Annotation of mandoc/apropos_db.c, Revision 1.1
1.1 ! schwarze 1: /* $Id: db.c,v 1.3 2011/11/09 23:31:36 kristaps Exp $ */
! 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:
! 30: #include "apropos_db.h"
! 31: #include "mandoc.h"
! 32:
! 33: enum match {
! 34: MATCH_REGEX,
! 35: MATCH_REGEXCASE,
! 36: MATCH_STR,
! 37: MATCH_STRCASE
! 38: };
! 39:
! 40: struct expr {
! 41: enum match match;
! 42: int mask;
! 43: char *v;
! 44: regex_t re;
! 45: };
! 46:
! 47: struct type {
! 48: int mask;
! 49: const char *name;
! 50: };
! 51:
! 52: static const struct type types[] = {
! 53: { TYPE_NAME, "name" },
! 54: { TYPE_FUNCTION, "func" },
! 55: { TYPE_UTILITY, "utility" },
! 56: { TYPE_INCLUDES, "incl" },
! 57: { TYPE_VARIABLE, "var" },
! 58: { TYPE_STANDARD, "stand" },
! 59: { TYPE_AUTHOR, "auth" },
! 60: { TYPE_CONFIG, "conf" },
! 61: { TYPE_DESC, "desc" },
! 62: { TYPE_XREF, "xref" },
! 63: { TYPE_PATH, "path" },
! 64: { TYPE_ENV, "env" },
! 65: { TYPE_ERR, "err" },
! 66: { INT_MAX, "all" },
! 67: { 0, NULL }
! 68: };
! 69:
! 70: static DB *btree_open(void);
! 71: static int btree_read(const DBT *, const struct mchars *, char **);
! 72: static int exprexec(const struct expr *, char *);
! 73: static DB *index_open(void);
! 74: static int index_read(const DBT *, const DBT *,
! 75: const struct mchars *, struct rec *);
! 76: static void norm_string(const char *,
! 77: const struct mchars *, char **);
! 78: static size_t norm_utf8(unsigned int, char[7]);
! 79:
! 80: /*
! 81: * Open the keyword mandoc-db database.
! 82: */
! 83: static DB *
! 84: btree_open(void)
! 85: {
! 86: BTREEINFO info;
! 87: DB *db;
! 88:
! 89: memset(&info, 0, sizeof(BTREEINFO));
! 90: info.flags = R_DUP;
! 91:
! 92: db = dbopen("mandoc.db", O_RDONLY, 0, DB_BTREE, &info);
! 93: if (NULL != db)
! 94: return(db);
! 95:
! 96: return(NULL);
! 97: }
! 98:
! 99: /*
! 100: * Read a keyword from the database and normalise it.
! 101: * Return 0 if the database is insane, else 1.
! 102: */
! 103: static int
! 104: btree_read(const DBT *v, const struct mchars *mc, char **buf)
! 105: {
! 106:
! 107: /* Sanity: are we nil-terminated? */
! 108:
! 109: assert(v->size > 0);
! 110: if ('\0' != ((char *)v->data)[(int)v->size - 1])
! 111: return(0);
! 112:
! 113: norm_string((char *)v->data, mc, buf);
! 114: return(1);
! 115: }
! 116:
! 117: /*
! 118: * Take a Unicode codepoint and produce its UTF-8 encoding.
! 119: * This isn't the best way to do this, but it works.
! 120: * The magic numbers are from the UTF-8 packaging.
! 121: * They're not as scary as they seem: read the UTF-8 spec for details.
! 122: */
! 123: static size_t
! 124: norm_utf8(unsigned int cp, char out[7])
! 125: {
! 126: size_t rc;
! 127:
! 128: rc = 0;
! 129:
! 130: if (cp <= 0x0000007F) {
! 131: rc = 1;
! 132: out[0] = (char)cp;
! 133: } else if (cp <= 0x000007FF) {
! 134: rc = 2;
! 135: out[0] = (cp >> 6 & 31) | 192;
! 136: out[1] = (cp & 63) | 128;
! 137: } else if (cp <= 0x0000FFFF) {
! 138: rc = 3;
! 139: out[0] = (cp >> 12 & 15) | 224;
! 140: out[1] = (cp >> 6 & 63) | 128;
! 141: out[2] = (cp & 63) | 128;
! 142: } else if (cp <= 0x001FFFFF) {
! 143: rc = 4;
! 144: out[0] = (cp >> 18 & 7) | 240;
! 145: out[1] = (cp >> 12 & 63) | 128;
! 146: out[2] = (cp >> 6 & 63) | 128;
! 147: out[3] = (cp & 63) | 128;
! 148: } else if (cp <= 0x03FFFFFF) {
! 149: rc = 5;
! 150: out[0] = (cp >> 24 & 3) | 248;
! 151: out[1] = (cp >> 18 & 63) | 128;
! 152: out[2] = (cp >> 12 & 63) | 128;
! 153: out[3] = (cp >> 6 & 63) | 128;
! 154: out[4] = (cp & 63) | 128;
! 155: } else if (cp <= 0x7FFFFFFF) {
! 156: rc = 6;
! 157: out[0] = (cp >> 30 & 1) | 252;
! 158: out[1] = (cp >> 24 & 63) | 128;
! 159: out[2] = (cp >> 18 & 63) | 128;
! 160: out[3] = (cp >> 12 & 63) | 128;
! 161: out[4] = (cp >> 6 & 63) | 128;
! 162: out[5] = (cp & 63) | 128;
! 163: } else
! 164: return(0);
! 165:
! 166: out[rc] = '\0';
! 167: return(rc);
! 168: }
! 169:
! 170: /*
! 171: * Normalise strings from the index and database.
! 172: * These strings are escaped as defined by mandoc_char(7) along with
! 173: * other goop in mandoc.h (e.g., soft hyphens).
! 174: * This function normalises these into a nice UTF-8 string.
! 175: * Returns 0 if the database is fucked.
! 176: */
! 177: static void
! 178: norm_string(const char *val, const struct mchars *mc, char **buf)
! 179: {
! 180: size_t sz, bsz;
! 181: char utfbuf[7];
! 182: const char *seq, *cpp;
! 183: int len, u, pos;
! 184: enum mandoc_esc esc;
! 185: static const char res[] = { '\\', '\t',
! 186: ASCII_NBRSP, ASCII_HYPH, '\0' };
! 187:
! 188: /* Pre-allocate by the length of the input */
! 189:
! 190: bsz = strlen(val) + 1;
! 191: *buf = mandoc_realloc(*buf, bsz);
! 192: pos = 0;
! 193:
! 194: while ('\0' != *val) {
! 195: /*
! 196: * Halt on the first escape sequence.
! 197: * This also halts on the end of string, in which case
! 198: * we just copy, fallthrough, and exit the loop.
! 199: */
! 200: if ((sz = strcspn(val, res)) > 0) {
! 201: memcpy(&(*buf)[pos], val, sz);
! 202: pos += (int)sz;
! 203: val += (int)sz;
! 204: }
! 205:
! 206: if (ASCII_HYPH == *val) {
! 207: (*buf)[pos++] = '-';
! 208: val++;
! 209: continue;
! 210: } else if ('\t' == *val || ASCII_NBRSP == *val) {
! 211: (*buf)[pos++] = ' ';
! 212: val++;
! 213: continue;
! 214: } else if ('\\' != *val)
! 215: break;
! 216:
! 217: /* Read past the slash. */
! 218:
! 219: val++;
! 220: u = 0;
! 221:
! 222: /*
! 223: * Parse the escape sequence and see if it's a
! 224: * predefined character or special character.
! 225: */
! 226:
! 227: esc = mandoc_escape(&val, &seq, &len);
! 228: if (ESCAPE_ERROR == esc)
! 229: break;
! 230:
! 231: /*
! 232: * XXX - this just does UTF-8, but we need to know
! 233: * beforehand whether we should do text substitution.
! 234: */
! 235:
! 236: switch (esc) {
! 237: case (ESCAPE_SPECIAL):
! 238: if (0 != (u = mchars_spec2cp(mc, seq, len)))
! 239: break;
! 240: /* FALLTHROUGH */
! 241: default:
! 242: continue;
! 243: }
! 244:
! 245: /*
! 246: * If we have a Unicode codepoint, try to convert that
! 247: * to a UTF-8 byte string.
! 248: */
! 249:
! 250: cpp = utfbuf;
! 251: if (0 == (sz = norm_utf8(u, utfbuf)))
! 252: continue;
! 253:
! 254: /* Copy the rendered glyph into the stream. */
! 255:
! 256: sz = strlen(cpp);
! 257: bsz += sz;
! 258:
! 259: *buf = mandoc_realloc(*buf, bsz);
! 260:
! 261: memcpy(&(*buf)[pos], cpp, sz);
! 262: pos += (int)sz;
! 263: }
! 264:
! 265: (*buf)[pos] = '\0';
! 266: }
! 267:
! 268: /*
! 269: * Open the filename-index mandoc-db database.
! 270: * Returns NULL if opening failed.
! 271: */
! 272: static DB *
! 273: index_open(void)
! 274: {
! 275: DB *db;
! 276:
! 277: db = dbopen("mandoc.index", O_RDONLY, 0, DB_RECNO, NULL);
! 278: if (NULL != db)
! 279: return(db);
! 280:
! 281: return(NULL);
! 282: }
! 283:
! 284: /*
! 285: * Safely unpack from an index file record into the structure.
! 286: * Returns 1 if an entry was unpacked, 0 if the database is insane.
! 287: */
! 288: static int
! 289: index_read(const DBT *key, const DBT *val,
! 290: const struct mchars *mc, struct rec *rec)
! 291: {
! 292: size_t left;
! 293: char *np, *cp;
! 294:
! 295: #define INDEX_BREAD(_dst) \
! 296: do { \
! 297: if (NULL == (np = memchr(cp, '\0', left))) \
! 298: return(0); \
! 299: norm_string(cp, mc, &(_dst)); \
! 300: left -= (np - cp) + 1; \
! 301: cp = np + 1; \
! 302: } while (/* CONSTCOND */ 0)
! 303:
! 304: left = val->size;
! 305: cp = (char *)val->data;
! 306:
! 307: rec->rec = *(recno_t *)key->data;
! 308:
! 309: INDEX_BREAD(rec->file);
! 310: INDEX_BREAD(rec->cat);
! 311: INDEX_BREAD(rec->title);
! 312: INDEX_BREAD(rec->arch);
! 313: INDEX_BREAD(rec->desc);
! 314: return(1);
! 315: }
! 316:
! 317: /*
! 318: * Search the mandocdb database for the expression "expr".
! 319: * Filter out by "opts".
! 320: * Call "res" with the results, which may be zero.
! 321: */
! 322: void
! 323: apropos_search(const struct opts *opts, const struct expr *expr,
! 324: void *arg, void (*res)(struct rec *, size_t, void *))
! 325: {
! 326: int i, len, root, leaf;
! 327: DBT key, val;
! 328: DB *btree, *idx;
! 329: struct mchars *mc;
! 330: int ch;
! 331: char *buf;
! 332: recno_t rec;
! 333: struct rec *recs;
! 334: struct rec srec;
! 335:
! 336: root = -1;
! 337: leaf = -1;
! 338: btree = NULL;
! 339: idx = NULL;
! 340: mc = NULL;
! 341: buf = NULL;
! 342: recs = NULL;
! 343: len = 0;
! 344:
! 345: memset(&srec, 0, sizeof(struct rec));
! 346:
! 347: /* XXX: error out with bad regexp? */
! 348:
! 349: mc = mchars_alloc();
! 350:
! 351: /* XXX: return fact that we've errored? */
! 352:
! 353: if (NULL == (btree = btree_open()))
! 354: goto out;
! 355: if (NULL == (idx = index_open()))
! 356: goto out;
! 357:
! 358: while (0 == (ch = (*btree->seq)(btree, &key, &val, R_NEXT))) {
! 359: /*
! 360: * Low-water mark for key and value.
! 361: * The key must have something in it, and the value must
! 362: * have the correct tags/recno mix.
! 363: */
! 364: if (key.size < 2 || 8 != val.size)
! 365: break;
! 366: if ( ! btree_read(&key, mc, &buf))
! 367: break;
! 368:
! 369: if ( ! exprexec(expr, buf))
! 370: continue;
! 371:
! 372: memcpy(&rec, val.data + 4, sizeof(recno_t));
! 373:
! 374: /*
! 375: * O(log n) scan for prior records. Since a record
! 376: * number is unbounded, this has decent performance over
! 377: * a complex hash function.
! 378: */
! 379:
! 380: for (leaf = root; leaf >= 0; )
! 381: if (rec > recs[leaf].rec && recs[leaf].rhs >= 0)
! 382: leaf = recs[leaf].rhs;
! 383: else if (rec < recs[leaf].rec && recs[leaf].lhs >= 0)
! 384: leaf = recs[leaf].lhs;
! 385: else
! 386: break;
! 387:
! 388: if (leaf >= 0 && recs[leaf].rec == rec)
! 389: continue;
! 390:
! 391: /*
! 392: * Now we actually extract the manpage's metadata from
! 393: * the index database.
! 394: */
! 395:
! 396: key.data = &rec;
! 397: key.size = sizeof(recno_t);
! 398:
! 399: if (0 != (*idx->get)(idx, &key, &val, 0))
! 400: break;
! 401:
! 402: srec.lhs = srec.rhs = -1;
! 403: if ( ! index_read(&key, &val, mc, &srec))
! 404: break;
! 405:
! 406: if (opts->cat && strcasecmp(opts->cat, srec.cat))
! 407: continue;
! 408: if (opts->arch && strcasecmp(opts->arch, srec.arch))
! 409: continue;
! 410:
! 411: recs = mandoc_realloc
! 412: (recs, (len + 1) * sizeof(struct rec));
! 413:
! 414: memcpy(&recs[len], &srec, sizeof(struct rec));
! 415:
! 416: /* Append to our tree. */
! 417:
! 418: if (leaf >= 0) {
! 419: if (rec > recs[leaf].rec)
! 420: recs[leaf].rhs = len;
! 421: else
! 422: recs[leaf].lhs = len;
! 423: } else
! 424: root = len;
! 425:
! 426: memset(&srec, 0, sizeof(struct rec));
! 427: len++;
! 428: }
! 429:
! 430: if (1 == ch)
! 431: (*res)(recs, len, arg);
! 432:
! 433: /* XXX: else? corrupt database error? */
! 434: out:
! 435: for (i = 0; i < len; i++) {
! 436: free(recs[i].file);
! 437: free(recs[i].cat);
! 438: free(recs[i].title);
! 439: free(recs[i].arch);
! 440: free(recs[i].desc);
! 441: }
! 442:
! 443: free(srec.file);
! 444: free(srec.cat);
! 445: free(srec.title);
! 446: free(srec.arch);
! 447: free(srec.desc);
! 448:
! 449: if (mc)
! 450: mchars_free(mc);
! 451: if (btree)
! 452: (*btree->close)(btree);
! 453: if (idx)
! 454: (*idx->close)(idx);
! 455:
! 456: free(buf);
! 457: free(recs);
! 458: }
! 459:
! 460: struct expr *
! 461: exprcomp(int cs, char *argv[], int argc)
! 462: {
! 463: struct expr *p;
! 464: struct expr e;
! 465: int i, pos, ch;
! 466:
! 467: pos = 0;
! 468:
! 469: if (pos > argc)
! 470: return(NULL);
! 471:
! 472: for (i = 0; 0 != types[i].mask; i++)
! 473: if (0 == strcmp(types[i].name, argv[pos]))
! 474: break;
! 475:
! 476: if (0 == (e.mask = types[i].mask))
! 477: return(NULL);
! 478:
! 479: if (++pos > argc--)
! 480: return(NULL);
! 481:
! 482: if ('-' != *argv[pos])
! 483: e.match = cs ? MATCH_STRCASE : MATCH_STR;
! 484: else if (0 == strcmp("-eq", argv[pos]))
! 485: e.match = cs ? MATCH_STRCASE : MATCH_STR;
! 486: else if (0 == strcmp("-ieq", argv[pos]))
! 487: e.match = MATCH_STRCASE;
! 488: else if (0 == strcmp("-re", argv[pos]))
! 489: e.match = cs ? MATCH_REGEXCASE : MATCH_REGEX;
! 490: else if (0 == strcmp("-ire", argv[pos]))
! 491: e.match = MATCH_REGEXCASE;
! 492: else
! 493: return(NULL);
! 494:
! 495: if ('-' == *argv[pos])
! 496: pos++;
! 497:
! 498: if (pos > argc--)
! 499: return(NULL);
! 500:
! 501: e.v = mandoc_strdup(argv[pos]);
! 502:
! 503: if (MATCH_REGEX == e.match || MATCH_REGEXCASE == e.match) {
! 504: ch = REG_EXTENDED | REG_NOSUB;
! 505: if (MATCH_REGEXCASE == e.match)
! 506: ch |= REG_ICASE;
! 507: if (regcomp(&e.re, e.v, ch))
! 508: return(NULL);
! 509: }
! 510:
! 511: p = mandoc_calloc(1, sizeof(struct expr));
! 512: memcpy(p, &e, sizeof(struct expr));
! 513: return(p);
! 514: }
! 515:
! 516: void
! 517: exprfree(struct expr *p)
! 518: {
! 519:
! 520: if (NULL == p)
! 521: return;
! 522:
! 523: if (MATCH_REGEX == p->match)
! 524: regfree(&p->re);
! 525:
! 526: free(p->v);
! 527: free(p);
! 528: }
! 529:
! 530: static int
! 531: exprexec(const struct expr *p, char *cp)
! 532: {
! 533:
! 534: if (MATCH_STR == p->match)
! 535: return(0 == strcmp(p->v, cp));
! 536: else if (MATCH_STRCASE == p->match)
! 537: return(0 == strcasecmp(p->v, cp));
! 538:
! 539: assert(MATCH_REGEX == p->match);
! 540: return(0 == regexec(&p->re, cp, 0, NULL, 0));
! 541: }
CVSweb