Annotation of mandoc/apropos_db.c, Revision 1.8
1.8 ! kristaps 1: /* $Id: apropos_db.c,v 1.7 2011/11/20 12:46:53 kristaps 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);
1.5 kristaps 118: static int btree_read(const DBT *,
119: const struct mchars *, char **);
120: static int expreval(const struct expr *, int *);
121: static void exprexec(const struct expr *,
1.6 kristaps 122: const char *, uint64_t, struct rec *);
1.5 kristaps 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);
128: static int index_read(const DBT *, const DBT *,
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,
! 136: struct mchars *);
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);
1.1 schwarze 151: if (NULL != db)
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.
179: * The magic numbers are from the UTF-8 packaging.
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;
244: static const char res[] = { '\\', '\t',
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:
290: /*
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
348: index_read(const DBT *key, const DBT *val,
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;
1.1 schwarze 367:
1.5 kristaps 368: INDEX_BREAD(rec->res.file);
369: INDEX_BREAD(rec->res.cat);
370: INDEX_BREAD(rec->res.title);
371: INDEX_BREAD(rec->res.arch);
372: INDEX_BREAD(rec->res.desc);
1.1 schwarze 373: return(1);
374: }
375:
376: /*
1.8 ! kristaps 377: * Search mandocdb databases in argv (size argc) for the expression
! 378: * "expr".
1.1 schwarze 379: * Filter out by "opts".
380: * Call "res" with the results, which may be zero.
1.5 kristaps 381: * Return 0 if there was a database error, else return 1.
1.1 schwarze 382: */
1.5 kristaps 383: int
1.8 ! kristaps 384: apropos_search(int argc, char *argv[], const struct opts *opts,
! 385: const struct expr *expr, size_t terms, void *arg,
1.5 kristaps 386: void (*res)(struct res *, size_t, void *))
1.1 schwarze 387: {
1.8 ! kristaps 388: struct rectree tree;
! 389: struct mchars *mc;
! 390: struct res *ress;
! 391: int i, mlen, rc;
! 392:
! 393: memset(&tree, 0, sizeof(struct rectree));
! 394:
! 395: mc = mchars_alloc();
! 396:
! 397: for (rc = 1, i = 0; rc && i < argc; i++) {
! 398: /* FIXME: ugly warning: we shouldn't get here! */
! 399: if (chdir(argv[i]))
! 400: continue;
! 401: rc = single_search(&tree, opts, expr, terms, mc);
! 402: /* FIXME: warn and continue... ? */
! 403: }
! 404:
! 405: /*
! 406: * Count the matching files
! 407: * and feed them to the output handler.
! 408: */
! 409:
! 410: for (mlen = i = 0; i < tree.len; i++)
! 411: if (tree.node[i].matched)
! 412: mlen++;
! 413:
! 414: ress = mandoc_malloc(mlen * sizeof(struct res));
! 415:
! 416: for (mlen = i = 0; i < tree.len; i++)
! 417: if (tree.node[i].matched)
! 418: memcpy(&ress[mlen++], &tree.node[i].res,
! 419: sizeof(struct res));
! 420:
! 421: (*res)(ress, mlen, arg);
! 422: free(ress);
! 423:
! 424: for (i = 0; i < tree.len; i++)
! 425: recfree(&tree.node[i]);
! 426:
! 427: free(tree.node);
! 428: mchars_free(mc);
! 429: return(rc);
! 430: }
! 431:
! 432: static int
! 433: single_search(struct rectree *tree, const struct opts *opts,
! 434: const struct expr *expr, size_t terms,
! 435: struct mchars *mc)
! 436: {
! 437: int root, leaf, ch;
1.6 kristaps 438: uint64_t mask;
1.1 schwarze 439: DBT key, val;
440: DB *btree, *idx;
441: char *buf;
442: recno_t rec;
1.5 kristaps 443: struct rec *rs;
444: struct rec r;
1.6 kristaps 445: struct db_val *vbuf;
1.1 schwarze 446:
447: root = -1;
448: leaf = -1;
449: btree = NULL;
450: idx = NULL;
451: buf = NULL;
1.8 ! kristaps 452: rs = tree->node;
1.1 schwarze 453:
1.5 kristaps 454: memset(&r, 0, sizeof(struct rec));
1.1 schwarze 455:
1.8 ! kristaps 456: if (NULL == (btree = btree_open()))
! 457: return(0);
1.1 schwarze 458:
1.8 ! kristaps 459: if (NULL == (idx = index_open())) {
! 460: (*btree->close)(btree);
! 461: return(0);
! 462: }
1.1 schwarze 463:
464: while (0 == (ch = (*btree->seq)(btree, &key, &val, R_NEXT))) {
1.6 kristaps 465: if (key.size < 2 || sizeof(struct db_val) != val.size)
466: break;
1.1 schwarze 467: if ( ! btree_read(&key, mc, &buf))
468: break;
469:
1.6 kristaps 470: vbuf = val.data;
471: rec = vbuf->rec;
472: mask = vbuf->mask;
1.5 kristaps 473:
474: /*
475: * See if this keyword record matches any of the
476: * expressions we have stored.
477: */
478: if ( ! exprmark(expr, buf, mask, NULL))
1.1 schwarze 479: continue;
480:
481: /*
482: * O(log n) scan for prior records. Since a record
483: * number is unbounded, this has decent performance over
484: * a complex hash function.
485: */
486:
487: for (leaf = root; leaf >= 0; )
1.5 kristaps 488: if (rec > rs[leaf].res.rec &&
489: rs[leaf].rhs >= 0)
490: leaf = rs[leaf].rhs;
491: else if (rec < rs[leaf].res.rec &&
492: rs[leaf].lhs >= 0)
493: leaf = rs[leaf].lhs;
1.1 schwarze 494: else
495: break;
496:
1.5 kristaps 497: /*
498: * If we find a record, see if it has already evaluated
499: * to true. If it has, great, just keep going. If not,
500: * try to evaluate it now and continue anyway.
501: */
502:
503: if (leaf >= 0 && rs[leaf].res.rec == rec) {
504: if (0 == rs[leaf].matched)
505: exprexec(expr, buf, mask, &rs[leaf]);
1.1 schwarze 506: continue;
1.5 kristaps 507: }
1.1 schwarze 508:
509: /*
1.5 kristaps 510: * We have a new file to examine.
511: * Extract the manpage's metadata from the index
512: * database, then begin partial evaluation.
1.1 schwarze 513: */
514:
515: key.data = &rec;
516: key.size = sizeof(recno_t);
517:
518: if (0 != (*idx->get)(idx, &key, &val, 0))
519: break;
520:
1.5 kristaps 521: r.lhs = r.rhs = -1;
522: if ( ! index_read(&key, &val, mc, &r))
1.1 schwarze 523: break;
524:
1.5 kristaps 525: /* XXX: this should be elsewhere, I guess? */
526:
527: if (opts->cat && strcasecmp(opts->cat, r.res.cat))
1.1 schwarze 528: continue;
1.5 kristaps 529: if (opts->arch && strcasecmp(opts->arch, r.res.arch))
1.1 schwarze 530: continue;
531:
1.8 ! kristaps 532: tree->node = rs = mandoc_realloc
! 533: (rs, (tree->len + 1) * sizeof(struct rec));
1.1 schwarze 534:
1.8 ! kristaps 535: memcpy(&rs[tree->len], &r, sizeof(struct rec));
! 536: rs[tree->len].matches =
! 537: mandoc_calloc(terms, sizeof(int));
1.1 schwarze 538:
1.8 ! kristaps 539: exprexec(expr, buf, mask, &rs[tree->len]);
1.1 schwarze 540: /* Append to our tree. */
541:
542: if (leaf >= 0) {
1.5 kristaps 543: if (rec > rs[leaf].res.rec)
1.8 ! kristaps 544: rs[leaf].rhs = tree->len;
1.1 schwarze 545: else
1.8 ! kristaps 546: rs[leaf].lhs = tree->len;
1.1 schwarze 547: } else
1.8 ! kristaps 548: root = tree->len;
1.1 schwarze 549:
1.5 kristaps 550: memset(&r, 0, sizeof(struct rec));
1.8 ! kristaps 551: tree->len++;
1.1 schwarze 552: }
1.5 kristaps 553:
1.8 ! kristaps 554: (*btree->close)(btree);
! 555: (*idx->close)(idx);
1.1 schwarze 556:
557: free(buf);
1.8 ! kristaps 558: return(1 == ch);
1.5 kristaps 559: }
560:
561: static void
562: recfree(struct rec *rec)
563: {
564:
565: free(rec->res.file);
566: free(rec->res.cat);
567: free(rec->res.title);
568: free(rec->res.arch);
569: free(rec->res.desc);
570:
571: free(rec->matches);
1.1 schwarze 572: }
573:
574: struct expr *
1.5 kristaps 575: exprcomp(int argc, char *argv[], size_t *tt)
1.1 schwarze 576: {
1.5 kristaps 577: int pos, lvl;
578: struct expr *e;
579:
580: pos = lvl = 0;
581: *tt = 0;
582:
583: e = exprexpr(argc, argv, &pos, &lvl, tt);
584:
585: if (0 == lvl && pos >= argc)
586: return(e);
587:
588: exprfree(e);
589: return(NULL);
590: }
591:
592: /*
593: * Compile an array of tokens into an expression.
594: * An informal expression grammar is defined in apropos(1).
595: * Return NULL if we fail doing so. All memory will be cleaned up.
596: * Return the root of the expression sequence if alright.
597: */
598: static struct expr *
1.8 ! kristaps 599: exprexpr(int argc, char **argv, int *pos, int *lvl, size_t *tt)
1.5 kristaps 600: {
601: struct expr *e, *first, *next;
602: int log;
603:
604: first = next = NULL;
605:
606: for ( ; *pos < argc; (*pos)++) {
607: e = next;
608:
609: /*
610: * Close out a subexpression.
611: */
612:
613: if (NULL != e && 0 == strcmp(")", argv[*pos])) {
614: if (--(*lvl) < 0)
615: goto err;
616: break;
617: }
618:
619: /*
620: * Small note: if we're just starting, don't let "-a"
621: * and "-o" be considered logical operators: they're
622: * just tokens unless pairwise joining, in which case we
623: * record their existence (or assume "OR").
624: */
625: log = 0;
626:
627: if (NULL != e && 0 == strcmp("-a", argv[*pos]))
628: log = 1;
629: else if (NULL != e && 0 == strcmp("-o", argv[*pos]))
630: log = 2;
631:
632: if (log > 0 && ++(*pos) >= argc)
633: goto err;
634:
635: /*
636: * Now we parse the term part. This can begin with
637: * "-i", in which case the expression is case
638: * insensitive.
639: */
640:
641: if (0 == strcmp("(", argv[*pos])) {
642: ++(*pos);
643: ++(*lvl);
644: next = mandoc_calloc(1, sizeof(struct expr));
645: next->cs = 1;
646: next->subexpr = exprexpr(argc, argv, pos, lvl, tt);
647: if (NULL == next->subexpr) {
648: free(next);
649: next = NULL;
650: }
651: } else if (0 == strcmp("-i", argv[*pos])) {
652: if (++(*pos) >= argc)
653: goto err;
654: next = exprterm(argv[*pos], 0);
655: } else
656: next = exprterm(argv[*pos], 1);
657:
658: if (NULL == next)
659: goto err;
660:
661: next->and = log == 1;
662: next->index = (int)(*tt)++;
663:
664: /* Append to our chain of expressions. */
665:
666: if (NULL == first) {
667: assert(NULL == e);
668: first = next;
669: } else {
670: assert(NULL != e);
671: e->next = next;
672: }
673: }
674:
675: return(first);
676: err:
677: exprfree(first);
678: return(NULL);
679: }
680:
681: /*
682: * Parse a terminal expression with the grammar as defined in
683: * apropos(1).
684: * Return NULL if we fail the parse.
685: */
686: static struct expr *
687: exprterm(char *buf, int cs)
688: {
689: struct expr e;
1.1 schwarze 690: struct expr *p;
1.3 schwarze 691: char *key;
1.5 kristaps 692: int i;
693:
694: memset(&e, 0, sizeof(struct expr));
1.1 schwarze 695:
1.5 kristaps 696: e.cs = cs;
1.1 schwarze 697:
1.5 kristaps 698: /* Choose regex or substring match. */
1.3 schwarze 699:
1.4 kristaps 700: if (NULL == (e.v = strpbrk(buf, "=~"))) {
1.3 schwarze 701: e.regex = 0;
1.4 kristaps 702: e.v = buf;
1.3 schwarze 703: } else {
704: e.regex = '~' == *e.v;
705: *e.v++ = '\0';
706: }
1.1 schwarze 707:
1.5 kristaps 708: /* Determine the record types to search for. */
1.3 schwarze 709:
710: e.mask = 0;
1.4 kristaps 711: if (buf < e.v) {
712: while (NULL != (key = strsep(&buf, ","))) {
1.3 schwarze 713: i = 0;
714: while (types[i].mask &&
1.4 kristaps 715: strcmp(types[i].name, key))
1.3 schwarze 716: i++;
717: e.mask |= types[i].mask;
718: }
719: }
720: if (0 == e.mask)
721: e.mask = TYPE_Nm | TYPE_Nd;
1.1 schwarze 722:
1.5 kristaps 723: if (e.regex) {
724: i = REG_EXTENDED | REG_NOSUB | cs ? 0 : REG_ICASE;
725: if (regcomp(&e.re, e.v, i))
726: return(NULL);
727: }
1.1 schwarze 728:
1.3 schwarze 729: e.v = mandoc_strdup(e.v);
1.1 schwarze 730:
731: p = mandoc_calloc(1, sizeof(struct expr));
732: memcpy(p, &e, sizeof(struct expr));
733: return(p);
734: }
735:
736: void
737: exprfree(struct expr *p)
738: {
1.5 kristaps 739: struct expr *pp;
740:
741: while (NULL != p) {
742: if (p->subexpr)
743: exprfree(p->subexpr);
744: if (p->regex)
745: regfree(&p->re);
746: free(p->v);
747: pp = p->next;
748: free(p);
749: p = pp;
750: }
751: }
1.1 schwarze 752:
1.5 kristaps 753: static int
1.6 kristaps 754: exprmark(const struct expr *p, const char *cp,
755: uint64_t mask, int *ms)
1.5 kristaps 756: {
757:
758: for ( ; p; p = p->next) {
759: if (p->subexpr) {
760: if (exprmark(p->subexpr, cp, mask, ms))
761: return(1);
762: continue;
763: } else if ( ! (mask & p->mask))
764: continue;
1.1 schwarze 765:
1.5 kristaps 766: if (p->regex) {
767: if (regexec(&p->re, cp, 0, NULL, 0))
768: continue;
769: } else if (p->cs) {
770: if (NULL == strstr(cp, p->v))
771: continue;
772: } else {
773: if (NULL == strcasestr(cp, p->v))
774: continue;
775: }
776:
777: if (NULL == ms)
778: return(1);
779: else
780: ms[p->index] = 1;
781: }
1.1 schwarze 782:
1.5 kristaps 783: return(0);
1.1 schwarze 784: }
785:
786: static int
1.5 kristaps 787: expreval(const struct expr *p, int *ms)
1.1 schwarze 788: {
1.5 kristaps 789: int match;
1.1 schwarze 790:
1.5 kristaps 791: /*
792: * AND has precedence over OR. Analysis is left-right, though
793: * it doesn't matter because there are no side-effects.
794: * Thus, step through pairwise ANDs and accumulate their Boolean
795: * evaluation. If we encounter a single true AND collection or
796: * standalone term, the whole expression is true (by definition
797: * of OR).
798: */
799:
800: for (match = 0; p && ! match; p = p->next) {
801: /* Evaluate a subexpression, if applicable. */
802: if (p->subexpr && ! ms[p->index])
803: ms[p->index] = expreval(p->subexpr, ms);
804:
805: match = ms[p->index];
806: for ( ; p->next && p->next->and; p = p->next) {
807: /* Evaluate a subexpression, if applicable. */
808: if (p->next->subexpr && ! ms[p->next->index])
809: ms[p->next->index] =
810: expreval(p->next->subexpr, ms);
811: match = match && ms[p->next->index];
812: }
813: }
814:
815: return(match);
816: }
817:
818: /*
819: * First, update the array of terms for which this expression evaluates
820: * to true.
821: * Second, logically evaluate all terms over the updated array of truth
822: * values.
823: * If this evaluates to true, mark the expression as satisfied.
824: */
825: static void
1.6 kristaps 826: exprexec(const struct expr *p, const char *cp,
827: uint64_t mask, struct rec *r)
1.5 kristaps 828: {
1.1 schwarze 829:
1.5 kristaps 830: assert(0 == r->matched);
831: exprmark(p, cp, mask, r->matches);
832: r->matched = expreval(p, r->matches);
1.1 schwarze 833: }
CVSweb