Annotation of mandoc/apropos_db.c, Revision 1.18
1.18 ! kristaps 1: /* $Id: apropos_db.c,v 1.17 2011/12/01 23:46:26 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 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:
1.8 kristaps 67: struct rectree {
68: struct rec *node; /* record array for dir tree */
69: int len; /* length of record array */
70: };
71:
1.1 schwarze 72: static const struct type types[] = {
1.7 kristaps 73: { TYPE_An, "An" },
74: { TYPE_Ar, "Ar" },
75: { TYPE_At, "At" },
76: { TYPE_Bsx, "Bsx" },
77: { TYPE_Bx, "Bx" },
78: { TYPE_Cd, "Cd" },
79: { TYPE_Cm, "Cm" },
80: { TYPE_Dv, "Dv" },
81: { TYPE_Dx, "Dx" },
82: { TYPE_Em, "Em" },
83: { TYPE_Er, "Er" },
84: { TYPE_Ev, "Ev" },
85: { TYPE_Fa, "Fa" },
86: { TYPE_Fl, "Fl" },
87: { TYPE_Fn, "Fn" },
88: { TYPE_Fn, "Fo" },
89: { TYPE_Ft, "Ft" },
90: { TYPE_Fx, "Fx" },
91: { TYPE_Ic, "Ic" },
92: { TYPE_In, "In" },
93: { TYPE_Lb, "Lb" },
94: { TYPE_Li, "Li" },
95: { TYPE_Lk, "Lk" },
96: { TYPE_Ms, "Ms" },
97: { TYPE_Mt, "Mt" },
98: { TYPE_Nd, "Nd" },
99: { TYPE_Nm, "Nm" },
100: { TYPE_Nx, "Nx" },
101: { TYPE_Ox, "Ox" },
102: { TYPE_Pa, "Pa" },
103: { TYPE_Rs, "Rs" },
104: { TYPE_Sh, "Sh" },
105: { TYPE_Ss, "Ss" },
106: { TYPE_St, "St" },
107: { TYPE_Sy, "Sy" },
108: { TYPE_Tn, "Tn" },
109: { TYPE_Va, "Va" },
110: { TYPE_Va, "Vt" },
111: { TYPE_Xr, "Xr" },
1.2 schwarze 112: { INT_MAX, "any" },
1.1 schwarze 113: { 0, NULL }
114: };
115:
116: static DB *btree_open(void);
1.17 kristaps 117: static int btree_read(const DBT *, const DBT *,
118: const struct mchars *,
119: struct db_val *, char **);
1.5 kristaps 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
1.17 kristaps 162: btree_read(const DBT *k, const DBT *v,
163: const struct mchars *mc,
164: struct db_val *dbv, char **buf)
1.1 schwarze 165: {
1.17 kristaps 166: const struct db_val *vp;
1.1 schwarze 167:
1.17 kristaps 168: /* Are our sizes sane? */
169: if (k->size < 2 || sizeof(struct db_val) != v->size)
170: return(0);
1.6 kristaps 171:
1.17 kristaps 172: /* Is our string nil-terminated? */
173: if ('\0' != ((const char *)k->data)[(int)k->size - 1])
1.1 schwarze 174: return(0);
175:
1.17 kristaps 176: vp = v->data;
177: norm_string((const char *)k->data, mc, buf);
1.18 ! kristaps 178: dbv->rec = betoh32(vp->rec);
! 179: dbv->mask = betoh64(vp->mask);
1.1 schwarze 180: return(1);
181: }
182:
183: /*
184: * Take a Unicode codepoint and produce its UTF-8 encoding.
185: * This isn't the best way to do this, but it works.
1.12 schwarze 186: * The magic numbers are from the UTF-8 packaging.
1.1 schwarze 187: * They're not as scary as they seem: read the UTF-8 spec for details.
188: */
189: static size_t
190: norm_utf8(unsigned int cp, char out[7])
191: {
192: size_t rc;
193:
194: rc = 0;
195:
196: if (cp <= 0x0000007F) {
197: rc = 1;
198: out[0] = (char)cp;
199: } else if (cp <= 0x000007FF) {
200: rc = 2;
201: out[0] = (cp >> 6 & 31) | 192;
202: out[1] = (cp & 63) | 128;
203: } else if (cp <= 0x0000FFFF) {
204: rc = 3;
205: out[0] = (cp >> 12 & 15) | 224;
206: out[1] = (cp >> 6 & 63) | 128;
207: out[2] = (cp & 63) | 128;
208: } else if (cp <= 0x001FFFFF) {
209: rc = 4;
210: out[0] = (cp >> 18 & 7) | 240;
211: out[1] = (cp >> 12 & 63) | 128;
212: out[2] = (cp >> 6 & 63) | 128;
213: out[3] = (cp & 63) | 128;
214: } else if (cp <= 0x03FFFFFF) {
215: rc = 5;
216: out[0] = (cp >> 24 & 3) | 248;
217: out[1] = (cp >> 18 & 63) | 128;
218: out[2] = (cp >> 12 & 63) | 128;
219: out[3] = (cp >> 6 & 63) | 128;
220: out[4] = (cp & 63) | 128;
221: } else if (cp <= 0x7FFFFFFF) {
222: rc = 6;
223: out[0] = (cp >> 30 & 1) | 252;
224: out[1] = (cp >> 24 & 63) | 128;
225: out[2] = (cp >> 18 & 63) | 128;
226: out[3] = (cp >> 12 & 63) | 128;
227: out[4] = (cp >> 6 & 63) | 128;
228: out[5] = (cp & 63) | 128;
229: } else
230: return(0);
231:
232: out[rc] = '\0';
233: return(rc);
234: }
235:
236: /*
237: * Normalise strings from the index and database.
238: * These strings are escaped as defined by mandoc_char(7) along with
239: * other goop in mandoc.h (e.g., soft hyphens).
240: * This function normalises these into a nice UTF-8 string.
241: * Returns 0 if the database is fucked.
242: */
243: static void
244: norm_string(const char *val, const struct mchars *mc, char **buf)
245: {
246: size_t sz, bsz;
247: char utfbuf[7];
248: const char *seq, *cpp;
249: int len, u, pos;
250: enum mandoc_esc esc;
1.12 schwarze 251: static const char res[] = { '\\', '\t',
1.1 schwarze 252: ASCII_NBRSP, ASCII_HYPH, '\0' };
253:
254: /* Pre-allocate by the length of the input */
255:
256: bsz = strlen(val) + 1;
257: *buf = mandoc_realloc(*buf, bsz);
258: pos = 0;
259:
260: while ('\0' != *val) {
261: /*
262: * Halt on the first escape sequence.
263: * This also halts on the end of string, in which case
264: * we just copy, fallthrough, and exit the loop.
265: */
266: if ((sz = strcspn(val, res)) > 0) {
267: memcpy(&(*buf)[pos], val, sz);
268: pos += (int)sz;
269: val += (int)sz;
270: }
271:
272: if (ASCII_HYPH == *val) {
273: (*buf)[pos++] = '-';
274: val++;
275: continue;
276: } else if ('\t' == *val || ASCII_NBRSP == *val) {
277: (*buf)[pos++] = ' ';
278: val++;
279: continue;
280: } else if ('\\' != *val)
281: break;
282:
283: /* Read past the slash. */
284:
285: val++;
286: u = 0;
287:
288: /*
289: * Parse the escape sequence and see if it's a
290: * predefined character or special character.
291: */
292:
293: esc = mandoc_escape(&val, &seq, &len);
294: if (ESCAPE_ERROR == esc)
295: break;
296:
1.12 schwarze 297: /*
1.1 schwarze 298: * XXX - this just does UTF-8, but we need to know
299: * beforehand whether we should do text substitution.
300: */
301:
302: switch (esc) {
303: case (ESCAPE_SPECIAL):
304: if (0 != (u = mchars_spec2cp(mc, seq, len)))
305: break;
306: /* FALLTHROUGH */
307: default:
308: continue;
309: }
310:
311: /*
312: * If we have a Unicode codepoint, try to convert that
313: * to a UTF-8 byte string.
314: */
315:
316: cpp = utfbuf;
317: if (0 == (sz = norm_utf8(u, utfbuf)))
318: continue;
319:
320: /* Copy the rendered glyph into the stream. */
321:
322: sz = strlen(cpp);
323: bsz += sz;
324:
325: *buf = mandoc_realloc(*buf, bsz);
326:
327: memcpy(&(*buf)[pos], cpp, sz);
328: pos += (int)sz;
329: }
330:
331: (*buf)[pos] = '\0';
332: }
333:
334: /*
335: * Open the filename-index mandoc-db database.
336: * Returns NULL if opening failed.
337: */
338: static DB *
339: index_open(void)
340: {
341: DB *db;
342:
1.2 schwarze 343: db = dbopen(MANDOC_IDX, O_RDONLY, 0, DB_RECNO, NULL);
1.1 schwarze 344: if (NULL != db)
345: return(db);
346:
347: return(NULL);
348: }
349:
350: /*
351: * Safely unpack from an index file record into the structure.
352: * Returns 1 if an entry was unpacked, 0 if the database is insane.
353: */
354: static int
1.11 kristaps 355: index_read(const DBT *key, const DBT *val, int index,
1.1 schwarze 356: const struct mchars *mc, struct rec *rec)
357: {
358: size_t left;
359: char *np, *cp;
360:
361: #define INDEX_BREAD(_dst) \
362: do { \
363: if (NULL == (np = memchr(cp, '\0', left))) \
364: return(0); \
365: norm_string(cp, mc, &(_dst)); \
366: left -= (np - cp) + 1; \
367: cp = np + 1; \
368: } while (/* CONSTCOND */ 0)
369:
370: left = val->size;
371: cp = (char *)val->data;
372:
1.5 kristaps 373: rec->res.rec = *(recno_t *)key->data;
1.11 kristaps 374: rec->res.volume = index;
1.1 schwarze 375:
1.14 schwarze 376: INDEX_BREAD(rec->res.type);
1.5 kristaps 377: INDEX_BREAD(rec->res.file);
378: INDEX_BREAD(rec->res.cat);
379: INDEX_BREAD(rec->res.title);
380: INDEX_BREAD(rec->res.arch);
381: INDEX_BREAD(rec->res.desc);
1.1 schwarze 382: return(1);
383: }
384:
385: /*
1.10 kristaps 386: * Search mandocdb databases in paths for expression "expr".
1.1 schwarze 387: * Filter out by "opts".
388: * Call "res" with the results, which may be zero.
1.5 kristaps 389: * Return 0 if there was a database error, else return 1.
1.1 schwarze 390: */
1.5 kristaps 391: int
1.10 kristaps 392: apropos_search(int pathsz, char **paths, const struct opts *opts,
1.12 schwarze 393: const struct expr *expr, size_t terms, void *arg,
1.5 kristaps 394: void (*res)(struct res *, size_t, void *))
1.1 schwarze 395: {
1.8 kristaps 396: struct rectree tree;
397: struct mchars *mc;
398: struct res *ress;
399: int i, mlen, rc;
400:
401: memset(&tree, 0, sizeof(struct rectree));
402:
1.10 kristaps 403: rc = 0;
1.8 kristaps 404: mc = mchars_alloc();
405:
1.10 kristaps 406: /*
407: * Main loop. Change into the directory containing manpage
408: * databases. Run our expession over each database in the set.
409: */
410:
411: for (i = 0; i < pathsz; i++) {
412: if (chdir(paths[i]))
1.8 kristaps 413: continue;
1.11 kristaps 414: if ( ! single_search(&tree, opts, expr, terms, mc, i))
1.10 kristaps 415: goto out;
1.8 kristaps 416: }
417:
418: /*
1.10 kristaps 419: * Count matching files, transfer to a "clean" array, then feed
420: * them to the output handler.
1.8 kristaps 421: */
422:
423: for (mlen = i = 0; i < tree.len; i++)
424: if (tree.node[i].matched)
425: mlen++;
426:
427: ress = mandoc_malloc(mlen * sizeof(struct res));
428:
429: for (mlen = i = 0; i < tree.len; i++)
430: if (tree.node[i].matched)
1.12 schwarze 431: memcpy(&ress[mlen++], &tree.node[i].res,
1.8 kristaps 432: sizeof(struct res));
433:
434: (*res)(ress, mlen, arg);
435: free(ress);
436:
1.10 kristaps 437: rc = 1;
438: out:
1.8 kristaps 439: for (i = 0; i < tree.len; i++)
440: recfree(&tree.node[i]);
441:
442: free(tree.node);
443: mchars_free(mc);
444: return(rc);
445: }
446:
447: static int
448: single_search(struct rectree *tree, const struct opts *opts,
449: const struct expr *expr, size_t terms,
1.11 kristaps 450: struct mchars *mc, int vol)
1.8 kristaps 451: {
452: int root, leaf, ch;
1.1 schwarze 453: DBT key, val;
454: DB *btree, *idx;
455: char *buf;
1.5 kristaps 456: struct rec *rs;
457: struct rec r;
1.17 kristaps 458: struct db_val vb;
1.1 schwarze 459:
460: root = -1;
461: leaf = -1;
462: btree = NULL;
463: idx = NULL;
464: buf = NULL;
1.8 kristaps 465: rs = tree->node;
1.1 schwarze 466:
1.5 kristaps 467: memset(&r, 0, sizeof(struct rec));
1.1 schwarze 468:
1.12 schwarze 469: if (NULL == (btree = btree_open()))
1.10 kristaps 470: return(1);
1.1 schwarze 471:
1.8 kristaps 472: if (NULL == (idx = index_open())) {
473: (*btree->close)(btree);
1.10 kristaps 474: return(1);
1.8 kristaps 475: }
1.1 schwarze 476:
477: while (0 == (ch = (*btree->seq)(btree, &key, &val, R_NEXT))) {
1.17 kristaps 478: if ( ! btree_read(&key, &val, mc, &vb, &buf))
1.1 schwarze 479: break;
480:
1.5 kristaps 481: /*
482: * See if this keyword record matches any of the
483: * expressions we have stored.
484: */
1.17 kristaps 485: if ( ! exprmark(expr, buf, vb.mask, NULL))
1.1 schwarze 486: continue;
487:
488: /*
489: * O(log n) scan for prior records. Since a record
490: * number is unbounded, this has decent performance over
491: * a complex hash function.
492: */
493:
494: for (leaf = root; leaf >= 0; )
1.17 kristaps 495: if (vb.rec > rs[leaf].res.rec &&
1.5 kristaps 496: rs[leaf].rhs >= 0)
497: leaf = rs[leaf].rhs;
1.17 kristaps 498: else if (vb.rec < rs[leaf].res.rec &&
1.5 kristaps 499: rs[leaf].lhs >= 0)
500: leaf = rs[leaf].lhs;
1.12 schwarze 501: else
1.1 schwarze 502: break;
503:
1.5 kristaps 504: /*
505: * If we find a record, see if it has already evaluated
506: * to true. If it has, great, just keep going. If not,
507: * try to evaluate it now and continue anyway.
508: */
509:
1.17 kristaps 510: if (leaf >= 0 && rs[leaf].res.rec == vb.rec) {
1.5 kristaps 511: if (0 == rs[leaf].matched)
1.17 kristaps 512: exprexec(expr, buf, vb.mask, &rs[leaf]);
1.1 schwarze 513: continue;
1.5 kristaps 514: }
1.1 schwarze 515:
516: /*
1.5 kristaps 517: * We have a new file to examine.
518: * Extract the manpage's metadata from the index
519: * database, then begin partial evaluation.
1.1 schwarze 520: */
521:
1.17 kristaps 522: key.data = &vb.rec;
1.1 schwarze 523: key.size = sizeof(recno_t);
524:
525: if (0 != (*idx->get)(idx, &key, &val, 0))
526: break;
527:
1.5 kristaps 528: r.lhs = r.rhs = -1;
1.11 kristaps 529: if ( ! index_read(&key, &val, vol, mc, &r))
1.1 schwarze 530: break;
531:
1.5 kristaps 532: /* XXX: this should be elsewhere, I guess? */
533:
534: if (opts->cat && strcasecmp(opts->cat, r.res.cat))
1.1 schwarze 535: continue;
1.5 kristaps 536: if (opts->arch && strcasecmp(opts->arch, r.res.arch))
1.1 schwarze 537: continue;
538:
1.8 kristaps 539: tree->node = rs = mandoc_realloc
540: (rs, (tree->len + 1) * sizeof(struct rec));
1.1 schwarze 541:
1.8 kristaps 542: memcpy(&rs[tree->len], &r, sizeof(struct rec));
1.12 schwarze 543: rs[tree->len].matches =
1.8 kristaps 544: mandoc_calloc(terms, sizeof(int));
1.1 schwarze 545:
1.17 kristaps 546: exprexec(expr, buf, vb.mask, &rs[tree->len]);
1.12 schwarze 547:
1.1 schwarze 548: /* Append to our tree. */
549:
550: if (leaf >= 0) {
1.17 kristaps 551: if (vb.rec > rs[leaf].res.rec)
1.8 kristaps 552: rs[leaf].rhs = tree->len;
1.1 schwarze 553: else
1.8 kristaps 554: rs[leaf].lhs = tree->len;
1.1 schwarze 555: } else
1.8 kristaps 556: root = tree->len;
1.12 schwarze 557:
1.5 kristaps 558: memset(&r, 0, sizeof(struct rec));
1.8 kristaps 559: tree->len++;
1.1 schwarze 560: }
1.12 schwarze 561:
1.8 kristaps 562: (*btree->close)(btree);
563: (*idx->close)(idx);
1.1 schwarze 564:
565: free(buf);
1.8 kristaps 566: return(1 == ch);
1.5 kristaps 567: }
568:
569: static void
570: recfree(struct rec *rec)
571: {
572:
1.17 kristaps 573: free(rec->res.type);
1.5 kristaps 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->subexpr = exprexpr(argc, argv, pos, lvl, tt);
698: if (NULL == next->subexpr) {
699: free(next);
700: next = NULL;
701: }
702: } else if (0 == strcmp("-i", argv[*pos])) {
703: if (++(*pos) >= argc)
704: goto err;
705: next = exprterm(argv[*pos], 0);
706: } else
707: next = exprterm(argv[*pos], 1);
708:
709: if (NULL == next)
710: goto err;
711:
712: next->and = log == 1;
713: next->index = (int)(*tt)++;
714:
715: /* Append to our chain of expressions. */
716:
717: if (NULL == first) {
718: assert(NULL == e);
719: first = next;
720: } else {
721: assert(NULL != e);
722: e->next = next;
723: }
724: }
725:
726: return(first);
727: err:
728: exprfree(first);
729: return(NULL);
730: }
731:
732: /*
733: * Parse a terminal expression with the grammar as defined in
734: * apropos(1).
735: * Return NULL if we fail the parse.
736: */
737: static struct expr *
738: exprterm(char *buf, int cs)
739: {
740: struct expr e;
1.1 schwarze 741: struct expr *p;
1.3 schwarze 742: char *key;
1.5 kristaps 743: int i;
744:
745: memset(&e, 0, sizeof(struct expr));
1.1 schwarze 746:
1.5 kristaps 747: /* Choose regex or substring match. */
1.3 schwarze 748:
1.4 kristaps 749: if (NULL == (e.v = strpbrk(buf, "=~"))) {
1.3 schwarze 750: e.regex = 0;
1.4 kristaps 751: e.v = buf;
1.3 schwarze 752: } else {
753: e.regex = '~' == *e.v;
754: *e.v++ = '\0';
755: }
1.1 schwarze 756:
1.5 kristaps 757: /* Determine the record types to search for. */
1.3 schwarze 758:
759: e.mask = 0;
1.4 kristaps 760: if (buf < e.v) {
761: while (NULL != (key = strsep(&buf, ","))) {
1.3 schwarze 762: i = 0;
763: while (types[i].mask &&
1.4 kristaps 764: strcmp(types[i].name, key))
1.3 schwarze 765: i++;
766: e.mask |= types[i].mask;
767: }
768: }
769: if (0 == e.mask)
770: e.mask = TYPE_Nm | TYPE_Nd;
1.1 schwarze 771:
1.5 kristaps 772: if (e.regex) {
1.13 kristaps 773: i = REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE);
1.5 kristaps 774: if (regcomp(&e.re, e.v, i))
775: return(NULL);
776: }
1.1 schwarze 777:
1.3 schwarze 778: e.v = mandoc_strdup(e.v);
1.1 schwarze 779:
780: p = mandoc_calloc(1, sizeof(struct expr));
781: memcpy(p, &e, sizeof(struct expr));
782: return(p);
783: }
784:
785: void
786: exprfree(struct expr *p)
787: {
1.5 kristaps 788: struct expr *pp;
1.12 schwarze 789:
1.5 kristaps 790: while (NULL != p) {
791: if (p->subexpr)
792: exprfree(p->subexpr);
793: if (p->regex)
794: regfree(&p->re);
795: free(p->v);
796: pp = p->next;
797: free(p);
798: p = pp;
799: }
800: }
1.1 schwarze 801:
1.5 kristaps 802: static int
1.12 schwarze 803: exprmark(const struct expr *p, const char *cp,
1.6 kristaps 804: uint64_t mask, int *ms)
1.5 kristaps 805: {
806:
807: for ( ; p; p = p->next) {
808: if (p->subexpr) {
809: if (exprmark(p->subexpr, cp, mask, ms))
810: return(1);
811: continue;
812: } else if ( ! (mask & p->mask))
813: continue;
1.1 schwarze 814:
1.5 kristaps 815: if (p->regex) {
816: if (regexec(&p->re, cp, 0, NULL, 0))
817: continue;
1.16 kristaps 818: } else if (NULL == strcasestr(cp, p->v))
819: continue;
1.5 kristaps 820:
821: if (NULL == ms)
822: return(1);
823: else
824: ms[p->index] = 1;
825: }
1.1 schwarze 826:
1.5 kristaps 827: return(0);
1.1 schwarze 828: }
829:
830: static int
1.5 kristaps 831: expreval(const struct expr *p, int *ms)
1.1 schwarze 832: {
1.5 kristaps 833: int match;
1.1 schwarze 834:
1.5 kristaps 835: /*
836: * AND has precedence over OR. Analysis is left-right, though
837: * it doesn't matter because there are no side-effects.
838: * Thus, step through pairwise ANDs and accumulate their Boolean
839: * evaluation. If we encounter a single true AND collection or
840: * standalone term, the whole expression is true (by definition
841: * of OR).
842: */
843:
844: for (match = 0; p && ! match; p = p->next) {
845: /* Evaluate a subexpression, if applicable. */
846: if (p->subexpr && ! ms[p->index])
847: ms[p->index] = expreval(p->subexpr, ms);
848:
849: match = ms[p->index];
850: for ( ; p->next && p->next->and; p = p->next) {
851: /* Evaluate a subexpression, if applicable. */
852: if (p->next->subexpr && ! ms[p->next->index])
1.12 schwarze 853: ms[p->next->index] =
1.5 kristaps 854: expreval(p->next->subexpr, ms);
855: match = match && ms[p->next->index];
856: }
857: }
858:
859: return(match);
860: }
861:
862: /*
863: * First, update the array of terms for which this expression evaluates
864: * to true.
865: * Second, logically evaluate all terms over the updated array of truth
866: * values.
867: * If this evaluates to true, mark the expression as satisfied.
868: */
869: static void
1.12 schwarze 870: exprexec(const struct expr *e, const char *cp,
1.6 kristaps 871: uint64_t mask, struct rec *r)
1.5 kristaps 872: {
1.1 schwarze 873:
1.5 kristaps 874: assert(0 == r->matched);
1.12 schwarze 875: exprmark(e, cp, mask, r->matches);
876: r->matched = expreval(e, r->matches);
1.1 schwarze 877: }
CVSweb