Annotation of mandoc/apropos_db.c, Revision 1.32.2.4
1.32.2.4! schwarze 1: /* $Id: apropos_db.c,v 1.32.2.3 2013/10/10 23:43:04 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"
52:
1.30 kristaps 53: #define RESFREE(_x) \
54: do { \
55: free((_x)->file); \
56: free((_x)->cat); \
57: free((_x)->title); \
58: free((_x)->arch); \
59: free((_x)->desc); \
60: free((_x)->matches); \
61: } while (/*CONSTCOND*/0)
1.5 kristaps 62:
1.1 schwarze 63: struct expr {
1.5 kristaps 64: int regex; /* is regex? */
65: int index; /* index in match array */
1.6 kristaps 66: uint64_t mask; /* type-mask */
1.5 kristaps 67: int and; /* is rhs of logical AND? */
68: char *v; /* search value */
69: regex_t re; /* compiled re, if regex */
70: struct expr *next; /* next in sequence */
71: struct expr *subexpr;
1.1 schwarze 72: };
73:
74: struct type {
1.6 kristaps 75: uint64_t mask;
1.1 schwarze 76: const char *name;
77: };
78:
1.8 kristaps 79: struct rectree {
1.30 kristaps 80: struct res *node; /* record array for dir tree */
1.8 kristaps 81: int len; /* length of record array */
82: };
83:
1.1 schwarze 84: static const struct type types[] = {
1.7 kristaps 85: { TYPE_An, "An" },
86: { TYPE_Ar, "Ar" },
87: { TYPE_At, "At" },
88: { TYPE_Bsx, "Bsx" },
89: { TYPE_Bx, "Bx" },
90: { TYPE_Cd, "Cd" },
91: { TYPE_Cm, "Cm" },
92: { TYPE_Dv, "Dv" },
93: { TYPE_Dx, "Dx" },
94: { TYPE_Em, "Em" },
95: { TYPE_Er, "Er" },
96: { TYPE_Ev, "Ev" },
97: { TYPE_Fa, "Fa" },
98: { TYPE_Fl, "Fl" },
99: { TYPE_Fn, "Fn" },
100: { TYPE_Fn, "Fo" },
101: { TYPE_Ft, "Ft" },
102: { TYPE_Fx, "Fx" },
103: { TYPE_Ic, "Ic" },
104: { TYPE_In, "In" },
105: { TYPE_Lb, "Lb" },
106: { TYPE_Li, "Li" },
107: { TYPE_Lk, "Lk" },
108: { TYPE_Ms, "Ms" },
109: { TYPE_Mt, "Mt" },
110: { TYPE_Nd, "Nd" },
111: { TYPE_Nm, "Nm" },
112: { TYPE_Nx, "Nx" },
113: { TYPE_Ox, "Ox" },
114: { TYPE_Pa, "Pa" },
115: { TYPE_Rs, "Rs" },
116: { TYPE_Sh, "Sh" },
117: { TYPE_Ss, "Ss" },
118: { TYPE_St, "St" },
119: { TYPE_Sy, "Sy" },
120: { TYPE_Tn, "Tn" },
121: { TYPE_Va, "Va" },
122: { TYPE_Va, "Vt" },
123: { TYPE_Xr, "Xr" },
1.25 kristaps 124: { UINT64_MAX, "any" },
1.1 schwarze 125: { 0, NULL }
126: };
127:
128: static DB *btree_open(void);
1.17 kristaps 129: static int btree_read(const DBT *, const DBT *,
1.28 schwarze 130: const struct mchars *,
131: uint64_t *, recno_t *, char **);
1.5 kristaps 132: static int expreval(const struct expr *, int *);
1.12 schwarze 133: static void exprexec(const struct expr *,
1.30 kristaps 134: const char *, uint64_t, struct res *);
1.12 schwarze 135: static int exprmark(const struct expr *,
1.6 kristaps 136: const char *, uint64_t, int *);
1.5 kristaps 137: static struct expr *exprexpr(int, char *[], int *, int *, size_t *);
138: static struct expr *exprterm(char *, int);
1.1 schwarze 139: static DB *index_open(void);
1.11 kristaps 140: static int index_read(const DBT *, const DBT *, int,
1.30 kristaps 141: const struct mchars *, struct res *);
1.1 schwarze 142: static void norm_string(const char *,
143: const struct mchars *, char **);
144: static size_t norm_utf8(unsigned int, char[7]);
1.8 kristaps 145: static int single_search(struct rectree *, const struct opts *,
146: const struct expr *, size_t terms,
1.11 kristaps 147: struct mchars *, int);
1.1 schwarze 148:
149: /*
150: * Open the keyword mandoc-db database.
151: */
152: static DB *
153: btree_open(void)
154: {
155: BTREEINFO info;
156: DB *db;
157:
158: memset(&info, 0, sizeof(BTREEINFO));
1.29 kristaps 159: info.lorder = 4321;
1.1 schwarze 160: info.flags = R_DUP;
161:
1.2 schwarze 162: db = dbopen(MANDOC_DB, O_RDONLY, 0, DB_BTREE, &info);
1.12 schwarze 163: if (NULL != db)
1.1 schwarze 164: return(db);
165:
166: return(NULL);
167: }
168:
169: /*
170: * Read a keyword from the database and normalise it.
171: * Return 0 if the database is insane, else 1.
172: */
173: static int
1.28 schwarze 174: btree_read(const DBT *k, const DBT *v, const struct mchars *mc,
175: uint64_t *mask, recno_t *rec, char **buf)
1.1 schwarze 176: {
1.28 schwarze 177: uint64_t vbuf[2];
1.1 schwarze 178:
1.17 kristaps 179: /* Are our sizes sane? */
1.28 schwarze 180: if (k->size < 2 || sizeof(vbuf) != v->size)
1.17 kristaps 181: return(0);
1.6 kristaps 182:
1.17 kristaps 183: /* Is our string nil-terminated? */
184: if ('\0' != ((const char *)k->data)[(int)k->size - 1])
1.1 schwarze 185: return(0);
186:
1.17 kristaps 187: norm_string((const char *)k->data, mc, buf);
1.28 schwarze 188: memcpy(vbuf, v->data, v->size);
189: *mask = betoh64(vbuf[0]);
190: *rec = betoh64(vbuf[1]);
1.1 schwarze 191: return(1);
192: }
193:
194: /*
195: * Take a Unicode codepoint and produce its UTF-8 encoding.
196: * This isn't the best way to do this, but it works.
1.12 schwarze 197: * The magic numbers are from the UTF-8 packaging.
1.1 schwarze 198: * They're not as scary as they seem: read the UTF-8 spec for details.
199: */
200: static size_t
201: norm_utf8(unsigned int cp, char out[7])
202: {
1.26 kristaps 203: int rc;
1.1 schwarze 204:
205: rc = 0;
206:
207: if (cp <= 0x0000007F) {
208: rc = 1;
209: out[0] = (char)cp;
210: } else if (cp <= 0x000007FF) {
211: rc = 2;
212: out[0] = (cp >> 6 & 31) | 192;
213: out[1] = (cp & 63) | 128;
214: } else if (cp <= 0x0000FFFF) {
215: rc = 3;
216: out[0] = (cp >> 12 & 15) | 224;
217: out[1] = (cp >> 6 & 63) | 128;
218: out[2] = (cp & 63) | 128;
219: } else if (cp <= 0x001FFFFF) {
220: rc = 4;
221: out[0] = (cp >> 18 & 7) | 240;
222: out[1] = (cp >> 12 & 63) | 128;
223: out[2] = (cp >> 6 & 63) | 128;
224: out[3] = (cp & 63) | 128;
225: } else if (cp <= 0x03FFFFFF) {
226: rc = 5;
227: out[0] = (cp >> 24 & 3) | 248;
228: out[1] = (cp >> 18 & 63) | 128;
229: out[2] = (cp >> 12 & 63) | 128;
230: out[3] = (cp >> 6 & 63) | 128;
231: out[4] = (cp & 63) | 128;
232: } else if (cp <= 0x7FFFFFFF) {
233: rc = 6;
234: out[0] = (cp >> 30 & 1) | 252;
235: out[1] = (cp >> 24 & 63) | 128;
236: out[2] = (cp >> 18 & 63) | 128;
237: out[3] = (cp >> 12 & 63) | 128;
238: out[4] = (cp >> 6 & 63) | 128;
239: out[5] = (cp & 63) | 128;
240: } else
241: return(0);
242:
243: out[rc] = '\0';
1.26 kristaps 244: return((size_t)rc);
1.1 schwarze 245: }
246:
247: /*
248: * Normalise strings from the index and database.
249: * These strings are escaped as defined by mandoc_char(7) along with
250: * other goop in mandoc.h (e.g., soft hyphens).
251: * This function normalises these into a nice UTF-8 string.
252: * Returns 0 if the database is fucked.
253: */
254: static void
255: norm_string(const char *val, const struct mchars *mc, char **buf)
256: {
257: size_t sz, bsz;
258: char utfbuf[7];
259: const char *seq, *cpp;
260: int len, u, pos;
261: enum mandoc_esc esc;
1.32.2.4! schwarze 262: static const char res[] = { '\\', '\t', ASCII_NBRSP,
! 263: ASCII_HYPH, ASCII_BREAK, '\0' };
1.1 schwarze 264:
265: /* Pre-allocate by the length of the input */
266:
267: bsz = strlen(val) + 1;
268: *buf = mandoc_realloc(*buf, bsz);
269: pos = 0;
270:
271: while ('\0' != *val) {
272: /*
273: * Halt on the first escape sequence.
274: * This also halts on the end of string, in which case
275: * we just copy, fallthrough, and exit the loop.
276: */
277: if ((sz = strcspn(val, res)) > 0) {
278: memcpy(&(*buf)[pos], val, sz);
279: pos += (int)sz;
280: val += (int)sz;
281: }
282:
1.32.2.4! schwarze 283: switch (*val) {
! 284: case (ASCII_HYPH):
1.1 schwarze 285: (*buf)[pos++] = '-';
286: val++;
287: continue;
1.32.2.4! schwarze 288: case ('\t'):
! 289: /* FALLTHROUGH */
! 290: case (ASCII_NBRSP):
1.1 schwarze 291: (*buf)[pos++] = ' ';
292: val++;
1.32.2.4! schwarze 293: /* FALLTHROUGH */
! 294: case (ASCII_BREAK):
1.1 schwarze 295: continue;
1.32.2.4! schwarze 296: default:
! 297: break;
! 298: }
! 299: if ('\\' != *val)
1.1 schwarze 300: break;
301:
302: /* Read past the slash. */
303:
304: val++;
305: u = 0;
306:
307: /*
308: * Parse the escape sequence and see if it's a
309: * predefined character or special character.
310: */
311:
312: esc = mandoc_escape(&val, &seq, &len);
313: if (ESCAPE_ERROR == esc)
314: break;
315:
1.12 schwarze 316: /*
1.1 schwarze 317: * XXX - this just does UTF-8, but we need to know
318: * beforehand whether we should do text substitution.
319: */
320:
321: switch (esc) {
322: case (ESCAPE_SPECIAL):
323: if (0 != (u = mchars_spec2cp(mc, seq, len)))
324: break;
325: /* FALLTHROUGH */
326: default:
327: continue;
328: }
329:
330: /*
331: * If we have a Unicode codepoint, try to convert that
332: * to a UTF-8 byte string.
333: */
334:
335: cpp = utfbuf;
336: if (0 == (sz = norm_utf8(u, utfbuf)))
337: continue;
338:
339: /* Copy the rendered glyph into the stream. */
340:
341: sz = strlen(cpp);
342: bsz += sz;
343:
344: *buf = mandoc_realloc(*buf, bsz);
345:
346: memcpy(&(*buf)[pos], cpp, sz);
347: pos += (int)sz;
348: }
349:
350: (*buf)[pos] = '\0';
351: }
352:
353: /*
354: * Open the filename-index mandoc-db database.
355: * Returns NULL if opening failed.
356: */
357: static DB *
358: index_open(void)
359: {
360: DB *db;
361:
1.2 schwarze 362: db = dbopen(MANDOC_IDX, O_RDONLY, 0, DB_RECNO, NULL);
1.1 schwarze 363: if (NULL != db)
364: return(db);
365:
366: return(NULL);
367: }
368:
369: /*
370: * Safely unpack from an index file record into the structure.
371: * Returns 1 if an entry was unpacked, 0 if the database is insane.
372: */
373: static int
1.11 kristaps 374: index_read(const DBT *key, const DBT *val, int index,
1.30 kristaps 375: const struct mchars *mc, struct res *rec)
1.1 schwarze 376: {
377: size_t left;
378: char *np, *cp;
1.24 kristaps 379: char type;
1.1 schwarze 380:
381: #define INDEX_BREAD(_dst) \
382: do { \
383: if (NULL == (np = memchr(cp, '\0', left))) \
384: return(0); \
385: norm_string(cp, mc, &(_dst)); \
386: left -= (np - cp) + 1; \
387: cp = np + 1; \
388: } while (/* CONSTCOND */ 0)
389:
1.24 kristaps 390: if (0 == (left = val->size))
391: return(0);
1.1 schwarze 392:
1.24 kristaps 393: cp = val->data;
1.27 schwarze 394: assert(sizeof(recno_t) == key->size);
1.30 kristaps 395: memcpy(&rec->rec, key->data, key->size);
396: rec->volume = index;
1.1 schwarze 397:
1.24 kristaps 398: if ('d' == (type = *cp++))
1.30 kristaps 399: rec->type = RESTYPE_MDOC;
1.24 kristaps 400: else if ('a' == type)
1.30 kristaps 401: rec->type = RESTYPE_MAN;
1.24 kristaps 402: else if ('c' == type)
1.30 kristaps 403: rec->type = RESTYPE_CAT;
1.24 kristaps 404: else
405: return(0);
406:
407: left--;
1.30 kristaps 408: INDEX_BREAD(rec->file);
409: INDEX_BREAD(rec->cat);
410: INDEX_BREAD(rec->title);
411: INDEX_BREAD(rec->arch);
412: INDEX_BREAD(rec->desc);
1.1 schwarze 413: return(1);
414: }
415:
416: /*
1.10 kristaps 417: * Search mandocdb databases in paths for expression "expr".
1.1 schwarze 418: * Filter out by "opts".
419: * Call "res" with the results, which may be zero.
1.5 kristaps 420: * Return 0 if there was a database error, else return 1.
1.1 schwarze 421: */
1.5 kristaps 422: int
1.10 kristaps 423: apropos_search(int pathsz, char **paths, const struct opts *opts,
1.12 schwarze 424: const struct expr *expr, size_t terms, void *arg,
1.30 kristaps 425: size_t *sz, struct res **resp,
1.5 kristaps 426: void (*res)(struct res *, size_t, void *))
1.1 schwarze 427: {
1.8 kristaps 428: struct rectree tree;
429: struct mchars *mc;
1.32.2.2 schwarze 430: int i;
1.8 kristaps 431:
432: memset(&tree, 0, sizeof(struct rectree));
433:
434: mc = mchars_alloc();
1.30 kristaps 435: *sz = 0;
436: *resp = NULL;
1.8 kristaps 437:
1.10 kristaps 438: /*
439: * Main loop. Change into the directory containing manpage
440: * databases. Run our expession over each database in the set.
441: */
442:
443: for (i = 0; i < pathsz; i++) {
1.32 kristaps 444: assert('/' == paths[i][0]);
1.10 kristaps 445: if (chdir(paths[i]))
1.8 kristaps 446: continue;
1.30 kristaps 447: if (single_search(&tree, opts, expr, terms, mc, i))
448: continue;
449:
450: resfree(tree.node, tree.len);
451: mchars_free(mc);
452: return(0);
1.8 kristaps 453: }
454:
1.30 kristaps 455: (*res)(tree.node, tree.len, arg);
456: *sz = tree.len;
457: *resp = tree.node;
1.8 kristaps 458: mchars_free(mc);
1.30 kristaps 459: return(1);
1.8 kristaps 460: }
461:
462: static int
463: single_search(struct rectree *tree, const struct opts *opts,
464: const struct expr *expr, size_t terms,
1.11 kristaps 465: struct mchars *mc, int vol)
1.8 kristaps 466: {
467: int root, leaf, ch;
1.1 schwarze 468: DBT key, val;
469: DB *btree, *idx;
470: char *buf;
1.30 kristaps 471: struct res *rs;
472: struct res r;
1.28 schwarze 473: uint64_t mask;
474: recno_t rec;
1.1 schwarze 475:
476: root = -1;
477: leaf = -1;
478: btree = NULL;
479: idx = NULL;
480: buf = NULL;
1.8 kristaps 481: rs = tree->node;
1.1 schwarze 482:
1.30 kristaps 483: memset(&r, 0, sizeof(struct res));
1.1 schwarze 484:
1.12 schwarze 485: if (NULL == (btree = btree_open()))
1.10 kristaps 486: return(1);
1.1 schwarze 487:
1.8 kristaps 488: if (NULL == (idx = index_open())) {
489: (*btree->close)(btree);
1.10 kristaps 490: return(1);
1.8 kristaps 491: }
1.1 schwarze 492:
493: while (0 == (ch = (*btree->seq)(btree, &key, &val, R_NEXT))) {
1.28 schwarze 494: if ( ! btree_read(&key, &val, mc, &mask, &rec, &buf))
1.1 schwarze 495: break;
496:
1.5 kristaps 497: /*
498: * See if this keyword record matches any of the
499: * expressions we have stored.
500: */
1.28 schwarze 501: if ( ! exprmark(expr, buf, mask, NULL))
1.1 schwarze 502: continue;
503:
504: /*
505: * O(log n) scan for prior records. Since a record
506: * number is unbounded, this has decent performance over
507: * a complex hash function.
508: */
509:
510: for (leaf = root; leaf >= 0; )
1.30 kristaps 511: if (rec > rs[leaf].rec &&
1.5 kristaps 512: rs[leaf].rhs >= 0)
513: leaf = rs[leaf].rhs;
1.30 kristaps 514: else if (rec < rs[leaf].rec &&
1.5 kristaps 515: rs[leaf].lhs >= 0)
516: leaf = rs[leaf].lhs;
1.12 schwarze 517: else
1.1 schwarze 518: break;
519:
1.5 kristaps 520: /*
521: * If we find a record, see if it has already evaluated
522: * to true. If it has, great, just keep going. If not,
523: * try to evaluate it now and continue anyway.
524: */
525:
1.30 kristaps 526: if (leaf >= 0 && rs[leaf].rec == rec) {
1.5 kristaps 527: if (0 == rs[leaf].matched)
1.28 schwarze 528: exprexec(expr, buf, mask, &rs[leaf]);
1.1 schwarze 529: continue;
1.5 kristaps 530: }
1.1 schwarze 531:
532: /*
1.5 kristaps 533: * We have a new file to examine.
534: * Extract the manpage's metadata from the index
535: * database, then begin partial evaluation.
1.1 schwarze 536: */
537:
1.28 schwarze 538: key.data = &rec;
1.1 schwarze 539: key.size = sizeof(recno_t);
540:
541: if (0 != (*idx->get)(idx, &key, &val, 0))
542: break;
543:
1.5 kristaps 544: r.lhs = r.rhs = -1;
1.11 kristaps 545: if ( ! index_read(&key, &val, vol, mc, &r))
1.1 schwarze 546: break;
547:
1.5 kristaps 548: /* XXX: this should be elsewhere, I guess? */
549:
1.30 kristaps 550: if (opts->cat && strcasecmp(opts->cat, r.cat))
1.1 schwarze 551: continue;
1.22 kristaps 552:
1.30 kristaps 553: if (opts->arch && *r.arch)
554: if (strcasecmp(opts->arch, r.arch))
1.22 kristaps 555: continue;
1.1 schwarze 556:
1.8 kristaps 557: tree->node = rs = mandoc_realloc
1.30 kristaps 558: (rs, (tree->len + 1) * sizeof(struct res));
1.1 schwarze 559:
1.30 kristaps 560: memcpy(&rs[tree->len], &r, sizeof(struct res));
561: memset(&r, 0, sizeof(struct res));
1.12 schwarze 562: rs[tree->len].matches =
1.8 kristaps 563: mandoc_calloc(terms, sizeof(int));
1.1 schwarze 564:
1.28 schwarze 565: exprexec(expr, buf, mask, &rs[tree->len]);
1.12 schwarze 566:
1.1 schwarze 567: /* Append to our tree. */
568:
569: if (leaf >= 0) {
1.30 kristaps 570: if (rec > rs[leaf].rec)
1.8 kristaps 571: rs[leaf].rhs = tree->len;
1.1 schwarze 572: else
1.8 kristaps 573: rs[leaf].lhs = tree->len;
1.1 schwarze 574: } else
1.8 kristaps 575: root = tree->len;
1.12 schwarze 576:
1.8 kristaps 577: tree->len++;
1.1 schwarze 578: }
1.12 schwarze 579:
1.8 kristaps 580: (*btree->close)(btree);
581: (*idx->close)(idx);
1.1 schwarze 582:
583: free(buf);
1.30 kristaps 584: RESFREE(&r);
1.8 kristaps 585: return(1 == ch);
1.5 kristaps 586: }
587:
1.30 kristaps 588: void
589: resfree(struct res *rec, size_t sz)
1.5 kristaps 590: {
1.30 kristaps 591: size_t i;
1.5 kristaps 592:
1.30 kristaps 593: for (i = 0; i < sz; i++)
594: RESFREE(&rec[i]);
595: free(rec);
1.1 schwarze 596: }
597:
1.13 kristaps 598: /*
599: * Compile a list of straight-up terms.
600: * The arguments are re-written into ~[[:<:]]term[[:>:]], or "term"
601: * surrounded by word boundaries, then pumped through exprterm().
602: * Terms are case-insensitive.
603: * This emulates whatis(1) behaviour.
604: */
605: struct expr *
606: termcomp(int argc, char *argv[], size_t *tt)
607: {
608: char *buf;
609: int pos;
610: struct expr *e, *next;
611: size_t sz;
612:
613: buf = NULL;
614: e = NULL;
615: *tt = 0;
616:
1.15 schwarze 617: for (pos = argc - 1; pos >= 0; pos--) {
618: sz = strlen(argv[pos]) + 18;
1.13 kristaps 619: buf = mandoc_realloc(buf, sz);
1.15 schwarze 620: strlcpy(buf, "Nm~[[:<:]]", sz);
1.13 kristaps 621: strlcat(buf, argv[pos], sz);
622: strlcat(buf, "[[:>:]]", sz);
623: if (NULL == (next = exprterm(buf, 0))) {
624: free(buf);
625: exprfree(e);
626: return(NULL);
627: }
1.15 schwarze 628: next->next = e;
1.13 kristaps 629: e = next;
630: (*tt)++;
631: }
632:
633: free(buf);
634: return(e);
635: }
636:
637: /*
638: * Compile a sequence of logical expressions.
639: * See apropos.1 for a grammar of this sequence.
640: */
1.1 schwarze 641: struct expr *
1.5 kristaps 642: exprcomp(int argc, char *argv[], size_t *tt)
1.1 schwarze 643: {
1.5 kristaps 644: int pos, lvl;
645: struct expr *e;
646:
647: pos = lvl = 0;
648: *tt = 0;
649:
650: e = exprexpr(argc, argv, &pos, &lvl, tt);
651:
652: if (0 == lvl && pos >= argc)
653: return(e);
654:
655: exprfree(e);
656: return(NULL);
657: }
658:
659: /*
660: * Compile an array of tokens into an expression.
661: * An informal expression grammar is defined in apropos(1).
662: * Return NULL if we fail doing so. All memory will be cleaned up.
663: * Return the root of the expression sequence if alright.
664: */
665: static struct expr *
1.9 kristaps 666: exprexpr(int argc, char *argv[], int *pos, int *lvl, size_t *tt)
1.5 kristaps 667: {
668: struct expr *e, *first, *next;
669: int log;
670:
671: first = next = NULL;
672:
673: for ( ; *pos < argc; (*pos)++) {
674: e = next;
675:
676: /*
677: * Close out a subexpression.
678: */
679:
680: if (NULL != e && 0 == strcmp(")", argv[*pos])) {
681: if (--(*lvl) < 0)
682: goto err;
683: break;
684: }
685:
686: /*
687: * Small note: if we're just starting, don't let "-a"
688: * and "-o" be considered logical operators: they're
689: * just tokens unless pairwise joining, in which case we
690: * record their existence (or assume "OR").
691: */
692: log = 0;
693:
694: if (NULL != e && 0 == strcmp("-a", argv[*pos]))
1.12 schwarze 695: log = 1;
1.5 kristaps 696: else if (NULL != e && 0 == strcmp("-o", argv[*pos]))
697: log = 2;
698:
699: if (log > 0 && ++(*pos) >= argc)
700: goto err;
701:
702: /*
703: * Now we parse the term part. This can begin with
704: * "-i", in which case the expression is case
705: * insensitive.
706: */
707:
708: if (0 == strcmp("(", argv[*pos])) {
709: ++(*pos);
710: ++(*lvl);
711: next = mandoc_calloc(1, sizeof(struct expr));
712: next->subexpr = exprexpr(argc, argv, pos, lvl, tt);
713: if (NULL == next->subexpr) {
714: free(next);
715: next = NULL;
716: }
717: } else if (0 == strcmp("-i", argv[*pos])) {
718: if (++(*pos) >= argc)
719: goto err;
720: next = exprterm(argv[*pos], 0);
721: } else
722: next = exprterm(argv[*pos], 1);
723:
724: if (NULL == next)
725: goto err;
726:
727: next->and = log == 1;
728: next->index = (int)(*tt)++;
729:
730: /* Append to our chain of expressions. */
731:
732: if (NULL == first) {
733: assert(NULL == e);
734: first = next;
735: } else {
736: assert(NULL != e);
737: e->next = next;
738: }
739: }
740:
741: return(first);
742: err:
743: exprfree(first);
744: return(NULL);
745: }
746:
747: /*
748: * Parse a terminal expression with the grammar as defined in
749: * apropos(1).
750: * Return NULL if we fail the parse.
751: */
752: static struct expr *
753: exprterm(char *buf, int cs)
754: {
755: struct expr e;
1.1 schwarze 756: struct expr *p;
1.3 schwarze 757: char *key;
1.5 kristaps 758: int i;
759:
760: memset(&e, 0, sizeof(struct expr));
1.1 schwarze 761:
1.5 kristaps 762: /* Choose regex or substring match. */
1.3 schwarze 763:
1.4 kristaps 764: if (NULL == (e.v = strpbrk(buf, "=~"))) {
1.3 schwarze 765: e.regex = 0;
1.4 kristaps 766: e.v = buf;
1.3 schwarze 767: } else {
768: e.regex = '~' == *e.v;
769: *e.v++ = '\0';
770: }
1.1 schwarze 771:
1.5 kristaps 772: /* Determine the record types to search for. */
1.3 schwarze 773:
774: e.mask = 0;
1.4 kristaps 775: if (buf < e.v) {
776: while (NULL != (key = strsep(&buf, ","))) {
1.3 schwarze 777: i = 0;
778: while (types[i].mask &&
1.4 kristaps 779: strcmp(types[i].name, key))
1.3 schwarze 780: i++;
781: e.mask |= types[i].mask;
782: }
783: }
784: if (0 == e.mask)
785: e.mask = TYPE_Nm | TYPE_Nd;
1.1 schwarze 786:
1.5 kristaps 787: if (e.regex) {
1.13 kristaps 788: i = REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE);
1.5 kristaps 789: if (regcomp(&e.re, e.v, i))
790: return(NULL);
791: }
1.1 schwarze 792:
1.3 schwarze 793: e.v = mandoc_strdup(e.v);
1.1 schwarze 794:
795: p = mandoc_calloc(1, sizeof(struct expr));
796: memcpy(p, &e, sizeof(struct expr));
797: return(p);
798: }
799:
800: void
801: exprfree(struct expr *p)
802: {
1.5 kristaps 803: struct expr *pp;
1.12 schwarze 804:
1.5 kristaps 805: while (NULL != p) {
806: if (p->subexpr)
807: exprfree(p->subexpr);
808: if (p->regex)
809: regfree(&p->re);
810: free(p->v);
811: pp = p->next;
812: free(p);
813: p = pp;
814: }
815: }
1.1 schwarze 816:
1.5 kristaps 817: static int
1.12 schwarze 818: exprmark(const struct expr *p, const char *cp,
1.6 kristaps 819: uint64_t mask, int *ms)
1.5 kristaps 820: {
821:
822: for ( ; p; p = p->next) {
823: if (p->subexpr) {
824: if (exprmark(p->subexpr, cp, mask, ms))
825: return(1);
826: continue;
827: } else if ( ! (mask & p->mask))
828: continue;
1.1 schwarze 829:
1.5 kristaps 830: if (p->regex) {
831: if (regexec(&p->re, cp, 0, NULL, 0))
832: continue;
1.16 kristaps 833: } else if (NULL == strcasestr(cp, p->v))
834: continue;
1.5 kristaps 835:
836: if (NULL == ms)
837: return(1);
838: else
839: ms[p->index] = 1;
840: }
1.1 schwarze 841:
1.5 kristaps 842: return(0);
1.1 schwarze 843: }
844:
845: static int
1.5 kristaps 846: expreval(const struct expr *p, int *ms)
1.1 schwarze 847: {
1.5 kristaps 848: int match;
1.1 schwarze 849:
1.5 kristaps 850: /*
851: * AND has precedence over OR. Analysis is left-right, though
852: * it doesn't matter because there are no side-effects.
853: * Thus, step through pairwise ANDs and accumulate their Boolean
854: * evaluation. If we encounter a single true AND collection or
855: * standalone term, the whole expression is true (by definition
856: * of OR).
857: */
858:
859: for (match = 0; p && ! match; p = p->next) {
860: /* Evaluate a subexpression, if applicable. */
861: if (p->subexpr && ! ms[p->index])
862: ms[p->index] = expreval(p->subexpr, ms);
863:
864: match = ms[p->index];
865: for ( ; p->next && p->next->and; p = p->next) {
866: /* Evaluate a subexpression, if applicable. */
867: if (p->next->subexpr && ! ms[p->next->index])
1.12 schwarze 868: ms[p->next->index] =
1.5 kristaps 869: expreval(p->next->subexpr, ms);
870: match = match && ms[p->next->index];
871: }
872: }
873:
874: return(match);
875: }
876:
877: /*
878: * First, update the array of terms for which this expression evaluates
879: * to true.
880: * Second, logically evaluate all terms over the updated array of truth
881: * values.
882: * If this evaluates to true, mark the expression as satisfied.
883: */
884: static void
1.12 schwarze 885: exprexec(const struct expr *e, const char *cp,
1.30 kristaps 886: uint64_t mask, struct res *r)
1.5 kristaps 887: {
1.1 schwarze 888:
1.5 kristaps 889: assert(0 == r->matched);
1.12 schwarze 890: exprmark(e, cp, mask, r->matches);
891: r->matched = expreval(e, r->matches);
1.1 schwarze 892: }
CVSweb