Annotation of mandoc/apropos_db.c, Revision 1.27
1.27 ! schwarze 1: /* $Id: apropos_db.c,v 1.26 2011/12/16 20:06:58 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" },
1.25 kristaps 120: { UINT64_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));
156: info.flags = R_DUP;
157:
1.2 schwarze 158: db = dbopen(MANDOC_DB, O_RDONLY, 0, DB_BTREE, &info);
1.12 schwarze 159: if (NULL != db)
1.1 schwarze 160: return(db);
161:
162: return(NULL);
163: }
164:
165: /*
166: * Read a keyword from the database and normalise it.
167: * Return 0 if the database is insane, else 1.
168: */
169: static int
1.17 kristaps 170: btree_read(const DBT *k, const DBT *v,
171: const struct mchars *mc,
172: struct db_val *dbv, char **buf)
1.1 schwarze 173: {
1.27 ! schwarze 174: struct db_val raw_dbv;
1.1 schwarze 175:
1.17 kristaps 176: /* Are our sizes sane? */
177: if (k->size < 2 || sizeof(struct db_val) != v->size)
178: return(0);
1.6 kristaps 179:
1.17 kristaps 180: /* Is our string nil-terminated? */
181: if ('\0' != ((const char *)k->data)[(int)k->size - 1])
1.1 schwarze 182: return(0);
183:
1.17 kristaps 184: norm_string((const char *)k->data, mc, buf);
1.27 ! schwarze 185: memcpy(&raw_dbv, v->data, v->size);
! 186: dbv->rec = betoh32(raw_dbv.rec);
! 187: dbv->mask = betoh64(raw_dbv.mask);
1.1 schwarze 188: return(1);
189: }
190:
191: /*
192: * Take a Unicode codepoint and produce its UTF-8 encoding.
193: * This isn't the best way to do this, but it works.
1.12 schwarze 194: * The magic numbers are from the UTF-8 packaging.
1.1 schwarze 195: * They're not as scary as they seem: read the UTF-8 spec for details.
196: */
197: static size_t
198: norm_utf8(unsigned int cp, char out[7])
199: {
1.26 kristaps 200: int rc;
1.1 schwarze 201:
202: rc = 0;
203:
204: if (cp <= 0x0000007F) {
205: rc = 1;
206: out[0] = (char)cp;
207: } else if (cp <= 0x000007FF) {
208: rc = 2;
209: out[0] = (cp >> 6 & 31) | 192;
210: out[1] = (cp & 63) | 128;
211: } else if (cp <= 0x0000FFFF) {
212: rc = 3;
213: out[0] = (cp >> 12 & 15) | 224;
214: out[1] = (cp >> 6 & 63) | 128;
215: out[2] = (cp & 63) | 128;
216: } else if (cp <= 0x001FFFFF) {
217: rc = 4;
218: out[0] = (cp >> 18 & 7) | 240;
219: out[1] = (cp >> 12 & 63) | 128;
220: out[2] = (cp >> 6 & 63) | 128;
221: out[3] = (cp & 63) | 128;
222: } else if (cp <= 0x03FFFFFF) {
223: rc = 5;
224: out[0] = (cp >> 24 & 3) | 248;
225: out[1] = (cp >> 18 & 63) | 128;
226: out[2] = (cp >> 12 & 63) | 128;
227: out[3] = (cp >> 6 & 63) | 128;
228: out[4] = (cp & 63) | 128;
229: } else if (cp <= 0x7FFFFFFF) {
230: rc = 6;
231: out[0] = (cp >> 30 & 1) | 252;
232: out[1] = (cp >> 24 & 63) | 128;
233: out[2] = (cp >> 18 & 63) | 128;
234: out[3] = (cp >> 12 & 63) | 128;
235: out[4] = (cp >> 6 & 63) | 128;
236: out[5] = (cp & 63) | 128;
237: } else
238: return(0);
239:
240: out[rc] = '\0';
1.26 kristaps 241: return((size_t)rc);
1.1 schwarze 242: }
243:
244: /*
245: * Normalise strings from the index and database.
246: * These strings are escaped as defined by mandoc_char(7) along with
247: * other goop in mandoc.h (e.g., soft hyphens).
248: * This function normalises these into a nice UTF-8 string.
249: * Returns 0 if the database is fucked.
250: */
251: static void
252: norm_string(const char *val, const struct mchars *mc, char **buf)
253: {
254: size_t sz, bsz;
255: char utfbuf[7];
256: const char *seq, *cpp;
257: int len, u, pos;
258: enum mandoc_esc esc;
1.12 schwarze 259: static const char res[] = { '\\', '\t',
1.1 schwarze 260: ASCII_NBRSP, ASCII_HYPH, '\0' };
261:
262: /* Pre-allocate by the length of the input */
263:
264: bsz = strlen(val) + 1;
265: *buf = mandoc_realloc(*buf, bsz);
266: pos = 0;
267:
268: while ('\0' != *val) {
269: /*
270: * Halt on the first escape sequence.
271: * This also halts on the end of string, in which case
272: * we just copy, fallthrough, and exit the loop.
273: */
274: if ((sz = strcspn(val, res)) > 0) {
275: memcpy(&(*buf)[pos], val, sz);
276: pos += (int)sz;
277: val += (int)sz;
278: }
279:
280: if (ASCII_HYPH == *val) {
281: (*buf)[pos++] = '-';
282: val++;
283: continue;
284: } else if ('\t' == *val || ASCII_NBRSP == *val) {
285: (*buf)[pos++] = ' ';
286: val++;
287: continue;
288: } else if ('\\' != *val)
289: break;
290:
291: /* Read past the slash. */
292:
293: val++;
294: u = 0;
295:
296: /*
297: * Parse the escape sequence and see if it's a
298: * predefined character or special character.
299: */
300:
301: esc = mandoc_escape(&val, &seq, &len);
302: if (ESCAPE_ERROR == esc)
303: break;
304:
1.12 schwarze 305: /*
1.1 schwarze 306: * XXX - this just does UTF-8, but we need to know
307: * beforehand whether we should do text substitution.
308: */
309:
310: switch (esc) {
311: case (ESCAPE_SPECIAL):
312: if (0 != (u = mchars_spec2cp(mc, seq, len)))
313: break;
314: /* FALLTHROUGH */
315: default:
316: continue;
317: }
318:
319: /*
320: * If we have a Unicode codepoint, try to convert that
321: * to a UTF-8 byte string.
322: */
323:
324: cpp = utfbuf;
325: if (0 == (sz = norm_utf8(u, utfbuf)))
326: continue;
327:
328: /* Copy the rendered glyph into the stream. */
329:
330: sz = strlen(cpp);
331: bsz += sz;
332:
333: *buf = mandoc_realloc(*buf, bsz);
334:
335: memcpy(&(*buf)[pos], cpp, sz);
336: pos += (int)sz;
337: }
338:
339: (*buf)[pos] = '\0';
340: }
341:
342: /*
343: * Open the filename-index mandoc-db database.
344: * Returns NULL if opening failed.
345: */
346: static DB *
347: index_open(void)
348: {
349: DB *db;
350:
1.2 schwarze 351: db = dbopen(MANDOC_IDX, O_RDONLY, 0, DB_RECNO, NULL);
1.1 schwarze 352: if (NULL != db)
353: return(db);
354:
355: return(NULL);
356: }
357:
358: /*
359: * Safely unpack from an index file record into the structure.
360: * Returns 1 if an entry was unpacked, 0 if the database is insane.
361: */
362: static int
1.11 kristaps 363: index_read(const DBT *key, const DBT *val, int index,
1.1 schwarze 364: const struct mchars *mc, struct rec *rec)
365: {
366: size_t left;
367: char *np, *cp;
1.24 kristaps 368: char type;
1.1 schwarze 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:
1.24 kristaps 379: if (0 == (left = val->size))
380: return(0);
1.1 schwarze 381:
1.24 kristaps 382: cp = val->data;
1.27 ! schwarze 383: assert(sizeof(recno_t) == key->size);
! 384: memcpy(&rec->res.rec, key->data, key->size);
1.11 kristaps 385: rec->res.volume = index;
1.1 schwarze 386:
1.24 kristaps 387: if ('d' == (type = *cp++))
388: rec->res.type = RESTYPE_MDOC;
389: else if ('a' == type)
390: rec->res.type = RESTYPE_MAN;
391: else if ('c' == type)
392: rec->res.type = RESTYPE_CAT;
393: else
394: return(0);
395:
396: left--;
1.5 kristaps 397: INDEX_BREAD(rec->res.file);
398: INDEX_BREAD(rec->res.cat);
399: INDEX_BREAD(rec->res.title);
400: INDEX_BREAD(rec->res.arch);
401: INDEX_BREAD(rec->res.desc);
1.1 schwarze 402: return(1);
403: }
404:
405: /*
1.10 kristaps 406: * Search mandocdb databases in paths for expression "expr".
1.1 schwarze 407: * Filter out by "opts".
408: * Call "res" with the results, which may be zero.
1.5 kristaps 409: * Return 0 if there was a database error, else return 1.
1.1 schwarze 410: */
1.5 kristaps 411: int
1.10 kristaps 412: apropos_search(int pathsz, char **paths, const struct opts *opts,
1.12 schwarze 413: const struct expr *expr, size_t terms, void *arg,
1.5 kristaps 414: void (*res)(struct res *, size_t, void *))
1.1 schwarze 415: {
1.8 kristaps 416: struct rectree tree;
417: struct mchars *mc;
418: struct res *ress;
419: int i, mlen, rc;
420:
421: memset(&tree, 0, sizeof(struct rectree));
422:
1.10 kristaps 423: rc = 0;
1.8 kristaps 424: mc = mchars_alloc();
425:
1.10 kristaps 426: /*
427: * Main loop. Change into the directory containing manpage
428: * databases. Run our expession over each database in the set.
429: */
430:
431: for (i = 0; i < pathsz; i++) {
432: if (chdir(paths[i]))
1.8 kristaps 433: continue;
1.11 kristaps 434: if ( ! single_search(&tree, opts, expr, terms, mc, i))
1.10 kristaps 435: goto out;
1.8 kristaps 436: }
437:
438: /*
1.10 kristaps 439: * Count matching files, transfer to a "clean" array, then feed
440: * them to the output handler.
1.8 kristaps 441: */
442:
443: for (mlen = i = 0; i < tree.len; i++)
444: if (tree.node[i].matched)
445: mlen++;
446:
447: ress = mandoc_malloc(mlen * sizeof(struct res));
448:
449: for (mlen = i = 0; i < tree.len; i++)
450: if (tree.node[i].matched)
1.12 schwarze 451: memcpy(&ress[mlen++], &tree.node[i].res,
1.8 kristaps 452: sizeof(struct res));
453:
454: (*res)(ress, mlen, arg);
455: free(ress);
456:
1.10 kristaps 457: rc = 1;
458: out:
1.8 kristaps 459: for (i = 0; i < tree.len; i++)
460: recfree(&tree.node[i]);
461:
462: free(tree.node);
463: mchars_free(mc);
464: return(rc);
465: }
466:
467: static int
468: single_search(struct rectree *tree, const struct opts *opts,
469: const struct expr *expr, size_t terms,
1.11 kristaps 470: struct mchars *mc, int vol)
1.8 kristaps 471: {
472: int root, leaf, ch;
1.1 schwarze 473: DBT key, val;
474: DB *btree, *idx;
475: char *buf;
1.5 kristaps 476: struct rec *rs;
477: struct rec r;
1.17 kristaps 478: struct db_val vb;
1.1 schwarze 479:
480: root = -1;
481: leaf = -1;
482: btree = NULL;
483: idx = NULL;
484: buf = NULL;
1.8 kristaps 485: rs = tree->node;
1.1 schwarze 486:
1.5 kristaps 487: memset(&r, 0, sizeof(struct rec));
1.1 schwarze 488:
1.12 schwarze 489: if (NULL == (btree = btree_open()))
1.10 kristaps 490: return(1);
1.1 schwarze 491:
1.8 kristaps 492: if (NULL == (idx = index_open())) {
493: (*btree->close)(btree);
1.10 kristaps 494: return(1);
1.8 kristaps 495: }
1.1 schwarze 496:
497: while (0 == (ch = (*btree->seq)(btree, &key, &val, R_NEXT))) {
1.17 kristaps 498: if ( ! btree_read(&key, &val, mc, &vb, &buf))
1.1 schwarze 499: break;
500:
1.5 kristaps 501: /*
502: * See if this keyword record matches any of the
503: * expressions we have stored.
504: */
1.17 kristaps 505: if ( ! exprmark(expr, buf, vb.mask, NULL))
1.1 schwarze 506: continue;
507:
508: /*
509: * O(log n) scan for prior records. Since a record
510: * number is unbounded, this has decent performance over
511: * a complex hash function.
512: */
513:
514: for (leaf = root; leaf >= 0; )
1.17 kristaps 515: if (vb.rec > rs[leaf].res.rec &&
1.5 kristaps 516: rs[leaf].rhs >= 0)
517: leaf = rs[leaf].rhs;
1.17 kristaps 518: else if (vb.rec < rs[leaf].res.rec &&
1.5 kristaps 519: rs[leaf].lhs >= 0)
520: leaf = rs[leaf].lhs;
1.12 schwarze 521: else
1.1 schwarze 522: break;
523:
1.5 kristaps 524: /*
525: * If we find a record, see if it has already evaluated
526: * to true. If it has, great, just keep going. If not,
527: * try to evaluate it now and continue anyway.
528: */
529:
1.17 kristaps 530: if (leaf >= 0 && rs[leaf].res.rec == vb.rec) {
1.5 kristaps 531: if (0 == rs[leaf].matched)
1.17 kristaps 532: exprexec(expr, buf, vb.mask, &rs[leaf]);
1.1 schwarze 533: continue;
1.5 kristaps 534: }
1.1 schwarze 535:
536: /*
1.5 kristaps 537: * We have a new file to examine.
538: * Extract the manpage's metadata from the index
539: * database, then begin partial evaluation.
1.1 schwarze 540: */
541:
1.17 kristaps 542: key.data = &vb.rec;
1.1 schwarze 543: key.size = sizeof(recno_t);
544:
545: if (0 != (*idx->get)(idx, &key, &val, 0))
546: break;
547:
1.5 kristaps 548: r.lhs = r.rhs = -1;
1.11 kristaps 549: if ( ! index_read(&key, &val, vol, mc, &r))
1.1 schwarze 550: break;
551:
1.5 kristaps 552: /* XXX: this should be elsewhere, I guess? */
553:
554: if (opts->cat && strcasecmp(opts->cat, r.res.cat))
1.1 schwarze 555: continue;
1.22 kristaps 556:
557: if (opts->arch && *r.res.arch)
558: if (strcasecmp(opts->arch, r.res.arch))
559: continue;
1.1 schwarze 560:
1.8 kristaps 561: tree->node = rs = mandoc_realloc
562: (rs, (tree->len + 1) * sizeof(struct rec));
1.1 schwarze 563:
1.8 kristaps 564: memcpy(&rs[tree->len], &r, sizeof(struct rec));
1.23 kristaps 565: memset(&r, 0, sizeof(struct rec));
1.12 schwarze 566: rs[tree->len].matches =
1.8 kristaps 567: mandoc_calloc(terms, sizeof(int));
1.1 schwarze 568:
1.17 kristaps 569: exprexec(expr, buf, vb.mask, &rs[tree->len]);
1.12 schwarze 570:
1.1 schwarze 571: /* Append to our tree. */
572:
573: if (leaf >= 0) {
1.17 kristaps 574: if (vb.rec > rs[leaf].res.rec)
1.8 kristaps 575: rs[leaf].rhs = tree->len;
1.1 schwarze 576: else
1.8 kristaps 577: rs[leaf].lhs = tree->len;
1.1 schwarze 578: } else
1.8 kristaps 579: root = tree->len;
1.12 schwarze 580:
1.8 kristaps 581: tree->len++;
1.1 schwarze 582: }
1.12 schwarze 583:
1.8 kristaps 584: (*btree->close)(btree);
585: (*idx->close)(idx);
1.1 schwarze 586:
587: free(buf);
1.23 kristaps 588: recfree(&r);
1.8 kristaps 589: return(1 == ch);
1.5 kristaps 590: }
591:
592: static void
593: recfree(struct rec *rec)
594: {
595:
596: free(rec->res.file);
597: free(rec->res.cat);
598: free(rec->res.title);
599: free(rec->res.arch);
600: free(rec->res.desc);
601:
602: free(rec->matches);
1.1 schwarze 603: }
604:
1.13 kristaps 605: /*
606: * Compile a list of straight-up terms.
607: * The arguments are re-written into ~[[:<:]]term[[:>:]], or "term"
608: * surrounded by word boundaries, then pumped through exprterm().
609: * Terms are case-insensitive.
610: * This emulates whatis(1) behaviour.
611: */
612: struct expr *
613: termcomp(int argc, char *argv[], size_t *tt)
614: {
615: char *buf;
616: int pos;
617: struct expr *e, *next;
618: size_t sz;
619:
620: buf = NULL;
621: e = NULL;
622: *tt = 0;
623:
1.15 schwarze 624: for (pos = argc - 1; pos >= 0; pos--) {
625: sz = strlen(argv[pos]) + 18;
1.13 kristaps 626: buf = mandoc_realloc(buf, sz);
1.15 schwarze 627: strlcpy(buf, "Nm~[[:<:]]", sz);
1.13 kristaps 628: strlcat(buf, argv[pos], sz);
629: strlcat(buf, "[[:>:]]", sz);
630: if (NULL == (next = exprterm(buf, 0))) {
631: free(buf);
632: exprfree(e);
633: return(NULL);
634: }
1.15 schwarze 635: next->next = e;
1.13 kristaps 636: e = next;
637: (*tt)++;
638: }
639:
640: free(buf);
641: return(e);
642: }
643:
644: /*
645: * Compile a sequence of logical expressions.
646: * See apropos.1 for a grammar of this sequence.
647: */
1.1 schwarze 648: struct expr *
1.5 kristaps 649: exprcomp(int argc, char *argv[], size_t *tt)
1.1 schwarze 650: {
1.5 kristaps 651: int pos, lvl;
652: struct expr *e;
653:
654: pos = lvl = 0;
655: *tt = 0;
656:
657: e = exprexpr(argc, argv, &pos, &lvl, tt);
658:
659: if (0 == lvl && pos >= argc)
660: return(e);
661:
662: exprfree(e);
663: return(NULL);
664: }
665:
666: /*
667: * Compile an array of tokens into an expression.
668: * An informal expression grammar is defined in apropos(1).
669: * Return NULL if we fail doing so. All memory will be cleaned up.
670: * Return the root of the expression sequence if alright.
671: */
672: static struct expr *
1.9 kristaps 673: exprexpr(int argc, char *argv[], int *pos, int *lvl, size_t *tt)
1.5 kristaps 674: {
675: struct expr *e, *first, *next;
676: int log;
677:
678: first = next = NULL;
679:
680: for ( ; *pos < argc; (*pos)++) {
681: e = next;
682:
683: /*
684: * Close out a subexpression.
685: */
686:
687: if (NULL != e && 0 == strcmp(")", argv[*pos])) {
688: if (--(*lvl) < 0)
689: goto err;
690: break;
691: }
692:
693: /*
694: * Small note: if we're just starting, don't let "-a"
695: * and "-o" be considered logical operators: they're
696: * just tokens unless pairwise joining, in which case we
697: * record their existence (or assume "OR").
698: */
699: log = 0;
700:
701: if (NULL != e && 0 == strcmp("-a", argv[*pos]))
1.12 schwarze 702: log = 1;
1.5 kristaps 703: else if (NULL != e && 0 == strcmp("-o", argv[*pos]))
704: log = 2;
705:
706: if (log > 0 && ++(*pos) >= argc)
707: goto err;
708:
709: /*
710: * Now we parse the term part. This can begin with
711: * "-i", in which case the expression is case
712: * insensitive.
713: */
714:
715: if (0 == strcmp("(", argv[*pos])) {
716: ++(*pos);
717: ++(*lvl);
718: next = mandoc_calloc(1, sizeof(struct expr));
719: next->subexpr = exprexpr(argc, argv, pos, lvl, tt);
720: if (NULL == next->subexpr) {
721: free(next);
722: next = NULL;
723: }
724: } else if (0 == strcmp("-i", argv[*pos])) {
725: if (++(*pos) >= argc)
726: goto err;
727: next = exprterm(argv[*pos], 0);
728: } else
729: next = exprterm(argv[*pos], 1);
730:
731: if (NULL == next)
732: goto err;
733:
734: next->and = log == 1;
735: next->index = (int)(*tt)++;
736:
737: /* Append to our chain of expressions. */
738:
739: if (NULL == first) {
740: assert(NULL == e);
741: first = next;
742: } else {
743: assert(NULL != e);
744: e->next = next;
745: }
746: }
747:
748: return(first);
749: err:
750: exprfree(first);
751: return(NULL);
752: }
753:
754: /*
755: * Parse a terminal expression with the grammar as defined in
756: * apropos(1).
757: * Return NULL if we fail the parse.
758: */
759: static struct expr *
760: exprterm(char *buf, int cs)
761: {
762: struct expr e;
1.1 schwarze 763: struct expr *p;
1.3 schwarze 764: char *key;
1.5 kristaps 765: int i;
766:
767: memset(&e, 0, sizeof(struct expr));
1.1 schwarze 768:
1.5 kristaps 769: /* Choose regex or substring match. */
1.3 schwarze 770:
1.4 kristaps 771: if (NULL == (e.v = strpbrk(buf, "=~"))) {
1.3 schwarze 772: e.regex = 0;
1.4 kristaps 773: e.v = buf;
1.3 schwarze 774: } else {
775: e.regex = '~' == *e.v;
776: *e.v++ = '\0';
777: }
1.1 schwarze 778:
1.5 kristaps 779: /* Determine the record types to search for. */
1.3 schwarze 780:
781: e.mask = 0;
1.4 kristaps 782: if (buf < e.v) {
783: while (NULL != (key = strsep(&buf, ","))) {
1.3 schwarze 784: i = 0;
785: while (types[i].mask &&
1.4 kristaps 786: strcmp(types[i].name, key))
1.3 schwarze 787: i++;
788: e.mask |= types[i].mask;
789: }
790: }
791: if (0 == e.mask)
792: e.mask = TYPE_Nm | TYPE_Nd;
1.1 schwarze 793:
1.5 kristaps 794: if (e.regex) {
1.13 kristaps 795: i = REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE);
1.5 kristaps 796: if (regcomp(&e.re, e.v, i))
797: return(NULL);
798: }
1.1 schwarze 799:
1.3 schwarze 800: e.v = mandoc_strdup(e.v);
1.1 schwarze 801:
802: p = mandoc_calloc(1, sizeof(struct expr));
803: memcpy(p, &e, sizeof(struct expr));
804: return(p);
805: }
806:
807: void
808: exprfree(struct expr *p)
809: {
1.5 kristaps 810: struct expr *pp;
1.12 schwarze 811:
1.5 kristaps 812: while (NULL != p) {
813: if (p->subexpr)
814: exprfree(p->subexpr);
815: if (p->regex)
816: regfree(&p->re);
817: free(p->v);
818: pp = p->next;
819: free(p);
820: p = pp;
821: }
822: }
1.1 schwarze 823:
1.5 kristaps 824: static int
1.12 schwarze 825: exprmark(const struct expr *p, const char *cp,
1.6 kristaps 826: uint64_t mask, int *ms)
1.5 kristaps 827: {
828:
829: for ( ; p; p = p->next) {
830: if (p->subexpr) {
831: if (exprmark(p->subexpr, cp, mask, ms))
832: return(1);
833: continue;
834: } else if ( ! (mask & p->mask))
835: continue;
1.1 schwarze 836:
1.5 kristaps 837: if (p->regex) {
838: if (regexec(&p->re, cp, 0, NULL, 0))
839: continue;
1.16 kristaps 840: } else if (NULL == strcasestr(cp, p->v))
841: continue;
1.5 kristaps 842:
843: if (NULL == ms)
844: return(1);
845: else
846: ms[p->index] = 1;
847: }
1.1 schwarze 848:
1.5 kristaps 849: return(0);
1.1 schwarze 850: }
851:
852: static int
1.5 kristaps 853: expreval(const struct expr *p, int *ms)
1.1 schwarze 854: {
1.5 kristaps 855: int match;
1.1 schwarze 856:
1.5 kristaps 857: /*
858: * AND has precedence over OR. Analysis is left-right, though
859: * it doesn't matter because there are no side-effects.
860: * Thus, step through pairwise ANDs and accumulate their Boolean
861: * evaluation. If we encounter a single true AND collection or
862: * standalone term, the whole expression is true (by definition
863: * of OR).
864: */
865:
866: for (match = 0; p && ! match; p = p->next) {
867: /* Evaluate a subexpression, if applicable. */
868: if (p->subexpr && ! ms[p->index])
869: ms[p->index] = expreval(p->subexpr, ms);
870:
871: match = ms[p->index];
872: for ( ; p->next && p->next->and; p = p->next) {
873: /* Evaluate a subexpression, if applicable. */
874: if (p->next->subexpr && ! ms[p->next->index])
1.12 schwarze 875: ms[p->next->index] =
1.5 kristaps 876: expreval(p->next->subexpr, ms);
877: match = match && ms[p->next->index];
878: }
879: }
880:
881: return(match);
882: }
883:
884: /*
885: * First, update the array of terms for which this expression evaluates
886: * to true.
887: * Second, logically evaluate all terms over the updated array of truth
888: * values.
889: * If this evaluates to true, mark the expression as satisfied.
890: */
891: static void
1.12 schwarze 892: exprexec(const struct expr *e, const char *cp,
1.6 kristaps 893: uint64_t mask, struct rec *r)
1.5 kristaps 894: {
1.1 schwarze 895:
1.5 kristaps 896: assert(0 == r->matched);
1.12 schwarze 897: exprmark(e, cp, mask, r->matches);
898: r->matched = expreval(e, r->matches);
1.1 schwarze 899: }
CVSweb