]> git.cameronkatri.com Git - mandoc.git/blob - apropos_db.c
Fix parsing of file names given on the command line; i broke it
[mandoc.git] / apropos_db.c
1 /* $Id: apropos_db.c,v 1.21 2011/12/03 18:47:09 kristaps Exp $ */
2 /*
3 * Copyright (c) 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2011 Ingo Schwarze <schwarze@openbsd.org>
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
22 #include <assert.h>
23 #include <fcntl.h>
24 #include <regex.h>
25 #include <stdarg.h>
26 #include <stdint.h>
27 #include <stdlib.h>
28 #include <string.h>
29 #include <unistd.h>
30
31 #if defined(__linux__)
32 # include <endian.h>
33 # include <db_185.h>
34 #elif defined(__APPLE__)
35 # include <libkern/OSByteOrder.h>
36 # include <db.h>
37 #else
38 # include <db.h>
39 #endif
40
41 #include "mandocdb.h"
42 #include "apropos_db.h"
43 #include "mandoc.h"
44
45 struct rec {
46 struct res res; /* resulting record info */
47 /*
48 * Maintain a binary tree for checking the uniqueness of `rec'
49 * when adding elements to the results array.
50 * Since the results array is dynamic, use offset in the array
51 * instead of a pointer to the structure.
52 */
53 int lhs;
54 int rhs;
55 int matched; /* expression is true */
56 int *matches; /* partial truth evaluations */
57 };
58
59 struct expr {
60 int regex; /* is regex? */
61 int index; /* index in match array */
62 uint64_t mask; /* type-mask */
63 int and; /* is rhs of logical AND? */
64 char *v; /* search value */
65 regex_t re; /* compiled re, if regex */
66 struct expr *next; /* next in sequence */
67 struct expr *subexpr;
68 };
69
70 struct type {
71 uint64_t mask;
72 const char *name;
73 };
74
75 struct rectree {
76 struct rec *node; /* record array for dir tree */
77 int len; /* length of record array */
78 };
79
80 static const struct type types[] = {
81 { TYPE_An, "An" },
82 { TYPE_Ar, "Ar" },
83 { TYPE_At, "At" },
84 { TYPE_Bsx, "Bsx" },
85 { TYPE_Bx, "Bx" },
86 { TYPE_Cd, "Cd" },
87 { TYPE_Cm, "Cm" },
88 { TYPE_Dv, "Dv" },
89 { TYPE_Dx, "Dx" },
90 { TYPE_Em, "Em" },
91 { TYPE_Er, "Er" },
92 { TYPE_Ev, "Ev" },
93 { TYPE_Fa, "Fa" },
94 { TYPE_Fl, "Fl" },
95 { TYPE_Fn, "Fn" },
96 { TYPE_Fn, "Fo" },
97 { TYPE_Ft, "Ft" },
98 { TYPE_Fx, "Fx" },
99 { TYPE_Ic, "Ic" },
100 { TYPE_In, "In" },
101 { TYPE_Lb, "Lb" },
102 { TYPE_Li, "Li" },
103 { TYPE_Lk, "Lk" },
104 { TYPE_Ms, "Ms" },
105 { TYPE_Mt, "Mt" },
106 { TYPE_Nd, "Nd" },
107 { TYPE_Nm, "Nm" },
108 { TYPE_Nx, "Nx" },
109 { TYPE_Ox, "Ox" },
110 { TYPE_Pa, "Pa" },
111 { TYPE_Rs, "Rs" },
112 { TYPE_Sh, "Sh" },
113 { TYPE_Ss, "Ss" },
114 { TYPE_St, "St" },
115 { TYPE_Sy, "Sy" },
116 { TYPE_Tn, "Tn" },
117 { TYPE_Va, "Va" },
118 { TYPE_Va, "Vt" },
119 { TYPE_Xr, "Xr" },
120 { INT_MAX, "any" },
121 { 0, NULL }
122 };
123
124 static DB *btree_open(void);
125 static int btree_read(const DBT *, const DBT *,
126 const struct mchars *,
127 struct db_val *, char **);
128 static int expreval(const struct expr *, int *);
129 static void exprexec(const struct expr *,
130 const char *, uint64_t, struct rec *);
131 static int exprmark(const struct expr *,
132 const char *, uint64_t, int *);
133 static struct expr *exprexpr(int, char *[], int *, int *, size_t *);
134 static struct expr *exprterm(char *, int);
135 static DB *index_open(void);
136 static int index_read(const DBT *, const DBT *, int,
137 const struct mchars *, struct rec *);
138 static void norm_string(const char *,
139 const struct mchars *, char **);
140 static size_t norm_utf8(unsigned int, char[7]);
141 static void recfree(struct rec *);
142 static int single_search(struct rectree *, const struct opts *,
143 const struct expr *, size_t terms,
144 struct mchars *, int);
145
146 /*
147 * Open the keyword mandoc-db database.
148 */
149 static DB *
150 btree_open(void)
151 {
152 BTREEINFO info;
153 DB *db;
154
155 memset(&info, 0, sizeof(BTREEINFO));
156 info.flags = R_DUP;
157
158 db = dbopen(MANDOC_DB, O_RDONLY, 0, DB_BTREE, &info);
159 if (NULL != db)
160 return(db);
161
162 return(NULL);
163 }
164
165 /*
166 * Read a keyword from the database and normalise it.
167 * Return 0 if the database is insane, else 1.
168 */
169 static int
170 btree_read(const DBT *k, const DBT *v,
171 const struct mchars *mc,
172 struct db_val *dbv, char **buf)
173 {
174 const struct db_val *vp;
175
176 /* Are our sizes sane? */
177 if (k->size < 2 || sizeof(struct db_val) != v->size)
178 return(0);
179
180 /* Is our string nil-terminated? */
181 if ('\0' != ((const char *)k->data)[(int)k->size - 1])
182 return(0);
183
184 vp = v->data;
185 norm_string((const char *)k->data, mc, buf);
186 dbv->rec = betoh32(vp->rec);
187 dbv->mask = betoh64(vp->mask);
188 return(1);
189 }
190
191 /*
192 * Take a Unicode codepoint and produce its UTF-8 encoding.
193 * This isn't the best way to do this, but it works.
194 * The magic numbers are from the UTF-8 packaging.
195 * They're not as scary as they seem: read the UTF-8 spec for details.
196 */
197 static size_t
198 norm_utf8(unsigned int cp, char out[7])
199 {
200 size_t rc;
201
202 rc = 0;
203
204 if (cp <= 0x0000007F) {
205 rc = 1;
206 out[0] = (char)cp;
207 } else if (cp <= 0x000007FF) {
208 rc = 2;
209 out[0] = (cp >> 6 & 31) | 192;
210 out[1] = (cp & 63) | 128;
211 } else if (cp <= 0x0000FFFF) {
212 rc = 3;
213 out[0] = (cp >> 12 & 15) | 224;
214 out[1] = (cp >> 6 & 63) | 128;
215 out[2] = (cp & 63) | 128;
216 } else if (cp <= 0x001FFFFF) {
217 rc = 4;
218 out[0] = (cp >> 18 & 7) | 240;
219 out[1] = (cp >> 12 & 63) | 128;
220 out[2] = (cp >> 6 & 63) | 128;
221 out[3] = (cp & 63) | 128;
222 } else if (cp <= 0x03FFFFFF) {
223 rc = 5;
224 out[0] = (cp >> 24 & 3) | 248;
225 out[1] = (cp >> 18 & 63) | 128;
226 out[2] = (cp >> 12 & 63) | 128;
227 out[3] = (cp >> 6 & 63) | 128;
228 out[4] = (cp & 63) | 128;
229 } else if (cp <= 0x7FFFFFFF) {
230 rc = 6;
231 out[0] = (cp >> 30 & 1) | 252;
232 out[1] = (cp >> 24 & 63) | 128;
233 out[2] = (cp >> 18 & 63) | 128;
234 out[3] = (cp >> 12 & 63) | 128;
235 out[4] = (cp >> 6 & 63) | 128;
236 out[5] = (cp & 63) | 128;
237 } else
238 return(0);
239
240 out[rc] = '\0';
241 return(rc);
242 }
243
244 /*
245 * Normalise strings from the index and database.
246 * These strings are escaped as defined by mandoc_char(7) along with
247 * other goop in mandoc.h (e.g., soft hyphens).
248 * This function normalises these into a nice UTF-8 string.
249 * Returns 0 if the database is fucked.
250 */
251 static void
252 norm_string(const char *val, const struct mchars *mc, char **buf)
253 {
254 size_t sz, bsz;
255 char utfbuf[7];
256 const char *seq, *cpp;
257 int len, u, pos;
258 enum mandoc_esc esc;
259 static const char res[] = { '\\', '\t',
260 ASCII_NBRSP, ASCII_HYPH, '\0' };
261
262 /* Pre-allocate by the length of the input */
263
264 bsz = strlen(val) + 1;
265 *buf = mandoc_realloc(*buf, bsz);
266 pos = 0;
267
268 while ('\0' != *val) {
269 /*
270 * Halt on the first escape sequence.
271 * This also halts on the end of string, in which case
272 * we just copy, fallthrough, and exit the loop.
273 */
274 if ((sz = strcspn(val, res)) > 0) {
275 memcpy(&(*buf)[pos], val, sz);
276 pos += (int)sz;
277 val += (int)sz;
278 }
279
280 if (ASCII_HYPH == *val) {
281 (*buf)[pos++] = '-';
282 val++;
283 continue;
284 } else if ('\t' == *val || ASCII_NBRSP == *val) {
285 (*buf)[pos++] = ' ';
286 val++;
287 continue;
288 } else if ('\\' != *val)
289 break;
290
291 /* Read past the slash. */
292
293 val++;
294 u = 0;
295
296 /*
297 * Parse the escape sequence and see if it's a
298 * predefined character or special character.
299 */
300
301 esc = mandoc_escape(&val, &seq, &len);
302 if (ESCAPE_ERROR == esc)
303 break;
304
305 /*
306 * XXX - this just does UTF-8, but we need to know
307 * beforehand whether we should do text substitution.
308 */
309
310 switch (esc) {
311 case (ESCAPE_SPECIAL):
312 if (0 != (u = mchars_spec2cp(mc, seq, len)))
313 break;
314 /* FALLTHROUGH */
315 default:
316 continue;
317 }
318
319 /*
320 * If we have a Unicode codepoint, try to convert that
321 * to a UTF-8 byte string.
322 */
323
324 cpp = utfbuf;
325 if (0 == (sz = norm_utf8(u, utfbuf)))
326 continue;
327
328 /* Copy the rendered glyph into the stream. */
329
330 sz = strlen(cpp);
331 bsz += sz;
332
333 *buf = mandoc_realloc(*buf, bsz);
334
335 memcpy(&(*buf)[pos], cpp, sz);
336 pos += (int)sz;
337 }
338
339 (*buf)[pos] = '\0';
340 }
341
342 /*
343 * Open the filename-index mandoc-db database.
344 * Returns NULL if opening failed.
345 */
346 static DB *
347 index_open(void)
348 {
349 DB *db;
350
351 db = dbopen(MANDOC_IDX, O_RDONLY, 0, DB_RECNO, NULL);
352 if (NULL != db)
353 return(db);
354
355 return(NULL);
356 }
357
358 /*
359 * Safely unpack from an index file record into the structure.
360 * Returns 1 if an entry was unpacked, 0 if the database is insane.
361 */
362 static int
363 index_read(const DBT *key, const DBT *val, int index,
364 const struct mchars *mc, struct rec *rec)
365 {
366 size_t left;
367 char *np, *cp;
368
369 #define INDEX_BREAD(_dst) \
370 do { \
371 if (NULL == (np = memchr(cp, '\0', left))) \
372 return(0); \
373 norm_string(cp, mc, &(_dst)); \
374 left -= (np - cp) + 1; \
375 cp = np + 1; \
376 } while (/* CONSTCOND */ 0)
377
378 left = val->size;
379 cp = (char *)val->data;
380
381 rec->res.rec = *(recno_t *)key->data;
382 rec->res.volume = index;
383
384 INDEX_BREAD(rec->res.type);
385 INDEX_BREAD(rec->res.file);
386 INDEX_BREAD(rec->res.cat);
387 INDEX_BREAD(rec->res.title);
388 INDEX_BREAD(rec->res.arch);
389 INDEX_BREAD(rec->res.desc);
390 return(1);
391 }
392
393 /*
394 * Search mandocdb databases in paths for expression "expr".
395 * Filter out by "opts".
396 * Call "res" with the results, which may be zero.
397 * Return 0 if there was a database error, else return 1.
398 */
399 int
400 apropos_search(int pathsz, char **paths, const struct opts *opts,
401 const struct expr *expr, size_t terms, void *arg,
402 void (*res)(struct res *, size_t, void *))
403 {
404 struct rectree tree;
405 struct mchars *mc;
406 struct res *ress;
407 int i, mlen, rc;
408
409 memset(&tree, 0, sizeof(struct rectree));
410
411 rc = 0;
412 mc = mchars_alloc();
413
414 /*
415 * Main loop. Change into the directory containing manpage
416 * databases. Run our expession over each database in the set.
417 */
418
419 for (i = 0; i < pathsz; i++) {
420 if (chdir(paths[i]))
421 continue;
422 if ( ! single_search(&tree, opts, expr, terms, mc, i))
423 goto out;
424 }
425
426 /*
427 * Count matching files, transfer to a "clean" array, then feed
428 * them to the output handler.
429 */
430
431 for (mlen = i = 0; i < tree.len; i++)
432 if (tree.node[i].matched)
433 mlen++;
434
435 ress = mandoc_malloc(mlen * sizeof(struct res));
436
437 for (mlen = i = 0; i < tree.len; i++)
438 if (tree.node[i].matched)
439 memcpy(&ress[mlen++], &tree.node[i].res,
440 sizeof(struct res));
441
442 (*res)(ress, mlen, arg);
443 free(ress);
444
445 rc = 1;
446 out:
447 for (i = 0; i < tree.len; i++)
448 recfree(&tree.node[i]);
449
450 free(tree.node);
451 mchars_free(mc);
452 return(rc);
453 }
454
455 static int
456 single_search(struct rectree *tree, const struct opts *opts,
457 const struct expr *expr, size_t terms,
458 struct mchars *mc, int vol)
459 {
460 int root, leaf, ch;
461 DBT key, val;
462 DB *btree, *idx;
463 char *buf;
464 struct rec *rs;
465 struct rec r;
466 struct db_val vb;
467
468 root = -1;
469 leaf = -1;
470 btree = NULL;
471 idx = NULL;
472 buf = NULL;
473 rs = tree->node;
474
475 memset(&r, 0, sizeof(struct rec));
476
477 if (NULL == (btree = btree_open()))
478 return(1);
479
480 if (NULL == (idx = index_open())) {
481 (*btree->close)(btree);
482 return(1);
483 }
484
485 while (0 == (ch = (*btree->seq)(btree, &key, &val, R_NEXT))) {
486 if ( ! btree_read(&key, &val, mc, &vb, &buf))
487 break;
488
489 /*
490 * See if this keyword record matches any of the
491 * expressions we have stored.
492 */
493 if ( ! exprmark(expr, buf, vb.mask, NULL))
494 continue;
495
496 /*
497 * O(log n) scan for prior records. Since a record
498 * number is unbounded, this has decent performance over
499 * a complex hash function.
500 */
501
502 for (leaf = root; leaf >= 0; )
503 if (vb.rec > rs[leaf].res.rec &&
504 rs[leaf].rhs >= 0)
505 leaf = rs[leaf].rhs;
506 else if (vb.rec < rs[leaf].res.rec &&
507 rs[leaf].lhs >= 0)
508 leaf = rs[leaf].lhs;
509 else
510 break;
511
512 /*
513 * If we find a record, see if it has already evaluated
514 * to true. If it has, great, just keep going. If not,
515 * try to evaluate it now and continue anyway.
516 */
517
518 if (leaf >= 0 && rs[leaf].res.rec == vb.rec) {
519 if (0 == rs[leaf].matched)
520 exprexec(expr, buf, vb.mask, &rs[leaf]);
521 continue;
522 }
523
524 /*
525 * We have a new file to examine.
526 * Extract the manpage's metadata from the index
527 * database, then begin partial evaluation.
528 */
529
530 key.data = &vb.rec;
531 key.size = sizeof(recno_t);
532
533 if (0 != (*idx->get)(idx, &key, &val, 0))
534 break;
535
536 r.lhs = r.rhs = -1;
537 if ( ! index_read(&key, &val, vol, mc, &r))
538 break;
539
540 /* XXX: this should be elsewhere, I guess? */
541
542 if (opts->cat && strcasecmp(opts->cat, r.res.cat))
543 continue;
544 if (opts->arch && strcasecmp(opts->arch, r.res.arch))
545 continue;
546
547 tree->node = rs = mandoc_realloc
548 (rs, (tree->len + 1) * sizeof(struct rec));
549
550 memcpy(&rs[tree->len], &r, sizeof(struct rec));
551 rs[tree->len].matches =
552 mandoc_calloc(terms, sizeof(int));
553
554 exprexec(expr, buf, vb.mask, &rs[tree->len]);
555
556 /* Append to our tree. */
557
558 if (leaf >= 0) {
559 if (vb.rec > rs[leaf].res.rec)
560 rs[leaf].rhs = tree->len;
561 else
562 rs[leaf].lhs = tree->len;
563 } else
564 root = tree->len;
565
566 memset(&r, 0, sizeof(struct rec));
567 tree->len++;
568 }
569
570 (*btree->close)(btree);
571 (*idx->close)(idx);
572
573 free(buf);
574 return(1 == ch);
575 }
576
577 static void
578 recfree(struct rec *rec)
579 {
580
581 free(rec->res.type);
582 free(rec->res.file);
583 free(rec->res.cat);
584 free(rec->res.title);
585 free(rec->res.arch);
586 free(rec->res.desc);
587
588 free(rec->matches);
589 }
590
591 /*
592 * Compile a list of straight-up terms.
593 * The arguments are re-written into ~[[:<:]]term[[:>:]], or "term"
594 * surrounded by word boundaries, then pumped through exprterm().
595 * Terms are case-insensitive.
596 * This emulates whatis(1) behaviour.
597 */
598 struct expr *
599 termcomp(int argc, char *argv[], size_t *tt)
600 {
601 char *buf;
602 int pos;
603 struct expr *e, *next;
604 size_t sz;
605
606 buf = NULL;
607 e = NULL;
608 *tt = 0;
609
610 for (pos = argc - 1; pos >= 0; pos--) {
611 sz = strlen(argv[pos]) + 18;
612 buf = mandoc_realloc(buf, sz);
613 strlcpy(buf, "Nm~[[:<:]]", sz);
614 strlcat(buf, argv[pos], sz);
615 strlcat(buf, "[[:>:]]", sz);
616 if (NULL == (next = exprterm(buf, 0))) {
617 free(buf);
618 exprfree(e);
619 return(NULL);
620 }
621 next->next = e;
622 e = next;
623 (*tt)++;
624 }
625
626 free(buf);
627 return(e);
628 }
629
630 /*
631 * Compile a sequence of logical expressions.
632 * See apropos.1 for a grammar of this sequence.
633 */
634 struct expr *
635 exprcomp(int argc, char *argv[], size_t *tt)
636 {
637 int pos, lvl;
638 struct expr *e;
639
640 pos = lvl = 0;
641 *tt = 0;
642
643 e = exprexpr(argc, argv, &pos, &lvl, tt);
644
645 if (0 == lvl && pos >= argc)
646 return(e);
647
648 exprfree(e);
649 return(NULL);
650 }
651
652 /*
653 * Compile an array of tokens into an expression.
654 * An informal expression grammar is defined in apropos(1).
655 * Return NULL if we fail doing so. All memory will be cleaned up.
656 * Return the root of the expression sequence if alright.
657 */
658 static struct expr *
659 exprexpr(int argc, char *argv[], int *pos, int *lvl, size_t *tt)
660 {
661 struct expr *e, *first, *next;
662 int log;
663
664 first = next = NULL;
665
666 for ( ; *pos < argc; (*pos)++) {
667 e = next;
668
669 /*
670 * Close out a subexpression.
671 */
672
673 if (NULL != e && 0 == strcmp(")", argv[*pos])) {
674 if (--(*lvl) < 0)
675 goto err;
676 break;
677 }
678
679 /*
680 * Small note: if we're just starting, don't let "-a"
681 * and "-o" be considered logical operators: they're
682 * just tokens unless pairwise joining, in which case we
683 * record their existence (or assume "OR").
684 */
685 log = 0;
686
687 if (NULL != e && 0 == strcmp("-a", argv[*pos]))
688 log = 1;
689 else if (NULL != e && 0 == strcmp("-o", argv[*pos]))
690 log = 2;
691
692 if (log > 0 && ++(*pos) >= argc)
693 goto err;
694
695 /*
696 * Now we parse the term part. This can begin with
697 * "-i", in which case the expression is case
698 * insensitive.
699 */
700
701 if (0 == strcmp("(", argv[*pos])) {
702 ++(*pos);
703 ++(*lvl);
704 next = mandoc_calloc(1, sizeof(struct expr));
705 next->subexpr = exprexpr(argc, argv, pos, lvl, tt);
706 if (NULL == next->subexpr) {
707 free(next);
708 next = NULL;
709 }
710 } else if (0 == strcmp("-i", argv[*pos])) {
711 if (++(*pos) >= argc)
712 goto err;
713 next = exprterm(argv[*pos], 0);
714 } else
715 next = exprterm(argv[*pos], 1);
716
717 if (NULL == next)
718 goto err;
719
720 next->and = log == 1;
721 next->index = (int)(*tt)++;
722
723 /* Append to our chain of expressions. */
724
725 if (NULL == first) {
726 assert(NULL == e);
727 first = next;
728 } else {
729 assert(NULL != e);
730 e->next = next;
731 }
732 }
733
734 return(first);
735 err:
736 exprfree(first);
737 return(NULL);
738 }
739
740 /*
741 * Parse a terminal expression with the grammar as defined in
742 * apropos(1).
743 * Return NULL if we fail the parse.
744 */
745 static struct expr *
746 exprterm(char *buf, int cs)
747 {
748 struct expr e;
749 struct expr *p;
750 char *key;
751 int i;
752
753 memset(&e, 0, sizeof(struct expr));
754
755 /* Choose regex or substring match. */
756
757 if (NULL == (e.v = strpbrk(buf, "=~"))) {
758 e.regex = 0;
759 e.v = buf;
760 } else {
761 e.regex = '~' == *e.v;
762 *e.v++ = '\0';
763 }
764
765 /* Determine the record types to search for. */
766
767 e.mask = 0;
768 if (buf < e.v) {
769 while (NULL != (key = strsep(&buf, ","))) {
770 i = 0;
771 while (types[i].mask &&
772 strcmp(types[i].name, key))
773 i++;
774 e.mask |= types[i].mask;
775 }
776 }
777 if (0 == e.mask)
778 e.mask = TYPE_Nm | TYPE_Nd;
779
780 if (e.regex) {
781 i = REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE);
782 if (regcomp(&e.re, e.v, i))
783 return(NULL);
784 }
785
786 e.v = mandoc_strdup(e.v);
787
788 p = mandoc_calloc(1, sizeof(struct expr));
789 memcpy(p, &e, sizeof(struct expr));
790 return(p);
791 }
792
793 void
794 exprfree(struct expr *p)
795 {
796 struct expr *pp;
797
798 while (NULL != p) {
799 if (p->subexpr)
800 exprfree(p->subexpr);
801 if (p->regex)
802 regfree(&p->re);
803 free(p->v);
804 pp = p->next;
805 free(p);
806 p = pp;
807 }
808 }
809
810 static int
811 exprmark(const struct expr *p, const char *cp,
812 uint64_t mask, int *ms)
813 {
814
815 for ( ; p; p = p->next) {
816 if (p->subexpr) {
817 if (exprmark(p->subexpr, cp, mask, ms))
818 return(1);
819 continue;
820 } else if ( ! (mask & p->mask))
821 continue;
822
823 if (p->regex) {
824 if (regexec(&p->re, cp, 0, NULL, 0))
825 continue;
826 } else if (NULL == strcasestr(cp, p->v))
827 continue;
828
829 if (NULL == ms)
830 return(1);
831 else
832 ms[p->index] = 1;
833 }
834
835 return(0);
836 }
837
838 static int
839 expreval(const struct expr *p, int *ms)
840 {
841 int match;
842
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])
861 ms[p->next->index] =
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
878 exprexec(const struct expr *e, const char *cp,
879 uint64_t mask, struct rec *r)
880 {
881
882 assert(0 == r->matched);
883 exprmark(e, cp, mask, r->matches);
884 r->matched = expreval(e, r->matches);
885 }