Annotation of mandoc/apropos_db.c, Revision 1.20
1.20 ! kristaps 1: /* $Id: apropos_db.c,v 1.19 2011/12/02 00:21:56 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: */
1.19 kristaps 18: #ifdef HAVE_CONFIG_H
19: #include "config.h"
20: #endif
21:
1.1 schwarze 22: #include <assert.h>
23: #include <fcntl.h>
24: #include <regex.h>
25: #include <stdarg.h>
1.6 kristaps 26: #include <stdint.h>
1.1 schwarze 27: #include <stdlib.h>
28: #include <string.h>
1.8 kristaps 29: #include <unistd.h>
1.1 schwarze 30:
1.19 kristaps 31: #if defined(__linux__)
32: # include <endian.h>
1.1 schwarze 33: # include <db_185.h>
1.19 kristaps 34: #elif defined(__APPLE__)
35: # include <libkern/OSByteOrder.h>
36: # include <db.h>
1.1 schwarze 37: #else
38: # include <db.h>
39: #endif
40:
1.2 schwarze 41: #include "mandocdb.h"
1.1 schwarze 42: #include "apropos_db.h"
43: #include "mandoc.h"
44:
1.5 kristaps 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:
1.1 schwarze 59: struct expr {
1.5 kristaps 60: int regex; /* is regex? */
61: int index; /* index in match array */
1.6 kristaps 62: uint64_t mask; /* type-mask */
1.5 kristaps 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;
1.1 schwarze 68: };
69:
70: struct type {
1.6 kristaps 71: uint64_t mask;
1.1 schwarze 72: const char *name;
73: };
74:
1.8 kristaps 75: struct rectree {
76: struct rec *node; /* record array for dir tree */
77: int len; /* length of record array */
78: };
79:
1.1 schwarze 80: static const struct type types[] = {
1.7 kristaps 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" },
1.2 schwarze 120: { INT_MAX, "any" },
1.1 schwarze 121: { 0, NULL }
122: };
123:
124: static DB *btree_open(void);
1.17 kristaps 125: static int btree_read(const DBT *, const DBT *,
126: const struct mchars *,
127: struct db_val *, char **);
1.5 kristaps 128: static int expreval(const struct expr *, int *);
1.12 schwarze 129: static void exprexec(const struct expr *,
1.6 kristaps 130: const char *, uint64_t, struct rec *);
1.12 schwarze 131: static int exprmark(const struct expr *,
1.6 kristaps 132: const char *, uint64_t, int *);
1.5 kristaps 133: static struct expr *exprexpr(int, char *[], int *, int *, size_t *);
134: static struct expr *exprterm(char *, int);
1.1 schwarze 135: static DB *index_open(void);
1.11 kristaps 136: static int index_read(const DBT *, const DBT *, int,
1.1 schwarze 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]);
1.5 kristaps 141: static void recfree(struct rec *);
1.8 kristaps 142: static int single_search(struct rectree *, const struct opts *,
143: const struct expr *, size_t terms,
1.11 kristaps 144: struct mchars *, int);
1.1 schwarze 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));
1.20 ! kristaps 156: info.lorder = 4321;
1.1 schwarze 157: info.flags = R_DUP;
158:
1.2 schwarze 159: db = dbopen(MANDOC_DB, O_RDONLY, 0, DB_BTREE, &info);
1.12 schwarze 160: if (NULL != db)
1.1 schwarze 161: return(db);
162:
163: return(NULL);
164: }
165:
166: /*
167: * Read a keyword from the database and normalise it.
168: * Return 0 if the database is insane, else 1.
169: */
170: static int
1.17 kristaps 171: btree_read(const DBT *k, const DBT *v,
172: const struct mchars *mc,
173: struct db_val *dbv, char **buf)
1.1 schwarze 174: {
1.17 kristaps 175: const struct db_val *vp;
1.1 schwarze 176:
1.17 kristaps 177: /* Are our sizes sane? */
178: if (k->size < 2 || sizeof(struct db_val) != v->size)
179: return(0);
1.6 kristaps 180:
1.17 kristaps 181: /* Is our string nil-terminated? */
182: if ('\0' != ((const char *)k->data)[(int)k->size - 1])
1.1 schwarze 183: return(0);
184:
1.17 kristaps 185: vp = v->data;
186: norm_string((const char *)k->data, mc, buf);
1.18 kristaps 187: dbv->rec = betoh32(vp->rec);
188: dbv->mask = betoh64(vp->mask);
1.1 schwarze 189: return(1);
190: }
191:
192: /*
193: * Take a Unicode codepoint and produce its UTF-8 encoding.
194: * This isn't the best way to do this, but it works.
1.12 schwarze 195: * The magic numbers are from the UTF-8 packaging.
1.1 schwarze 196: * They're not as scary as they seem: read the UTF-8 spec for details.
197: */
198: static size_t
199: norm_utf8(unsigned int cp, char out[7])
200: {
201: size_t rc;
202:
203: rc = 0;
204:
205: if (cp <= 0x0000007F) {
206: rc = 1;
207: out[0] = (char)cp;
208: } else if (cp <= 0x000007FF) {
209: rc = 2;
210: out[0] = (cp >> 6 & 31) | 192;
211: out[1] = (cp & 63) | 128;
212: } else if (cp <= 0x0000FFFF) {
213: rc = 3;
214: out[0] = (cp >> 12 & 15) | 224;
215: out[1] = (cp >> 6 & 63) | 128;
216: out[2] = (cp & 63) | 128;
217: } else if (cp <= 0x001FFFFF) {
218: rc = 4;
219: out[0] = (cp >> 18 & 7) | 240;
220: out[1] = (cp >> 12 & 63) | 128;
221: out[2] = (cp >> 6 & 63) | 128;
222: out[3] = (cp & 63) | 128;
223: } else if (cp <= 0x03FFFFFF) {
224: rc = 5;
225: out[0] = (cp >> 24 & 3) | 248;
226: out[1] = (cp >> 18 & 63) | 128;
227: out[2] = (cp >> 12 & 63) | 128;
228: out[3] = (cp >> 6 & 63) | 128;
229: out[4] = (cp & 63) | 128;
230: } else if (cp <= 0x7FFFFFFF) {
231: rc = 6;
232: out[0] = (cp >> 30 & 1) | 252;
233: out[1] = (cp >> 24 & 63) | 128;
234: out[2] = (cp >> 18 & 63) | 128;
235: out[3] = (cp >> 12 & 63) | 128;
236: out[4] = (cp >> 6 & 63) | 128;
237: out[5] = (cp & 63) | 128;
238: } else
239: return(0);
240:
241: out[rc] = '\0';
242: return(rc);
243: }
244:
245: /*
246: * Normalise strings from the index and database.
247: * These strings are escaped as defined by mandoc_char(7) along with
248: * other goop in mandoc.h (e.g., soft hyphens).
249: * This function normalises these into a nice UTF-8 string.
250: * Returns 0 if the database is fucked.
251: */
252: static void
253: norm_string(const char *val, const struct mchars *mc, char **buf)
254: {
255: size_t sz, bsz;
256: char utfbuf[7];
257: const char *seq, *cpp;
258: int len, u, pos;
259: enum mandoc_esc esc;
1.12 schwarze 260: static const char res[] = { '\\', '\t',
1.1 schwarze 261: ASCII_NBRSP, ASCII_HYPH, '\0' };
262:
263: /* Pre-allocate by the length of the input */
264:
265: bsz = strlen(val) + 1;
266: *buf = mandoc_realloc(*buf, bsz);
267: pos = 0;
268:
269: while ('\0' != *val) {
270: /*
271: * Halt on the first escape sequence.
272: * This also halts on the end of string, in which case
273: * we just copy, fallthrough, and exit the loop.
274: */
275: if ((sz = strcspn(val, res)) > 0) {
276: memcpy(&(*buf)[pos], val, sz);
277: pos += (int)sz;
278: val += (int)sz;
279: }
280:
281: if (ASCII_HYPH == *val) {
282: (*buf)[pos++] = '-';
283: val++;
284: continue;
285: } else if ('\t' == *val || ASCII_NBRSP == *val) {
286: (*buf)[pos++] = ' ';
287: val++;
288: continue;
289: } else if ('\\' != *val)
290: break;
291:
292: /* Read past the slash. */
293:
294: val++;
295: u = 0;
296:
297: /*
298: * Parse the escape sequence and see if it's a
299: * predefined character or special character.
300: */
301:
302: esc = mandoc_escape(&val, &seq, &len);
303: if (ESCAPE_ERROR == esc)
304: break;
305:
1.12 schwarze 306: /*
1.1 schwarze 307: * XXX - this just does UTF-8, but we need to know
308: * beforehand whether we should do text substitution.
309: */
310:
311: switch (esc) {
312: case (ESCAPE_SPECIAL):
313: if (0 != (u = mchars_spec2cp(mc, seq, len)))
314: break;
315: /* FALLTHROUGH */
316: default:
317: continue;
318: }
319:
320: /*
321: * If we have a Unicode codepoint, try to convert that
322: * to a UTF-8 byte string.
323: */
324:
325: cpp = utfbuf;
326: if (0 == (sz = norm_utf8(u, utfbuf)))
327: continue;
328:
329: /* Copy the rendered glyph into the stream. */
330:
331: sz = strlen(cpp);
332: bsz += sz;
333:
334: *buf = mandoc_realloc(*buf, bsz);
335:
336: memcpy(&(*buf)[pos], cpp, sz);
337: pos += (int)sz;
338: }
339:
340: (*buf)[pos] = '\0';
341: }
342:
343: /*
344: * Open the filename-index mandoc-db database.
345: * Returns NULL if opening failed.
346: */
347: static DB *
348: index_open(void)
349: {
350: DB *db;
351:
1.2 schwarze 352: db = dbopen(MANDOC_IDX, O_RDONLY, 0, DB_RECNO, NULL);
1.1 schwarze 353: if (NULL != db)
354: return(db);
355:
356: return(NULL);
357: }
358:
359: /*
360: * Safely unpack from an index file record into the structure.
361: * Returns 1 if an entry was unpacked, 0 if the database is insane.
362: */
363: static int
1.11 kristaps 364: index_read(const DBT *key, const DBT *val, int index,
1.1 schwarze 365: const struct mchars *mc, struct rec *rec)
366: {
367: size_t left;
368: char *np, *cp;
369:
370: #define INDEX_BREAD(_dst) \
371: do { \
372: if (NULL == (np = memchr(cp, '\0', left))) \
373: return(0); \
374: norm_string(cp, mc, &(_dst)); \
375: left -= (np - cp) + 1; \
376: cp = np + 1; \
377: } while (/* CONSTCOND */ 0)
378:
379: left = val->size;
380: cp = (char *)val->data;
381:
1.5 kristaps 382: rec->res.rec = *(recno_t *)key->data;
1.11 kristaps 383: rec->res.volume = index;
1.1 schwarze 384:
1.14 schwarze 385: INDEX_BREAD(rec->res.type);
1.5 kristaps 386: INDEX_BREAD(rec->res.file);
387: INDEX_BREAD(rec->res.cat);
388: INDEX_BREAD(rec->res.title);
389: INDEX_BREAD(rec->res.arch);
390: INDEX_BREAD(rec->res.desc);
1.1 schwarze 391: return(1);
392: }
393:
394: /*
1.10 kristaps 395: * Search mandocdb databases in paths for expression "expr".
1.1 schwarze 396: * Filter out by "opts".
397: * Call "res" with the results, which may be zero.
1.5 kristaps 398: * Return 0 if there was a database error, else return 1.
1.1 schwarze 399: */
1.5 kristaps 400: int
1.10 kristaps 401: apropos_search(int pathsz, char **paths, const struct opts *opts,
1.12 schwarze 402: const struct expr *expr, size_t terms, void *arg,
1.5 kristaps 403: void (*res)(struct res *, size_t, void *))
1.1 schwarze 404: {
1.8 kristaps 405: struct rectree tree;
406: struct mchars *mc;
407: struct res *ress;
408: int i, mlen, rc;
409:
410: memset(&tree, 0, sizeof(struct rectree));
411:
1.10 kristaps 412: rc = 0;
1.8 kristaps 413: mc = mchars_alloc();
414:
1.10 kristaps 415: /*
416: * Main loop. Change into the directory containing manpage
417: * databases. Run our expession over each database in the set.
418: */
419:
420: for (i = 0; i < pathsz; i++) {
421: if (chdir(paths[i]))
1.8 kristaps 422: continue;
1.11 kristaps 423: if ( ! single_search(&tree, opts, expr, terms, mc, i))
1.10 kristaps 424: goto out;
1.8 kristaps 425: }
426:
427: /*
1.10 kristaps 428: * Count matching files, transfer to a "clean" array, then feed
429: * them to the output handler.
1.8 kristaps 430: */
431:
432: for (mlen = i = 0; i < tree.len; i++)
433: if (tree.node[i].matched)
434: mlen++;
435:
436: ress = mandoc_malloc(mlen * sizeof(struct res));
437:
438: for (mlen = i = 0; i < tree.len; i++)
439: if (tree.node[i].matched)
1.12 schwarze 440: memcpy(&ress[mlen++], &tree.node[i].res,
1.8 kristaps 441: sizeof(struct res));
442:
443: (*res)(ress, mlen, arg);
444: free(ress);
445:
1.10 kristaps 446: rc = 1;
447: out:
1.8 kristaps 448: for (i = 0; i < tree.len; i++)
449: recfree(&tree.node[i]);
450:
451: free(tree.node);
452: mchars_free(mc);
453: return(rc);
454: }
455:
456: static int
457: single_search(struct rectree *tree, const struct opts *opts,
458: const struct expr *expr, size_t terms,
1.11 kristaps 459: struct mchars *mc, int vol)
1.8 kristaps 460: {
461: int root, leaf, ch;
1.1 schwarze 462: DBT key, val;
463: DB *btree, *idx;
464: char *buf;
1.5 kristaps 465: struct rec *rs;
466: struct rec r;
1.17 kristaps 467: struct db_val vb;
1.1 schwarze 468:
469: root = -1;
470: leaf = -1;
471: btree = NULL;
472: idx = NULL;
473: buf = NULL;
1.8 kristaps 474: rs = tree->node;
1.1 schwarze 475:
1.5 kristaps 476: memset(&r, 0, sizeof(struct rec));
1.1 schwarze 477:
1.12 schwarze 478: if (NULL == (btree = btree_open()))
1.10 kristaps 479: return(1);
1.1 schwarze 480:
1.8 kristaps 481: if (NULL == (idx = index_open())) {
482: (*btree->close)(btree);
1.10 kristaps 483: return(1);
1.8 kristaps 484: }
1.1 schwarze 485:
486: while (0 == (ch = (*btree->seq)(btree, &key, &val, R_NEXT))) {
1.17 kristaps 487: if ( ! btree_read(&key, &val, mc, &vb, &buf))
1.1 schwarze 488: break;
489:
1.5 kristaps 490: /*
491: * See if this keyword record matches any of the
492: * expressions we have stored.
493: */
1.17 kristaps 494: if ( ! exprmark(expr, buf, vb.mask, NULL))
1.1 schwarze 495: continue;
496:
497: /*
498: * O(log n) scan for prior records. Since a record
499: * number is unbounded, this has decent performance over
500: * a complex hash function.
501: */
502:
503: for (leaf = root; leaf >= 0; )
1.17 kristaps 504: if (vb.rec > rs[leaf].res.rec &&
1.5 kristaps 505: rs[leaf].rhs >= 0)
506: leaf = rs[leaf].rhs;
1.17 kristaps 507: else if (vb.rec < rs[leaf].res.rec &&
1.5 kristaps 508: rs[leaf].lhs >= 0)
509: leaf = rs[leaf].lhs;
1.12 schwarze 510: else
1.1 schwarze 511: break;
512:
1.5 kristaps 513: /*
514: * If we find a record, see if it has already evaluated
515: * to true. If it has, great, just keep going. If not,
516: * try to evaluate it now and continue anyway.
517: */
518:
1.17 kristaps 519: if (leaf >= 0 && rs[leaf].res.rec == vb.rec) {
1.5 kristaps 520: if (0 == rs[leaf].matched)
1.17 kristaps 521: exprexec(expr, buf, vb.mask, &rs[leaf]);
1.1 schwarze 522: continue;
1.5 kristaps 523: }
1.1 schwarze 524:
525: /*
1.5 kristaps 526: * We have a new file to examine.
527: * Extract the manpage's metadata from the index
528: * database, then begin partial evaluation.
1.1 schwarze 529: */
530:
1.17 kristaps 531: key.data = &vb.rec;
1.1 schwarze 532: key.size = sizeof(recno_t);
533:
534: if (0 != (*idx->get)(idx, &key, &val, 0))
535: break;
536:
1.5 kristaps 537: r.lhs = r.rhs = -1;
1.11 kristaps 538: if ( ! index_read(&key, &val, vol, mc, &r))
1.1 schwarze 539: break;
540:
1.5 kristaps 541: /* XXX: this should be elsewhere, I guess? */
542:
543: if (opts->cat && strcasecmp(opts->cat, r.res.cat))
1.1 schwarze 544: continue;
1.5 kristaps 545: if (opts->arch && strcasecmp(opts->arch, r.res.arch))
1.1 schwarze 546: continue;
547:
1.8 kristaps 548: tree->node = rs = mandoc_realloc
549: (rs, (tree->len + 1) * sizeof(struct rec));
1.1 schwarze 550:
1.8 kristaps 551: memcpy(&rs[tree->len], &r, sizeof(struct rec));
1.12 schwarze 552: rs[tree->len].matches =
1.8 kristaps 553: mandoc_calloc(terms, sizeof(int));
1.1 schwarze 554:
1.17 kristaps 555: exprexec(expr, buf, vb.mask, &rs[tree->len]);
1.12 schwarze 556:
1.1 schwarze 557: /* Append to our tree. */
558:
559: if (leaf >= 0) {
1.17 kristaps 560: if (vb.rec > rs[leaf].res.rec)
1.8 kristaps 561: rs[leaf].rhs = tree->len;
1.1 schwarze 562: else
1.8 kristaps 563: rs[leaf].lhs = tree->len;
1.1 schwarze 564: } else
1.8 kristaps 565: root = tree->len;
1.12 schwarze 566:
1.5 kristaps 567: memset(&r, 0, sizeof(struct rec));
1.8 kristaps 568: tree->len++;
1.1 schwarze 569: }
1.12 schwarze 570:
1.8 kristaps 571: (*btree->close)(btree);
572: (*idx->close)(idx);
1.1 schwarze 573:
574: free(buf);
1.8 kristaps 575: return(1 == ch);
1.5 kristaps 576: }
577:
578: static void
579: recfree(struct rec *rec)
580: {
581:
1.17 kristaps 582: free(rec->res.type);
1.5 kristaps 583: free(rec->res.file);
584: free(rec->res.cat);
585: free(rec->res.title);
586: free(rec->res.arch);
587: free(rec->res.desc);
588:
589: free(rec->matches);
1.1 schwarze 590: }
591:
1.13 kristaps 592: /*
593: * Compile a list of straight-up terms.
594: * The arguments are re-written into ~[[:<:]]term[[:>:]], or "term"
595: * surrounded by word boundaries, then pumped through exprterm().
596: * Terms are case-insensitive.
597: * This emulates whatis(1) behaviour.
598: */
599: struct expr *
600: termcomp(int argc, char *argv[], size_t *tt)
601: {
602: char *buf;
603: int pos;
604: struct expr *e, *next;
605: size_t sz;
606:
607: buf = NULL;
608: e = NULL;
609: *tt = 0;
610:
1.15 schwarze 611: for (pos = argc - 1; pos >= 0; pos--) {
612: sz = strlen(argv[pos]) + 18;
1.13 kristaps 613: buf = mandoc_realloc(buf, sz);
1.15 schwarze 614: strlcpy(buf, "Nm~[[:<:]]", sz);
1.13 kristaps 615: strlcat(buf, argv[pos], sz);
616: strlcat(buf, "[[:>:]]", sz);
617: if (NULL == (next = exprterm(buf, 0))) {
618: free(buf);
619: exprfree(e);
620: return(NULL);
621: }
1.15 schwarze 622: next->next = e;
1.13 kristaps 623: e = next;
624: (*tt)++;
625: }
626:
627: free(buf);
628: return(e);
629: }
630:
631: /*
632: * Compile a sequence of logical expressions.
633: * See apropos.1 for a grammar of this sequence.
634: */
1.1 schwarze 635: struct expr *
1.5 kristaps 636: exprcomp(int argc, char *argv[], size_t *tt)
1.1 schwarze 637: {
1.5 kristaps 638: int pos, lvl;
639: struct expr *e;
640:
641: pos = lvl = 0;
642: *tt = 0;
643:
644: e = exprexpr(argc, argv, &pos, &lvl, tt);
645:
646: if (0 == lvl && pos >= argc)
647: return(e);
648:
649: exprfree(e);
650: return(NULL);
651: }
652:
653: /*
654: * Compile an array of tokens into an expression.
655: * An informal expression grammar is defined in apropos(1).
656: * Return NULL if we fail doing so. All memory will be cleaned up.
657: * Return the root of the expression sequence if alright.
658: */
659: static struct expr *
1.9 kristaps 660: exprexpr(int argc, char *argv[], int *pos, int *lvl, size_t *tt)
1.5 kristaps 661: {
662: struct expr *e, *first, *next;
663: int log;
664:
665: first = next = NULL;
666:
667: for ( ; *pos < argc; (*pos)++) {
668: e = next;
669:
670: /*
671: * Close out a subexpression.
672: */
673:
674: if (NULL != e && 0 == strcmp(")", argv[*pos])) {
675: if (--(*lvl) < 0)
676: goto err;
677: break;
678: }
679:
680: /*
681: * Small note: if we're just starting, don't let "-a"
682: * and "-o" be considered logical operators: they're
683: * just tokens unless pairwise joining, in which case we
684: * record their existence (or assume "OR").
685: */
686: log = 0;
687:
688: if (NULL != e && 0 == strcmp("-a", argv[*pos]))
1.12 schwarze 689: log = 1;
1.5 kristaps 690: else if (NULL != e && 0 == strcmp("-o", argv[*pos]))
691: log = 2;
692:
693: if (log > 0 && ++(*pos) >= argc)
694: goto err;
695:
696: /*
697: * Now we parse the term part. This can begin with
698: * "-i", in which case the expression is case
699: * insensitive.
700: */
701:
702: if (0 == strcmp("(", argv[*pos])) {
703: ++(*pos);
704: ++(*lvl);
705: next = mandoc_calloc(1, sizeof(struct expr));
706: next->subexpr = exprexpr(argc, argv, pos, lvl, tt);
707: if (NULL == next->subexpr) {
708: free(next);
709: next = NULL;
710: }
711: } else if (0 == strcmp("-i", argv[*pos])) {
712: if (++(*pos) >= argc)
713: goto err;
714: next = exprterm(argv[*pos], 0);
715: } else
716: next = exprterm(argv[*pos], 1);
717:
718: if (NULL == next)
719: goto err;
720:
721: next->and = log == 1;
722: next->index = (int)(*tt)++;
723:
724: /* Append to our chain of expressions. */
725:
726: if (NULL == first) {
727: assert(NULL == e);
728: first = next;
729: } else {
730: assert(NULL != e);
731: e->next = next;
732: }
733: }
734:
735: return(first);
736: err:
737: exprfree(first);
738: return(NULL);
739: }
740:
741: /*
742: * Parse a terminal expression with the grammar as defined in
743: * apropos(1).
744: * Return NULL if we fail the parse.
745: */
746: static struct expr *
747: exprterm(char *buf, int cs)
748: {
749: struct expr e;
1.1 schwarze 750: struct expr *p;
1.3 schwarze 751: char *key;
1.5 kristaps 752: int i;
753:
754: memset(&e, 0, sizeof(struct expr));
1.1 schwarze 755:
1.5 kristaps 756: /* Choose regex or substring match. */
1.3 schwarze 757:
1.4 kristaps 758: if (NULL == (e.v = strpbrk(buf, "=~"))) {
1.3 schwarze 759: e.regex = 0;
1.4 kristaps 760: e.v = buf;
1.3 schwarze 761: } else {
762: e.regex = '~' == *e.v;
763: *e.v++ = '\0';
764: }
1.1 schwarze 765:
1.5 kristaps 766: /* Determine the record types to search for. */
1.3 schwarze 767:
768: e.mask = 0;
1.4 kristaps 769: if (buf < e.v) {
770: while (NULL != (key = strsep(&buf, ","))) {
1.3 schwarze 771: i = 0;
772: while (types[i].mask &&
1.4 kristaps 773: strcmp(types[i].name, key))
1.3 schwarze 774: i++;
775: e.mask |= types[i].mask;
776: }
777: }
778: if (0 == e.mask)
779: e.mask = TYPE_Nm | TYPE_Nd;
1.1 schwarze 780:
1.5 kristaps 781: if (e.regex) {
1.13 kristaps 782: i = REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE);
1.5 kristaps 783: if (regcomp(&e.re, e.v, i))
784: return(NULL);
785: }
1.1 schwarze 786:
1.3 schwarze 787: e.v = mandoc_strdup(e.v);
1.1 schwarze 788:
789: p = mandoc_calloc(1, sizeof(struct expr));
790: memcpy(p, &e, sizeof(struct expr));
791: return(p);
792: }
793:
794: void
795: exprfree(struct expr *p)
796: {
1.5 kristaps 797: struct expr *pp;
1.12 schwarze 798:
1.5 kristaps 799: while (NULL != p) {
800: if (p->subexpr)
801: exprfree(p->subexpr);
802: if (p->regex)
803: regfree(&p->re);
804: free(p->v);
805: pp = p->next;
806: free(p);
807: p = pp;
808: }
809: }
1.1 schwarze 810:
1.5 kristaps 811: static int
1.12 schwarze 812: exprmark(const struct expr *p, const char *cp,
1.6 kristaps 813: uint64_t mask, int *ms)
1.5 kristaps 814: {
815:
816: for ( ; p; p = p->next) {
817: if (p->subexpr) {
818: if (exprmark(p->subexpr, cp, mask, ms))
819: return(1);
820: continue;
821: } else if ( ! (mask & p->mask))
822: continue;
1.1 schwarze 823:
1.5 kristaps 824: if (p->regex) {
825: if (regexec(&p->re, cp, 0, NULL, 0))
826: continue;
1.16 kristaps 827: } else if (NULL == strcasestr(cp, p->v))
828: continue;
1.5 kristaps 829:
830: if (NULL == ms)
831: return(1);
832: else
833: ms[p->index] = 1;
834: }
1.1 schwarze 835:
1.5 kristaps 836: return(0);
1.1 schwarze 837: }
838:
839: static int
1.5 kristaps 840: expreval(const struct expr *p, int *ms)
1.1 schwarze 841: {
1.5 kristaps 842: int match;
1.1 schwarze 843:
1.5 kristaps 844: /*
845: * AND has precedence over OR. Analysis is left-right, though
846: * it doesn't matter because there are no side-effects.
847: * Thus, step through pairwise ANDs and accumulate their Boolean
848: * evaluation. If we encounter a single true AND collection or
849: * standalone term, the whole expression is true (by definition
850: * of OR).
851: */
852:
853: for (match = 0; p && ! match; p = p->next) {
854: /* Evaluate a subexpression, if applicable. */
855: if (p->subexpr && ! ms[p->index])
856: ms[p->index] = expreval(p->subexpr, ms);
857:
858: match = ms[p->index];
859: for ( ; p->next && p->next->and; p = p->next) {
860: /* Evaluate a subexpression, if applicable. */
861: if (p->next->subexpr && ! ms[p->next->index])
1.12 schwarze 862: ms[p->next->index] =
1.5 kristaps 863: expreval(p->next->subexpr, ms);
864: match = match && ms[p->next->index];
865: }
866: }
867:
868: return(match);
869: }
870:
871: /*
872: * First, update the array of terms for which this expression evaluates
873: * to true.
874: * Second, logically evaluate all terms over the updated array of truth
875: * values.
876: * If this evaluates to true, mark the expression as satisfied.
877: */
878: static void
1.12 schwarze 879: exprexec(const struct expr *e, const char *cp,
1.6 kristaps 880: uint64_t mask, struct rec *r)
1.5 kristaps 881: {
1.1 schwarze 882:
1.5 kristaps 883: assert(0 == r->matched);
1.12 schwarze 884: exprmark(e, cp, mask, r->matches);
885: r->matched = expreval(e, r->matches);
1.1 schwarze 886: }
CVSweb