Annotation of src/usr.bin/mandoc/mansearch.c, Revision 1.39
1.39 ! schwarze 1: /* $OpenBSD: mansearch.c,v 1.38 2014/11/27 01:57:42 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.4 schwarze 4: * Copyright (c) 2013, 2014 Ingo Schwarze <schwarze@openbsd.org>
1.1 schwarze 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: */
1.33 schwarze 18:
1.19 schwarze 19: #include <sys/mman.h>
1.33 schwarze 20: #include <sys/types.h>
21:
1.1 schwarze 22: #include <assert.h>
1.39 ! schwarze 23: #include <errno.h>
1.1 schwarze 24: #include <fcntl.h>
25: #include <getopt.h>
26: #include <limits.h>
27: #include <regex.h>
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:
35: #include <ohash.h>
36: #include <sqlite3.h>
37:
38: #include "mandoc.h"
1.14 schwarze 39: #include "mandoc_aux.h"
1.1 schwarze 40: #include "manpath.h"
41: #include "mansearch.h"
42:
1.11 schwarze 43: extern int mansearch_keymax;
44: extern const char *const mansearch_keynames[];
45:
1.1 schwarze 46: #define SQL_BIND_TEXT(_db, _s, _i, _v) \
47: do { if (SQLITE_OK != sqlite3_bind_text \
48: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
49: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
50: } while (0)
51: #define SQL_BIND_INT64(_db, _s, _i, _v) \
52: do { if (SQLITE_OK != sqlite3_bind_int64 \
53: ((_s), (_i)++, (_v))) \
54: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
55: } while (0)
56: #define SQL_BIND_BLOB(_db, _s, _i, _v) \
57: do { if (SQLITE_OK != sqlite3_bind_blob \
58: ((_s), (_i)++, (&_v), sizeof(_v), SQLITE_STATIC)) \
59: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
60: } while (0)
61:
62: struct expr {
1.28 schwarze 63: regex_t regexp; /* compiled regexp, if applicable */
64: const char *substr; /* to search for, if applicable */
65: struct expr *next; /* next in sequence */
1.24 schwarze 66: uint64_t bits; /* type-mask */
1.28 schwarze 67: int equal; /* equality, not subsring match */
1.4 schwarze 68: int open; /* opening parentheses before */
69: int and; /* logical AND before */
70: int close; /* closing parentheses after */
1.1 schwarze 71: };
72:
73: struct match {
1.22 schwarze 74: uint64_t pageid; /* identifier in database */
1.37 schwarze 75: uint64_t bits; /* name type mask */
1.17 schwarze 76: char *desc; /* manual page description */
1.35 schwarze 77: int form; /* bit field: formatted, zipped? */
1.1 schwarze 78: };
79:
1.8 schwarze 80: static void buildnames(struct manpage *, sqlite3 *,
1.10 schwarze 81: sqlite3_stmt *, uint64_t,
82: const char *, int form);
1.3 schwarze 83: static char *buildoutput(sqlite3 *, sqlite3_stmt *,
84: uint64_t, uint64_t);
1.1 schwarze 85: static void *hash_alloc(size_t, void *);
1.27 espie 86: static void hash_free(void *, void *);
87: static void *hash_calloc(size_t, size_t, void *);
1.24 schwarze 88: static struct expr *exprcomp(const struct mansearch *,
1.1 schwarze 89: int, char *[]);
90: static void exprfree(struct expr *);
1.6 schwarze 91: static struct expr *exprspec(struct expr *, uint64_t,
92: const char *, const char *);
1.1 schwarze 93: static struct expr *exprterm(const struct mansearch *, char *, int);
1.29 schwarze 94: static int manpage_compare(const void *, const void *);
1.4 schwarze 95: static void sql_append(char **sql, size_t *sz,
96: const char *newstr, int count);
1.1 schwarze 97: static void sql_match(sqlite3_context *context,
98: int argc, sqlite3_value **argv);
99: static void sql_regexp(sqlite3_context *context,
100: int argc, sqlite3_value **argv);
1.6 schwarze 101: static char *sql_statement(const struct expr *);
1.19 schwarze 102:
1.24 schwarze 103:
1.19 schwarze 104: int
105: mansearch_setup(int start)
106: {
107: static void *pagecache;
108: int c;
109:
110: #define PC_PAGESIZE 1280
111: #define PC_NUMPAGES 256
112:
113: if (start) {
114: if (NULL != pagecache) {
115: fprintf(stderr, "pagecache already enabled\n");
116: return((int)MANDOCLEVEL_BADARG);
117: }
118:
119: pagecache = mmap(NULL, PC_PAGESIZE * PC_NUMPAGES,
1.31 schwarze 120: PROT_READ | PROT_WRITE,
121: MAP_SHARED | MAP_ANON, -1, 0);
1.19 schwarze 122:
123: if (MAP_FAILED == pagecache) {
124: perror("mmap");
125: pagecache = NULL;
126: return((int)MANDOCLEVEL_SYSERR);
127: }
128:
129: c = sqlite3_config(SQLITE_CONFIG_PAGECACHE,
130: pagecache, PC_PAGESIZE, PC_NUMPAGES);
131:
132: if (SQLITE_OK == c)
133: return((int)MANDOCLEVEL_OK);
134:
135: fprintf(stderr, "pagecache: %s\n", sqlite3_errstr(c));
136:
137: } else if (NULL == pagecache) {
138: fprintf(stderr, "pagecache missing\n");
139: return((int)MANDOCLEVEL_BADARG);
140: }
141:
142: if (-1 == munmap(pagecache, PC_PAGESIZE * PC_NUMPAGES)) {
143: perror("munmap");
144: pagecache = NULL;
145: return((int)MANDOCLEVEL_SYSERR);
146: }
147:
148: pagecache = NULL;
149: return((int)MANDOCLEVEL_OK);
150: }
1.1 schwarze 151:
152: int
153: mansearch(const struct mansearch *search,
1.3 schwarze 154: const struct manpaths *paths,
155: int argc, char *argv[],
1.1 schwarze 156: struct manpage **res, size_t *sz)
157: {
1.11 schwarze 158: int fd, rc, c, indexbit;
1.22 schwarze 159: int64_t pageid;
1.11 schwarze 160: uint64_t outbit, iterbit;
1.1 schwarze 161: char buf[PATH_MAX];
1.2 schwarze 162: char *sql;
1.1 schwarze 163: struct manpage *mpage;
164: struct expr *e, *ep;
165: sqlite3 *db;
1.3 schwarze 166: sqlite3_stmt *s, *s2;
1.1 schwarze 167: struct match *mp;
168: struct ohash_info info;
169: struct ohash htab;
170: unsigned int idx;
171: size_t i, j, cur, maxres;
172:
1.27 espie 173: info.calloc = hash_calloc;
1.1 schwarze 174: info.alloc = hash_alloc;
1.27 espie 175: info.free = hash_free;
1.22 schwarze 176: info.key_offset = offsetof(struct match, pageid);
1.1 schwarze 177:
178: *sz = cur = maxres = 0;
179: sql = NULL;
180: *res = NULL;
181: fd = -1;
182: e = NULL;
183: rc = 0;
184:
185: if (0 == argc)
186: goto out;
187: if (NULL == (e = exprcomp(search, argc, argv)))
188: goto out;
189:
1.3 schwarze 190: outbit = 0;
1.33 schwarze 191: if (NULL != search->outkey) {
1.11 schwarze 192: for (indexbit = 0, iterbit = 1;
193: indexbit < mansearch_keymax;
194: indexbit++, iterbit <<= 1) {
1.33 schwarze 195: if (0 == strcasecmp(search->outkey,
1.11 schwarze 196: mansearch_keynames[indexbit])) {
197: outbit = iterbit;
1.3 schwarze 198: break;
199: }
200: }
201: }
202:
1.1 schwarze 203: /*
204: * Save a descriptor to the current working directory.
205: * Since pathnames in the "paths" variable might be relative,
206: * and we'll be chdir()ing into them, we need to keep a handle
207: * on our current directory from which to start the chdir().
208: */
209:
210: if (NULL == getcwd(buf, PATH_MAX)) {
1.20 schwarze 211: perror("getcwd");
1.1 schwarze 212: goto out;
213: } else if (-1 == (fd = open(buf, O_RDONLY, 0))) {
214: perror(buf);
215: goto out;
216: }
217:
1.6 schwarze 218: sql = sql_statement(e);
1.1 schwarze 219:
220: /*
221: * Loop over the directories (containing databases) for us to
222: * search.
223: * Don't let missing/bad databases/directories phase us.
224: * In each, try to open the resident database and, if it opens,
225: * scan it for our match expression.
226: */
227:
228: for (i = 0; i < paths->sz; i++) {
229: if (-1 == fchdir(fd)) {
230: perror(buf);
231: free(*res);
232: break;
233: } else if (-1 == chdir(paths->paths[i])) {
234: perror(paths->paths[i]);
235: continue;
1.24 schwarze 236: }
1.1 schwarze 237:
1.24 schwarze 238: c = sqlite3_open_v2(MANDOC_DB, &db,
239: SQLITE_OPEN_READONLY, NULL);
1.1 schwarze 240:
241: if (SQLITE_OK != c) {
1.39 ! schwarze 242: fprintf(stderr, "%s/%s: %s\n",
! 243: paths->paths[i], MANDOC_DB, strerror(errno));
1.1 schwarze 244: sqlite3_close(db);
245: continue;
246: }
247:
248: /*
249: * Define the SQL functions for substring
250: * and regular expression matching.
251: */
252:
253: c = sqlite3_create_function(db, "match", 2,
1.21 schwarze 254: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
255: NULL, sql_match, NULL, NULL);
1.1 schwarze 256: assert(SQLITE_OK == c);
257: c = sqlite3_create_function(db, "regexp", 2,
1.21 schwarze 258: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
259: NULL, sql_regexp, NULL, NULL);
1.1 schwarze 260: assert(SQLITE_OK == c);
261:
262: j = 1;
263: c = sqlite3_prepare_v2(db, sql, -1, &s, NULL);
264: if (SQLITE_OK != c)
265: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
266:
267: for (ep = e; NULL != ep; ep = ep->next) {
268: if (NULL == ep->substr) {
269: SQL_BIND_BLOB(db, s, j, ep->regexp);
270: } else
271: SQL_BIND_TEXT(db, s, j, ep->substr);
1.18 schwarze 272: if (0 == ((TYPE_Nd | TYPE_Nm) & ep->bits))
1.17 schwarze 273: SQL_BIND_INT64(db, s, j, ep->bits);
1.1 schwarze 274: }
275:
276: memset(&htab, 0, sizeof(struct ohash));
277: ohash_init(&htab, 4, &info);
278:
279: /*
280: * Hash each entry on its [unique] document identifier.
281: * This is a uint64_t.
282: * Instead of using a hash function, simply convert the
283: * uint64_t to a uint32_t, the hash value's type.
284: * This gives good performance and preserves the
285: * distribution of buckets in the table.
286: */
287: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.22 schwarze 288: pageid = sqlite3_column_int64(s, 2);
1.24 schwarze 289: idx = ohash_lookup_memory(&htab,
290: (char *)&pageid, sizeof(uint64_t),
291: (uint32_t)pageid);
1.1 schwarze 292:
293: if (NULL != ohash_find(&htab, idx))
294: continue;
295:
296: mp = mandoc_calloc(1, sizeof(struct match));
1.22 schwarze 297: mp->pageid = pageid;
1.17 schwarze 298: mp->form = sqlite3_column_int(s, 1);
1.37 schwarze 299: mp->bits = sqlite3_column_int64(s, 3);
1.17 schwarze 300: if (TYPE_Nd == outbit)
1.30 schwarze 301: mp->desc = mandoc_strdup((const char *)
1.17 schwarze 302: sqlite3_column_text(s, 0));
1.1 schwarze 303: ohash_insert(&htab, idx, mp);
304: }
305:
306: if (SQLITE_DONE != c)
307: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
308:
309: sqlite3_finalize(s);
310:
1.24 schwarze 311: c = sqlite3_prepare_v2(db,
1.25 schwarze 312: "SELECT sec, arch, name, pageid FROM mlinks "
313: "WHERE pageid=? ORDER BY sec, arch, name",
1.1 schwarze 314: -1, &s, NULL);
315: if (SQLITE_OK != c)
316: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
317:
1.3 schwarze 318: c = sqlite3_prepare_v2(db,
1.25 schwarze 319: "SELECT bits, key, pageid FROM keys "
320: "WHERE pageid=? AND bits & ?",
1.3 schwarze 321: -1, &s2, NULL);
322: if (SQLITE_OK != c)
323: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
324:
1.1 schwarze 325: for (mp = ohash_first(&htab, &idx);
326: NULL != mp;
327: mp = ohash_next(&htab, &idx)) {
328: if (cur + 1 > maxres) {
329: maxres += 1024;
1.26 schwarze 330: *res = mandoc_reallocarray(*res,
331: maxres, sizeof(struct manpage));
1.1 schwarze 332: }
333: mpage = *res + cur;
1.34 schwarze 334: mpage->ipath = i;
1.37 schwarze 335: mpage->bits = mp->bits;
1.29 schwarze 336: mpage->sec = 10;
1.1 schwarze 337: mpage->form = mp->form;
1.22 schwarze 338: buildnames(mpage, db, s, mp->pageid,
1.10 schwarze 339: paths->paths[i], mp->form);
1.17 schwarze 340: mpage->output = TYPE_Nd & outbit ?
341: mp->desc : outbit ?
1.22 schwarze 342: buildoutput(db, s2, mp->pageid, outbit) : NULL;
1.1 schwarze 343:
344: free(mp);
345: cur++;
346: }
347:
348: sqlite3_finalize(s);
1.3 schwarze 349: sqlite3_finalize(s2);
1.1 schwarze 350: sqlite3_close(db);
351: ohash_delete(&htab);
1.36 schwarze 352:
353: /*
354: * In man(1) mode, prefer matches in earlier trees
355: * over matches in later trees.
356: */
357:
358: if (cur && search->firstmatch)
359: break;
1.1 schwarze 360: }
1.29 schwarze 361: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.1 schwarze 362: rc = 1;
363: out:
1.20 schwarze 364: if (-1 != fd) {
365: if (-1 == fchdir(fd))
366: perror(buf);
367: close(fd);
368: }
1.1 schwarze 369: exprfree(e);
370: free(sql);
371: *sz = cur;
372: return(rc);
1.2 schwarze 373: }
374:
1.33 schwarze 375: void
376: mansearch_free(struct manpage *res, size_t sz)
377: {
378: size_t i;
379:
380: for (i = 0; i < sz; i++) {
381: free(res[i].file);
382: free(res[i].names);
383: free(res[i].output);
384: }
385: free(res);
386: }
387:
1.29 schwarze 388: static int
389: manpage_compare(const void *vp1, const void *vp2)
390: {
391: const struct manpage *mp1, *mp2;
392: int diff;
393:
394: mp1 = vp1;
395: mp2 = vp2;
1.37 schwarze 396: return( (diff = mp2->bits - mp1->bits) ? diff :
397: (diff = mp1->sec - mp2->sec) ? diff :
398: strcasecmp(mp1->names, mp2->names));
1.29 schwarze 399: }
400:
1.8 schwarze 401: static void
402: buildnames(struct manpage *mpage, sqlite3 *db, sqlite3_stmt *s,
1.22 schwarze 403: uint64_t pageid, const char *path, int form)
1.2 schwarze 404: {
1.13 schwarze 405: char *newnames, *prevsec, *prevarch;
1.10 schwarze 406: const char *oldnames, *sep1, *name, *sec, *sep2, *arch, *fsec;
1.2 schwarze 407: size_t i;
408: int c;
409:
1.16 schwarze 410: mpage->file = NULL;
1.8 schwarze 411: mpage->names = NULL;
1.13 schwarze 412: prevsec = prevarch = NULL;
1.2 schwarze 413: i = 1;
1.22 schwarze 414: SQL_BIND_INT64(db, s, i, pageid);
1.2 schwarze 415: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.8 schwarze 416:
1.13 schwarze 417: /* Decide whether we already have some names. */
1.8 schwarze 418:
419: if (NULL == mpage->names) {
1.2 schwarze 420: oldnames = "";
421: sep1 = "";
422: } else {
1.8 schwarze 423: oldnames = mpage->names;
1.2 schwarze 424: sep1 = ", ";
425: }
1.13 schwarze 426:
427: /* Fetch the next name. */
428:
1.30 schwarze 429: sec = (const char *)sqlite3_column_text(s, 0);
430: arch = (const char *)sqlite3_column_text(s, 1);
431: name = (const char *)sqlite3_column_text(s, 2);
1.29 schwarze 432:
433: /* Remember the first section found. */
434:
435: if (9 < mpage->sec && '1' <= *sec && '9' >= *sec)
436: mpage->sec = (*sec - '1') + 1;
1.13 schwarze 437:
438: /* If the section changed, append the old one. */
439:
440: if (NULL != prevsec &&
441: (strcmp(sec, prevsec) ||
442: strcmp(arch, prevarch))) {
443: sep2 = '\0' == *prevarch ? "" : "/";
1.15 schwarze 444: mandoc_asprintf(&newnames, "%s(%s%s%s)",
445: oldnames, prevsec, sep2, prevarch);
1.13 schwarze 446: free(mpage->names);
447: oldnames = mpage->names = newnames;
448: free(prevsec);
449: free(prevarch);
450: prevsec = prevarch = NULL;
451: }
452:
453: /* Save the new section, to append it later. */
454:
455: if (NULL == prevsec) {
456: prevsec = mandoc_strdup(sec);
457: prevarch = mandoc_strdup(arch);
458: }
459:
460: /* Append the new name. */
461:
1.15 schwarze 462: mandoc_asprintf(&newnames, "%s%s%s",
463: oldnames, sep1, name);
1.8 schwarze 464: free(mpage->names);
465: mpage->names = newnames;
466:
467: /* Also save the first file name encountered. */
468:
1.38 schwarze 469: if (mpage->file != NULL)
1.8 schwarze 470: continue;
471:
1.35 schwarze 472: if (form & FORM_SRC) {
1.10 schwarze 473: sep1 = "man";
474: fsec = sec;
475: } else {
476: sep1 = "cat";
477: fsec = "0";
478: }
1.38 schwarze 479: sep2 = *arch == '\0' ? "" : "/";
480: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.%s",
481: path, sep1, sec, sep2, arch, name, fsec);
1.2 schwarze 482: }
1.38 schwarze 483: if (c != SQLITE_DONE)
1.2 schwarze 484: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
485: sqlite3_reset(s);
1.13 schwarze 486:
487: /* Append one final section to the names. */
488:
1.38 schwarze 489: if (prevsec != NULL) {
490: sep2 = *prevarch == '\0' ? "" : "/";
1.15 schwarze 491: mandoc_asprintf(&newnames, "%s(%s%s%s)",
492: mpage->names, prevsec, sep2, prevarch);
1.13 schwarze 493: free(mpage->names);
494: mpage->names = newnames;
495: free(prevsec);
496: free(prevarch);
497: }
1.3 schwarze 498: }
499:
500: static char *
1.22 schwarze 501: buildoutput(sqlite3 *db, sqlite3_stmt *s, uint64_t pageid, uint64_t outbit)
1.3 schwarze 502: {
503: char *output, *newoutput;
504: const char *oldoutput, *sep1, *data;
505: size_t i;
506: int c;
507:
508: output = NULL;
509: i = 1;
1.22 schwarze 510: SQL_BIND_INT64(db, s, i, pageid);
1.3 schwarze 511: SQL_BIND_INT64(db, s, i, outbit);
512: while (SQLITE_ROW == (c = sqlite3_step(s))) {
513: if (NULL == output) {
514: oldoutput = "";
515: sep1 = "";
516: } else {
517: oldoutput = output;
518: sep1 = " # ";
519: }
1.30 schwarze 520: data = (const char *)sqlite3_column_text(s, 1);
1.15 schwarze 521: mandoc_asprintf(&newoutput, "%s%s%s",
522: oldoutput, sep1, data);
1.3 schwarze 523: free(output);
524: output = newoutput;
525: }
526: if (SQLITE_DONE != c)
527: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
528: sqlite3_reset(s);
529: return(output);
1.1 schwarze 530: }
531:
532: /*
533: * Implement substring match as an application-defined SQL function.
534: * Using the SQL LIKE or GLOB operators instead would be a bad idea
535: * because that would require escaping metacharacters in the string
536: * being searched for.
537: */
538: static void
539: sql_match(sqlite3_context *context, int argc, sqlite3_value **argv)
540: {
541:
542: assert(2 == argc);
543: sqlite3_result_int(context, NULL != strcasestr(
544: (const char *)sqlite3_value_text(argv[1]),
545: (const char *)sqlite3_value_text(argv[0])));
546: }
547:
548: /*
549: * Implement regular expression match
550: * as an application-defined SQL function.
551: */
552: static void
553: sql_regexp(sqlite3_context *context, int argc, sqlite3_value **argv)
554: {
555:
556: assert(2 == argc);
557: sqlite3_result_int(context, !regexec(
558: (regex_t *)sqlite3_value_blob(argv[0]),
559: (const char *)sqlite3_value_text(argv[1]),
560: 0, NULL, 0));
561: }
562:
1.4 schwarze 563: static void
564: sql_append(char **sql, size_t *sz, const char *newstr, int count)
565: {
566: size_t newsz;
567:
568: newsz = 1 < count ? (size_t)count : strlen(newstr);
569: *sql = mandoc_realloc(*sql, *sz + newsz + 1);
570: if (1 < count)
571: memset(*sql + *sz, *newstr, (size_t)count);
572: else
573: memcpy(*sql + *sz, newstr, newsz);
574: *sz += newsz;
575: (*sql)[*sz] = '\0';
576: }
577:
1.1 schwarze 578: /*
579: * Prepare the search SQL statement.
580: */
581: static char *
1.6 schwarze 582: sql_statement(const struct expr *e)
1.1 schwarze 583: {
584: char *sql;
585: size_t sz;
1.4 schwarze 586: int needop;
1.1 schwarze 587:
1.37 schwarze 588: sql = mandoc_strdup(e->equal ?
589: "SELECT desc, form, pageid, bits "
590: "FROM mpages NATURAL JOIN names WHERE " :
591: "SELECT desc, form, pageid, 0 FROM mpages WHERE ");
1.1 schwarze 592: sz = strlen(sql);
593:
1.4 schwarze 594: for (needop = 0; NULL != e; e = e->next) {
595: if (e->and)
596: sql_append(&sql, &sz, " AND ", 1);
597: else if (needop)
598: sql_append(&sql, &sz, " OR ", 1);
599: if (e->open)
600: sql_append(&sql, &sz, "(", e->open);
1.17 schwarze 601: sql_append(&sql, &sz,
602: TYPE_Nd & e->bits
603: ? (NULL == e->substr
604: ? "desc REGEXP ?"
605: : "desc MATCH ?")
1.18 schwarze 606: : TYPE_Nm == e->bits
607: ? (NULL == e->substr
1.22 schwarze 608: ? "pageid IN (SELECT pageid FROM names "
1.18 schwarze 609: "WHERE name REGEXP ?)"
1.28 schwarze 610: : e->equal
1.37 schwarze 611: ? "name = ? "
1.22 schwarze 612: : "pageid IN (SELECT pageid FROM names "
1.18 schwarze 613: "WHERE name MATCH ?)")
1.17 schwarze 614: : (NULL == e->substr
1.22 schwarze 615: ? "pageid IN (SELECT pageid FROM keys "
1.17 schwarze 616: "WHERE key REGEXP ? AND bits & ?)"
1.22 schwarze 617: : "pageid IN (SELECT pageid FROM keys "
1.17 schwarze 618: "WHERE key MATCH ? AND bits & ?)"), 1);
1.4 schwarze 619: if (e->close)
620: sql_append(&sql, &sz, ")", e->close);
621: needop = 1;
1.1 schwarze 622: }
623:
624: return(sql);
625: }
626:
627: /*
628: * Compile a set of string tokens into an expression.
629: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
630: * "(", "foo=bar", etc.).
631: */
632: static struct expr *
633: exprcomp(const struct mansearch *search, int argc, char *argv[])
634: {
1.18 schwarze 635: uint64_t mask;
1.4 schwarze 636: int i, toopen, logic, igncase, toclose;
1.18 schwarze 637: struct expr *first, *prev, *cur, *next;
1.1 schwarze 638:
639: first = cur = NULL;
1.6 schwarze 640: logic = igncase = toclose = 0;
1.23 schwarze 641: toopen = NULL != search->sec || NULL != search->arch;
1.1 schwarze 642:
643: for (i = 0; i < argc; i++) {
1.4 schwarze 644: if (0 == strcmp("(", argv[i])) {
645: if (igncase)
646: goto fail;
647: toopen++;
648: toclose++;
649: continue;
650: } else if (0 == strcmp(")", argv[i])) {
651: if (toopen || logic || igncase || NULL == cur)
652: goto fail;
653: cur->close++;
654: if (0 > --toclose)
655: goto fail;
656: continue;
657: } else if (0 == strcmp("-a", argv[i])) {
658: if (toopen || logic || igncase || NULL == cur)
659: goto fail;
660: logic = 1;
661: continue;
662: } else if (0 == strcmp("-o", argv[i])) {
663: if (toopen || logic || igncase || NULL == cur)
664: goto fail;
665: logic = 2;
666: continue;
667: } else if (0 == strcmp("-i", argv[i])) {
668: if (igncase)
669: goto fail;
670: igncase = 1;
671: continue;
1.1 schwarze 672: }
1.4 schwarze 673: next = exprterm(search, argv[i], !igncase);
674: if (NULL == next)
675: goto fail;
1.17 schwarze 676: if (NULL == first)
677: first = next;
678: else
1.1 schwarze 679: cur->next = next;
1.18 schwarze 680: prev = cur = next;
1.17 schwarze 681:
682: /*
683: * Searching for descriptions must be split out
684: * because they are stored in the mpages table,
685: * not in the keys table.
686: */
687:
1.18 schwarze 688: for (mask = TYPE_Nm; mask <= TYPE_Nd; mask <<= 1) {
689: if (mask & cur->bits && ~mask & cur->bits) {
690: next = mandoc_calloc(1,
691: sizeof(struct expr));
692: memcpy(next, cur, sizeof(struct expr));
693: prev->open = 1;
694: cur->bits = mask;
695: cur->next = next;
696: cur = next;
697: cur->bits &= ~mask;
698: }
699: }
700: prev->and = (1 == logic);
701: prev->open += toopen;
702: if (cur != prev)
1.17 schwarze 703: cur->close = 1;
1.18 schwarze 704:
1.4 schwarze 705: toopen = logic = igncase = 0;
1.1 schwarze 706: }
1.6 schwarze 707: if (toopen || logic || igncase || toclose)
708: goto fail;
709:
1.23 schwarze 710: if (NULL != search->sec || NULL != search->arch)
711: cur->close++;
712: if (NULL != search->arch)
713: cur = exprspec(cur, TYPE_arch, search->arch, "^(%s|any)$");
714: if (NULL != search->sec)
715: exprspec(cur, TYPE_sec, search->sec, "^%s$");
1.6 schwarze 716:
717: return(first);
718:
1.4 schwarze 719: fail:
720: if (NULL != first)
721: exprfree(first);
722: return(NULL);
1.1 schwarze 723: }
724:
725: static struct expr *
1.6 schwarze 726: exprspec(struct expr *cur, uint64_t key, const char *value,
727: const char *format)
728: {
729: char errbuf[BUFSIZ];
730: char *cp;
731: int irc;
732:
1.15 schwarze 733: mandoc_asprintf(&cp, format, value);
1.6 schwarze 734: cur->next = mandoc_calloc(1, sizeof(struct expr));
735: cur = cur->next;
736: cur->and = 1;
737: cur->bits = key;
738: if (0 != (irc = regcomp(&cur->regexp, cp,
739: REG_EXTENDED | REG_NOSUB | REG_ICASE))) {
740: regerror(irc, &cur->regexp, errbuf, sizeof(errbuf));
741: fprintf(stderr, "regcomp: %s\n", errbuf);
742: cur->substr = value;
743: }
744: free(cp);
745: return(cur);
746: }
747:
748: static struct expr *
1.1 schwarze 749: exprterm(const struct mansearch *search, char *buf, int cs)
750: {
1.6 schwarze 751: char errbuf[BUFSIZ];
1.1 schwarze 752: struct expr *e;
1.28 schwarze 753: char *key, *val;
1.11 schwarze 754: uint64_t iterbit;
755: int i, irc;
1.1 schwarze 756:
757: if ('\0' == *buf)
758: return(NULL);
759:
760: e = mandoc_calloc(1, sizeof(struct expr));
761:
1.33 schwarze 762: if (search->argmode == ARG_NAME) {
763: e->bits = TYPE_Nm;
1.1 schwarze 764: e->substr = buf;
1.28 schwarze 765: e->equal = 1;
1.1 schwarze 766: return(e);
767: }
768:
769: /*
1.33 schwarze 770: * Separate macro keys from search string.
771: * If needed, request regular expression handling
772: * by setting e->substr to NULL.
1.1 schwarze 773: */
774:
1.33 schwarze 775: if (search->argmode == ARG_WORD) {
776: e->bits = TYPE_Nm;
777: e->substr = NULL;
778: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", buf);
779: cs = 0;
780: } else if ((val = strpbrk(buf, "=~")) == NULL) {
781: e->bits = TYPE_Nm | TYPE_Nd;
1.1 schwarze 782: e->substr = buf;
1.28 schwarze 783: } else {
784: if (val == buf)
1.33 schwarze 785: e->bits = TYPE_Nm | TYPE_Nd;
1.28 schwarze 786: if ('=' == *val)
787: e->substr = val + 1;
788: *val++ = '\0';
1.12 schwarze 789: if (NULL != strstr(buf, "arch"))
790: cs = 0;
1.28 schwarze 791: }
792:
793: /* Compile regular expressions. */
794:
795: if (NULL == e->substr) {
796: irc = regcomp(&e->regexp, val,
797: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.33 schwarze 798: if (search->argmode == ARG_WORD)
1.28 schwarze 799: free(val);
800: if (irc) {
1.6 schwarze 801: regerror(irc, &e->regexp, errbuf, sizeof(errbuf));
802: fprintf(stderr, "regcomp: %s\n", errbuf);
1.1 schwarze 803: free(e);
804: return(NULL);
805: }
1.28 schwarze 806: }
807:
808: if (e->bits)
809: return(e);
1.1 schwarze 810:
811: /*
812: * Parse out all possible fields.
813: * If the field doesn't resolve, bail.
814: */
815:
816: while (NULL != (key = strsep(&buf, ","))) {
817: if ('\0' == *key)
818: continue;
1.11 schwarze 819: for (i = 0, iterbit = 1;
820: i < mansearch_keymax;
821: i++, iterbit <<= 1) {
822: if (0 == strcasecmp(key,
823: mansearch_keynames[i])) {
824: e->bits |= iterbit;
825: break;
826: }
827: }
828: if (i == mansearch_keymax) {
829: if (strcasecmp(key, "any")) {
830: free(e);
831: return(NULL);
832: }
833: e->bits |= ~0ULL;
1.1 schwarze 834: }
835: }
836:
837: return(e);
838: }
839:
840: static void
841: exprfree(struct expr *p)
842: {
843: struct expr *pp;
844:
845: while (NULL != p) {
846: pp = p->next;
847: free(p);
848: p = pp;
849: }
850: }
851:
852: static void *
1.27 espie 853: hash_calloc(size_t nmemb, size_t sz, void *arg)
1.1 schwarze 854: {
855:
1.27 espie 856: return(mandoc_calloc(nmemb, sz));
1.1 schwarze 857: }
858:
859: static void *
860: hash_alloc(size_t sz, void *arg)
861: {
862:
863: return(mandoc_malloc(sz));
864: }
865:
866: static void
1.27 espie 867: hash_free(void *p, void *arg)
1.1 schwarze 868: {
869:
870: free(p);
871: }