Annotation of mandoc/apropos_db.c, Revision 1.15
1.15 ! schwarze 1: /* $Id: apropos_db.c,v 1.14 2011/11/27 23:11:37 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2011 Kristaps Dzonsons <kristaps@bsd.lv>
1.3 schwarze 4: * Copyright (c) 2011 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: */
18: #include <assert.h>
19: #include <fcntl.h>
20: #include <regex.h>
21: #include <stdarg.h>
1.6 kristaps 22: #include <stdint.h>
1.1 schwarze 23: #include <stdlib.h>
24: #include <string.h>
1.8 kristaps 25: #include <unistd.h>
1.1 schwarze 26:
27: #ifdef __linux__
28: # include <db_185.h>
29: #else
30: # include <db.h>
31: #endif
32:
1.2 schwarze 33: #include "mandocdb.h"
1.1 schwarze 34: #include "apropos_db.h"
35: #include "mandoc.h"
36:
1.5 kristaps 37: struct rec {
38: struct res res; /* resulting record info */
39: /*
40: * Maintain a binary tree for checking the uniqueness of `rec'
41: * when adding elements to the results array.
42: * Since the results array is dynamic, use offset in the array
43: * instead of a pointer to the structure.
44: */
45: int lhs;
46: int rhs;
47: int matched; /* expression is true */
48: int *matches; /* partial truth evaluations */
49: };
50:
1.1 schwarze 51: struct expr {
1.5 kristaps 52: int regex; /* is regex? */
53: int index; /* index in match array */
1.6 kristaps 54: uint64_t mask; /* type-mask */
1.5 kristaps 55: int cs; /* is case-sensitive? */
56: int and; /* is rhs of logical AND? */
57: char *v; /* search value */
58: regex_t re; /* compiled re, if regex */
59: struct expr *next; /* next in sequence */
60: struct expr *subexpr;
1.1 schwarze 61: };
62:
63: struct type {
1.6 kristaps 64: uint64_t mask;
1.1 schwarze 65: const char *name;
66: };
67:
1.8 kristaps 68: struct rectree {
69: struct rec *node; /* record array for dir tree */
70: int len; /* length of record array */
71: };
72:
1.1 schwarze 73: static const struct type types[] = {
1.7 kristaps 74: { TYPE_An, "An" },
75: { TYPE_Ar, "Ar" },
76: { TYPE_At, "At" },
77: { TYPE_Bsx, "Bsx" },
78: { TYPE_Bx, "Bx" },
79: { TYPE_Cd, "Cd" },
80: { TYPE_Cm, "Cm" },
81: { TYPE_Dv, "Dv" },
82: { TYPE_Dx, "Dx" },
83: { TYPE_Em, "Em" },
84: { TYPE_Er, "Er" },
85: { TYPE_Ev, "Ev" },
86: { TYPE_Fa, "Fa" },
87: { TYPE_Fl, "Fl" },
88: { TYPE_Fn, "Fn" },
89: { TYPE_Fn, "Fo" },
90: { TYPE_Ft, "Ft" },
91: { TYPE_Fx, "Fx" },
92: { TYPE_Ic, "Ic" },
93: { TYPE_In, "In" },
94: { TYPE_Lb, "Lb" },
95: { TYPE_Li, "Li" },
96: { TYPE_Lk, "Lk" },
97: { TYPE_Ms, "Ms" },
98: { TYPE_Mt, "Mt" },
99: { TYPE_Nd, "Nd" },
100: { TYPE_Nm, "Nm" },
101: { TYPE_Nx, "Nx" },
102: { TYPE_Ox, "Ox" },
103: { TYPE_Pa, "Pa" },
104: { TYPE_Rs, "Rs" },
105: { TYPE_Sh, "Sh" },
106: { TYPE_Ss, "Ss" },
107: { TYPE_St, "St" },
108: { TYPE_Sy, "Sy" },
109: { TYPE_Tn, "Tn" },
110: { TYPE_Va, "Va" },
111: { TYPE_Va, "Vt" },
112: { TYPE_Xr, "Xr" },
1.2 schwarze 113: { INT_MAX, "any" },
1.1 schwarze 114: { 0, NULL }
115: };
116:
117: static DB *btree_open(void);
1.12 schwarze 118: static int btree_read(const DBT *,
1.5 kristaps 119: const struct mchars *, char **);
120: static int expreval(const struct expr *, int *);
1.12 schwarze 121: static void exprexec(const struct expr *,
1.6 kristaps 122: const char *, uint64_t, struct rec *);
1.12 schwarze 123: static int exprmark(const struct expr *,
1.6 kristaps 124: const char *, uint64_t, int *);
1.5 kristaps 125: static struct expr *exprexpr(int, char *[], int *, int *, size_t *);
126: static struct expr *exprterm(char *, int);
1.1 schwarze 127: static DB *index_open(void);
1.11 kristaps 128: static int index_read(const DBT *, const DBT *, int,
1.1 schwarze 129: const struct mchars *, struct rec *);
130: static void norm_string(const char *,
131: const struct mchars *, char **);
132: static size_t norm_utf8(unsigned int, char[7]);
1.5 kristaps 133: static void recfree(struct rec *);
1.8 kristaps 134: static int single_search(struct rectree *, const struct opts *,
135: const struct expr *, size_t terms,
1.11 kristaps 136: struct mchars *, int);
1.1 schwarze 137:
138: /*
139: * Open the keyword mandoc-db database.
140: */
141: static DB *
142: btree_open(void)
143: {
144: BTREEINFO info;
145: DB *db;
146:
147: memset(&info, 0, sizeof(BTREEINFO));
148: info.flags = R_DUP;
149:
1.2 schwarze 150: db = dbopen(MANDOC_DB, O_RDONLY, 0, DB_BTREE, &info);
1.12 schwarze 151: if (NULL != db)
1.1 schwarze 152: return(db);
153:
154: return(NULL);
155: }
156:
157: /*
158: * Read a keyword from the database and normalise it.
159: * Return 0 if the database is insane, else 1.
160: */
161: static int
162: btree_read(const DBT *v, const struct mchars *mc, char **buf)
163: {
164:
165: /* Sanity: are we nil-terminated? */
166:
167: assert(v->size > 0);
1.6 kristaps 168:
1.1 schwarze 169: if ('\0' != ((char *)v->data)[(int)v->size - 1])
170: return(0);
171:
172: norm_string((char *)v->data, mc, buf);
173: return(1);
174: }
175:
176: /*
177: * Take a Unicode codepoint and produce its UTF-8 encoding.
178: * This isn't the best way to do this, but it works.
1.12 schwarze 179: * The magic numbers are from the UTF-8 packaging.
1.1 schwarze 180: * They're not as scary as they seem: read the UTF-8 spec for details.
181: */
182: static size_t
183: norm_utf8(unsigned int cp, char out[7])
184: {
185: size_t rc;
186:
187: rc = 0;
188:
189: if (cp <= 0x0000007F) {
190: rc = 1;
191: out[0] = (char)cp;
192: } else if (cp <= 0x000007FF) {
193: rc = 2;
194: out[0] = (cp >> 6 & 31) | 192;
195: out[1] = (cp & 63) | 128;
196: } else if (cp <= 0x0000FFFF) {
197: rc = 3;
198: out[0] = (cp >> 12 & 15) | 224;
199: out[1] = (cp >> 6 & 63) | 128;
200: out[2] = (cp & 63) | 128;
201: } else if (cp <= 0x001FFFFF) {
202: rc = 4;
203: out[0] = (cp >> 18 & 7) | 240;
204: out[1] = (cp >> 12 & 63) | 128;
205: out[2] = (cp >> 6 & 63) | 128;
206: out[3] = (cp & 63) | 128;
207: } else if (cp <= 0x03FFFFFF) {
208: rc = 5;
209: out[0] = (cp >> 24 & 3) | 248;
210: out[1] = (cp >> 18 & 63) | 128;
211: out[2] = (cp >> 12 & 63) | 128;
212: out[3] = (cp >> 6 & 63) | 128;
213: out[4] = (cp & 63) | 128;
214: } else if (cp <= 0x7FFFFFFF) {
215: rc = 6;
216: out[0] = (cp >> 30 & 1) | 252;
217: out[1] = (cp >> 24 & 63) | 128;
218: out[2] = (cp >> 18 & 63) | 128;
219: out[3] = (cp >> 12 & 63) | 128;
220: out[4] = (cp >> 6 & 63) | 128;
221: out[5] = (cp & 63) | 128;
222: } else
223: return(0);
224:
225: out[rc] = '\0';
226: return(rc);
227: }
228:
229: /*
230: * Normalise strings from the index and database.
231: * These strings are escaped as defined by mandoc_char(7) along with
232: * other goop in mandoc.h (e.g., soft hyphens).
233: * This function normalises these into a nice UTF-8 string.
234: * Returns 0 if the database is fucked.
235: */
236: static void
237: norm_string(const char *val, const struct mchars *mc, char **buf)
238: {
239: size_t sz, bsz;
240: char utfbuf[7];
241: const char *seq, *cpp;
242: int len, u, pos;
243: enum mandoc_esc esc;
1.12 schwarze 244: static const char res[] = { '\\', '\t',
1.1 schwarze 245: ASCII_NBRSP, ASCII_HYPH, '\0' };
246:
247: /* Pre-allocate by the length of the input */
248:
249: bsz = strlen(val) + 1;
250: *buf = mandoc_realloc(*buf, bsz);
251: pos = 0;
252:
253: while ('\0' != *val) {
254: /*
255: * Halt on the first escape sequence.
256: * This also halts on the end of string, in which case
257: * we just copy, fallthrough, and exit the loop.
258: */
259: if ((sz = strcspn(val, res)) > 0) {
260: memcpy(&(*buf)[pos], val, sz);
261: pos += (int)sz;
262: val += (int)sz;
263: }
264:
265: if (ASCII_HYPH == *val) {
266: (*buf)[pos++] = '-';
267: val++;
268: continue;
269: } else if ('\t' == *val || ASCII_NBRSP == *val) {
270: (*buf)[pos++] = ' ';
271: val++;
272: continue;
273: } else if ('\\' != *val)
274: break;
275:
276: /* Read past the slash. */
277:
278: val++;
279: u = 0;
280:
281: /*
282: * Parse the escape sequence and see if it's a
283: * predefined character or special character.
284: */
285:
286: esc = mandoc_escape(&val, &seq, &len);
287: if (ESCAPE_ERROR == esc)
288: break;
289:
1.12 schwarze 290: /*
1.1 schwarze 291: * XXX - this just does UTF-8, but we need to know
292: * beforehand whether we should do text substitution.
293: */
294:
295: switch (esc) {
296: case (ESCAPE_SPECIAL):
297: if (0 != (u = mchars_spec2cp(mc, seq, len)))
298: break;
299: /* FALLTHROUGH */
300: default:
301: continue;
302: }
303:
304: /*
305: * If we have a Unicode codepoint, try to convert that
306: * to a UTF-8 byte string.
307: */
308:
309: cpp = utfbuf;
310: if (0 == (sz = norm_utf8(u, utfbuf)))
311: continue;
312:
313: /* Copy the rendered glyph into the stream. */
314:
315: sz = strlen(cpp);
316: bsz += sz;
317:
318: *buf = mandoc_realloc(*buf, bsz);
319:
320: memcpy(&(*buf)[pos], cpp, sz);
321: pos += (int)sz;
322: }
323:
324: (*buf)[pos] = '\0';
325: }
326:
327: /*
328: * Open the filename-index mandoc-db database.
329: * Returns NULL if opening failed.
330: */
331: static DB *
332: index_open(void)
333: {
334: DB *db;
335:
1.2 schwarze 336: db = dbopen(MANDOC_IDX, O_RDONLY, 0, DB_RECNO, NULL);
1.1 schwarze 337: if (NULL != db)
338: return(db);
339:
340: return(NULL);
341: }
342:
343: /*
344: * Safely unpack from an index file record into the structure.
345: * Returns 1 if an entry was unpacked, 0 if the database is insane.
346: */
347: static int
1.11 kristaps 348: index_read(const DBT *key, const DBT *val, int index,
1.1 schwarze 349: const struct mchars *mc, struct rec *rec)
350: {
351: size_t left;
352: char *np, *cp;
353:
354: #define INDEX_BREAD(_dst) \
355: do { \
356: if (NULL == (np = memchr(cp, '\0', left))) \
357: return(0); \
358: norm_string(cp, mc, &(_dst)); \
359: left -= (np - cp) + 1; \
360: cp = np + 1; \
361: } while (/* CONSTCOND */ 0)
362:
363: left = val->size;
364: cp = (char *)val->data;
365:
1.5 kristaps 366: rec->res.rec = *(recno_t *)key->data;
1.11 kristaps 367: rec->res.volume = index;
1.1 schwarze 368:
1.14 schwarze 369: INDEX_BREAD(rec->res.type);
1.5 kristaps 370: INDEX_BREAD(rec->res.file);
371: INDEX_BREAD(rec->res.cat);
372: INDEX_BREAD(rec->res.title);
373: INDEX_BREAD(rec->res.arch);
374: INDEX_BREAD(rec->res.desc);
1.1 schwarze 375: return(1);
376: }
377:
378: /*
1.10 kristaps 379: * Search mandocdb databases in paths for expression "expr".
1.1 schwarze 380: * Filter out by "opts".
381: * Call "res" with the results, which may be zero.
1.5 kristaps 382: * Return 0 if there was a database error, else return 1.
1.1 schwarze 383: */
1.5 kristaps 384: int
1.10 kristaps 385: apropos_search(int pathsz, char **paths, const struct opts *opts,
1.12 schwarze 386: const struct expr *expr, size_t terms, void *arg,
1.5 kristaps 387: void (*res)(struct res *, size_t, void *))
1.1 schwarze 388: {
1.8 kristaps 389: struct rectree tree;
390: struct mchars *mc;
391: struct res *ress;
392: int i, mlen, rc;
393:
394: memset(&tree, 0, sizeof(struct rectree));
395:
1.10 kristaps 396: rc = 0;
1.8 kristaps 397: mc = mchars_alloc();
398:
1.10 kristaps 399: /*
400: * Main loop. Change into the directory containing manpage
401: * databases. Run our expession over each database in the set.
402: */
403:
404: for (i = 0; i < pathsz; i++) {
405: if (chdir(paths[i]))
1.8 kristaps 406: continue;
1.11 kristaps 407: if ( ! single_search(&tree, opts, expr, terms, mc, i))
1.10 kristaps 408: goto out;
1.8 kristaps 409: }
410:
411: /*
1.10 kristaps 412: * Count matching files, transfer to a "clean" array, then feed
413: * them to the output handler.
1.8 kristaps 414: */
415:
416: for (mlen = i = 0; i < tree.len; i++)
417: if (tree.node[i].matched)
418: mlen++;
419:
420: ress = mandoc_malloc(mlen * sizeof(struct res));
421:
422: for (mlen = i = 0; i < tree.len; i++)
423: if (tree.node[i].matched)
1.12 schwarze 424: memcpy(&ress[mlen++], &tree.node[i].res,
1.8 kristaps 425: sizeof(struct res));
426:
427: (*res)(ress, mlen, arg);
428: free(ress);
429:
1.10 kristaps 430: rc = 1;
431: out:
1.8 kristaps 432: for (i = 0; i < tree.len; i++)
433: recfree(&tree.node[i]);
434:
435: free(tree.node);
436: mchars_free(mc);
437: return(rc);
438: }
439:
440: static int
441: single_search(struct rectree *tree, const struct opts *opts,
442: const struct expr *expr, size_t terms,
1.11 kristaps 443: struct mchars *mc, int vol)
1.8 kristaps 444: {
445: int root, leaf, ch;
1.6 kristaps 446: uint64_t mask;
1.1 schwarze 447: DBT key, val;
448: DB *btree, *idx;
449: char *buf;
450: recno_t rec;
1.5 kristaps 451: struct rec *rs;
452: struct rec r;
1.6 kristaps 453: struct db_val *vbuf;
1.1 schwarze 454:
455: root = -1;
456: leaf = -1;
457: btree = NULL;
458: idx = NULL;
459: buf = NULL;
1.8 kristaps 460: rs = tree->node;
1.1 schwarze 461:
1.5 kristaps 462: memset(&r, 0, sizeof(struct rec));
1.1 schwarze 463:
1.12 schwarze 464: if (NULL == (btree = btree_open()))
1.10 kristaps 465: return(1);
1.1 schwarze 466:
1.8 kristaps 467: if (NULL == (idx = index_open())) {
468: (*btree->close)(btree);
1.10 kristaps 469: return(1);
1.8 kristaps 470: }
1.1 schwarze 471:
472: while (0 == (ch = (*btree->seq)(btree, &key, &val, R_NEXT))) {
1.12 schwarze 473: if (key.size < 2 || sizeof(struct db_val) != val.size)
474: break;
1.1 schwarze 475: if ( ! btree_read(&key, mc, &buf))
476: break;
477:
1.6 kristaps 478: vbuf = val.data;
479: rec = vbuf->rec;
480: mask = vbuf->mask;
1.5 kristaps 481:
482: /*
483: * See if this keyword record matches any of the
484: * expressions we have stored.
485: */
486: if ( ! exprmark(expr, buf, mask, NULL))
1.1 schwarze 487: continue;
488:
489: /*
490: * O(log n) scan for prior records. Since a record
491: * number is unbounded, this has decent performance over
492: * a complex hash function.
493: */
494:
495: for (leaf = root; leaf >= 0; )
1.12 schwarze 496: if (rec > rs[leaf].res.rec &&
1.5 kristaps 497: rs[leaf].rhs >= 0)
498: leaf = rs[leaf].rhs;
1.12 schwarze 499: else if (rec < rs[leaf].res.rec &&
1.5 kristaps 500: rs[leaf].lhs >= 0)
501: leaf = rs[leaf].lhs;
1.12 schwarze 502: else
1.1 schwarze 503: break;
504:
1.5 kristaps 505: /*
506: * If we find a record, see if it has already evaluated
507: * to true. If it has, great, just keep going. If not,
508: * try to evaluate it now and continue anyway.
509: */
510:
511: if (leaf >= 0 && rs[leaf].res.rec == rec) {
512: if (0 == rs[leaf].matched)
513: exprexec(expr, buf, mask, &rs[leaf]);
1.1 schwarze 514: continue;
1.5 kristaps 515: }
1.1 schwarze 516:
517: /*
1.5 kristaps 518: * We have a new file to examine.
519: * Extract the manpage's metadata from the index
520: * database, then begin partial evaluation.
1.1 schwarze 521: */
522:
523: key.data = &rec;
524: key.size = sizeof(recno_t);
525:
526: if (0 != (*idx->get)(idx, &key, &val, 0))
527: break;
528:
1.5 kristaps 529: r.lhs = r.rhs = -1;
1.11 kristaps 530: if ( ! index_read(&key, &val, vol, mc, &r))
1.1 schwarze 531: break;
532:
1.5 kristaps 533: /* XXX: this should be elsewhere, I guess? */
534:
535: if (opts->cat && strcasecmp(opts->cat, r.res.cat))
1.1 schwarze 536: continue;
1.5 kristaps 537: if (opts->arch && strcasecmp(opts->arch, r.res.arch))
1.1 schwarze 538: continue;
539:
1.8 kristaps 540: tree->node = rs = mandoc_realloc
541: (rs, (tree->len + 1) * sizeof(struct rec));
1.1 schwarze 542:
1.8 kristaps 543: memcpy(&rs[tree->len], &r, sizeof(struct rec));
1.12 schwarze 544: rs[tree->len].matches =
1.8 kristaps 545: mandoc_calloc(terms, sizeof(int));
1.1 schwarze 546:
1.12 schwarze 547: exprexec(expr, buf, mask, &rs[tree->len]);
548:
1.1 schwarze 549: /* Append to our tree. */
550:
551: if (leaf >= 0) {
1.5 kristaps 552: if (rec > rs[leaf].res.rec)
1.8 kristaps 553: rs[leaf].rhs = tree->len;
1.1 schwarze 554: else
1.8 kristaps 555: rs[leaf].lhs = tree->len;
1.1 schwarze 556: } else
1.8 kristaps 557: root = tree->len;
1.12 schwarze 558:
1.5 kristaps 559: memset(&r, 0, sizeof(struct rec));
1.8 kristaps 560: tree->len++;
1.1 schwarze 561: }
1.12 schwarze 562:
1.8 kristaps 563: (*btree->close)(btree);
564: (*idx->close)(idx);
1.1 schwarze 565:
566: free(buf);
1.8 kristaps 567: return(1 == ch);
1.5 kristaps 568: }
569:
570: static void
571: recfree(struct rec *rec)
572: {
573:
574: free(rec->res.file);
575: free(rec->res.cat);
576: free(rec->res.title);
577: free(rec->res.arch);
578: free(rec->res.desc);
579:
580: free(rec->matches);
1.1 schwarze 581: }
582:
1.13 kristaps 583: /*
584: * Compile a list of straight-up terms.
585: * The arguments are re-written into ~[[:<:]]term[[:>:]], or "term"
586: * surrounded by word boundaries, then pumped through exprterm().
587: * Terms are case-insensitive.
588: * This emulates whatis(1) behaviour.
589: */
590: struct expr *
591: termcomp(int argc, char *argv[], size_t *tt)
592: {
593: char *buf;
594: int pos;
595: struct expr *e, *next;
596: size_t sz;
597:
598: buf = NULL;
599: e = NULL;
600: *tt = 0;
601:
1.15 ! schwarze 602: for (pos = argc - 1; pos >= 0; pos--) {
! 603: sz = strlen(argv[pos]) + 18;
1.13 kristaps 604: buf = mandoc_realloc(buf, sz);
1.15 ! schwarze 605: strlcpy(buf, "Nm~[[:<:]]", sz);
1.13 kristaps 606: strlcat(buf, argv[pos], sz);
607: strlcat(buf, "[[:>:]]", sz);
608: if (NULL == (next = exprterm(buf, 0))) {
609: free(buf);
610: exprfree(e);
611: return(NULL);
612: }
1.15 ! schwarze 613: next->next = e;
1.13 kristaps 614: e = next;
615: (*tt)++;
616: }
617:
618: free(buf);
619: return(e);
620: }
621:
622: /*
623: * Compile a sequence of logical expressions.
624: * See apropos.1 for a grammar of this sequence.
625: */
1.1 schwarze 626: struct expr *
1.5 kristaps 627: exprcomp(int argc, char *argv[], size_t *tt)
1.1 schwarze 628: {
1.5 kristaps 629: int pos, lvl;
630: struct expr *e;
631:
632: pos = lvl = 0;
633: *tt = 0;
634:
635: e = exprexpr(argc, argv, &pos, &lvl, tt);
636:
637: if (0 == lvl && pos >= argc)
638: return(e);
639:
640: exprfree(e);
641: return(NULL);
642: }
643:
644: /*
645: * Compile an array of tokens into an expression.
646: * An informal expression grammar is defined in apropos(1).
647: * Return NULL if we fail doing so. All memory will be cleaned up.
648: * Return the root of the expression sequence if alright.
649: */
650: static struct expr *
1.9 kristaps 651: exprexpr(int argc, char *argv[], int *pos, int *lvl, size_t *tt)
1.5 kristaps 652: {
653: struct expr *e, *first, *next;
654: int log;
655:
656: first = next = NULL;
657:
658: for ( ; *pos < argc; (*pos)++) {
659: e = next;
660:
661: /*
662: * Close out a subexpression.
663: */
664:
665: if (NULL != e && 0 == strcmp(")", argv[*pos])) {
666: if (--(*lvl) < 0)
667: goto err;
668: break;
669: }
670:
671: /*
672: * Small note: if we're just starting, don't let "-a"
673: * and "-o" be considered logical operators: they're
674: * just tokens unless pairwise joining, in which case we
675: * record their existence (or assume "OR").
676: */
677: log = 0;
678:
679: if (NULL != e && 0 == strcmp("-a", argv[*pos]))
1.12 schwarze 680: log = 1;
1.5 kristaps 681: else if (NULL != e && 0 == strcmp("-o", argv[*pos]))
682: log = 2;
683:
684: if (log > 0 && ++(*pos) >= argc)
685: goto err;
686:
687: /*
688: * Now we parse the term part. This can begin with
689: * "-i", in which case the expression is case
690: * insensitive.
691: */
692:
693: if (0 == strcmp("(", argv[*pos])) {
694: ++(*pos);
695: ++(*lvl);
696: next = mandoc_calloc(1, sizeof(struct expr));
697: next->cs = 1;
698: next->subexpr = exprexpr(argc, argv, pos, lvl, tt);
699: if (NULL == next->subexpr) {
700: free(next);
701: next = NULL;
702: }
703: } else if (0 == strcmp("-i", argv[*pos])) {
704: if (++(*pos) >= argc)
705: goto err;
706: next = exprterm(argv[*pos], 0);
707: } else
708: next = exprterm(argv[*pos], 1);
709:
710: if (NULL == next)
711: goto err;
712:
713: next->and = log == 1;
714: next->index = (int)(*tt)++;
715:
716: /* Append to our chain of expressions. */
717:
718: if (NULL == first) {
719: assert(NULL == e);
720: first = next;
721: } else {
722: assert(NULL != e);
723: e->next = next;
724: }
725: }
726:
727: return(first);
728: err:
729: exprfree(first);
730: return(NULL);
731: }
732:
733: /*
734: * Parse a terminal expression with the grammar as defined in
735: * apropos(1).
736: * Return NULL if we fail the parse.
737: */
738: static struct expr *
739: exprterm(char *buf, int cs)
740: {
741: struct expr e;
1.1 schwarze 742: struct expr *p;
1.3 schwarze 743: char *key;
1.5 kristaps 744: int i;
745:
746: memset(&e, 0, sizeof(struct expr));
1.1 schwarze 747:
1.5 kristaps 748: e.cs = cs;
1.1 schwarze 749:
1.5 kristaps 750: /* Choose regex or substring match. */
1.3 schwarze 751:
1.4 kristaps 752: if (NULL == (e.v = strpbrk(buf, "=~"))) {
1.3 schwarze 753: e.regex = 0;
1.4 kristaps 754: e.v = buf;
1.3 schwarze 755: } else {
756: e.regex = '~' == *e.v;
757: *e.v++ = '\0';
758: }
1.1 schwarze 759:
1.5 kristaps 760: /* Determine the record types to search for. */
1.3 schwarze 761:
762: e.mask = 0;
1.4 kristaps 763: if (buf < e.v) {
764: while (NULL != (key = strsep(&buf, ","))) {
1.3 schwarze 765: i = 0;
766: while (types[i].mask &&
1.4 kristaps 767: strcmp(types[i].name, key))
1.3 schwarze 768: i++;
769: e.mask |= types[i].mask;
770: }
771: }
772: if (0 == e.mask)
773: e.mask = TYPE_Nm | TYPE_Nd;
1.1 schwarze 774:
1.5 kristaps 775: if (e.regex) {
1.13 kristaps 776: i = REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE);
1.5 kristaps 777: if (regcomp(&e.re, e.v, i))
778: return(NULL);
779: }
1.1 schwarze 780:
1.3 schwarze 781: e.v = mandoc_strdup(e.v);
1.1 schwarze 782:
783: p = mandoc_calloc(1, sizeof(struct expr));
784: memcpy(p, &e, sizeof(struct expr));
785: return(p);
786: }
787:
788: void
789: exprfree(struct expr *p)
790: {
1.5 kristaps 791: struct expr *pp;
1.12 schwarze 792:
1.5 kristaps 793: while (NULL != p) {
794: if (p->subexpr)
795: exprfree(p->subexpr);
796: if (p->regex)
797: regfree(&p->re);
798: free(p->v);
799: pp = p->next;
800: free(p);
801: p = pp;
802: }
803: }
1.1 schwarze 804:
1.5 kristaps 805: static int
1.12 schwarze 806: exprmark(const struct expr *p, const char *cp,
1.6 kristaps 807: uint64_t mask, int *ms)
1.5 kristaps 808: {
809:
810: for ( ; p; p = p->next) {
811: if (p->subexpr) {
812: if (exprmark(p->subexpr, cp, mask, ms))
813: return(1);
814: continue;
815: } else if ( ! (mask & p->mask))
816: continue;
1.1 schwarze 817:
1.5 kristaps 818: if (p->regex) {
819: if (regexec(&p->re, cp, 0, NULL, 0))
820: continue;
821: } else if (p->cs) {
822: if (NULL == strstr(cp, p->v))
823: continue;
824: } else {
825: if (NULL == strcasestr(cp, p->v))
826: continue;
827: }
828:
829: if (NULL == ms)
830: return(1);
831: else
832: ms[p->index] = 1;
833: }
1.1 schwarze 834:
1.5 kristaps 835: return(0);
1.1 schwarze 836: }
837:
838: static int
1.5 kristaps 839: expreval(const struct expr *p, int *ms)
1.1 schwarze 840: {
1.5 kristaps 841: int match;
1.1 schwarze 842:
1.5 kristaps 843: /*
844: * AND has precedence over OR. Analysis is left-right, though
845: * it doesn't matter because there are no side-effects.
846: * Thus, step through pairwise ANDs and accumulate their Boolean
847: * evaluation. If we encounter a single true AND collection or
848: * standalone term, the whole expression is true (by definition
849: * of OR).
850: */
851:
852: for (match = 0; p && ! match; p = p->next) {
853: /* Evaluate a subexpression, if applicable. */
854: if (p->subexpr && ! ms[p->index])
855: ms[p->index] = expreval(p->subexpr, ms);
856:
857: match = ms[p->index];
858: for ( ; p->next && p->next->and; p = p->next) {
859: /* Evaluate a subexpression, if applicable. */
860: if (p->next->subexpr && ! ms[p->next->index])
1.12 schwarze 861: ms[p->next->index] =
1.5 kristaps 862: expreval(p->next->subexpr, ms);
863: match = match && ms[p->next->index];
864: }
865: }
866:
867: return(match);
868: }
869:
870: /*
871: * First, update the array of terms for which this expression evaluates
872: * to true.
873: * Second, logically evaluate all terms over the updated array of truth
874: * values.
875: * If this evaluates to true, mark the expression as satisfied.
876: */
877: static void
1.12 schwarze 878: exprexec(const struct expr *e, const char *cp,
1.6 kristaps 879: uint64_t mask, struct rec *r)
1.5 kristaps 880: {
1.1 schwarze 881:
1.5 kristaps 882: assert(0 == r->matched);
1.12 schwarze 883: exprmark(e, cp, mask, r->matches);
884: r->matched = expreval(e, r->matches);
1.1 schwarze 885: }
CVSweb