Annotation of mandoc/apropos_db.c, Revision 1.7
1.7 ! kristaps 1: /* $Id: apropos_db.c,v 1.6 2011/11/20 12:39:08 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>
25:
26: #ifdef __linux__
27: # include <db_185.h>
28: #else
29: # include <db.h>
30: #endif
31:
1.2 schwarze 32: #include "mandocdb.h"
1.1 schwarze 33: #include "apropos_db.h"
34: #include "mandoc.h"
35:
1.5 kristaps 36: struct rec {
37: struct res res; /* resulting record info */
38: /*
39: * Maintain a binary tree for checking the uniqueness of `rec'
40: * when adding elements to the results array.
41: * Since the results array is dynamic, use offset in the array
42: * instead of a pointer to the structure.
43: */
44: int lhs;
45: int rhs;
46: int matched; /* expression is true */
47: int *matches; /* partial truth evaluations */
48: };
49:
1.1 schwarze 50: struct expr {
1.5 kristaps 51: int regex; /* is regex? */
52: int index; /* index in match array */
1.6 kristaps 53: uint64_t mask; /* type-mask */
1.5 kristaps 54: int cs; /* is case-sensitive? */
55: int and; /* is rhs of logical AND? */
56: char *v; /* search value */
57: regex_t re; /* compiled re, if regex */
58: struct expr *next; /* next in sequence */
59: struct expr *subexpr;
1.1 schwarze 60: };
61:
62: struct type {
1.6 kristaps 63: uint64_t mask;
1.1 schwarze 64: const char *name;
65: };
66:
67: static const struct type types[] = {
1.7 ! kristaps 68: { TYPE_An, "An" },
! 69: { TYPE_Ar, "Ar" },
! 70: { TYPE_At, "At" },
! 71: { TYPE_Bsx, "Bsx" },
! 72: { TYPE_Bx, "Bx" },
! 73: { TYPE_Cd, "Cd" },
! 74: { TYPE_Cm, "Cm" },
! 75: { TYPE_Dv, "Dv" },
! 76: { TYPE_Dx, "Dx" },
! 77: { TYPE_Em, "Em" },
! 78: { TYPE_Er, "Er" },
! 79: { TYPE_Ev, "Ev" },
! 80: { TYPE_Fa, "Fa" },
! 81: { TYPE_Fl, "Fl" },
! 82: { TYPE_Fn, "Fn" },
! 83: { TYPE_Fn, "Fo" },
! 84: { TYPE_Ft, "Ft" },
! 85: { TYPE_Fx, "Fx" },
! 86: { TYPE_Ic, "Ic" },
! 87: { TYPE_In, "In" },
! 88: { TYPE_Lb, "Lb" },
! 89: { TYPE_Li, "Li" },
! 90: { TYPE_Lk, "Lk" },
! 91: { TYPE_Ms, "Ms" },
! 92: { TYPE_Mt, "Mt" },
! 93: { TYPE_Nd, "Nd" },
! 94: { TYPE_Nm, "Nm" },
! 95: { TYPE_Nx, "Nx" },
! 96: { TYPE_Ox, "Ox" },
! 97: { TYPE_Pa, "Pa" },
! 98: { TYPE_Rs, "Rs" },
! 99: { TYPE_Sh, "Sh" },
! 100: { TYPE_Ss, "Ss" },
! 101: { TYPE_St, "St" },
! 102: { TYPE_Sy, "Sy" },
! 103: { TYPE_Tn, "Tn" },
! 104: { TYPE_Va, "Va" },
! 105: { TYPE_Va, "Vt" },
! 106: { TYPE_Xr, "Xr" },
1.2 schwarze 107: { INT_MAX, "any" },
1.1 schwarze 108: { 0, NULL }
109: };
110:
111: static DB *btree_open(void);
1.5 kristaps 112: static int btree_read(const DBT *,
113: const struct mchars *, char **);
114: static int expreval(const struct expr *, int *);
115: static void exprexec(const struct expr *,
1.6 kristaps 116: const char *, uint64_t, struct rec *);
1.5 kristaps 117: static int exprmark(const struct expr *,
1.6 kristaps 118: const char *, uint64_t, int *);
1.5 kristaps 119: static struct expr *exprexpr(int, char *[], int *, int *, size_t *);
120: static struct expr *exprterm(char *, int);
1.1 schwarze 121: static DB *index_open(void);
122: static int index_read(const DBT *, const DBT *,
123: const struct mchars *, struct rec *);
124: static void norm_string(const char *,
125: const struct mchars *, char **);
126: static size_t norm_utf8(unsigned int, char[7]);
1.5 kristaps 127: static void recfree(struct rec *);
1.1 schwarze 128:
129: /*
130: * Open the keyword mandoc-db database.
131: */
132: static DB *
133: btree_open(void)
134: {
135: BTREEINFO info;
136: DB *db;
137:
138: memset(&info, 0, sizeof(BTREEINFO));
139: info.flags = R_DUP;
140:
1.2 schwarze 141: db = dbopen(MANDOC_DB, O_RDONLY, 0, DB_BTREE, &info);
1.1 schwarze 142: if (NULL != db)
143: return(db);
144:
145: return(NULL);
146: }
147:
148: /*
149: * Read a keyword from the database and normalise it.
150: * Return 0 if the database is insane, else 1.
151: */
152: static int
153: btree_read(const DBT *v, const struct mchars *mc, char **buf)
154: {
155:
156: /* Sanity: are we nil-terminated? */
157:
158: assert(v->size > 0);
1.6 kristaps 159:
1.1 schwarze 160: if ('\0' != ((char *)v->data)[(int)v->size - 1])
161: return(0);
162:
163: norm_string((char *)v->data, mc, buf);
164: return(1);
165: }
166:
167: /*
168: * Take a Unicode codepoint and produce its UTF-8 encoding.
169: * This isn't the best way to do this, but it works.
170: * The magic numbers are from the UTF-8 packaging.
171: * They're not as scary as they seem: read the UTF-8 spec for details.
172: */
173: static size_t
174: norm_utf8(unsigned int cp, char out[7])
175: {
176: size_t rc;
177:
178: rc = 0;
179:
180: if (cp <= 0x0000007F) {
181: rc = 1;
182: out[0] = (char)cp;
183: } else if (cp <= 0x000007FF) {
184: rc = 2;
185: out[0] = (cp >> 6 & 31) | 192;
186: out[1] = (cp & 63) | 128;
187: } else if (cp <= 0x0000FFFF) {
188: rc = 3;
189: out[0] = (cp >> 12 & 15) | 224;
190: out[1] = (cp >> 6 & 63) | 128;
191: out[2] = (cp & 63) | 128;
192: } else if (cp <= 0x001FFFFF) {
193: rc = 4;
194: out[0] = (cp >> 18 & 7) | 240;
195: out[1] = (cp >> 12 & 63) | 128;
196: out[2] = (cp >> 6 & 63) | 128;
197: out[3] = (cp & 63) | 128;
198: } else if (cp <= 0x03FFFFFF) {
199: rc = 5;
200: out[0] = (cp >> 24 & 3) | 248;
201: out[1] = (cp >> 18 & 63) | 128;
202: out[2] = (cp >> 12 & 63) | 128;
203: out[3] = (cp >> 6 & 63) | 128;
204: out[4] = (cp & 63) | 128;
205: } else if (cp <= 0x7FFFFFFF) {
206: rc = 6;
207: out[0] = (cp >> 30 & 1) | 252;
208: out[1] = (cp >> 24 & 63) | 128;
209: out[2] = (cp >> 18 & 63) | 128;
210: out[3] = (cp >> 12 & 63) | 128;
211: out[4] = (cp >> 6 & 63) | 128;
212: out[5] = (cp & 63) | 128;
213: } else
214: return(0);
215:
216: out[rc] = '\0';
217: return(rc);
218: }
219:
220: /*
221: * Normalise strings from the index and database.
222: * These strings are escaped as defined by mandoc_char(7) along with
223: * other goop in mandoc.h (e.g., soft hyphens).
224: * This function normalises these into a nice UTF-8 string.
225: * Returns 0 if the database is fucked.
226: */
227: static void
228: norm_string(const char *val, const struct mchars *mc, char **buf)
229: {
230: size_t sz, bsz;
231: char utfbuf[7];
232: const char *seq, *cpp;
233: int len, u, pos;
234: enum mandoc_esc esc;
235: static const char res[] = { '\\', '\t',
236: ASCII_NBRSP, ASCII_HYPH, '\0' };
237:
238: /* Pre-allocate by the length of the input */
239:
240: bsz = strlen(val) + 1;
241: *buf = mandoc_realloc(*buf, bsz);
242: pos = 0;
243:
244: while ('\0' != *val) {
245: /*
246: * Halt on the first escape sequence.
247: * This also halts on the end of string, in which case
248: * we just copy, fallthrough, and exit the loop.
249: */
250: if ((sz = strcspn(val, res)) > 0) {
251: memcpy(&(*buf)[pos], val, sz);
252: pos += (int)sz;
253: val += (int)sz;
254: }
255:
256: if (ASCII_HYPH == *val) {
257: (*buf)[pos++] = '-';
258: val++;
259: continue;
260: } else if ('\t' == *val || ASCII_NBRSP == *val) {
261: (*buf)[pos++] = ' ';
262: val++;
263: continue;
264: } else if ('\\' != *val)
265: break;
266:
267: /* Read past the slash. */
268:
269: val++;
270: u = 0;
271:
272: /*
273: * Parse the escape sequence and see if it's a
274: * predefined character or special character.
275: */
276:
277: esc = mandoc_escape(&val, &seq, &len);
278: if (ESCAPE_ERROR == esc)
279: break;
280:
281: /*
282: * XXX - this just does UTF-8, but we need to know
283: * beforehand whether we should do text substitution.
284: */
285:
286: switch (esc) {
287: case (ESCAPE_SPECIAL):
288: if (0 != (u = mchars_spec2cp(mc, seq, len)))
289: break;
290: /* FALLTHROUGH */
291: default:
292: continue;
293: }
294:
295: /*
296: * If we have a Unicode codepoint, try to convert that
297: * to a UTF-8 byte string.
298: */
299:
300: cpp = utfbuf;
301: if (0 == (sz = norm_utf8(u, utfbuf)))
302: continue;
303:
304: /* Copy the rendered glyph into the stream. */
305:
306: sz = strlen(cpp);
307: bsz += sz;
308:
309: *buf = mandoc_realloc(*buf, bsz);
310:
311: memcpy(&(*buf)[pos], cpp, sz);
312: pos += (int)sz;
313: }
314:
315: (*buf)[pos] = '\0';
316: }
317:
318: /*
319: * Open the filename-index mandoc-db database.
320: * Returns NULL if opening failed.
321: */
322: static DB *
323: index_open(void)
324: {
325: DB *db;
326:
1.2 schwarze 327: db = dbopen(MANDOC_IDX, O_RDONLY, 0, DB_RECNO, NULL);
1.1 schwarze 328: if (NULL != db)
329: return(db);
330:
331: return(NULL);
332: }
333:
334: /*
335: * Safely unpack from an index file record into the structure.
336: * Returns 1 if an entry was unpacked, 0 if the database is insane.
337: */
338: static int
339: index_read(const DBT *key, const DBT *val,
340: const struct mchars *mc, struct rec *rec)
341: {
342: size_t left;
343: char *np, *cp;
344:
345: #define INDEX_BREAD(_dst) \
346: do { \
347: if (NULL == (np = memchr(cp, '\0', left))) \
348: return(0); \
349: norm_string(cp, mc, &(_dst)); \
350: left -= (np - cp) + 1; \
351: cp = np + 1; \
352: } while (/* CONSTCOND */ 0)
353:
354: left = val->size;
355: cp = (char *)val->data;
356:
1.5 kristaps 357: rec->res.rec = *(recno_t *)key->data;
1.1 schwarze 358:
1.5 kristaps 359: INDEX_BREAD(rec->res.file);
360: INDEX_BREAD(rec->res.cat);
361: INDEX_BREAD(rec->res.title);
362: INDEX_BREAD(rec->res.arch);
363: INDEX_BREAD(rec->res.desc);
1.1 schwarze 364: return(1);
365: }
366:
367: /*
368: * Search the mandocdb database for the expression "expr".
369: * Filter out by "opts".
370: * Call "res" with the results, which may be zero.
1.5 kristaps 371: * Return 0 if there was a database error, else return 1.
1.1 schwarze 372: */
1.5 kristaps 373: int
1.1 schwarze 374: apropos_search(const struct opts *opts, const struct expr *expr,
1.5 kristaps 375: size_t terms, void *arg,
376: void (*res)(struct res *, size_t, void *))
1.1 schwarze 377: {
1.6 kristaps 378: int i, rsz, root, leaf, mlen, rc, ch;
379: uint64_t mask;
1.1 schwarze 380: DBT key, val;
381: DB *btree, *idx;
382: struct mchars *mc;
383: char *buf;
384: recno_t rec;
1.5 kristaps 385: struct rec *rs;
386: struct res *ress;
387: struct rec r;
1.6 kristaps 388: struct db_val *vbuf;
1.1 schwarze 389:
1.5 kristaps 390: rc = 0;
1.1 schwarze 391: root = -1;
392: leaf = -1;
393: btree = NULL;
394: idx = NULL;
395: mc = NULL;
396: buf = NULL;
1.5 kristaps 397: rs = NULL;
398: rsz = 0;
1.1 schwarze 399:
1.5 kristaps 400: memset(&r, 0, sizeof(struct rec));
1.1 schwarze 401:
402: mc = mchars_alloc();
403:
404: if (NULL == (btree = btree_open()))
405: goto out;
406: if (NULL == (idx = index_open()))
407: goto out;
408:
409: while (0 == (ch = (*btree->seq)(btree, &key, &val, R_NEXT))) {
1.6 kristaps 410: if (key.size < 2 || sizeof(struct db_val) != val.size)
411: break;
1.1 schwarze 412: if ( ! btree_read(&key, mc, &buf))
413: break;
414:
1.6 kristaps 415: vbuf = val.data;
416: rec = vbuf->rec;
417: mask = vbuf->mask;
1.5 kristaps 418:
419: /*
420: * See if this keyword record matches any of the
421: * expressions we have stored.
422: */
423: if ( ! exprmark(expr, buf, mask, NULL))
1.1 schwarze 424: continue;
425:
426: /*
427: * O(log n) scan for prior records. Since a record
428: * number is unbounded, this has decent performance over
429: * a complex hash function.
430: */
431:
432: for (leaf = root; leaf >= 0; )
1.5 kristaps 433: if (rec > rs[leaf].res.rec &&
434: rs[leaf].rhs >= 0)
435: leaf = rs[leaf].rhs;
436: else if (rec < rs[leaf].res.rec &&
437: rs[leaf].lhs >= 0)
438: leaf = rs[leaf].lhs;
1.1 schwarze 439: else
440: break;
441:
1.5 kristaps 442: /*
443: * If we find a record, see if it has already evaluated
444: * to true. If it has, great, just keep going. If not,
445: * try to evaluate it now and continue anyway.
446: */
447:
448: if (leaf >= 0 && rs[leaf].res.rec == rec) {
449: if (0 == rs[leaf].matched)
450: exprexec(expr, buf, mask, &rs[leaf]);
1.1 schwarze 451: continue;
1.5 kristaps 452: }
1.1 schwarze 453:
454: /*
1.5 kristaps 455: * We have a new file to examine.
456: * Extract the manpage's metadata from the index
457: * database, then begin partial evaluation.
1.1 schwarze 458: */
459:
460: key.data = &rec;
461: key.size = sizeof(recno_t);
462:
463: if (0 != (*idx->get)(idx, &key, &val, 0))
464: break;
465:
1.5 kristaps 466: r.lhs = r.rhs = -1;
467: if ( ! index_read(&key, &val, mc, &r))
1.1 schwarze 468: break;
469:
1.5 kristaps 470: /* XXX: this should be elsewhere, I guess? */
471:
472: if (opts->cat && strcasecmp(opts->cat, r.res.cat))
1.1 schwarze 473: continue;
1.5 kristaps 474: if (opts->arch && strcasecmp(opts->arch, r.res.arch))
1.1 schwarze 475: continue;
476:
1.5 kristaps 477: rs = mandoc_realloc
478: (rs, (rsz + 1) * sizeof(struct rec));
1.1 schwarze 479:
1.5 kristaps 480: memcpy(&rs[rsz], &r, sizeof(struct rec));
481: rs[rsz].matches = mandoc_calloc(terms, sizeof(int));
1.1 schwarze 482:
1.5 kristaps 483: exprexec(expr, buf, mask, &rs[rsz]);
1.1 schwarze 484: /* Append to our tree. */
485:
486: if (leaf >= 0) {
1.5 kristaps 487: if (rec > rs[leaf].res.rec)
488: rs[leaf].rhs = rsz;
1.1 schwarze 489: else
1.5 kristaps 490: rs[leaf].lhs = rsz;
1.1 schwarze 491: } else
1.5 kristaps 492: root = rsz;
1.1 schwarze 493:
1.5 kristaps 494: memset(&r, 0, sizeof(struct rec));
495: rsz++;
1.1 schwarze 496: }
1.5 kristaps 497:
498: /*
499: * If we haven't encountered any database errors, then construct
500: * an array of results and push them to the caller.
501: */
1.1 schwarze 502:
1.5 kristaps 503: if (1 == ch) {
504: for (mlen = i = 0; i < rsz; i++)
505: if (rs[i].matched)
506: mlen++;
507: ress = mandoc_malloc(mlen * sizeof(struct res));
508: for (mlen = i = 0; i < rsz; i++)
509: if (rs[i].matched)
510: memcpy(&ress[mlen++], &rs[i].res,
511: sizeof(struct res));
512: (*res)(ress, mlen, arg);
513: free(ress);
514: rc = 1;
515: }
1.1 schwarze 516:
517: out:
1.5 kristaps 518: for (i = 0; i < rsz; i++)
519: recfree(&rs[i]);
1.1 schwarze 520:
1.5 kristaps 521: recfree(&r);
1.1 schwarze 522:
523: if (mc)
524: mchars_free(mc);
525: if (btree)
526: (*btree->close)(btree);
527: if (idx)
528: (*idx->close)(idx);
529:
530: free(buf);
1.5 kristaps 531: free(rs);
532: return(rc);
533: }
534:
535: static void
536: recfree(struct rec *rec)
537: {
538:
539: free(rec->res.file);
540: free(rec->res.cat);
541: free(rec->res.title);
542: free(rec->res.arch);
543: free(rec->res.desc);
544:
545: free(rec->matches);
1.1 schwarze 546: }
547:
548: struct expr *
1.5 kristaps 549: exprcomp(int argc, char *argv[], size_t *tt)
1.1 schwarze 550: {
1.5 kristaps 551: int pos, lvl;
552: struct expr *e;
553:
554: pos = lvl = 0;
555: *tt = 0;
556:
557: e = exprexpr(argc, argv, &pos, &lvl, tt);
558:
559: if (0 == lvl && pos >= argc)
560: return(e);
561:
562: exprfree(e);
563: return(NULL);
564: }
565:
566: /*
567: * Compile an array of tokens into an expression.
568: * An informal expression grammar is defined in apropos(1).
569: * Return NULL if we fail doing so. All memory will be cleaned up.
570: * Return the root of the expression sequence if alright.
571: */
572: static struct expr *
573: exprexpr(int argc, char *argv[], int *pos, int *lvl, size_t *tt)
574: {
575: struct expr *e, *first, *next;
576: int log;
577:
578: first = next = NULL;
579:
580: for ( ; *pos < argc; (*pos)++) {
581: e = next;
582:
583: /*
584: * Close out a subexpression.
585: */
586:
587: if (NULL != e && 0 == strcmp(")", argv[*pos])) {
588: if (--(*lvl) < 0)
589: goto err;
590: break;
591: }
592:
593: /*
594: * Small note: if we're just starting, don't let "-a"
595: * and "-o" be considered logical operators: they're
596: * just tokens unless pairwise joining, in which case we
597: * record their existence (or assume "OR").
598: */
599: log = 0;
600:
601: if (NULL != e && 0 == strcmp("-a", argv[*pos]))
602: log = 1;
603: else if (NULL != e && 0 == strcmp("-o", argv[*pos]))
604: log = 2;
605:
606: if (log > 0 && ++(*pos) >= argc)
607: goto err;
608:
609: /*
610: * Now we parse the term part. This can begin with
611: * "-i", in which case the expression is case
612: * insensitive.
613: */
614:
615: if (0 == strcmp("(", argv[*pos])) {
616: ++(*pos);
617: ++(*lvl);
618: next = mandoc_calloc(1, sizeof(struct expr));
619: next->cs = 1;
620: next->subexpr = exprexpr(argc, argv, pos, lvl, tt);
621: if (NULL == next->subexpr) {
622: free(next);
623: next = NULL;
624: }
625: } else if (0 == strcmp("-i", argv[*pos])) {
626: if (++(*pos) >= argc)
627: goto err;
628: next = exprterm(argv[*pos], 0);
629: } else
630: next = exprterm(argv[*pos], 1);
631:
632: if (NULL == next)
633: goto err;
634:
635: next->and = log == 1;
636: next->index = (int)(*tt)++;
637:
638: /* Append to our chain of expressions. */
639:
640: if (NULL == first) {
641: assert(NULL == e);
642: first = next;
643: } else {
644: assert(NULL != e);
645: e->next = next;
646: }
647: }
648:
649: return(first);
650: err:
651: exprfree(first);
652: return(NULL);
653: }
654:
655: /*
656: * Parse a terminal expression with the grammar as defined in
657: * apropos(1).
658: * Return NULL if we fail the parse.
659: */
660: static struct expr *
661: exprterm(char *buf, int cs)
662: {
663: struct expr e;
1.1 schwarze 664: struct expr *p;
1.3 schwarze 665: char *key;
1.5 kristaps 666: int i;
667:
668: memset(&e, 0, sizeof(struct expr));
1.1 schwarze 669:
1.5 kristaps 670: e.cs = cs;
1.1 schwarze 671:
1.5 kristaps 672: /* Choose regex or substring match. */
1.3 schwarze 673:
1.4 kristaps 674: if (NULL == (e.v = strpbrk(buf, "=~"))) {
1.3 schwarze 675: e.regex = 0;
1.4 kristaps 676: e.v = buf;
1.3 schwarze 677: } else {
678: e.regex = '~' == *e.v;
679: *e.v++ = '\0';
680: }
1.1 schwarze 681:
1.5 kristaps 682: /* Determine the record types to search for. */
1.3 schwarze 683:
684: e.mask = 0;
1.4 kristaps 685: if (buf < e.v) {
686: while (NULL != (key = strsep(&buf, ","))) {
1.3 schwarze 687: i = 0;
688: while (types[i].mask &&
1.4 kristaps 689: strcmp(types[i].name, key))
1.3 schwarze 690: i++;
691: e.mask |= types[i].mask;
692: }
693: }
694: if (0 == e.mask)
695: e.mask = TYPE_Nm | TYPE_Nd;
1.1 schwarze 696:
1.5 kristaps 697: if (e.regex) {
698: i = REG_EXTENDED | REG_NOSUB | cs ? 0 : REG_ICASE;
699: if (regcomp(&e.re, e.v, i))
700: return(NULL);
701: }
1.1 schwarze 702:
1.3 schwarze 703: e.v = mandoc_strdup(e.v);
1.1 schwarze 704:
705: p = mandoc_calloc(1, sizeof(struct expr));
706: memcpy(p, &e, sizeof(struct expr));
707: return(p);
708: }
709:
710: void
711: exprfree(struct expr *p)
712: {
1.5 kristaps 713: struct expr *pp;
714:
715: while (NULL != p) {
716: if (p->subexpr)
717: exprfree(p->subexpr);
718: if (p->regex)
719: regfree(&p->re);
720: free(p->v);
721: pp = p->next;
722: free(p);
723: p = pp;
724: }
725: }
1.1 schwarze 726:
1.5 kristaps 727: static int
1.6 kristaps 728: exprmark(const struct expr *p, const char *cp,
729: uint64_t mask, int *ms)
1.5 kristaps 730: {
731:
732: for ( ; p; p = p->next) {
733: if (p->subexpr) {
734: if (exprmark(p->subexpr, cp, mask, ms))
735: return(1);
736: continue;
737: } else if ( ! (mask & p->mask))
738: continue;
1.1 schwarze 739:
1.5 kristaps 740: if (p->regex) {
741: if (regexec(&p->re, cp, 0, NULL, 0))
742: continue;
743: } else if (p->cs) {
744: if (NULL == strstr(cp, p->v))
745: continue;
746: } else {
747: if (NULL == strcasestr(cp, p->v))
748: continue;
749: }
750:
751: if (NULL == ms)
752: return(1);
753: else
754: ms[p->index] = 1;
755: }
1.1 schwarze 756:
1.5 kristaps 757: return(0);
1.1 schwarze 758: }
759:
760: static int
1.5 kristaps 761: expreval(const struct expr *p, int *ms)
1.1 schwarze 762: {
1.5 kristaps 763: int match;
1.1 schwarze 764:
1.5 kristaps 765: /*
766: * AND has precedence over OR. Analysis is left-right, though
767: * it doesn't matter because there are no side-effects.
768: * Thus, step through pairwise ANDs and accumulate their Boolean
769: * evaluation. If we encounter a single true AND collection or
770: * standalone term, the whole expression is true (by definition
771: * of OR).
772: */
773:
774: for (match = 0; p && ! match; p = p->next) {
775: /* Evaluate a subexpression, if applicable. */
776: if (p->subexpr && ! ms[p->index])
777: ms[p->index] = expreval(p->subexpr, ms);
778:
779: match = ms[p->index];
780: for ( ; p->next && p->next->and; p = p->next) {
781: /* Evaluate a subexpression, if applicable. */
782: if (p->next->subexpr && ! ms[p->next->index])
783: ms[p->next->index] =
784: expreval(p->next->subexpr, ms);
785: match = match && ms[p->next->index];
786: }
787: }
788:
789: return(match);
790: }
791:
792: /*
793: * First, update the array of terms for which this expression evaluates
794: * to true.
795: * Second, logically evaluate all terms over the updated array of truth
796: * values.
797: * If this evaluates to true, mark the expression as satisfied.
798: */
799: static void
1.6 kristaps 800: exprexec(const struct expr *p, const char *cp,
801: uint64_t mask, struct rec *r)
1.5 kristaps 802: {
1.1 schwarze 803:
1.5 kristaps 804: assert(0 == r->matched);
805: exprmark(p, cp, mask, r->matches);
806: r->matched = expreval(p, r->matches);
1.1 schwarze 807: }
CVSweb