1.6 schwarze 212: if (NULL == getcwd(buf, PATH_MAX)) {
1.30 schwarze 213: perror("getcwd");
1.1 kristaps 214: goto out;
215: } else if (-1 == (fd = open(buf, O_RDONLY, 0))) {
216: perror(buf);
217: goto out;
218: }
219:
1.15 schwarze 220: sql = sql_statement(e);
1.1 kristaps 221:
222: /*
223: * Loop over the directories (containing databases) for us to
224: * search.
225: * Don't let missing/bad databases/directories phase us.
226: * In each, try to open the resident database and, if it opens,
227: * scan it for our match expression.
228: */
229:
230: for (i = 0; i < paths->sz; i++) {
231: if (-1 == fchdir(fd)) {
232: perror(buf);
233: free(*res);
234: break;
235: } else if (-1 == chdir(paths->paths[i])) {
236: perror(paths->paths[i]);
237: continue;
1.34 schwarze 238: }
1.1 kristaps 239:
1.34 schwarze 240: c = sqlite3_open_v2(MANDOC_DB, &db,
241: SQLITE_OPEN_READONLY, NULL);
1.1 kristaps 242:
1.2 kristaps 243: if (SQLITE_OK != c) {
1.1 kristaps 244: perror(MANDOC_DB);
245: sqlite3_close(db);
246: continue;
247: }
248:
1.8 schwarze 249: /*
250: * Define the SQL functions for substring
251: * and regular expression matching.
252: */
1.7 schwarze 253:
254: c = sqlite3_create_function(db, "match", 2,
1.31 schwarze 255: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
256: NULL, sql_match, NULL, NULL);
1.8 schwarze 257: assert(SQLITE_OK == c);
258: c = sqlite3_create_function(db, "regexp", 2,
1.31 schwarze 259: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
260: NULL, sql_regexp, NULL, NULL);
1.8 schwarze 261: assert(SQLITE_OK == c);
1.7 schwarze 262:
1.1 kristaps 263: j = 1;
1.2 kristaps 264: c = sqlite3_prepare_v2(db, sql, -1, &s, NULL);
265: if (SQLITE_OK != c)
266: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
1.1 kristaps 267:
268: for (ep = e; NULL != ep; ep = ep->next) {
1.8 schwarze 269: if (NULL == ep->substr) {
270: SQL_BIND_BLOB(db, s, j, ep->regexp);
271: } else
272: SQL_BIND_TEXT(db, s, j, ep->substr);
1.27 schwarze 273: if (0 == ((TYPE_Nd | TYPE_Nm) & ep->bits))
1.26 schwarze 274: SQL_BIND_INT64(db, s, j, ep->bits);
1.1 kristaps 275: }
276:
277: memset(&htab, 0, sizeof(struct ohash));
278: ohash_init(&htab, 4, &info);
279:
280: /*
281: * Hash each entry on its [unique] document identifier.
282: * This is a uint64_t.
283: * Instead of using a hash function, simply convert the
284: * uint64_t to a uint32_t, the hash value's type.
285: * This gives good performance and preserves the
286: * distribution of buckets in the table.
287: */
1.2 kristaps 288: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.32 schwarze 289: pageid = sqlite3_column_int64(s, 2);
1.34 schwarze 290: idx = ohash_lookup_memory(&htab,
291: (char *)&pageid, sizeof(uint64_t),
292: (uint32_t)pageid);
1.1 kristaps 293:
294: if (NULL != ohash_find(&htab, idx))
295: continue;
296:
297: mp = mandoc_calloc(1, sizeof(struct match));
1.32 schwarze 298: mp->pageid = pageid;
1.26 schwarze 299: mp->form = sqlite3_column_int(s, 1);
300: if (TYPE_Nd == outbit)
301: mp->desc = mandoc_strdup(
302: sqlite3_column_text(s, 0));
1.1 kristaps 303: ohash_insert(&htab, idx, mp);
304: }
305:
1.2 kristaps 306: if (SQLITE_DONE != c)
307: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
308:
1.1 kristaps 309: sqlite3_finalize(s);
1.10 schwarze 310:
1.34 schwarze 311: c = sqlite3_prepare_v2(db,
1.35 schwarze 312: "SELECT sec, arch, name, pageid FROM mlinks "
313: "WHERE pageid=? ORDER BY sec, arch, name",
1.10 schwarze 314: -1, &s, NULL);
315: if (SQLITE_OK != c)
316: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
1.1 kristaps 317:
1.12 schwarze 318: c = sqlite3_prepare_v2(db,
1.35 schwarze 319: "SELECT bits, key, pageid FROM keys "
320: "WHERE pageid=? AND bits & ?",
1.12 schwarze 321: -1, &s2, NULL);
322: if (SQLITE_OK != c)
323: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
324:
1.1 kristaps 325: for (mp = ohash_first(&htab, &idx);
326: NULL != mp;
327: mp = ohash_next(&htab, &idx)) {
328: if (cur + 1 > maxres) {
329: maxres += 1024;
1.36 schwarze 330: *res = mandoc_reallocarray(*res,
331: maxres, sizeof(struct manpage));
1.1 kristaps 332: }
1.10 schwarze 333: mpage = *res + cur;
334: mpage->form = mp->form;
1.32 schwarze 335: buildnames(mpage, db, s, mp->pageid,
1.19 schwarze 336: paths->paths[i], mp->form);
1.26 schwarze 337: mpage->output = TYPE_Nd & outbit ?
338: mp->desc : outbit ?
1.32 schwarze 339: buildoutput(db, s2, mp->pageid, outbit) : NULL;
1.10 schwarze 340:
1.1 kristaps 341: free(mp);
342: cur++;
343: }
1.10 schwarze 344:
345: sqlite3_finalize(s);
1.12 schwarze 346: sqlite3_finalize(s2);
1.10 schwarze 347: sqlite3_close(db);
1.1 kristaps 348: ohash_delete(&htab);
349: }
1.2 kristaps 350: rc = 1;
1.1 kristaps 351: out:
1.30 schwarze 352: if (-1 != fd) {
353: if (-1 == fchdir(fd))
354: perror(buf);
355: close(fd);
356: }
1.1 kristaps 357: exprfree(e);
358: free(sql);
359: *sz = cur;
1.2 kristaps 360: return(rc);
1.11 schwarze 361: }
362:
1.17 schwarze 363: static void
364: buildnames(struct manpage *mpage, sqlite3 *db, sqlite3_stmt *s,
1.32 schwarze 365: uint64_t pageid, const char *path, int form)
1.11 schwarze 366: {
1.22 schwarze 367: char *newnames, *prevsec, *prevarch;
1.19 schwarze 368: const char *oldnames, *sep1, *name, *sec, *sep2, *arch, *fsec;
1.11 schwarze 369: size_t i;
370: int c;
371:
1.25 schwarze 372: mpage->file = NULL;
1.17 schwarze 373: mpage->names = NULL;
1.22 schwarze 374: prevsec = prevarch = NULL;
1.11 schwarze 375: i = 1;
1.32 schwarze 376: SQL_BIND_INT64(db, s, i, pageid);
1.11 schwarze 377: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.17 schwarze 378:
1.22 schwarze 379: /* Decide whether we already have some names. */
1.17 schwarze 380:
381: if (NULL == mpage->names) {
1.11 schwarze 382: oldnames = "";
383: sep1 = "";
384: } else {
1.17 schwarze 385: oldnames = mpage->names;
1.11 schwarze 386: sep1 = ", ";
387: }
1.22 schwarze 388:
389: /* Fetch the next name. */
390:
1.19 schwarze 391: sec = sqlite3_column_text(s, 0);
392: arch = sqlite3_column_text(s, 1);
393: name = sqlite3_column_text(s, 2);
1.22 schwarze 394:
395: /* If the section changed, append the old one. */
396:
397: if (NULL != prevsec &&
398: (strcmp(sec, prevsec) ||
399: strcmp(arch, prevarch))) {
400: sep2 = '\0' == *prevarch ? "" : "/";
1.24 schwarze 401: mandoc_asprintf(&newnames, "%s(%s%s%s)",
402: oldnames, prevsec, sep2, prevarch);
1.22 schwarze 403: free(mpage->names);
404: oldnames = mpage->names = newnames;
405: free(prevsec);
406: free(prevarch);
407: prevsec = prevarch = NULL;
408: }
409:
410: /* Save the new section, to append it later. */
411:
412: if (NULL == prevsec) {
413: prevsec = mandoc_strdup(sec);
414: prevarch = mandoc_strdup(arch);
415: }
416:
417: /* Append the new name. */
418:
1.24 schwarze 419: mandoc_asprintf(&newnames, "%s%s%s",
420: oldnames, sep1, name);
1.17 schwarze 421: free(mpage->names);
422: mpage->names = newnames;
423:
424: /* Also save the first file name encountered. */
425:
426: if (NULL != mpage->file)
427: continue;
428:
1.19 schwarze 429: if (form) {
430: sep1 = "man";
431: fsec = sec;
432: } else {
433: sep1 = "cat";
434: fsec = "0";
435: }
1.22 schwarze 436: sep2 = '\0' == *arch ? "" : "/";
1.24 schwarze 437: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.%s",
438: path, sep1, sec, sep2, arch, name, fsec);
1.11 schwarze 439: }
440: if (SQLITE_DONE != c)
441: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
442: sqlite3_reset(s);
1.22 schwarze 443:
444: /* Append one final section to the names. */
445:
446: if (NULL != prevsec) {
447: sep2 = '\0' == *prevarch ? "" : "/";
1.24 schwarze 448: mandoc_asprintf(&newnames, "%s(%s%s%s)",
449: mpage->names, prevsec, sep2, prevarch);
1.22 schwarze 450: free(mpage->names);
451: mpage->names = newnames;
452: free(prevsec);
453: free(prevarch);
454: }
1.12 schwarze 455: }
456:
457: static char *
1.32 schwarze 458: buildoutput(sqlite3 *db, sqlite3_stmt *s, uint64_t pageid, uint64_t outbit)
1.12 schwarze 459: {
460: char *output, *newoutput;
461: const char *oldoutput, *sep1, *data;
462: size_t i;
463: int c;
464:
465: output = NULL;
466: i = 1;
1.32 schwarze 467: SQL_BIND_INT64(db, s, i, pageid);
1.12 schwarze 468: SQL_BIND_INT64(db, s, i, outbit);
469: while (SQLITE_ROW == (c = sqlite3_step(s))) {
470: if (NULL == output) {
471: oldoutput = "";
472: sep1 = "";
473: } else {
474: oldoutput = output;
475: sep1 = " # ";
476: }
477: data = sqlite3_column_text(s, 1);
1.24 schwarze 478: mandoc_asprintf(&newoutput, "%s%s%s",
479: oldoutput, sep1, data);
1.12 schwarze 480: free(output);
481: output = newoutput;
482: }
483: if (SQLITE_DONE != c)
484: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
485: sqlite3_reset(s);
486: return(output);
1.1 kristaps 487: }
488:
489: /*
1.7 schwarze 490: * Implement substring match as an application-defined SQL function.
491: * Using the SQL LIKE or GLOB operators instead would be a bad idea
492: * because that would require escaping metacharacters in the string
493: * being searched for.
494: */
495: static void
496: sql_match(sqlite3_context *context, int argc, sqlite3_value **argv)
497: {
498:
499: assert(2 == argc);
500: sqlite3_result_int(context, NULL != strcasestr(
501: (const char *)sqlite3_value_text(argv[1]),
502: (const char *)sqlite3_value_text(argv[0])));
503: }
504:
505: /*
1.8 schwarze 506: * Implement regular expression match
507: * as an application-defined SQL function.
508: */
509: static void
510: sql_regexp(sqlite3_context *context, int argc, sqlite3_value **argv)
511: {
512:
513: assert(2 == argc);
514: sqlite3_result_int(context, !regexec(
515: (regex_t *)sqlite3_value_blob(argv[0]),
516: (const char *)sqlite3_value_text(argv[1]),
517: 0, NULL, 0));
518: }
519:
1.13 schwarze 520: static void
521: sql_append(char **sql, size_t *sz, const char *newstr, int count)
522: {
523: size_t newsz;
524:
525: newsz = 1 < count ? (size_t)count : strlen(newstr);
526: *sql = mandoc_realloc(*sql, *sz + newsz + 1);
527: if (1 < count)
528: memset(*sql + *sz, *newstr, (size_t)count);
529: else
530: memcpy(*sql + *sz, newstr, newsz);
531: *sz += newsz;
532: (*sql)[*sz] = '\0';
533: }
534:
1.8 schwarze 535: /*
1.1 kristaps 536: * Prepare the search SQL statement.
537: */
538: static char *
1.15 schwarze 539: sql_statement(const struct expr *e)
1.1 kristaps 540: {
541: char *sql;
542: size_t sz;
1.13 schwarze 543: int needop;
1.1 kristaps 544:
1.35 schwarze 545: sql = mandoc_strdup(
546: "SELECT desc, form, pageid FROM mpages WHERE ");
1.1 kristaps 547: sz = strlen(sql);
1.2 kristaps 548:
1.13 schwarze 549: for (needop = 0; NULL != e; e = e->next) {
550: if (e->and)
551: sql_append(&sql, &sz, " AND ", 1);
552: else if (needop)
553: sql_append(&sql, &sz, " OR ", 1);
554: if (e->open)
555: sql_append(&sql, &sz, "(", e->open);
1.26 schwarze 556: sql_append(&sql, &sz,
557: TYPE_Nd & e->bits
558: ? (NULL == e->substr
559: ? "desc REGEXP ?"
560: : "desc MATCH ?")
1.27 schwarze 561: : TYPE_Nm == e->bits
562: ? (NULL == e->substr
1.32 schwarze 563: ? "pageid IN (SELECT pageid FROM names "
1.27 schwarze 564: "WHERE name REGEXP ?)"
1.38 ! schwarze 565: : e->equal
! 566: ? "pageid IN (SELECT pageid FROM names "
! 567: "WHERE name = ?)"
1.32 schwarze 568: : "pageid IN (SELECT pageid FROM names "
1.27 schwarze 569: "WHERE name MATCH ?)")
1.26 schwarze 570: : (NULL == e->substr
1.32 schwarze 571: ? "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 572: "WHERE key REGEXP ? AND bits & ?)"
1.32 schwarze 573: : "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 574: "WHERE key MATCH ? AND bits & ?)"), 1);
1.13 schwarze 575: if (e->close)
576: sql_append(&sql, &sz, ")", e->close);
577: needop = 1;
1.1 kristaps 578: }
579:
580: return(sql);
581: }
582:
583: /*
584: * Compile a set of string tokens into an expression.
585: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
586: * "(", "foo=bar", etc.).
587: */
588: static struct expr *
1.5 kristaps 589: exprcomp(const struct mansearch *search, int argc, char *argv[])
1.1 kristaps 590: {
1.27 schwarze 591: uint64_t mask;
1.13 schwarze 592: int i, toopen, logic, igncase, toclose;
1.27 schwarze 593: struct expr *first, *prev, *cur, *next;
1.1 kristaps 594:
595: first = cur = NULL;
1.15 schwarze 596: logic = igncase = toclose = 0;
1.33 schwarze 597: toopen = NULL != search->sec || NULL != search->arch;
1.1 kristaps 598:
599: for (i = 0; i < argc; i++) {
1.13 schwarze 600: if (0 == strcmp("(", argv[i])) {
601: if (igncase)
602: goto fail;
603: toopen++;
604: toclose++;
605: continue;
606: } else if (0 == strcmp(")", argv[i])) {
607: if (toopen || logic || igncase || NULL == cur)
608: goto fail;
609: cur->close++;
610: if (0 > --toclose)
611: goto fail;
612: continue;
613: } else if (0 == strcmp("-a", argv[i])) {
614: if (toopen || logic || igncase || NULL == cur)
615: goto fail;
616: logic = 1;
617: continue;
618: } else if (0 == strcmp("-o", argv[i])) {
619: if (toopen || logic || igncase || NULL == cur)
620: goto fail;
621: logic = 2;
622: continue;
623: } else if (0 == strcmp("-i", argv[i])) {
624: if (igncase)
625: goto fail;
626: igncase = 1;
627: continue;
1.1 kristaps 628: }
1.13 schwarze 629: next = exprterm(search, argv[i], !igncase);
630: if (NULL == next)
631: goto fail;
1.26 schwarze 632: if (NULL == first)
633: first = next;
634: else
1.1 kristaps 635: cur->next = next;
1.27 schwarze 636: prev = cur = next;
1.26 schwarze 637:
638: /*
639: * Searching for descriptions must be split out
640: * because they are stored in the mpages table,
641: * not in the keys table.
642: */
643:
1.27 schwarze 644: for (mask = TYPE_Nm; mask <= TYPE_Nd; mask <<= 1) {
645: if (mask & cur->bits && ~mask & cur->bits) {
646: next = mandoc_calloc(1,
647: sizeof(struct expr));
648: memcpy(next, cur, sizeof(struct expr));
649: prev->open = 1;
650: cur->bits = mask;
651: cur->next = next;
652: cur = next;
653: cur->bits &= ~mask;
654: }
655: }
656: prev->and = (1 == logic);
657: prev->open += toopen;
658: if (cur != prev)
1.26 schwarze 659: cur->close = 1;
1.27 schwarze 660:
1.13 schwarze 661: toopen = logic = igncase = 0;
1.1 kristaps 662: }
1.15 schwarze 663: if (toopen || logic || igncase || toclose)
664: goto fail;
665:
1.33 schwarze 666: if (NULL != search->sec || NULL != search->arch)
667: cur->close++;
668: if (NULL != search->arch)
669: cur = exprspec(cur, TYPE_arch, search->arch, "^(%s|any)$");
670: if (NULL != search->sec)
671: exprspec(cur, TYPE_sec, search->sec, "^%s$");
1.15 schwarze 672:
673: return(first);
674:
1.13 schwarze 675: fail:
676: if (NULL != first)
677: exprfree(first);
678: return(NULL);
1.1 kristaps 679: }
680:
681: static struct expr *
1.15 schwarze 682: exprspec(struct expr *cur, uint64_t key, const char *value,
683: const char *format)
684: {
685: char errbuf[BUFSIZ];
686: char *cp;
687: int irc;
688:
1.24 schwarze 689: mandoc_asprintf(&cp, format, value);
1.15 schwarze 690: cur->next = mandoc_calloc(1, sizeof(struct expr));
691: cur = cur->next;
692: cur->and = 1;
693: cur->bits = key;
694: if (0 != (irc = regcomp(&cur->regexp, cp,
695: REG_EXTENDED | REG_NOSUB | REG_ICASE))) {
696: regerror(irc, &cur->regexp, errbuf, sizeof(errbuf));
697: fprintf(stderr, "regcomp: %s\n", errbuf);
698: cur->substr = value;
699: }
700: free(cp);
701: return(cur);
702: }
703:
704: static struct expr *
1.8 schwarze 705: exprterm(const struct mansearch *search, char *buf, int cs)
1.1 kristaps 706: {
1.15 schwarze 707: char errbuf[BUFSIZ];
1.1 kristaps 708: struct expr *e;
1.38 ! schwarze 709: char *key, *val;
1.20 schwarze 710: uint64_t iterbit;
711: int i, irc;
1.1 kristaps 712:
713: if ('\0' == *buf)
714: return(NULL);
715:
716: e = mandoc_calloc(1, sizeof(struct expr));
717:
1.38 ! schwarze 718: if (MANSEARCH_MAN & search->flags) {
! 719: e->bits = search->deftype;
1.8 schwarze 720: e->substr = buf;
1.38 ! schwarze 721: e->equal = 1;
1.5 kristaps 722: return(e);
723: }
724:
1.1 kristaps 725: /*
1.38 ! schwarze 726: * Look for an '=' or '~' operator,
! 727: * unless forced to some fixed macro keys.
1.1 kristaps 728: */
729:
1.38 ! schwarze 730: if (MANSEARCH_WHATIS & search->flags)
! 731: val = NULL;
! 732: else
! 733: val = strpbrk(buf, "=~");
! 734:
! 735: if (NULL == val) {
! 736: e->bits = search->deftype;
1.8 schwarze 737: e->substr = buf;
1.1 kristaps 738:
1.38 ! schwarze 739: /*
! 740: * Found an operator.
! 741: * Regexp search is requested by !e->substr.
! 742: */
! 743:
! 744: } else {
! 745: if (val == buf)
! 746: e->bits = search->deftype;
! 747: if ('=' == *val)
! 748: e->substr = val + 1;
! 749: *val++ = '\0';
1.21 schwarze 750: if (NULL != strstr(buf, "arch"))
751: cs = 0;
1.38 ! schwarze 752: }
! 753:
! 754: /* Compile regular expressions. */
! 755:
! 756: if (MANSEARCH_WHATIS & search->flags) {
! 757: e->substr = NULL;
! 758: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", buf);
! 759: }
! 760:
! 761: if (NULL == e->substr) {
! 762: irc = regcomp(&e->regexp, val,
! 763: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
! 764: if (MANSEARCH_WHATIS & search->flags)
! 765: free(val);
! 766: if (irc) {
1.15 schwarze 767: regerror(irc, &e->regexp, errbuf, sizeof(errbuf));
768: fprintf(stderr, "regcomp: %s\n", errbuf);
1.8 schwarze 769: free(e);
770: return(NULL);
771: }
1.38 ! schwarze 772: }
! 773:
! 774: if (e->bits)
! 775: return(e);
1.1 kristaps 776:
777: /*
778: * Parse out all possible fields.
779: * If the field doesn't resolve, bail.
780: */
781:
782: while (NULL != (key = strsep(&buf, ","))) {
783: if ('\0' == *key)
784: continue;
1.20 schwarze 785: for (i = 0, iterbit = 1;
786: i < mansearch_keymax;
787: i++, iterbit <<= 1) {
788: if (0 == strcasecmp(key,
789: mansearch_keynames[i])) {
790: e->bits |= iterbit;
791: break;
792: }
793: }
794: if (i == mansearch_keymax) {
795: if (strcasecmp(key, "any")) {
796: free(e);
797: return(NULL);
798: }
799: e->bits |= ~0ULL;
1.1 kristaps 800: }
801: }
802:
803: return(e);
804: }
805:
806: static void
807: exprfree(struct expr *p)
808: {
809: struct expr *pp;
810:
811: while (NULL != p) {
812: pp = p->next;
813: free(p);
814: p = pp;
815: }
816: }
817:
818: static void *
1.37 schwarze 819: hash_calloc(size_t nmemb, size_t sz, void *arg)
1.1 kristaps 820: {
821:
1.37 schwarze 822: return(mandoc_calloc(nmemb, sz));
1.1 kristaps 823: }
824:
825: static void *
826: hash_alloc(size_t sz, void *arg)
827: {
828:
829: return(mandoc_malloc(sz));
830: }
831:
832: static void
1.37 schwarze 833: hash_free(void *p, void *arg)
1.1 kristaps 834: {
835:
836: free(p);
837: }
CVSweb