Annotation of mandoc/mansearch.c, Revision 1.45
1.45 ! schwarze 1: /* $Id: mansearch.c,v 1.44 2014/08/16 19:00:01 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: #include "config.h"
19:
1.28 schwarze 20: #include <sys/mman.h>
1.43 schwarze 21: #include <sys/types.h>
22:
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:
1.44 schwarze 35: #if 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>
1.40 schwarze 41: #ifndef SQLITE_DETERMINISTIC
42: #define SQLITE_DETERMINISTIC 0
43: #endif
1.1 kristaps 44:
45: #include "mandoc.h"
1.23 schwarze 46: #include "mandoc_aux.h"
1.1 kristaps 47: #include "manpath.h"
48: #include "mansearch.h"
49:
1.20 schwarze 50: extern int mansearch_keymax;
51: extern const char *const mansearch_keynames[];
52:
1.3 kristaps 53: #define SQL_BIND_TEXT(_db, _s, _i, _v) \
1.8 schwarze 54: do { if (SQLITE_OK != sqlite3_bind_text \
1.2 kristaps 55: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
1.8 schwarze 56: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
57: } while (0)
1.3 kristaps 58: #define SQL_BIND_INT64(_db, _s, _i, _v) \
1.8 schwarze 59: do { if (SQLITE_OK != sqlite3_bind_int64 \
1.2 kristaps 60: ((_s), (_i)++, (_v))) \
1.8 schwarze 61: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
62: } while (0)
63: #define SQL_BIND_BLOB(_db, _s, _i, _v) \
64: do { if (SQLITE_OK != sqlite3_bind_blob \
65: ((_s), (_i)++, (&_v), sizeof(_v), SQLITE_STATIC)) \
66: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
67: } while (0)
1.2 kristaps 68:
1.1 kristaps 69: struct expr {
1.38 schwarze 70: regex_t regexp; /* compiled regexp, if applicable */
71: const char *substr; /* to search for, if applicable */
72: struct expr *next; /* next in sequence */
1.34 schwarze 73: uint64_t bits; /* type-mask */
1.38 schwarze 74: int equal; /* equality, not subsring match */
1.13 schwarze 75: int open; /* opening parentheses before */
76: int and; /* logical AND before */
77: int close; /* closing parentheses after */
1.1 kristaps 78: };
79:
80: struct match {
1.32 schwarze 81: uint64_t pageid; /* identifier in database */
1.26 schwarze 82: char *desc; /* manual page description */
1.1 kristaps 83: int form; /* 0 == catpage */
84: };
85:
1.17 schwarze 86: static void buildnames(struct manpage *, sqlite3 *,
1.19 schwarze 87: sqlite3_stmt *, uint64_t,
88: const char *, int form);
1.12 schwarze 89: static char *buildoutput(sqlite3 *, sqlite3_stmt *,
90: uint64_t, uint64_t);
1.1 kristaps 91: static void *hash_alloc(size_t, void *);
1.37 schwarze 92: static void hash_free(void *, void *);
93: static void *hash_calloc(size_t, size_t, void *);
1.34 schwarze 94: static struct expr *exprcomp(const struct mansearch *,
1.5 kristaps 95: int, char *[]);
1.1 kristaps 96: static void exprfree(struct expr *);
1.15 schwarze 97: static struct expr *exprspec(struct expr *, uint64_t,
98: const char *, const char *);
1.8 schwarze 99: static struct expr *exprterm(const struct mansearch *, char *, int);
1.39 schwarze 100: static int manpage_compare(const void *, const void *);
1.13 schwarze 101: static void sql_append(char **sql, size_t *sz,
102: const char *newstr, int count);
1.7 schwarze 103: static void sql_match(sqlite3_context *context,
104: int argc, sqlite3_value **argv);
1.8 schwarze 105: static void sql_regexp(sqlite3_context *context,
106: int argc, sqlite3_value **argv);
1.15 schwarze 107: static char *sql_statement(const struct expr *);
1.28 schwarze 108:
1.34 schwarze 109:
1.28 schwarze 110: int
111: mansearch_setup(int start)
112: {
113: static void *pagecache;
114: int c;
115:
116: #define PC_PAGESIZE 1280
117: #define PC_NUMPAGES 256
118:
119: if (start) {
120: if (NULL != pagecache) {
121: fprintf(stderr, "pagecache already enabled\n");
122: return((int)MANDOCLEVEL_BADARG);
123: }
124:
125: pagecache = mmap(NULL, PC_PAGESIZE * PC_NUMPAGES,
1.42 schwarze 126: PROT_READ | PROT_WRITE,
127: MAP_SHARED | MAP_ANON, -1, 0);
1.28 schwarze 128:
129: if (MAP_FAILED == pagecache) {
130: perror("mmap");
131: pagecache = NULL;
132: return((int)MANDOCLEVEL_SYSERR);
133: }
134:
135: c = sqlite3_config(SQLITE_CONFIG_PAGECACHE,
136: pagecache, PC_PAGESIZE, PC_NUMPAGES);
137:
138: if (SQLITE_OK == c)
139: return((int)MANDOCLEVEL_OK);
140:
141: fprintf(stderr, "pagecache: %s\n", sqlite3_errstr(c));
142:
143: } else if (NULL == pagecache) {
144: fprintf(stderr, "pagecache missing\n");
145: return((int)MANDOCLEVEL_BADARG);
146: }
147:
148: if (-1 == munmap(pagecache, PC_PAGESIZE * PC_NUMPAGES)) {
149: perror("munmap");
150: pagecache = NULL;
151: return((int)MANDOCLEVEL_SYSERR);
152: }
153:
154: pagecache = NULL;
155: return((int)MANDOCLEVEL_OK);
156: }
1.1 kristaps 157:
158: int
1.5 kristaps 159: mansearch(const struct mansearch *search,
1.12 schwarze 160: const struct manpaths *paths,
161: int argc, char *argv[],
1.1 kristaps 162: struct manpage **res, size_t *sz)
163: {
1.20 schwarze 164: int fd, rc, c, indexbit;
1.32 schwarze 165: int64_t pageid;
1.20 schwarze 166: uint64_t outbit, iterbit;
1.6 schwarze 167: char buf[PATH_MAX];
1.11 schwarze 168: char *sql;
1.10 schwarze 169: struct manpage *mpage;
1.1 kristaps 170: struct expr *e, *ep;
171: sqlite3 *db;
1.12 schwarze 172: sqlite3_stmt *s, *s2;
1.1 kristaps 173: struct match *mp;
174: struct ohash_info info;
175: struct ohash htab;
176: unsigned int idx;
177: size_t i, j, cur, maxres;
178:
1.37 schwarze 179: info.calloc = hash_calloc;
1.1 kristaps 180: info.alloc = hash_alloc;
1.37 schwarze 181: info.free = hash_free;
1.32 schwarze 182: info.key_offset = offsetof(struct match, pageid);
1.1 kristaps 183:
1.2 kristaps 184: *sz = cur = maxres = 0;
1.1 kristaps 185: sql = NULL;
186: *res = NULL;
187: fd = -1;
188: e = NULL;
1.2 kristaps 189: rc = 0;
1.1 kristaps 190:
191: if (0 == argc)
192: goto out;
1.5 kristaps 193: if (NULL == (e = exprcomp(search, argc, argv)))
1.1 kristaps 194: goto out;
195:
1.12 schwarze 196: outbit = 0;
1.45 ! schwarze 197: if (NULL != search->outkey) {
1.20 schwarze 198: for (indexbit = 0, iterbit = 1;
199: indexbit < mansearch_keymax;
200: indexbit++, iterbit <<= 1) {
1.45 ! schwarze 201: if (0 == strcasecmp(search->outkey,
1.20 schwarze 202: mansearch_keynames[indexbit])) {
203: outbit = iterbit;
1.12 schwarze 204: break;
205: }
206: }
207: }
208:
1.1 kristaps 209: /*
210: * Save a descriptor to the current working directory.
211: * Since pathnames in the "paths" variable might be relative,
212: * and we'll be chdir()ing into them, we need to keep a handle
213: * on our current directory from which to start the chdir().
214: */
215:
1.6 schwarze 216: if (NULL == getcwd(buf, PATH_MAX)) {
1.30 schwarze 217: perror("getcwd");
1.1 kristaps 218: goto out;
219: } else if (-1 == (fd = open(buf, O_RDONLY, 0))) {
220: perror(buf);
221: goto out;
222: }
223:
1.15 schwarze 224: sql = sql_statement(e);
1.1 kristaps 225:
226: /*
227: * Loop over the directories (containing databases) for us to
228: * search.
229: * Don't let missing/bad databases/directories phase us.
230: * In each, try to open the resident database and, if it opens,
231: * scan it for our match expression.
232: */
233:
234: for (i = 0; i < paths->sz; i++) {
235: if (-1 == fchdir(fd)) {
236: perror(buf);
237: free(*res);
238: break;
239: } else if (-1 == chdir(paths->paths[i])) {
240: perror(paths->paths[i]);
241: continue;
1.34 schwarze 242: }
1.1 kristaps 243:
1.34 schwarze 244: c = sqlite3_open_v2(MANDOC_DB, &db,
245: SQLITE_OPEN_READONLY, NULL);
1.1 kristaps 246:
1.2 kristaps 247: if (SQLITE_OK != c) {
1.1 kristaps 248: perror(MANDOC_DB);
249: sqlite3_close(db);
250: continue;
251: }
252:
1.8 schwarze 253: /*
254: * Define the SQL functions for substring
255: * and regular expression matching.
256: */
1.7 schwarze 257:
258: c = sqlite3_create_function(db, "match", 2,
1.31 schwarze 259: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
260: NULL, sql_match, NULL, NULL);
1.8 schwarze 261: assert(SQLITE_OK == c);
262: c = sqlite3_create_function(db, "regexp", 2,
1.31 schwarze 263: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
264: NULL, sql_regexp, NULL, NULL);
1.8 schwarze 265: assert(SQLITE_OK == c);
1.7 schwarze 266:
1.1 kristaps 267: j = 1;
1.2 kristaps 268: c = sqlite3_prepare_v2(db, sql, -1, &s, NULL);
269: if (SQLITE_OK != c)
270: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
1.1 kristaps 271:
272: for (ep = e; NULL != ep; ep = ep->next) {
1.8 schwarze 273: if (NULL == ep->substr) {
274: SQL_BIND_BLOB(db, s, j, ep->regexp);
275: } else
276: SQL_BIND_TEXT(db, s, j, ep->substr);
1.27 schwarze 277: if (0 == ((TYPE_Nd | TYPE_Nm) & ep->bits))
1.26 schwarze 278: SQL_BIND_INT64(db, s, j, ep->bits);
1.1 kristaps 279: }
280:
281: memset(&htab, 0, sizeof(struct ohash));
282: ohash_init(&htab, 4, &info);
283:
284: /*
285: * Hash each entry on its [unique] document identifier.
286: * This is a uint64_t.
287: * Instead of using a hash function, simply convert the
288: * uint64_t to a uint32_t, the hash value's type.
289: * This gives good performance and preserves the
290: * distribution of buckets in the table.
291: */
1.2 kristaps 292: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.32 schwarze 293: pageid = sqlite3_column_int64(s, 2);
1.34 schwarze 294: idx = ohash_lookup_memory(&htab,
295: (char *)&pageid, sizeof(uint64_t),
296: (uint32_t)pageid);
1.1 kristaps 297:
298: if (NULL != ohash_find(&htab, idx))
299: continue;
300:
301: mp = mandoc_calloc(1, sizeof(struct match));
1.32 schwarze 302: mp->pageid = pageid;
1.26 schwarze 303: mp->form = sqlite3_column_int(s, 1);
304: if (TYPE_Nd == outbit)
1.41 schwarze 305: mp->desc = mandoc_strdup((const char *)
1.26 schwarze 306: sqlite3_column_text(s, 0));
1.1 kristaps 307: ohash_insert(&htab, idx, mp);
308: }
309:
1.2 kristaps 310: if (SQLITE_DONE != c)
311: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
312:
1.1 kristaps 313: sqlite3_finalize(s);
1.10 schwarze 314:
1.34 schwarze 315: c = sqlite3_prepare_v2(db,
1.35 schwarze 316: "SELECT sec, arch, name, pageid FROM mlinks "
317: "WHERE pageid=? ORDER BY sec, arch, name",
1.10 schwarze 318: -1, &s, NULL);
319: if (SQLITE_OK != c)
320: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
1.1 kristaps 321:
1.12 schwarze 322: c = sqlite3_prepare_v2(db,
1.35 schwarze 323: "SELECT bits, key, pageid FROM keys "
324: "WHERE pageid=? AND bits & ?",
1.12 schwarze 325: -1, &s2, NULL);
326: if (SQLITE_OK != c)
327: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
328:
1.1 kristaps 329: for (mp = ohash_first(&htab, &idx);
330: NULL != mp;
331: mp = ohash_next(&htab, &idx)) {
332: if (cur + 1 > maxres) {
333: maxres += 1024;
1.36 schwarze 334: *res = mandoc_reallocarray(*res,
335: maxres, sizeof(struct manpage));
1.1 kristaps 336: }
1.10 schwarze 337: mpage = *res + cur;
1.39 schwarze 338: mpage->sec = 10;
1.10 schwarze 339: mpage->form = mp->form;
1.32 schwarze 340: buildnames(mpage, db, s, mp->pageid,
1.19 schwarze 341: paths->paths[i], mp->form);
1.26 schwarze 342: mpage->output = TYPE_Nd & outbit ?
343: mp->desc : outbit ?
1.32 schwarze 344: buildoutput(db, s2, mp->pageid, outbit) : NULL;
1.10 schwarze 345:
1.1 kristaps 346: free(mp);
347: cur++;
348: }
1.10 schwarze 349:
350: sqlite3_finalize(s);
1.12 schwarze 351: sqlite3_finalize(s2);
1.10 schwarze 352: sqlite3_close(db);
1.1 kristaps 353: ohash_delete(&htab);
354: }
1.39 schwarze 355: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.2 kristaps 356: rc = 1;
1.1 kristaps 357: out:
1.30 schwarze 358: if (-1 != fd) {
359: if (-1 == fchdir(fd))
360: perror(buf);
361: close(fd);
362: }
1.1 kristaps 363: exprfree(e);
364: free(sql);
365: *sz = cur;
1.2 kristaps 366: return(rc);
1.11 schwarze 367: }
368:
1.45 ! schwarze 369: void
! 370: mansearch_free(struct manpage *res, size_t sz)
! 371: {
! 372: size_t i;
! 373:
! 374: for (i = 0; i < sz; i++) {
! 375: free(res[i].file);
! 376: free(res[i].names);
! 377: free(res[i].output);
! 378: }
! 379: free(res);
! 380: }
! 381:
1.39 schwarze 382: static int
383: manpage_compare(const void *vp1, const void *vp2)
384: {
385: const struct manpage *mp1, *mp2;
386: int diff;
387:
388: mp1 = vp1;
389: mp2 = vp2;
390: diff = mp1->sec - mp2->sec;
391: return(diff ? diff : strcasecmp(mp1->names, mp2->names));
392: }
393:
1.17 schwarze 394: static void
395: buildnames(struct manpage *mpage, sqlite3 *db, sqlite3_stmt *s,
1.32 schwarze 396: uint64_t pageid, const char *path, int form)
1.11 schwarze 397: {
1.22 schwarze 398: char *newnames, *prevsec, *prevarch;
1.19 schwarze 399: const char *oldnames, *sep1, *name, *sec, *sep2, *arch, *fsec;
1.11 schwarze 400: size_t i;
401: int c;
402:
1.25 schwarze 403: mpage->file = NULL;
1.17 schwarze 404: mpage->names = NULL;
1.22 schwarze 405: prevsec = prevarch = NULL;
1.11 schwarze 406: i = 1;
1.32 schwarze 407: SQL_BIND_INT64(db, s, i, pageid);
1.11 schwarze 408: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.17 schwarze 409:
1.22 schwarze 410: /* Decide whether we already have some names. */
1.17 schwarze 411:
412: if (NULL == mpage->names) {
1.11 schwarze 413: oldnames = "";
414: sep1 = "";
415: } else {
1.17 schwarze 416: oldnames = mpage->names;
1.11 schwarze 417: sep1 = ", ";
418: }
1.22 schwarze 419:
420: /* Fetch the next name. */
421:
1.41 schwarze 422: sec = (const char *)sqlite3_column_text(s, 0);
423: arch = (const char *)sqlite3_column_text(s, 1);
424: name = (const char *)sqlite3_column_text(s, 2);
1.39 schwarze 425:
426: /* Remember the first section found. */
427:
428: if (9 < mpage->sec && '1' <= *sec && '9' >= *sec)
429: mpage->sec = (*sec - '1') + 1;
1.22 schwarze 430:
431: /* If the section changed, append the old one. */
432:
433: if (NULL != prevsec &&
434: (strcmp(sec, prevsec) ||
435: strcmp(arch, prevarch))) {
436: sep2 = '\0' == *prevarch ? "" : "/";
1.24 schwarze 437: mandoc_asprintf(&newnames, "%s(%s%s%s)",
438: oldnames, prevsec, sep2, prevarch);
1.22 schwarze 439: free(mpage->names);
440: oldnames = mpage->names = newnames;
441: free(prevsec);
442: free(prevarch);
443: prevsec = prevarch = NULL;
444: }
445:
446: /* Save the new section, to append it later. */
447:
448: if (NULL == prevsec) {
449: prevsec = mandoc_strdup(sec);
450: prevarch = mandoc_strdup(arch);
451: }
452:
453: /* Append the new name. */
454:
1.24 schwarze 455: mandoc_asprintf(&newnames, "%s%s%s",
456: oldnames, sep1, name);
1.17 schwarze 457: free(mpage->names);
458: mpage->names = newnames;
459:
460: /* Also save the first file name encountered. */
461:
462: if (NULL != mpage->file)
463: continue;
464:
1.19 schwarze 465: if (form) {
466: sep1 = "man";
467: fsec = sec;
468: } else {
469: sep1 = "cat";
470: fsec = "0";
471: }
1.22 schwarze 472: sep2 = '\0' == *arch ? "" : "/";
1.24 schwarze 473: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.%s",
474: path, sep1, sec, sep2, arch, name, fsec);
1.11 schwarze 475: }
476: if (SQLITE_DONE != c)
477: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
478: sqlite3_reset(s);
1.22 schwarze 479:
480: /* Append one final section to the names. */
481:
482: if (NULL != prevsec) {
483: sep2 = '\0' == *prevarch ? "" : "/";
1.24 schwarze 484: mandoc_asprintf(&newnames, "%s(%s%s%s)",
485: mpage->names, prevsec, sep2, prevarch);
1.22 schwarze 486: free(mpage->names);
487: mpage->names = newnames;
488: free(prevsec);
489: free(prevarch);
490: }
1.12 schwarze 491: }
492:
493: static char *
1.32 schwarze 494: buildoutput(sqlite3 *db, sqlite3_stmt *s, uint64_t pageid, uint64_t outbit)
1.12 schwarze 495: {
496: char *output, *newoutput;
497: const char *oldoutput, *sep1, *data;
498: size_t i;
499: int c;
500:
501: output = NULL;
502: i = 1;
1.32 schwarze 503: SQL_BIND_INT64(db, s, i, pageid);
1.12 schwarze 504: SQL_BIND_INT64(db, s, i, outbit);
505: while (SQLITE_ROW == (c = sqlite3_step(s))) {
506: if (NULL == output) {
507: oldoutput = "";
508: sep1 = "";
509: } else {
510: oldoutput = output;
511: sep1 = " # ";
512: }
1.41 schwarze 513: data = (const char *)sqlite3_column_text(s, 1);
1.24 schwarze 514: mandoc_asprintf(&newoutput, "%s%s%s",
515: oldoutput, sep1, data);
1.12 schwarze 516: free(output);
517: output = newoutput;
518: }
519: if (SQLITE_DONE != c)
520: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
521: sqlite3_reset(s);
522: return(output);
1.1 kristaps 523: }
524:
525: /*
1.7 schwarze 526: * Implement substring match as an application-defined SQL function.
527: * Using the SQL LIKE or GLOB operators instead would be a bad idea
528: * because that would require escaping metacharacters in the string
529: * being searched for.
530: */
531: static void
532: sql_match(sqlite3_context *context, int argc, sqlite3_value **argv)
533: {
534:
535: assert(2 == argc);
536: sqlite3_result_int(context, NULL != strcasestr(
537: (const char *)sqlite3_value_text(argv[1]),
538: (const char *)sqlite3_value_text(argv[0])));
539: }
540:
541: /*
1.8 schwarze 542: * Implement regular expression match
543: * as an application-defined SQL function.
544: */
545: static void
546: sql_regexp(sqlite3_context *context, int argc, sqlite3_value **argv)
547: {
548:
549: assert(2 == argc);
550: sqlite3_result_int(context, !regexec(
551: (regex_t *)sqlite3_value_blob(argv[0]),
552: (const char *)sqlite3_value_text(argv[1]),
553: 0, NULL, 0));
554: }
555:
1.13 schwarze 556: static void
557: sql_append(char **sql, size_t *sz, const char *newstr, int count)
558: {
559: size_t newsz;
560:
561: newsz = 1 < count ? (size_t)count : strlen(newstr);
562: *sql = mandoc_realloc(*sql, *sz + newsz + 1);
563: if (1 < count)
564: memset(*sql + *sz, *newstr, (size_t)count);
565: else
566: memcpy(*sql + *sz, newstr, newsz);
567: *sz += newsz;
568: (*sql)[*sz] = '\0';
569: }
570:
1.8 schwarze 571: /*
1.1 kristaps 572: * Prepare the search SQL statement.
573: */
574: static char *
1.15 schwarze 575: sql_statement(const struct expr *e)
1.1 kristaps 576: {
577: char *sql;
578: size_t sz;
1.13 schwarze 579: int needop;
1.1 kristaps 580:
1.35 schwarze 581: sql = mandoc_strdup(
582: "SELECT desc, form, pageid FROM mpages WHERE ");
1.1 kristaps 583: sz = strlen(sql);
1.2 kristaps 584:
1.13 schwarze 585: for (needop = 0; NULL != e; e = e->next) {
586: if (e->and)
587: sql_append(&sql, &sz, " AND ", 1);
588: else if (needop)
589: sql_append(&sql, &sz, " OR ", 1);
590: if (e->open)
591: sql_append(&sql, &sz, "(", e->open);
1.26 schwarze 592: sql_append(&sql, &sz,
593: TYPE_Nd & e->bits
594: ? (NULL == e->substr
595: ? "desc REGEXP ?"
596: : "desc MATCH ?")
1.27 schwarze 597: : TYPE_Nm == e->bits
598: ? (NULL == e->substr
1.32 schwarze 599: ? "pageid IN (SELECT pageid FROM names "
1.27 schwarze 600: "WHERE name REGEXP ?)"
1.38 schwarze 601: : e->equal
602: ? "pageid IN (SELECT pageid FROM names "
603: "WHERE name = ?)"
1.32 schwarze 604: : "pageid IN (SELECT pageid FROM names "
1.27 schwarze 605: "WHERE name MATCH ?)")
1.26 schwarze 606: : (NULL == e->substr
1.32 schwarze 607: ? "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 608: "WHERE key REGEXP ? AND bits & ?)"
1.32 schwarze 609: : "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 610: "WHERE key MATCH ? AND bits & ?)"), 1);
1.13 schwarze 611: if (e->close)
612: sql_append(&sql, &sz, ")", e->close);
613: needop = 1;
1.1 kristaps 614: }
615:
616: return(sql);
617: }
618:
619: /*
620: * Compile a set of string tokens into an expression.
621: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
622: * "(", "foo=bar", etc.).
623: */
624: static struct expr *
1.5 kristaps 625: exprcomp(const struct mansearch *search, int argc, char *argv[])
1.1 kristaps 626: {
1.27 schwarze 627: uint64_t mask;
1.13 schwarze 628: int i, toopen, logic, igncase, toclose;
1.27 schwarze 629: struct expr *first, *prev, *cur, *next;
1.1 kristaps 630:
631: first = cur = NULL;
1.15 schwarze 632: logic = igncase = toclose = 0;
1.33 schwarze 633: toopen = NULL != search->sec || NULL != search->arch;
1.1 kristaps 634:
635: for (i = 0; i < argc; i++) {
1.13 schwarze 636: if (0 == strcmp("(", argv[i])) {
637: if (igncase)
638: goto fail;
639: toopen++;
640: toclose++;
641: continue;
642: } else if (0 == strcmp(")", argv[i])) {
643: if (toopen || logic || igncase || NULL == cur)
644: goto fail;
645: cur->close++;
646: if (0 > --toclose)
647: goto fail;
648: continue;
649: } else if (0 == strcmp("-a", argv[i])) {
650: if (toopen || logic || igncase || NULL == cur)
651: goto fail;
652: logic = 1;
653: continue;
654: } else if (0 == strcmp("-o", argv[i])) {
655: if (toopen || logic || igncase || NULL == cur)
656: goto fail;
657: logic = 2;
658: continue;
659: } else if (0 == strcmp("-i", argv[i])) {
660: if (igncase)
661: goto fail;
662: igncase = 1;
663: continue;
1.1 kristaps 664: }
1.13 schwarze 665: next = exprterm(search, argv[i], !igncase);
666: if (NULL == next)
667: goto fail;
1.26 schwarze 668: if (NULL == first)
669: first = next;
670: else
1.1 kristaps 671: cur->next = next;
1.27 schwarze 672: prev = cur = next;
1.26 schwarze 673:
674: /*
675: * Searching for descriptions must be split out
676: * because they are stored in the mpages table,
677: * not in the keys table.
678: */
679:
1.27 schwarze 680: for (mask = TYPE_Nm; mask <= TYPE_Nd; mask <<= 1) {
681: if (mask & cur->bits && ~mask & cur->bits) {
682: next = mandoc_calloc(1,
683: sizeof(struct expr));
684: memcpy(next, cur, sizeof(struct expr));
685: prev->open = 1;
686: cur->bits = mask;
687: cur->next = next;
688: cur = next;
689: cur->bits &= ~mask;
690: }
691: }
692: prev->and = (1 == logic);
693: prev->open += toopen;
694: if (cur != prev)
1.26 schwarze 695: cur->close = 1;
1.27 schwarze 696:
1.13 schwarze 697: toopen = logic = igncase = 0;
1.1 kristaps 698: }
1.15 schwarze 699: if (toopen || logic || igncase || toclose)
700: goto fail;
701:
1.33 schwarze 702: if (NULL != search->sec || NULL != search->arch)
703: cur->close++;
704: if (NULL != search->arch)
705: cur = exprspec(cur, TYPE_arch, search->arch, "^(%s|any)$");
706: if (NULL != search->sec)
707: exprspec(cur, TYPE_sec, search->sec, "^%s$");
1.15 schwarze 708:
709: return(first);
710:
1.13 schwarze 711: fail:
712: if (NULL != first)
713: exprfree(first);
714: return(NULL);
1.1 kristaps 715: }
716:
717: static struct expr *
1.15 schwarze 718: exprspec(struct expr *cur, uint64_t key, const char *value,
719: const char *format)
720: {
721: char errbuf[BUFSIZ];
722: char *cp;
723: int irc;
724:
1.24 schwarze 725: mandoc_asprintf(&cp, format, value);
1.15 schwarze 726: cur->next = mandoc_calloc(1, sizeof(struct expr));
727: cur = cur->next;
728: cur->and = 1;
729: cur->bits = key;
730: if (0 != (irc = regcomp(&cur->regexp, cp,
731: REG_EXTENDED | REG_NOSUB | REG_ICASE))) {
732: regerror(irc, &cur->regexp, errbuf, sizeof(errbuf));
733: fprintf(stderr, "regcomp: %s\n", errbuf);
734: cur->substr = value;
735: }
736: free(cp);
737: return(cur);
738: }
739:
740: static struct expr *
1.8 schwarze 741: exprterm(const struct mansearch *search, char *buf, int cs)
1.1 kristaps 742: {
1.15 schwarze 743: char errbuf[BUFSIZ];
1.1 kristaps 744: struct expr *e;
1.38 schwarze 745: char *key, *val;
1.20 schwarze 746: uint64_t iterbit;
747: int i, irc;
1.1 kristaps 748:
749: if ('\0' == *buf)
750: return(NULL);
751:
752: e = mandoc_calloc(1, sizeof(struct expr));
753:
1.45 ! schwarze 754: if (search->argmode == ARG_NAME) {
! 755: e->bits = TYPE_Nm;
1.8 schwarze 756: e->substr = buf;
1.38 schwarze 757: e->equal = 1;
1.5 kristaps 758: return(e);
759: }
760:
1.1 kristaps 761: /*
1.45 ! schwarze 762: * Separate macro keys from search string.
! 763: * If needed, request regular expression handling
! 764: * by setting e->substr to NULL.
1.1 kristaps 765: */
766:
1.45 ! schwarze 767: if (search->argmode == ARG_WORD) {
! 768: e->bits = TYPE_Nm;
! 769: e->substr = NULL;
! 770: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", buf);
! 771: } else if ((val = strpbrk(buf, "=~")) == NULL) {
! 772: e->bits = TYPE_Nm | TYPE_Nd;
1.8 schwarze 773: e->substr = buf;
1.38 schwarze 774: } else {
775: if (val == buf)
1.45 ! schwarze 776: e->bits = TYPE_Nm | TYPE_Nd;
1.38 schwarze 777: if ('=' == *val)
778: e->substr = val + 1;
779: *val++ = '\0';
1.21 schwarze 780: if (NULL != strstr(buf, "arch"))
781: cs = 0;
1.38 schwarze 782: }
783:
784: /* Compile regular expressions. */
785:
786: if (NULL == e->substr) {
787: irc = regcomp(&e->regexp, val,
788: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.45 ! schwarze 789: if (search->argmode == ARG_WORD)
1.38 schwarze 790: free(val);
791: if (irc) {
1.15 schwarze 792: regerror(irc, &e->regexp, errbuf, sizeof(errbuf));
793: fprintf(stderr, "regcomp: %s\n", errbuf);
1.8 schwarze 794: free(e);
795: return(NULL);
796: }
1.38 schwarze 797: }
798:
799: if (e->bits)
800: return(e);
1.1 kristaps 801:
802: /*
803: * Parse out all possible fields.
804: * If the field doesn't resolve, bail.
805: */
806:
807: while (NULL != (key = strsep(&buf, ","))) {
808: if ('\0' == *key)
809: continue;
1.20 schwarze 810: for (i = 0, iterbit = 1;
811: i < mansearch_keymax;
812: i++, iterbit <<= 1) {
813: if (0 == strcasecmp(key,
814: mansearch_keynames[i])) {
815: e->bits |= iterbit;
816: break;
817: }
818: }
819: if (i == mansearch_keymax) {
820: if (strcasecmp(key, "any")) {
821: free(e);
822: return(NULL);
823: }
824: e->bits |= ~0ULL;
1.1 kristaps 825: }
826: }
827:
828: return(e);
829: }
830:
831: static void
832: exprfree(struct expr *p)
833: {
834: struct expr *pp;
835:
836: while (NULL != p) {
837: pp = p->next;
838: free(p);
839: p = pp;
840: }
841: }
842:
843: static void *
1.37 schwarze 844: hash_calloc(size_t nmemb, size_t sz, void *arg)
1.1 kristaps 845: {
846:
1.37 schwarze 847: return(mandoc_calloc(nmemb, sz));
1.1 kristaps 848: }
849:
850: static void *
851: hash_alloc(size_t sz, void *arg)
852: {
853:
854: return(mandoc_malloc(sz));
855: }
856:
857: static void
1.37 schwarze 858: hash_free(void *p, void *arg)
1.1 kristaps 859: {
860:
861: free(p);
862: }
CVSweb