Annotation of mandoc/apropos_db.c, Revision 1.2
1.2 ! schwarze 1: /* $Id: apropos_db.c,v 1.1 2011/11/13 10:12:05 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4: *
5: * Permission to use, copy, modify, and distribute this software for any
6: * purpose with or without fee is hereby granted, provided that the above
7: * copyright notice and this permission notice appear in all copies.
8: *
9: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
10: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
11: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
12: * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
13: * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
14: * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
15: * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
16: */
17: #include <assert.h>
18: #include <fcntl.h>
19: #include <regex.h>
20: #include <stdarg.h>
21: #include <stdlib.h>
22: #include <string.h>
23:
24: #ifdef __linux__
25: # include <db_185.h>
26: #else
27: # include <db.h>
28: #endif
29:
1.2 ! schwarze 30: #include "mandocdb.h"
1.1 schwarze 31: #include "apropos_db.h"
32: #include "mandoc.h"
33:
34: enum match {
35: MATCH_REGEX,
36: MATCH_REGEXCASE,
37: MATCH_STR,
38: MATCH_STRCASE
39: };
40:
41: struct expr {
42: enum match match;
43: int mask;
44: char *v;
45: regex_t re;
46: };
47:
48: struct type {
49: int mask;
50: const char *name;
51: };
52:
53: static const struct type types[] = {
1.2 ! schwarze 54: { TYPE_An, "An" },
! 55: { TYPE_Cd, "Cd" },
! 56: { TYPE_Er, "Er" },
! 57: { TYPE_Ev, "Ev" },
! 58: { TYPE_Fn, "Fn" },
! 59: { TYPE_Fn, "Fo" },
! 60: { TYPE_In, "In" },
! 61: { TYPE_Nd, "Nd" },
! 62: { TYPE_Nm, "Nm" },
! 63: { TYPE_Pa, "Pa" },
! 64: { TYPE_St, "St" },
! 65: { TYPE_Va, "Va" },
! 66: { TYPE_Va, "Vt" },
! 67: { TYPE_Xr, "Xr" },
! 68: { INT_MAX, "any" },
1.1 schwarze 69: { 0, NULL }
70: };
71:
72: static DB *btree_open(void);
73: static int btree_read(const DBT *, const struct mchars *, char **);
74: static int exprexec(const struct expr *, char *);
75: static DB *index_open(void);
76: static int index_read(const DBT *, const DBT *,
77: const struct mchars *, struct rec *);
78: static void norm_string(const char *,
79: const struct mchars *, char **);
80: static size_t norm_utf8(unsigned int, char[7]);
81:
82: /*
83: * Open the keyword mandoc-db database.
84: */
85: static DB *
86: btree_open(void)
87: {
88: BTREEINFO info;
89: DB *db;
90:
91: memset(&info, 0, sizeof(BTREEINFO));
92: info.flags = R_DUP;
93:
1.2 ! schwarze 94: db = dbopen(MANDOC_DB, O_RDONLY, 0, DB_BTREE, &info);
1.1 schwarze 95: if (NULL != db)
96: return(db);
97:
98: return(NULL);
99: }
100:
101: /*
102: * Read a keyword from the database and normalise it.
103: * Return 0 if the database is insane, else 1.
104: */
105: static int
106: btree_read(const DBT *v, const struct mchars *mc, char **buf)
107: {
108:
109: /* Sanity: are we nil-terminated? */
110:
111: assert(v->size > 0);
112: if ('\0' != ((char *)v->data)[(int)v->size - 1])
113: return(0);
114:
115: norm_string((char *)v->data, mc, buf);
116: return(1);
117: }
118:
119: /*
120: * Take a Unicode codepoint and produce its UTF-8 encoding.
121: * This isn't the best way to do this, but it works.
122: * The magic numbers are from the UTF-8 packaging.
123: * They're not as scary as they seem: read the UTF-8 spec for details.
124: */
125: static size_t
126: norm_utf8(unsigned int cp, char out[7])
127: {
128: size_t rc;
129:
130: rc = 0;
131:
132: if (cp <= 0x0000007F) {
133: rc = 1;
134: out[0] = (char)cp;
135: } else if (cp <= 0x000007FF) {
136: rc = 2;
137: out[0] = (cp >> 6 & 31) | 192;
138: out[1] = (cp & 63) | 128;
139: } else if (cp <= 0x0000FFFF) {
140: rc = 3;
141: out[0] = (cp >> 12 & 15) | 224;
142: out[1] = (cp >> 6 & 63) | 128;
143: out[2] = (cp & 63) | 128;
144: } else if (cp <= 0x001FFFFF) {
145: rc = 4;
146: out[0] = (cp >> 18 & 7) | 240;
147: out[1] = (cp >> 12 & 63) | 128;
148: out[2] = (cp >> 6 & 63) | 128;
149: out[3] = (cp & 63) | 128;
150: } else if (cp <= 0x03FFFFFF) {
151: rc = 5;
152: out[0] = (cp >> 24 & 3) | 248;
153: out[1] = (cp >> 18 & 63) | 128;
154: out[2] = (cp >> 12 & 63) | 128;
155: out[3] = (cp >> 6 & 63) | 128;
156: out[4] = (cp & 63) | 128;
157: } else if (cp <= 0x7FFFFFFF) {
158: rc = 6;
159: out[0] = (cp >> 30 & 1) | 252;
160: out[1] = (cp >> 24 & 63) | 128;
161: out[2] = (cp >> 18 & 63) | 128;
162: out[3] = (cp >> 12 & 63) | 128;
163: out[4] = (cp >> 6 & 63) | 128;
164: out[5] = (cp & 63) | 128;
165: } else
166: return(0);
167:
168: out[rc] = '\0';
169: return(rc);
170: }
171:
172: /*
173: * Normalise strings from the index and database.
174: * These strings are escaped as defined by mandoc_char(7) along with
175: * other goop in mandoc.h (e.g., soft hyphens).
176: * This function normalises these into a nice UTF-8 string.
177: * Returns 0 if the database is fucked.
178: */
179: static void
180: norm_string(const char *val, const struct mchars *mc, char **buf)
181: {
182: size_t sz, bsz;
183: char utfbuf[7];
184: const char *seq, *cpp;
185: int len, u, pos;
186: enum mandoc_esc esc;
187: static const char res[] = { '\\', '\t',
188: ASCII_NBRSP, ASCII_HYPH, '\0' };
189:
190: /* Pre-allocate by the length of the input */
191:
192: bsz = strlen(val) + 1;
193: *buf = mandoc_realloc(*buf, bsz);
194: pos = 0;
195:
196: while ('\0' != *val) {
197: /*
198: * Halt on the first escape sequence.
199: * This also halts on the end of string, in which case
200: * we just copy, fallthrough, and exit the loop.
201: */
202: if ((sz = strcspn(val, res)) > 0) {
203: memcpy(&(*buf)[pos], val, sz);
204: pos += (int)sz;
205: val += (int)sz;
206: }
207:
208: if (ASCII_HYPH == *val) {
209: (*buf)[pos++] = '-';
210: val++;
211: continue;
212: } else if ('\t' == *val || ASCII_NBRSP == *val) {
213: (*buf)[pos++] = ' ';
214: val++;
215: continue;
216: } else if ('\\' != *val)
217: break;
218:
219: /* Read past the slash. */
220:
221: val++;
222: u = 0;
223:
224: /*
225: * Parse the escape sequence and see if it's a
226: * predefined character or special character.
227: */
228:
229: esc = mandoc_escape(&val, &seq, &len);
230: if (ESCAPE_ERROR == esc)
231: break;
232:
233: /*
234: * XXX - this just does UTF-8, but we need to know
235: * beforehand whether we should do text substitution.
236: */
237:
238: switch (esc) {
239: case (ESCAPE_SPECIAL):
240: if (0 != (u = mchars_spec2cp(mc, seq, len)))
241: break;
242: /* FALLTHROUGH */
243: default:
244: continue;
245: }
246:
247: /*
248: * If we have a Unicode codepoint, try to convert that
249: * to a UTF-8 byte string.
250: */
251:
252: cpp = utfbuf;
253: if (0 == (sz = norm_utf8(u, utfbuf)))
254: continue;
255:
256: /* Copy the rendered glyph into the stream. */
257:
258: sz = strlen(cpp);
259: bsz += sz;
260:
261: *buf = mandoc_realloc(*buf, bsz);
262:
263: memcpy(&(*buf)[pos], cpp, sz);
264: pos += (int)sz;
265: }
266:
267: (*buf)[pos] = '\0';
268: }
269:
270: /*
271: * Open the filename-index mandoc-db database.
272: * Returns NULL if opening failed.
273: */
274: static DB *
275: index_open(void)
276: {
277: DB *db;
278:
1.2 ! schwarze 279: db = dbopen(MANDOC_IDX, O_RDONLY, 0, DB_RECNO, NULL);
1.1 schwarze 280: if (NULL != db)
281: return(db);
282:
283: return(NULL);
284: }
285:
286: /*
287: * Safely unpack from an index file record into the structure.
288: * Returns 1 if an entry was unpacked, 0 if the database is insane.
289: */
290: static int
291: index_read(const DBT *key, const DBT *val,
292: const struct mchars *mc, struct rec *rec)
293: {
294: size_t left;
295: char *np, *cp;
296:
297: #define INDEX_BREAD(_dst) \
298: do { \
299: if (NULL == (np = memchr(cp, '\0', left))) \
300: return(0); \
301: norm_string(cp, mc, &(_dst)); \
302: left -= (np - cp) + 1; \
303: cp = np + 1; \
304: } while (/* CONSTCOND */ 0)
305:
306: left = val->size;
307: cp = (char *)val->data;
308:
309: rec->rec = *(recno_t *)key->data;
310:
311: INDEX_BREAD(rec->file);
312: INDEX_BREAD(rec->cat);
313: INDEX_BREAD(rec->title);
314: INDEX_BREAD(rec->arch);
315: INDEX_BREAD(rec->desc);
316: return(1);
317: }
318:
319: /*
320: * Search the mandocdb database for the expression "expr".
321: * Filter out by "opts".
322: * Call "res" with the results, which may be zero.
323: */
324: void
325: apropos_search(const struct opts *opts, const struct expr *expr,
326: void *arg, void (*res)(struct rec *, size_t, void *))
327: {
328: int i, len, root, leaf;
329: DBT key, val;
330: DB *btree, *idx;
331: struct mchars *mc;
332: int ch;
333: char *buf;
334: recno_t rec;
335: struct rec *recs;
336: struct rec srec;
337:
338: root = -1;
339: leaf = -1;
340: btree = NULL;
341: idx = NULL;
342: mc = NULL;
343: buf = NULL;
344: recs = NULL;
345: len = 0;
346:
347: memset(&srec, 0, sizeof(struct rec));
348:
349: /* XXX: error out with bad regexp? */
350:
351: mc = mchars_alloc();
352:
353: /* XXX: return fact that we've errored? */
354:
355: if (NULL == (btree = btree_open()))
356: goto out;
357: if (NULL == (idx = index_open()))
358: goto out;
359:
360: while (0 == (ch = (*btree->seq)(btree, &key, &val, R_NEXT))) {
361: /*
362: * Low-water mark for key and value.
363: * The key must have something in it, and the value must
364: * have the correct tags/recno mix.
365: */
366: if (key.size < 2 || 8 != val.size)
367: break;
368: if ( ! btree_read(&key, mc, &buf))
369: break;
370:
371: if ( ! exprexec(expr, buf))
372: continue;
373:
374: memcpy(&rec, val.data + 4, sizeof(recno_t));
375:
376: /*
377: * O(log n) scan for prior records. Since a record
378: * number is unbounded, this has decent performance over
379: * a complex hash function.
380: */
381:
382: for (leaf = root; leaf >= 0; )
383: if (rec > recs[leaf].rec && recs[leaf].rhs >= 0)
384: leaf = recs[leaf].rhs;
385: else if (rec < recs[leaf].rec && recs[leaf].lhs >= 0)
386: leaf = recs[leaf].lhs;
387: else
388: break;
389:
390: if (leaf >= 0 && recs[leaf].rec == rec)
391: continue;
392:
393: /*
394: * Now we actually extract the manpage's metadata from
395: * the index database.
396: */
397:
398: key.data = &rec;
399: key.size = sizeof(recno_t);
400:
401: if (0 != (*idx->get)(idx, &key, &val, 0))
402: break;
403:
404: srec.lhs = srec.rhs = -1;
405: if ( ! index_read(&key, &val, mc, &srec))
406: break;
407:
408: if (opts->cat && strcasecmp(opts->cat, srec.cat))
409: continue;
410: if (opts->arch && strcasecmp(opts->arch, srec.arch))
411: continue;
412:
413: recs = mandoc_realloc
414: (recs, (len + 1) * sizeof(struct rec));
415:
416: memcpy(&recs[len], &srec, sizeof(struct rec));
417:
418: /* Append to our tree. */
419:
420: if (leaf >= 0) {
421: if (rec > recs[leaf].rec)
422: recs[leaf].rhs = len;
423: else
424: recs[leaf].lhs = len;
425: } else
426: root = len;
427:
428: memset(&srec, 0, sizeof(struct rec));
429: len++;
430: }
431:
432: if (1 == ch)
433: (*res)(recs, len, arg);
434:
435: /* XXX: else? corrupt database error? */
436: out:
437: for (i = 0; i < len; i++) {
438: free(recs[i].file);
439: free(recs[i].cat);
440: free(recs[i].title);
441: free(recs[i].arch);
442: free(recs[i].desc);
443: }
444:
445: free(srec.file);
446: free(srec.cat);
447: free(srec.title);
448: free(srec.arch);
449: free(srec.desc);
450:
451: if (mc)
452: mchars_free(mc);
453: if (btree)
454: (*btree->close)(btree);
455: if (idx)
456: (*idx->close)(idx);
457:
458: free(buf);
459: free(recs);
460: }
461:
462: struct expr *
463: exprcomp(int cs, char *argv[], int argc)
464: {
465: struct expr *p;
466: struct expr e;
467: int i, pos, ch;
468:
469: pos = 0;
470:
471: if (pos > argc)
472: return(NULL);
473:
474: for (i = 0; 0 != types[i].mask; i++)
475: if (0 == strcmp(types[i].name, argv[pos]))
476: break;
477:
478: if (0 == (e.mask = types[i].mask))
479: return(NULL);
480:
481: if (++pos > argc--)
482: return(NULL);
483:
484: if ('-' != *argv[pos])
485: e.match = cs ? MATCH_STRCASE : MATCH_STR;
486: else if (0 == strcmp("-eq", argv[pos]))
487: e.match = cs ? MATCH_STRCASE : MATCH_STR;
488: else if (0 == strcmp("-ieq", argv[pos]))
489: e.match = MATCH_STRCASE;
490: else if (0 == strcmp("-re", argv[pos]))
491: e.match = cs ? MATCH_REGEXCASE : MATCH_REGEX;
492: else if (0 == strcmp("-ire", argv[pos]))
493: e.match = MATCH_REGEXCASE;
494: else
495: return(NULL);
496:
497: if ('-' == *argv[pos])
498: pos++;
499:
500: if (pos > argc--)
501: return(NULL);
502:
503: e.v = mandoc_strdup(argv[pos]);
504:
505: if (MATCH_REGEX == e.match || MATCH_REGEXCASE == e.match) {
506: ch = REG_EXTENDED | REG_NOSUB;
507: if (MATCH_REGEXCASE == e.match)
508: ch |= REG_ICASE;
509: if (regcomp(&e.re, e.v, ch))
510: return(NULL);
511: }
512:
513: p = mandoc_calloc(1, sizeof(struct expr));
514: memcpy(p, &e, sizeof(struct expr));
515: return(p);
516: }
517:
518: void
519: exprfree(struct expr *p)
520: {
521:
522: if (NULL == p)
523: return;
524:
525: if (MATCH_REGEX == p->match)
526: regfree(&p->re);
527:
528: free(p->v);
529: free(p);
530: }
531:
532: static int
533: exprexec(const struct expr *p, char *cp)
534: {
535:
536: if (MATCH_STR == p->match)
537: return(0 == strcmp(p->v, cp));
538: else if (MATCH_STRCASE == p->match)
539: return(0 == strcasecmp(p->v, cp));
540:
541: assert(MATCH_REGEX == p->match);
542: return(0 == regexec(&p->re, cp, 0, NULL, 0));
543: }
CVSweb