Annotation of mandoc/apropos_db.c, Revision 1.32.2.6
1.32.2.6! schwarze 1: /* $Id: apropos_db.c,v 1.32.2.5 2014/03/23 12:04:54 schwarze Exp $ */
1.1 schwarze 2: /*
1.31 kristaps 3: * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.32.2.4 schwarze 4: * Copyright (c) 2011, 2014 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:
1.32 kristaps 22: #include <sys/param.h>
23:
1.1 schwarze 24: #include <assert.h>
25: #include <fcntl.h>
26: #include <regex.h>
27: #include <stdarg.h>
1.6 kristaps 28: #include <stdint.h>
1.1 schwarze 29: #include <stdlib.h>
30: #include <string.h>
1.8 kristaps 31: #include <unistd.h>
1.1 schwarze 32:
1.32.2.3 schwarze 33: #if defined(__APPLE__)
1.19 kristaps 34: # include <libkern/OSByteOrder.h>
1.32.2.3 schwarze 35: #elif defined(__linux__)
36: # include <endian.h>
37: #elif defined(__sun)
38: # include <sys/byteorder.h>
1.1 schwarze 39: #else
1.32.2.1 schwarze 40: # include <sys/endian.h>
1.32.2.3 schwarze 41: #endif
42:
43: #if defined(__linux__) || defined(__sun)
44: # include <db_185.h>
45: #else
1.1 schwarze 46: # include <db.h>
47: #endif
48:
1.2 schwarze 49: #include "mandocdb.h"
1.1 schwarze 50: #include "apropos_db.h"
51: #include "mandoc.h"
1.32.2.5 schwarze 52: #include "mandoc_aux.h"
1.1 schwarze 53:
1.30 kristaps 54: #define RESFREE(_x) \
55: do { \
56: free((_x)->file); \
57: free((_x)->cat); \
58: free((_x)->title); \
59: free((_x)->arch); \
60: free((_x)->desc); \
61: free((_x)->matches); \
62: } while (/*CONSTCOND*/0)
1.5 kristaps 63:
1.1 schwarze 64: struct expr {
1.5 kristaps 65: int regex; /* is regex? */
66: int index; /* index in match array */
1.6 kristaps 67: uint64_t mask; /* type-mask */
1.5 kristaps 68: int and; /* is rhs of logical AND? */
69: char *v; /* search value */
70: regex_t re; /* compiled re, if regex */
71: struct expr *next; /* next in sequence */
72: struct expr *subexpr;
1.1 schwarze 73: };
74:
75: struct type {
1.6 kristaps 76: uint64_t mask;
1.1 schwarze 77: const char *name;
78: };
79:
1.8 kristaps 80: struct rectree {
1.30 kristaps 81: struct res *node; /* record array for dir tree */
1.8 kristaps 82: int len; /* length of record array */
83: };
84:
1.1 schwarze 85: static const struct type types[] = {
1.7 kristaps 86: { TYPE_An, "An" },
87: { TYPE_Ar, "Ar" },
88: { TYPE_At, "At" },
89: { TYPE_Bsx, "Bsx" },
90: { TYPE_Bx, "Bx" },
91: { TYPE_Cd, "Cd" },
92: { TYPE_Cm, "Cm" },
93: { TYPE_Dv, "Dv" },
94: { TYPE_Dx, "Dx" },
95: { TYPE_Em, "Em" },
96: { TYPE_Er, "Er" },
97: { TYPE_Ev, "Ev" },
98: { TYPE_Fa, "Fa" },
99: { TYPE_Fl, "Fl" },
100: { TYPE_Fn, "Fn" },
101: { TYPE_Fn, "Fo" },
102: { TYPE_Ft, "Ft" },
103: { TYPE_Fx, "Fx" },
104: { TYPE_Ic, "Ic" },
105: { TYPE_In, "In" },
106: { TYPE_Lb, "Lb" },
107: { TYPE_Li, "Li" },
108: { TYPE_Lk, "Lk" },
109: { TYPE_Ms, "Ms" },
110: { TYPE_Mt, "Mt" },
111: { TYPE_Nd, "Nd" },
112: { TYPE_Nm, "Nm" },
113: { TYPE_Nx, "Nx" },
114: { TYPE_Ox, "Ox" },
115: { TYPE_Pa, "Pa" },
116: { TYPE_Rs, "Rs" },
117: { TYPE_Sh, "Sh" },
118: { TYPE_Ss, "Ss" },
119: { TYPE_St, "St" },
120: { TYPE_Sy, "Sy" },
121: { TYPE_Tn, "Tn" },
122: { TYPE_Va, "Va" },
123: { TYPE_Va, "Vt" },
124: { TYPE_Xr, "Xr" },
1.25 kristaps 125: { UINT64_MAX, "any" },
1.1 schwarze 126: { 0, NULL }
127: };
128:
129: static DB *btree_open(void);
1.17 kristaps 130: static int btree_read(const DBT *, const DBT *,
1.28 schwarze 131: const struct mchars *,
132: uint64_t *, recno_t *, char **);
1.5 kristaps 133: static int expreval(const struct expr *, int *);
1.12 schwarze 134: static void exprexec(const struct expr *,
1.30 kristaps 135: const char *, uint64_t, struct res *);
1.12 schwarze 136: static int exprmark(const struct expr *,
1.6 kristaps 137: const char *, uint64_t, int *);
1.5 kristaps 138: static struct expr *exprexpr(int, char *[], int *, int *, size_t *);
139: static struct expr *exprterm(char *, int);
1.1 schwarze 140: static DB *index_open(void);
1.11 kristaps 141: static int index_read(const DBT *, const DBT *, int,
1.30 kristaps 142: const struct mchars *, struct res *);
1.1 schwarze 143: static void norm_string(const char *,
144: const struct mchars *, char **);
145: static size_t norm_utf8(unsigned int, char[7]);
1.8 kristaps 146: static int single_search(struct rectree *, const struct opts *,
147: const struct expr *, size_t terms,
1.11 kristaps 148: struct mchars *, int);
1.1 schwarze 149:
150: /*
151: * Open the keyword mandoc-db database.
152: */
153: static DB *
154: btree_open(void)
155: {
156: BTREEINFO info;
157: DB *db;
158:
159: memset(&info, 0, sizeof(BTREEINFO));
1.29 kristaps 160: info.lorder = 4321;
1.1 schwarze 161: info.flags = R_DUP;
162:
1.2 schwarze 163: db = dbopen(MANDOC_DB, O_RDONLY, 0, DB_BTREE, &info);
1.12 schwarze 164: if (NULL != db)
1.1 schwarze 165: return(db);
166:
167: return(NULL);
168: }
169:
170: /*
171: * Read a keyword from the database and normalise it.
172: * Return 0 if the database is insane, else 1.
173: */
174: static int
1.28 schwarze 175: btree_read(const DBT *k, const DBT *v, const struct mchars *mc,
176: uint64_t *mask, recno_t *rec, char **buf)
1.1 schwarze 177: {
1.28 schwarze 178: uint64_t vbuf[2];
1.1 schwarze 179:
1.17 kristaps 180: /* Are our sizes sane? */
1.28 schwarze 181: if (k->size < 2 || sizeof(vbuf) != v->size)
1.17 kristaps 182: return(0);
1.6 kristaps 183:
1.17 kristaps 184: /* Is our string nil-terminated? */
185: if ('\0' != ((const char *)k->data)[(int)k->size - 1])
1.1 schwarze 186: return(0);
187:
1.17 kristaps 188: norm_string((const char *)k->data, mc, buf);
1.28 schwarze 189: memcpy(vbuf, v->data, v->size);
190: *mask = betoh64(vbuf[0]);
191: *rec = betoh64(vbuf[1]);
1.1 schwarze 192: return(1);
193: }
194:
195: /*
196: * Take a Unicode codepoint and produce its UTF-8 encoding.
197: * This isn't the best way to do this, but it works.
1.12 schwarze 198: * The magic numbers are from the UTF-8 packaging.
1.1 schwarze 199: * They're not as scary as they seem: read the UTF-8 spec for details.
200: */
201: static size_t
202: norm_utf8(unsigned int cp, char out[7])
203: {
1.26 kristaps 204: int rc;
1.1 schwarze 205:
206: rc = 0;
207:
208: if (cp <= 0x0000007F) {
209: rc = 1;
210: out[0] = (char)cp;
211: } else if (cp <= 0x000007FF) {
212: rc = 2;
213: out[0] = (cp >> 6 & 31) | 192;
214: out[1] = (cp & 63) | 128;
215: } else if (cp <= 0x0000FFFF) {
216: rc = 3;
217: out[0] = (cp >> 12 & 15) | 224;
218: out[1] = (cp >> 6 & 63) | 128;
219: out[2] = (cp & 63) | 128;
220: } else if (cp <= 0x001FFFFF) {
221: rc = 4;
222: out[0] = (cp >> 18 & 7) | 240;
223: out[1] = (cp >> 12 & 63) | 128;
224: out[2] = (cp >> 6 & 63) | 128;
225: out[3] = (cp & 63) | 128;
226: } else if (cp <= 0x03FFFFFF) {
227: rc = 5;
228: out[0] = (cp >> 24 & 3) | 248;
229: out[1] = (cp >> 18 & 63) | 128;
230: out[2] = (cp >> 12 & 63) | 128;
231: out[3] = (cp >> 6 & 63) | 128;
232: out[4] = (cp & 63) | 128;
233: } else if (cp <= 0x7FFFFFFF) {
234: rc = 6;
235: out[0] = (cp >> 30 & 1) | 252;
236: out[1] = (cp >> 24 & 63) | 128;
237: out[2] = (cp >> 18 & 63) | 128;
238: out[3] = (cp >> 12 & 63) | 128;
239: out[4] = (cp >> 6 & 63) | 128;
240: out[5] = (cp & 63) | 128;
241: } else
242: return(0);
243:
244: out[rc] = '\0';
1.26 kristaps 245: return((size_t)rc);
1.1 schwarze 246: }
247:
248: /*
249: * Normalise strings from the index and database.
250: * These strings are escaped as defined by mandoc_char(7) along with
251: * other goop in mandoc.h (e.g., soft hyphens).
252: * This function normalises these into a nice UTF-8 string.
253: * Returns 0 if the database is fucked.
254: */
255: static void
256: norm_string(const char *val, const struct mchars *mc, char **buf)
257: {
258: size_t sz, bsz;
259: char utfbuf[7];
260: const char *seq, *cpp;
261: int len, u, pos;
262: enum mandoc_esc esc;
1.32.2.4 schwarze 263: static const char res[] = { '\\', '\t', ASCII_NBRSP,
264: ASCII_HYPH, ASCII_BREAK, '\0' };
1.1 schwarze 265:
266: /* Pre-allocate by the length of the input */
267:
268: bsz = strlen(val) + 1;
269: *buf = mandoc_realloc(*buf, bsz);
270: pos = 0;
271:
272: while ('\0' != *val) {
273: /*
274: * Halt on the first escape sequence.
275: * This also halts on the end of string, in which case
276: * we just copy, fallthrough, and exit the loop.
277: */
278: if ((sz = strcspn(val, res)) > 0) {
279: memcpy(&(*buf)[pos], val, sz);
280: pos += (int)sz;
281: val += (int)sz;
282: }
283:
1.32.2.4 schwarze 284: switch (*val) {
285: case (ASCII_HYPH):
1.1 schwarze 286: (*buf)[pos++] = '-';
287: val++;
288: continue;
1.32.2.4 schwarze 289: case ('\t'):
290: /* FALLTHROUGH */
291: case (ASCII_NBRSP):
1.1 schwarze 292: (*buf)[pos++] = ' ';
293: val++;
1.32.2.4 schwarze 294: /* FALLTHROUGH */
295: case (ASCII_BREAK):
1.1 schwarze 296: continue;
1.32.2.4 schwarze 297: default:
298: break;
299: }
300: if ('\\' != *val)
1.1 schwarze 301: break;
302:
303: /* Read past the slash. */
304:
305: val++;
306: u = 0;
307:
308: /*
309: * Parse the escape sequence and see if it's a
310: * predefined character or special character.
311: */
312:
313: esc = mandoc_escape(&val, &seq, &len);
314: if (ESCAPE_ERROR == esc)
315: break;
316:
1.12 schwarze 317: /*
1.1 schwarze 318: * XXX - this just does UTF-8, but we need to know
319: * beforehand whether we should do text substitution.
320: */
321:
322: switch (esc) {
323: case (ESCAPE_SPECIAL):
324: if (0 != (u = mchars_spec2cp(mc, seq, len)))
325: break;
326: /* FALLTHROUGH */
327: default:
328: continue;
329: }
330:
331: /*
332: * If we have a Unicode codepoint, try to convert that
333: * to a UTF-8 byte string.
334: */
335:
336: cpp = utfbuf;
337: if (0 == (sz = norm_utf8(u, utfbuf)))
338: continue;
339:
340: /* Copy the rendered glyph into the stream. */
341:
342: sz = strlen(cpp);
343: bsz += sz;
344:
345: *buf = mandoc_realloc(*buf, bsz);
346:
347: memcpy(&(*buf)[pos], cpp, sz);
348: pos += (int)sz;
349: }
350:
351: (*buf)[pos] = '\0';
352: }
353:
354: /*
355: * Open the filename-index mandoc-db database.
356: * Returns NULL if opening failed.
357: */
358: static DB *
359: index_open(void)
360: {
361: DB *db;
362:
1.2 schwarze 363: db = dbopen(MANDOC_IDX, O_RDONLY, 0, DB_RECNO, NULL);
1.1 schwarze 364: if (NULL != db)
365: return(db);
366:
367: return(NULL);
368: }
369:
370: /*
371: * Safely unpack from an index file record into the structure.
372: * Returns 1 if an entry was unpacked, 0 if the database is insane.
373: */
374: static int
1.11 kristaps 375: index_read(const DBT *key, const DBT *val, int index,
1.30 kristaps 376: const struct mchars *mc, struct res *rec)
1.1 schwarze 377: {
378: size_t left;
379: char *np, *cp;
1.24 kristaps 380: char type;
1.1 schwarze 381:
382: #define INDEX_BREAD(_dst) \
383: do { \
384: if (NULL == (np = memchr(cp, '\0', left))) \
385: return(0); \
386: norm_string(cp, mc, &(_dst)); \
387: left -= (np - cp) + 1; \
388: cp = np + 1; \
389: } while (/* CONSTCOND */ 0)
390:
1.24 kristaps 391: if (0 == (left = val->size))
392: return(0);
1.1 schwarze 393:
1.24 kristaps 394: cp = val->data;
1.27 schwarze 395: assert(sizeof(recno_t) == key->size);
1.30 kristaps 396: memcpy(&rec->rec, key->data, key->size);
397: rec->volume = index;
1.1 schwarze 398:
1.24 kristaps 399: if ('d' == (type = *cp++))
1.30 kristaps 400: rec->type = RESTYPE_MDOC;
1.24 kristaps 401: else if ('a' == type)
1.30 kristaps 402: rec->type = RESTYPE_MAN;
1.24 kristaps 403: else if ('c' == type)
1.30 kristaps 404: rec->type = RESTYPE_CAT;
1.24 kristaps 405: else
406: return(0);
407:
408: left--;
1.30 kristaps 409: INDEX_BREAD(rec->file);
410: INDEX_BREAD(rec->cat);
411: INDEX_BREAD(rec->title);
412: INDEX_BREAD(rec->arch);
413: INDEX_BREAD(rec->desc);
1.1 schwarze 414: return(1);
415: }
416:
417: /*
1.10 kristaps 418: * Search mandocdb databases in paths for expression "expr".
1.1 schwarze 419: * Filter out by "opts".
420: * Call "res" with the results, which may be zero.
1.5 kristaps 421: * Return 0 if there was a database error, else return 1.
1.1 schwarze 422: */
1.5 kristaps 423: int
1.10 kristaps 424: apropos_search(int pathsz, char **paths, const struct opts *opts,
1.12 schwarze 425: const struct expr *expr, size_t terms, void *arg,
1.30 kristaps 426: size_t *sz, struct res **resp,
1.5 kristaps 427: void (*res)(struct res *, size_t, void *))
1.1 schwarze 428: {
1.8 kristaps 429: struct rectree tree;
430: struct mchars *mc;
1.32.2.2 schwarze 431: int i;
1.8 kristaps 432:
433: memset(&tree, 0, sizeof(struct rectree));
434:
435: mc = mchars_alloc();
1.30 kristaps 436: *sz = 0;
437: *resp = NULL;
1.8 kristaps 438:
1.10 kristaps 439: /*
440: * Main loop. Change into the directory containing manpage
441: * databases. Run our expession over each database in the set.
442: */
443:
444: for (i = 0; i < pathsz; i++) {
1.32 kristaps 445: assert('/' == paths[i][0]);
1.10 kristaps 446: if (chdir(paths[i]))
1.8 kristaps 447: continue;
1.30 kristaps 448: if (single_search(&tree, opts, expr, terms, mc, i))
449: continue;
450:
451: resfree(tree.node, tree.len);
452: mchars_free(mc);
453: return(0);
1.8 kristaps 454: }
455:
1.30 kristaps 456: (*res)(tree.node, tree.len, arg);
457: *sz = tree.len;
458: *resp = tree.node;
1.8 kristaps 459: mchars_free(mc);
1.30 kristaps 460: return(1);
1.8 kristaps 461: }
462:
463: static int
464: single_search(struct rectree *tree, const struct opts *opts,
465: const struct expr *expr, size_t terms,
1.11 kristaps 466: struct mchars *mc, int vol)
1.8 kristaps 467: {
468: int root, leaf, ch;
1.1 schwarze 469: DBT key, val;
470: DB *btree, *idx;
471: char *buf;
1.30 kristaps 472: struct res *rs;
473: struct res r;
1.28 schwarze 474: uint64_t mask;
475: recno_t rec;
1.1 schwarze 476:
477: root = -1;
478: leaf = -1;
479: btree = NULL;
480: idx = NULL;
481: buf = NULL;
1.8 kristaps 482: rs = tree->node;
1.1 schwarze 483:
1.30 kristaps 484: memset(&r, 0, sizeof(struct res));
1.1 schwarze 485:
1.12 schwarze 486: if (NULL == (btree = btree_open()))
1.10 kristaps 487: return(1);
1.1 schwarze 488:
1.8 kristaps 489: if (NULL == (idx = index_open())) {
490: (*btree->close)(btree);
1.10 kristaps 491: return(1);
1.8 kristaps 492: }
1.1 schwarze 493:
494: while (0 == (ch = (*btree->seq)(btree, &key, &val, R_NEXT))) {
1.28 schwarze 495: if ( ! btree_read(&key, &val, mc, &mask, &rec, &buf))
1.1 schwarze 496: break;
497:
1.5 kristaps 498: /*
499: * See if this keyword record matches any of the
500: * expressions we have stored.
501: */
1.28 schwarze 502: if ( ! exprmark(expr, buf, mask, NULL))
1.1 schwarze 503: continue;
504:
505: /*
506: * O(log n) scan for prior records. Since a record
507: * number is unbounded, this has decent performance over
508: * a complex hash function.
509: */
510:
511: for (leaf = root; leaf >= 0; )
1.30 kristaps 512: if (rec > rs[leaf].rec &&
1.5 kristaps 513: rs[leaf].rhs >= 0)
514: leaf = rs[leaf].rhs;
1.30 kristaps 515: else if (rec < rs[leaf].rec &&
1.5 kristaps 516: rs[leaf].lhs >= 0)
517: leaf = rs[leaf].lhs;
1.12 schwarze 518: else
1.1 schwarze 519: break;
520:
1.5 kristaps 521: /*
522: * If we find a record, see if it has already evaluated
523: * to true. If it has, great, just keep going. If not,
524: * try to evaluate it now and continue anyway.
525: */
526:
1.30 kristaps 527: if (leaf >= 0 && rs[leaf].rec == rec) {
1.5 kristaps 528: if (0 == rs[leaf].matched)
1.28 schwarze 529: exprexec(expr, buf, mask, &rs[leaf]);
1.1 schwarze 530: continue;
1.5 kristaps 531: }
1.1 schwarze 532:
533: /*
1.5 kristaps 534: * We have a new file to examine.
535: * Extract the manpage's metadata from the index
536: * database, then begin partial evaluation.
1.1 schwarze 537: */
538:
1.28 schwarze 539: key.data = &rec;
1.1 schwarze 540: key.size = sizeof(recno_t);
541:
542: if (0 != (*idx->get)(idx, &key, &val, 0))
543: break;
544:
1.5 kristaps 545: r.lhs = r.rhs = -1;
1.11 kristaps 546: if ( ! index_read(&key, &val, vol, mc, &r))
1.1 schwarze 547: break;
548:
1.5 kristaps 549: /* XXX: this should be elsewhere, I guess? */
550:
1.30 kristaps 551: if (opts->cat && strcasecmp(opts->cat, r.cat))
1.1 schwarze 552: continue;
1.22 kristaps 553:
1.30 kristaps 554: if (opts->arch && *r.arch)
555: if (strcasecmp(opts->arch, r.arch))
1.22 kristaps 556: continue;
1.1 schwarze 557:
1.32.2.6! schwarze 558: tree->node = rs = mandoc_reallocarray(rs,
! 559: tree->len + 1, sizeof(struct res));
1.1 schwarze 560:
1.30 kristaps 561: memcpy(&rs[tree->len], &r, sizeof(struct res));
562: memset(&r, 0, sizeof(struct res));
1.12 schwarze 563: rs[tree->len].matches =
1.8 kristaps 564: mandoc_calloc(terms, sizeof(int));
1.1 schwarze 565:
1.28 schwarze 566: exprexec(expr, buf, mask, &rs[tree->len]);
1.12 schwarze 567:
1.1 schwarze 568: /* Append to our tree. */
569:
570: if (leaf >= 0) {
1.30 kristaps 571: if (rec > rs[leaf].rec)
1.8 kristaps 572: rs[leaf].rhs = tree->len;
1.1 schwarze 573: else
1.8 kristaps 574: rs[leaf].lhs = tree->len;
1.1 schwarze 575: } else
1.8 kristaps 576: root = tree->len;
1.12 schwarze 577:
1.8 kristaps 578: tree->len++;
1.1 schwarze 579: }
1.12 schwarze 580:
1.8 kristaps 581: (*btree->close)(btree);
582: (*idx->close)(idx);
1.1 schwarze 583:
584: free(buf);
1.30 kristaps 585: RESFREE(&r);
1.8 kristaps 586: return(1 == ch);
1.5 kristaps 587: }
588:
1.30 kristaps 589: void
590: resfree(struct res *rec, size_t sz)
1.5 kristaps 591: {
1.30 kristaps 592: size_t i;
1.5 kristaps 593:
1.30 kristaps 594: for (i = 0; i < sz; i++)
595: RESFREE(&rec[i]);
596: free(rec);
1.1 schwarze 597: }
598:
1.13 kristaps 599: /*
600: * Compile a list of straight-up terms.
601: * The arguments are re-written into ~[[:<:]]term[[:>:]], or "term"
602: * surrounded by word boundaries, then pumped through exprterm().
603: * Terms are case-insensitive.
604: * This emulates whatis(1) behaviour.
605: */
606: struct expr *
607: termcomp(int argc, char *argv[], size_t *tt)
608: {
609: char *buf;
610: int pos;
611: struct expr *e, *next;
612: size_t sz;
613:
614: buf = NULL;
615: e = NULL;
616: *tt = 0;
617:
1.15 schwarze 618: for (pos = argc - 1; pos >= 0; pos--) {
619: sz = strlen(argv[pos]) + 18;
1.13 kristaps 620: buf = mandoc_realloc(buf, sz);
1.15 schwarze 621: strlcpy(buf, "Nm~[[:<:]]", sz);
1.13 kristaps 622: strlcat(buf, argv[pos], sz);
623: strlcat(buf, "[[:>:]]", sz);
624: if (NULL == (next = exprterm(buf, 0))) {
625: free(buf);
626: exprfree(e);
627: return(NULL);
628: }
1.15 schwarze 629: next->next = e;
1.13 kristaps 630: e = next;
631: (*tt)++;
632: }
633:
634: free(buf);
635: return(e);
636: }
637:
638: /*
639: * Compile a sequence of logical expressions.
640: * See apropos.1 for a grammar of this sequence.
641: */
1.1 schwarze 642: struct expr *
1.5 kristaps 643: exprcomp(int argc, char *argv[], size_t *tt)
1.1 schwarze 644: {
1.5 kristaps 645: int pos, lvl;
646: struct expr *e;
647:
648: pos = lvl = 0;
649: *tt = 0;
650:
651: e = exprexpr(argc, argv, &pos, &lvl, tt);
652:
653: if (0 == lvl && pos >= argc)
654: return(e);
655:
656: exprfree(e);
657: return(NULL);
658: }
659:
660: /*
661: * Compile an array of tokens into an expression.
662: * An informal expression grammar is defined in apropos(1).
663: * Return NULL if we fail doing so. All memory will be cleaned up.
664: * Return the root of the expression sequence if alright.
665: */
666: static struct expr *
1.9 kristaps 667: exprexpr(int argc, char *argv[], int *pos, int *lvl, size_t *tt)
1.5 kristaps 668: {
669: struct expr *e, *first, *next;
670: int log;
671:
672: first = next = NULL;
673:
674: for ( ; *pos < argc; (*pos)++) {
675: e = next;
676:
677: /*
678: * Close out a subexpression.
679: */
680:
681: if (NULL != e && 0 == strcmp(")", argv[*pos])) {
682: if (--(*lvl) < 0)
683: goto err;
684: break;
685: }
686:
687: /*
688: * Small note: if we're just starting, don't let "-a"
689: * and "-o" be considered logical operators: they're
690: * just tokens unless pairwise joining, in which case we
691: * record their existence (or assume "OR").
692: */
693: log = 0;
694:
695: if (NULL != e && 0 == strcmp("-a", argv[*pos]))
1.12 schwarze 696: log = 1;
1.5 kristaps 697: else if (NULL != e && 0 == strcmp("-o", argv[*pos]))
698: log = 2;
699:
700: if (log > 0 && ++(*pos) >= argc)
701: goto err;
702:
703: /*
704: * Now we parse the term part. This can begin with
705: * "-i", in which case the expression is case
706: * insensitive.
707: */
708:
709: if (0 == strcmp("(", argv[*pos])) {
710: ++(*pos);
711: ++(*lvl);
712: next = mandoc_calloc(1, sizeof(struct expr));
713: next->subexpr = exprexpr(argc, argv, pos, lvl, tt);
714: if (NULL == next->subexpr) {
715: free(next);
716: next = NULL;
717: }
718: } else if (0 == strcmp("-i", argv[*pos])) {
719: if (++(*pos) >= argc)
720: goto err;
721: next = exprterm(argv[*pos], 0);
722: } else
723: next = exprterm(argv[*pos], 1);
724:
725: if (NULL == next)
726: goto err;
727:
728: next->and = log == 1;
729: next->index = (int)(*tt)++;
730:
731: /* Append to our chain of expressions. */
732:
733: if (NULL == first) {
734: assert(NULL == e);
735: first = next;
736: } else {
737: assert(NULL != e);
738: e->next = next;
739: }
740: }
741:
742: return(first);
743: err:
744: exprfree(first);
745: return(NULL);
746: }
747:
748: /*
749: * Parse a terminal expression with the grammar as defined in
750: * apropos(1).
751: * Return NULL if we fail the parse.
752: */
753: static struct expr *
754: exprterm(char *buf, int cs)
755: {
756: struct expr e;
1.1 schwarze 757: struct expr *p;
1.3 schwarze 758: char *key;
1.5 kristaps 759: int i;
760:
761: memset(&e, 0, sizeof(struct expr));
1.1 schwarze 762:
1.5 kristaps 763: /* Choose regex or substring match. */
1.3 schwarze 764:
1.4 kristaps 765: if (NULL == (e.v = strpbrk(buf, "=~"))) {
1.3 schwarze 766: e.regex = 0;
1.4 kristaps 767: e.v = buf;
1.3 schwarze 768: } else {
769: e.regex = '~' == *e.v;
770: *e.v++ = '\0';
771: }
1.1 schwarze 772:
1.5 kristaps 773: /* Determine the record types to search for. */
1.3 schwarze 774:
775: e.mask = 0;
1.4 kristaps 776: if (buf < e.v) {
777: while (NULL != (key = strsep(&buf, ","))) {
1.3 schwarze 778: i = 0;
779: while (types[i].mask &&
1.4 kristaps 780: strcmp(types[i].name, key))
1.3 schwarze 781: i++;
782: e.mask |= types[i].mask;
783: }
784: }
785: if (0 == e.mask)
786: e.mask = TYPE_Nm | TYPE_Nd;
1.1 schwarze 787:
1.5 kristaps 788: if (e.regex) {
1.13 kristaps 789: i = REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE);
1.5 kristaps 790: if (regcomp(&e.re, e.v, i))
791: return(NULL);
792: }
1.1 schwarze 793:
1.3 schwarze 794: e.v = mandoc_strdup(e.v);
1.1 schwarze 795:
796: p = mandoc_calloc(1, sizeof(struct expr));
797: memcpy(p, &e, sizeof(struct expr));
798: return(p);
799: }
800:
801: void
802: exprfree(struct expr *p)
803: {
1.5 kristaps 804: struct expr *pp;
1.12 schwarze 805:
1.5 kristaps 806: while (NULL != p) {
807: if (p->subexpr)
808: exprfree(p->subexpr);
809: if (p->regex)
810: regfree(&p->re);
811: free(p->v);
812: pp = p->next;
813: free(p);
814: p = pp;
815: }
816: }
1.1 schwarze 817:
1.5 kristaps 818: static int
1.12 schwarze 819: exprmark(const struct expr *p, const char *cp,
1.6 kristaps 820: uint64_t mask, int *ms)
1.5 kristaps 821: {
822:
823: for ( ; p; p = p->next) {
824: if (p->subexpr) {
825: if (exprmark(p->subexpr, cp, mask, ms))
826: return(1);
827: continue;
828: } else if ( ! (mask & p->mask))
829: continue;
1.1 schwarze 830:
1.5 kristaps 831: if (p->regex) {
832: if (regexec(&p->re, cp, 0, NULL, 0))
833: continue;
1.16 kristaps 834: } else if (NULL == strcasestr(cp, p->v))
835: continue;
1.5 kristaps 836:
837: if (NULL == ms)
838: return(1);
839: else
840: ms[p->index] = 1;
841: }
1.1 schwarze 842:
1.5 kristaps 843: return(0);
1.1 schwarze 844: }
845:
846: static int
1.5 kristaps 847: expreval(const struct expr *p, int *ms)
1.1 schwarze 848: {
1.5 kristaps 849: int match;
1.1 schwarze 850:
1.5 kristaps 851: /*
852: * AND has precedence over OR. Analysis is left-right, though
853: * it doesn't matter because there are no side-effects.
854: * Thus, step through pairwise ANDs and accumulate their Boolean
855: * evaluation. If we encounter a single true AND collection or
856: * standalone term, the whole expression is true (by definition
857: * of OR).
858: */
859:
860: for (match = 0; p && ! match; p = p->next) {
861: /* Evaluate a subexpression, if applicable. */
862: if (p->subexpr && ! ms[p->index])
863: ms[p->index] = expreval(p->subexpr, ms);
864:
865: match = ms[p->index];
866: for ( ; p->next && p->next->and; p = p->next) {
867: /* Evaluate a subexpression, if applicable. */
868: if (p->next->subexpr && ! ms[p->next->index])
1.12 schwarze 869: ms[p->next->index] =
1.5 kristaps 870: expreval(p->next->subexpr, ms);
871: match = match && ms[p->next->index];
872: }
873: }
874:
875: return(match);
876: }
877:
878: /*
879: * First, update the array of terms for which this expression evaluates
880: * to true.
881: * Second, logically evaluate all terms over the updated array of truth
882: * values.
883: * If this evaluates to true, mark the expression as satisfied.
884: */
885: static void
1.12 schwarze 886: exprexec(const struct expr *e, const char *cp,
1.30 kristaps 887: uint64_t mask, struct res *r)
1.5 kristaps 888: {
1.1 schwarze 889:
1.5 kristaps 890: assert(0 == r->matched);
1.12 schwarze 891: exprmark(e, cp, mask, r->matches);
892: r->matched = expreval(e, r->matches);
1.1 schwarze 893: }
CVSweb