Annotation of mandoc/mandocdb.c, Revision 1.75
1.75 ! schwarze 1: /* $Id: mandocdb.c,v 1.74 2013/10/27 16:09:44 schwarze Exp $ */
1.1 kristaps 2: /*
1.48 schwarze 3: * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.56 schwarze 4: * Copyright (c) 2011, 2012, 2013 Ingo Schwarze <schwarze@openbsd.org>
1.1 kristaps 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: */
18: #ifdef HAVE_CONFIG_H
19: #include "config.h"
20: #endif
21:
1.50 kristaps 22: #include <sys/stat.h>
1.1 kristaps 23:
24: #include <assert.h>
1.43 kristaps 25: #include <ctype.h>
1.63 schwarze 26: #include <errno.h>
1.1 kristaps 27: #include <fcntl.h>
1.50 kristaps 28: #include <fts.h>
1.1 kristaps 29: #include <getopt.h>
1.58 schwarze 30: #include <limits.h>
1.50 kristaps 31: #include <stddef.h>
1.59 schwarze 32: #include <stdio.h>
1.1 kristaps 33: #include <stdint.h>
34: #include <stdlib.h>
35: #include <string.h>
1.17 schwarze 36: #include <unistd.h>
1.1 kristaps 37:
1.53 kristaps 38: #ifdef HAVE_OHASH
1.50 kristaps 39: #include <ohash.h>
1.53 kristaps 40: #else
41: #include "compat_ohash.h"
42: #endif
1.50 kristaps 43: #include <sqlite3.h>
1.1 kristaps 44:
1.50 kristaps 45: #include "mdoc.h"
1.1 kristaps 46: #include "man.h"
47: #include "mandoc.h"
1.10 kristaps 48: #include "manpath.h"
1.55 kristaps 49: #include "mansearch.h"
1.1 kristaps 50:
1.52 kristaps 51: #define SQL_EXEC(_v) \
52: if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
53: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
54: #define SQL_BIND_TEXT(_s, _i, _v) \
55: if (SQLITE_OK != sqlite3_bind_text \
56: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
57: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
58: #define SQL_BIND_INT(_s, _i, _v) \
59: if (SQLITE_OK != sqlite3_bind_int \
60: ((_s), (_i)++, (_v))) \
61: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
62: #define SQL_BIND_INT64(_s, _i, _v) \
63: if (SQLITE_OK != sqlite3_bind_int64 \
64: ((_s), (_i)++, (_v))) \
65: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
66: #define SQL_STEP(_s) \
67: if (SQLITE_DONE != sqlite3_step((_s))) \
68: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
69:
1.50 kristaps 70: enum op {
71: OP_DEFAULT = 0, /* new dbs from dir list or default config */
72: OP_CONFFILE, /* new databases from custom config file */
73: OP_UPDATE, /* delete/add entries in existing database */
74: OP_DELETE, /* delete entries from existing database */
75: OP_TEST /* change no databases, report potential problems */
1.38 schwarze 76: };
77:
1.50 kristaps 78: enum form {
79: FORM_SRC, /* format is -man or -mdoc */
80: FORM_CAT, /* format is cat */
81: FORM_NONE /* format is unknown */
82: };
1.38 schwarze 83:
1.50 kristaps 84: struct str {
85: char *utf8; /* key in UTF-8 form */
86: const struct of *of; /* if set, the owning parse */
87: uint64_t mask; /* bitmask in sequence */
1.51 kristaps 88: char key[]; /* the string itself */
1.38 schwarze 89: };
90:
1.50 kristaps 91: struct id {
92: ino_t ino;
93: dev_t dev;
94: };
1.5 kristaps 95:
1.3 kristaps 96: struct of {
1.50 kristaps 97: struct id id; /* used for hashing routine */
98: struct of *next; /* next in ofs */
99: enum form dform; /* path-cued form */
100: enum form sform; /* suffix-cued form */
1.58 schwarze 101: char file[PATH_MAX]; /* filename rel. to manpath */
1.66 schwarze 102: char *desc; /* parsed description */
103: char *name; /* name (from filename) (not empty) */
104: char *sec; /* suffix-cued section (or empty) */
105: char *dsec; /* path-cued section (or empty) */
106: char *arch; /* path-cued arch. (or empty) */
1.3 kristaps 107: };
108:
1.69 schwarze 109: struct title {
110: char *title; /* name(sec/arch) given inside the file */
111: char *file; /* file name in case of mismatch */
112: };
113:
1.50 kristaps 114: enum stmt {
115: STMT_DELETE = 0, /* delete manpage */
116: STMT_INSERT_DOC, /* insert manpage */
117: STMT_INSERT_KEY, /* insert parsed key */
118: STMT__MAX
1.1 kristaps 119: };
120:
1.50 kristaps 121: typedef int (*mdoc_fp)(struct of *, const struct mdoc_node *);
1.1 kristaps 122:
1.50 kristaps 123: struct mdoc_handler {
124: mdoc_fp fp; /* optional handler */
125: uint64_t mask; /* set unless handler returns 0 */
1.1 kristaps 126: };
127:
1.59 schwarze 128: static void dbclose(int);
129: static void dbindex(struct mchars *, int, const struct of *);
130: static int dbopen(int);
131: static void dbprune(void);
1.50 kristaps 132: static void fileadd(struct of *);
133: static int filecheck(const char *);
1.59 schwarze 134: static void filescan(const char *);
1.50 kristaps 135: static void *hash_alloc(size_t, void *);
136: static void hash_free(void *, size_t, void *);
137: static void *hash_halloc(size_t, void *);
138: static void inoadd(const struct stat *, struct of *);
139: static int inocheck(const struct stat *);
1.59 schwarze 140: static void ofadd(int, const char *, const char *, const char *,
141: const char *, const char *, const struct stat *);
1.50 kristaps 142: static void offree(void);
1.71 schwarze 143: static void ofmerge(struct mchars *, struct mparse *, int);
1.59 schwarze 144: static void parse_catpage(struct of *);
1.61 schwarze 145: static void parse_man(struct of *, const struct man_node *);
1.50 kristaps 146: static void parse_mdoc(struct of *, const struct mdoc_node *);
147: static int parse_mdoc_body(struct of *, const struct mdoc_node *);
148: static int parse_mdoc_head(struct of *, const struct mdoc_node *);
149: static int parse_mdoc_Fd(struct of *, const struct mdoc_node *);
150: static int parse_mdoc_Fn(struct of *, const struct mdoc_node *);
151: static int parse_mdoc_In(struct of *, const struct mdoc_node *);
152: static int parse_mdoc_Nd(struct of *, const struct mdoc_node *);
153: static int parse_mdoc_Nm(struct of *, const struct mdoc_node *);
154: static int parse_mdoc_Sh(struct of *, const struct mdoc_node *);
155: static int parse_mdoc_St(struct of *, const struct mdoc_node *);
156: static int parse_mdoc_Xr(struct of *, const struct mdoc_node *);
1.59 schwarze 157: static int set_basedir(const char *);
1.50 kristaps 158: static void putkey(const struct of *,
159: const char *, uint64_t);
160: static void putkeys(const struct of *,
1.64 schwarze 161: const char *, size_t, uint64_t);
1.50 kristaps 162: static void putmdockey(const struct of *,
163: const struct mdoc_node *, uint64_t);
1.59 schwarze 164: static void say(const char *, const char *, ...);
165: static int treescan(void);
1.50 kristaps 166: static size_t utf8(unsigned int, char [7]);
167: static void utf8key(struct mchars *, struct str *);
168:
169: static char *progname;
170: static int use_all; /* use all found files */
171: static int nodb; /* no database changes */
172: static int verb; /* print what we're doing */
173: static int warnings; /* warn about crap */
1.59 schwarze 174: static int exitcode; /* to be returned by main */
1.50 kristaps 175: static enum op op; /* operational mode */
1.59 schwarze 176: static char basedir[PATH_MAX]; /* current base directory */
1.50 kristaps 177: static struct ohash inos; /* table of inodes/devices */
178: static struct ohash filenames; /* table of filenames */
179: static struct ohash strings; /* table of all strings */
180: static struct of *ofs = NULL; /* vector of files to parse */
181: static sqlite3 *db = NULL; /* current database */
182: static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
1.25 schwarze 183:
184: static const struct mdoc_handler mdocs[MDOC_MAX] = {
1.68 schwarze 185: { NULL, 0 }, /* Ap */
186: { NULL, 0 }, /* Dd */
187: { NULL, 0 }, /* Dt */
188: { NULL, 0 }, /* Os */
189: { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
190: { parse_mdoc_head, TYPE_Ss }, /* Ss */
191: { NULL, 0 }, /* Pp */
192: { NULL, 0 }, /* D1 */
193: { NULL, 0 }, /* Dl */
194: { NULL, 0 }, /* Bd */
195: { NULL, 0 }, /* Ed */
196: { NULL, 0 }, /* Bl */
197: { NULL, 0 }, /* El */
198: { NULL, 0 }, /* It */
199: { NULL, 0 }, /* Ad */
200: { NULL, TYPE_An }, /* An */
201: { NULL, TYPE_Ar }, /* Ar */
202: { NULL, TYPE_Cd }, /* Cd */
203: { NULL, TYPE_Cm }, /* Cm */
204: { NULL, TYPE_Dv }, /* Dv */
205: { NULL, TYPE_Er }, /* Er */
206: { NULL, TYPE_Ev }, /* Ev */
207: { NULL, 0 }, /* Ex */
208: { NULL, TYPE_Fa }, /* Fa */
209: { parse_mdoc_Fd, 0 }, /* Fd */
210: { NULL, TYPE_Fl }, /* Fl */
211: { parse_mdoc_Fn, 0 }, /* Fn */
212: { NULL, TYPE_Ft }, /* Ft */
213: { NULL, TYPE_Ic }, /* Ic */
214: { parse_mdoc_In, TYPE_In }, /* In */
215: { NULL, TYPE_Li }, /* Li */
216: { parse_mdoc_Nd, TYPE_Nd }, /* Nd */
217: { parse_mdoc_Nm, TYPE_Nm }, /* Nm */
218: { NULL, 0 }, /* Op */
219: { NULL, 0 }, /* Ot */
220: { NULL, TYPE_Pa }, /* Pa */
221: { NULL, 0 }, /* Rv */
222: { parse_mdoc_St, 0 }, /* St */
223: { NULL, TYPE_Va }, /* Va */
224: { parse_mdoc_body, TYPE_Va }, /* Vt */
225: { parse_mdoc_Xr, 0 }, /* Xr */
226: { NULL, 0 }, /* %A */
227: { NULL, 0 }, /* %B */
228: { NULL, 0 }, /* %D */
229: { NULL, 0 }, /* %I */
230: { NULL, 0 }, /* %J */
231: { NULL, 0 }, /* %N */
232: { NULL, 0 }, /* %O */
233: { NULL, 0 }, /* %P */
234: { NULL, 0 }, /* %R */
235: { NULL, 0 }, /* %T */
236: { NULL, 0 }, /* %V */
237: { NULL, 0 }, /* Ac */
238: { NULL, 0 }, /* Ao */
239: { NULL, 0 }, /* Aq */
240: { NULL, TYPE_At }, /* At */
241: { NULL, 0 }, /* Bc */
242: { NULL, 0 }, /* Bf */
243: { NULL, 0 }, /* Bo */
244: { NULL, 0 }, /* Bq */
245: { NULL, TYPE_Bsx }, /* Bsx */
246: { NULL, TYPE_Bx }, /* Bx */
247: { NULL, 0 }, /* Db */
248: { NULL, 0 }, /* Dc */
249: { NULL, 0 }, /* Do */
250: { NULL, 0 }, /* Dq */
251: { NULL, 0 }, /* Ec */
252: { NULL, 0 }, /* Ef */
253: { NULL, TYPE_Em }, /* Em */
254: { NULL, 0 }, /* Eo */
255: { NULL, TYPE_Fx }, /* Fx */
256: { NULL, TYPE_Ms }, /* Ms */
257: { NULL, 0 }, /* No */
258: { NULL, 0 }, /* Ns */
259: { NULL, TYPE_Nx }, /* Nx */
260: { NULL, TYPE_Ox }, /* Ox */
261: { NULL, 0 }, /* Pc */
262: { NULL, 0 }, /* Pf */
263: { NULL, 0 }, /* Po */
264: { NULL, 0 }, /* Pq */
265: { NULL, 0 }, /* Qc */
266: { NULL, 0 }, /* Ql */
267: { NULL, 0 }, /* Qo */
268: { NULL, 0 }, /* Qq */
269: { NULL, 0 }, /* Re */
270: { NULL, 0 }, /* Rs */
271: { NULL, 0 }, /* Sc */
272: { NULL, 0 }, /* So */
273: { NULL, 0 }, /* Sq */
274: { NULL, 0 }, /* Sm */
275: { NULL, 0 }, /* Sx */
276: { NULL, TYPE_Sy }, /* Sy */
277: { NULL, TYPE_Tn }, /* Tn */
278: { NULL, 0 }, /* Ux */
279: { NULL, 0 }, /* Xc */
280: { NULL, 0 }, /* Xo */
281: { parse_mdoc_head, 0 }, /* Fo */
282: { NULL, 0 }, /* Fc */
283: { NULL, 0 }, /* Oo */
284: { NULL, 0 }, /* Oc */
285: { NULL, 0 }, /* Bk */
286: { NULL, 0 }, /* Ek */
287: { NULL, 0 }, /* Bt */
288: { NULL, 0 }, /* Hf */
289: { NULL, 0 }, /* Fr */
290: { NULL, 0 }, /* Ud */
291: { NULL, TYPE_Lb }, /* Lb */
292: { NULL, 0 }, /* Lp */
293: { NULL, TYPE_Lk }, /* Lk */
294: { NULL, TYPE_Mt }, /* Mt */
295: { NULL, 0 }, /* Brq */
296: { NULL, 0 }, /* Bro */
297: { NULL, 0 }, /* Brc */
298: { NULL, 0 }, /* %C */
299: { NULL, 0 }, /* Es */
300: { NULL, 0 }, /* En */
301: { NULL, TYPE_Dx }, /* Dx */
302: { NULL, 0 }, /* %Q */
303: { NULL, 0 }, /* br */
304: { NULL, 0 }, /* sp */
305: { NULL, 0 }, /* %U */
306: { NULL, 0 }, /* Ta */
1.1 kristaps 307: };
308:
309: int
310: main(int argc, char *argv[])
311: {
1.59 schwarze 312: int ch, i;
1.50 kristaps 313: size_t j, sz;
1.59 schwarze 314: const char *path_arg;
1.50 kristaps 315: struct mchars *mc;
316: struct manpaths dirs;
317: struct mparse *mp;
1.71 schwarze 318: struct ohash_info ino_info, filename_info;
1.50 kristaps 319:
320: memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
321: memset(&dirs, 0, sizeof(struct manpaths));
322:
1.71 schwarze 323: ino_info.alloc = filename_info.alloc = hash_alloc;
324: ino_info.halloc = filename_info.halloc = hash_halloc;
325: ino_info.hfree = filename_info.hfree = hash_free;
1.50 kristaps 326:
327: ino_info.key_offset = offsetof(struct of, id);
328: filename_info.key_offset = offsetof(struct of, file);
1.1 kristaps 329:
330: progname = strrchr(argv[0], '/');
331: if (progname == NULL)
332: progname = argv[0];
333: else
334: ++progname;
335:
1.50 kristaps 336: /*
337: * We accept a few different invocations.
338: * The CHECKOP macro makes sure that invocation styles don't
339: * clobber each other.
340: */
341: #define CHECKOP(_op, _ch) do \
342: if (OP_DEFAULT != (_op)) { \
343: fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \
344: goto usage; \
345: } while (/*CONSTCOND*/0)
1.10 kristaps 346:
1.59 schwarze 347: path_arg = NULL;
1.38 schwarze 348: op = OP_DEFAULT;
1.1 kristaps 349:
1.50 kristaps 350: while (-1 != (ch = getopt(argc, argv, "aC:d:ntu:vW")))
1.1 kristaps 351: switch (ch) {
1.12 schwarze 352: case ('a'):
353: use_all = 1;
354: break;
1.34 schwarze 355: case ('C'):
1.50 kristaps 356: CHECKOP(op, ch);
1.59 schwarze 357: path_arg = optarg;
1.38 schwarze 358: op = OP_CONFFILE;
1.34 schwarze 359: break;
1.5 kristaps 360: case ('d'):
1.50 kristaps 361: CHECKOP(op, ch);
1.59 schwarze 362: path_arg = optarg;
1.5 kristaps 363: op = OP_UPDATE;
364: break;
1.50 kristaps 365: case ('n'):
366: nodb = 1;
367: break;
1.38 schwarze 368: case ('t'):
1.50 kristaps 369: CHECKOP(op, ch);
1.38 schwarze 370: dup2(STDOUT_FILENO, STDERR_FILENO);
371: op = OP_TEST;
1.50 kristaps 372: nodb = warnings = 1;
1.38 schwarze 373: break;
1.5 kristaps 374: case ('u'):
1.50 kristaps 375: CHECKOP(op, ch);
1.59 schwarze 376: path_arg = optarg;
1.5 kristaps 377: op = OP_DELETE;
378: break;
379: case ('v'):
380: verb++;
381: break;
1.38 schwarze 382: case ('W'):
383: warnings = 1;
384: break;
1.1 kristaps 385: default:
1.38 schwarze 386: goto usage;
1.1 kristaps 387: }
388:
389: argc -= optind;
390: argv += optind;
391:
1.38 schwarze 392: if (OP_CONFFILE == op && argc > 0) {
1.50 kristaps 393: fprintf(stderr, "-C: Too many arguments\n");
1.38 schwarze 394: goto usage;
395: }
396:
1.59 schwarze 397: exitcode = (int)MANDOCLEVEL_OK;
1.50 kristaps 398: mp = mparse_alloc(MPARSE_AUTO,
399: MANDOCLEVEL_FATAL, NULL, NULL, NULL);
400: mc = mchars_alloc();
401:
402: ohash_init(&inos, 6, &ino_info);
403: ohash_init(&filenames, 6, &filename_info);
404:
405: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
406: /*
407: * Force processing all files.
408: */
409: use_all = 1;
1.59 schwarze 410:
1.50 kristaps 411: /*
412: * All of these deal with a specific directory.
413: * Jump into that directory then collect files specified
414: * on the command-line.
415: */
1.59 schwarze 416: if (0 == set_basedir(path_arg))
1.50 kristaps 417: goto out;
418: for (i = 0; i < argc; i++)
1.59 schwarze 419: filescan(argv[i]);
420: if (0 == dbopen(1))
1.50 kristaps 421: goto out;
422: if (OP_TEST != op)
1.59 schwarze 423: dbprune();
1.50 kristaps 424: if (OP_DELETE != op)
1.71 schwarze 425: ofmerge(mc, mp, 0);
1.59 schwarze 426: dbclose(1);
1.50 kristaps 427: } else {
428: /*
429: * If we have arguments, use them as our manpaths.
430: * If we don't, grok from manpath(1) or however else
431: * manpath_parse() wants to do it.
432: */
433: if (argc > 0) {
434: dirs.paths = mandoc_calloc
435: (argc, sizeof(char *));
436: dirs.sz = (size_t)argc;
437: for (i = 0; i < argc; i++)
438: dirs.paths[i] = mandoc_strdup(argv[i]);
439: } else
1.59 schwarze 440: manpath_parse(&dirs, path_arg, NULL, NULL);
1.50 kristaps 441:
442: /*
1.69 schwarze 443: * First scan the tree rooted at a base directory, then
444: * build a new database and finally move it into place.
1.50 kristaps 445: * Ignore zero-length directories and strip trailing
446: * slashes.
447: */
448: for (j = 0; j < dirs.sz; j++) {
449: sz = strlen(dirs.paths[j]);
450: if (sz && '/' == dirs.paths[j][sz - 1])
451: dirs.paths[j][--sz] = '\0';
452: if (0 == sz)
453: continue;
1.66 schwarze 454:
455: if (j) {
456: ohash_init(&inos, 6, &ino_info);
457: ohash_init(&filenames, 6, &filename_info);
458: }
459:
1.59 schwarze 460: if (0 == set_basedir(dirs.paths[j]))
1.50 kristaps 461: goto out;
1.59 schwarze 462: if (0 == treescan())
1.50 kristaps 463: goto out;
1.59 schwarze 464: if (0 == set_basedir(dirs.paths[j]))
1.50 kristaps 465: goto out;
1.59 schwarze 466: if (0 == dbopen(0))
1.50 kristaps 467: goto out;
1.52 kristaps 468:
1.71 schwarze 469: ofmerge(mc, mp, warnings && !use_all);
1.59 schwarze 470: dbclose(0);
1.66 schwarze 471:
472: if (j + 1 < dirs.sz) {
473: ohash_delete(&inos);
474: ohash_delete(&filenames);
475: offree();
476: }
1.50 kristaps 477: }
478: }
479: out:
1.59 schwarze 480: set_basedir(NULL);
1.50 kristaps 481: manpath_free(&dirs);
482: mchars_free(mc);
483: mparse_free(mp);
484: ohash_delete(&inos);
485: ohash_delete(&filenames);
486: offree();
1.59 schwarze 487: return(exitcode);
1.50 kristaps 488: usage:
489: fprintf(stderr, "usage: %s [-anvW] [-C file]\n"
490: " %s [-anvW] dir ...\n"
491: " %s [-nvW] -d dir [file ...]\n"
492: " %s [-nvW] -u dir [file ...]\n"
493: " %s -t file ...\n",
494: progname, progname, progname,
495: progname, progname);
496:
1.59 schwarze 497: return((int)MANDOCLEVEL_BADARG);
1.50 kristaps 498: }
499:
500: /*
1.59 schwarze 501: * Scan a directory tree rooted at "basedir" for manpages.
1.50 kristaps 502: * We use fts(), scanning directory parts along the way for clues to our
503: * section and architecture.
504: *
505: * If use_all has been specified, grok all files.
506: * If not, sanitise paths to the following:
507: *
508: * [./]man*[/<arch>]/<name>.<section>
509: * or
510: * [./]cat<section>[/<arch>]/<name>.0
511: *
512: * TODO: accomodate for multi-language directories.
513: */
514: static int
1.59 schwarze 515: treescan(void)
1.50 kristaps 516: {
517: FTS *f;
518: FTSENT *ff;
519: int dform;
520: char *sec;
1.66 schwarze 521: const char *dsec, *arch, *cp, *path;
1.50 kristaps 522: const char *argv[2];
523:
524: argv[0] = ".";
525: argv[1] = (char *)NULL;
526:
527: /*
528: * Walk through all components under the directory, using the
529: * logical descent of files.
530: */
531: f = fts_open((char * const *)argv, FTS_LOGICAL, NULL);
532: if (NULL == f) {
1.59 schwarze 533: exitcode = (int)MANDOCLEVEL_SYSERR;
534: say("", NULL);
1.50 kristaps 535: return(0);
536: }
537:
538: dsec = arch = NULL;
539: dform = FORM_NONE;
540:
541: while (NULL != (ff = fts_read(f))) {
542: path = ff->fts_path + 2;
543: /*
544: * If we're a regular file, add an "of" by using the
545: * stored directory data and handling the filename.
546: * Disallow duplicate (hard-linked) files.
547: */
548: if (FTS_F == ff->fts_info) {
1.60 schwarze 549: if (0 == strcmp(path, MANDOC_DB))
550: continue;
1.50 kristaps 551: if ( ! use_all && ff->fts_level < 2) {
1.56 schwarze 552: if (warnings)
1.59 schwarze 553: say(path, "Extraneous file");
1.50 kristaps 554: continue;
555: } else if (inocheck(ff->fts_statp)) {
1.56 schwarze 556: if (warnings)
1.59 schwarze 557: say(path, "Duplicate file");
1.50 kristaps 558: continue;
1.60 schwarze 559: } else if (NULL == (sec =
560: strrchr(ff->fts_name, '.'))) {
561: if ( ! use_all) {
1.56 schwarze 562: if (warnings)
1.60 schwarze 563: say(path,
564: "No filename suffix");
1.50 kristaps 565: continue;
566: }
1.60 schwarze 567: } else if (0 == strcmp(++sec, "html")) {
568: if (warnings)
569: say(path, "Skip html");
570: continue;
571: } else if (0 == strcmp(sec, "gz")) {
572: if (warnings)
573: say(path, "Skip gz");
574: continue;
575: } else if (0 == strcmp(sec, "ps")) {
576: if (warnings)
577: say(path, "Skip ps");
578: continue;
579: } else if (0 == strcmp(sec, "pdf")) {
580: if (warnings)
581: say(path, "Skip pdf");
582: continue;
583: } else if ( ! use_all &&
584: ((FORM_SRC == dform && strcmp(sec, dsec)) ||
585: (FORM_CAT == dform && strcmp(sec, "0")))) {
586: if (warnings)
587: say(path, "Wrong filename suffix");
588: continue;
1.66 schwarze 589: } else
1.60 schwarze 590: sec[-1] = '\0';
1.66 schwarze 591: ofadd(dform, path, ff->fts_name, dsec, sec,
592: arch, ff->fts_statp);
1.50 kristaps 593: continue;
594: } else if (FTS_D != ff->fts_info &&
1.60 schwarze 595: FTS_DP != ff->fts_info) {
596: if (warnings)
597: say(path, "Not a regular file");
1.50 kristaps 598: continue;
1.60 schwarze 599: }
1.1 kristaps 600:
1.50 kristaps 601: switch (ff->fts_level) {
602: case (0):
603: /* Ignore the root directory. */
604: break;
605: case (1):
606: /*
607: * This might contain manX/ or catX/.
608: * Try to infer this from the name.
609: * If we're not in use_all, enforce it.
610: */
611: dsec = NULL;
612: dform = FORM_NONE;
613: cp = ff->fts_name;
614: if (FTS_DP == ff->fts_info)
615: break;
1.1 kristaps 616:
1.50 kristaps 617: if (0 == strncmp(cp, "man", 3)) {
618: dform = FORM_SRC;
1.66 schwarze 619: dsec = cp + 3;
1.50 kristaps 620: } else if (0 == strncmp(cp, "cat", 3)) {
621: dform = FORM_CAT;
1.66 schwarze 622: dsec = cp + 3;
1.50 kristaps 623: }
1.1 kristaps 624:
1.50 kristaps 625: if (NULL != dsec || use_all)
626: break;
1.1 kristaps 627:
1.56 schwarze 628: if (warnings)
1.59 schwarze 629: say(path, "Unknown directory part");
1.50 kristaps 630: fts_set(f, ff, FTS_SKIP);
631: break;
632: case (2):
633: /*
634: * Possibly our architecture.
635: * If we're descending, keep tabs on it.
636: */
637: arch = NULL;
638: if (FTS_DP != ff->fts_info && NULL != dsec)
1.66 schwarze 639: arch = ff->fts_name;
1.50 kristaps 640: break;
641: default:
642: if (FTS_DP == ff->fts_info || use_all)
643: break;
1.56 schwarze 644: if (warnings)
1.59 schwarze 645: say(path, "Extraneous directory part");
1.50 kristaps 646: fts_set(f, ff, FTS_SKIP);
647: break;
1.5 kristaps 648: }
1.50 kristaps 649: }
650:
651: fts_close(f);
652: return(1);
653: }
1.5 kristaps 654:
1.50 kristaps 655: /*
656: * Add a file to the file vector.
657: * Do not verify that it's a "valid" looking manpage (we'll do that
658: * later).
659: *
660: * Try to infer the manual section, architecture, and page name from the
661: * path, assuming it looks like
662: *
663: * [./]man*[/<arch>]/<name>.<section>
664: * or
665: * [./]cat<section>[/<arch>]/<name>.0
666: *
667: * Stuff this information directly into the "of" vector.
668: * See treescan() for the fts(3) version of this.
669: */
670: static void
1.59 schwarze 671: filescan(const char *file)
1.50 kristaps 672: {
1.59 schwarze 673: char buf[PATH_MAX];
1.50 kristaps 674: const char *sec, *arch, *name, *dsec;
1.59 schwarze 675: char *p, *start;
1.50 kristaps 676: int dform;
677: struct stat st;
1.5 kristaps 678:
1.50 kristaps 679: assert(use_all);
1.5 kristaps 680:
1.50 kristaps 681: if (0 == strncmp(file, "./", 2))
682: file += 2;
1.5 kristaps 683:
1.59 schwarze 684: if (NULL == realpath(file, buf)) {
685: exitcode = (int)MANDOCLEVEL_BADARG;
686: say(file, NULL);
687: return;
1.72 schwarze 688: } else if (OP_TEST != op && strstr(buf, basedir) != buf) {
1.59 schwarze 689: exitcode = (int)MANDOCLEVEL_BADARG;
690: say("", "%s: outside base directory", buf);
691: return;
692: } else if (-1 == stat(buf, &st)) {
693: exitcode = (int)MANDOCLEVEL_BADARG;
694: say(file, NULL);
1.50 kristaps 695: return;
696: } else if ( ! (S_IFREG & st.st_mode)) {
1.59 schwarze 697: exitcode = (int)MANDOCLEVEL_BADARG;
698: say(file, "Not a regular file");
1.50 kristaps 699: return;
700: } else if (inocheck(&st)) {
1.56 schwarze 701: if (warnings)
1.59 schwarze 702: say(file, "Duplicate file");
1.50 kristaps 703: return;
704: }
1.59 schwarze 705: start = buf + strlen(basedir);
1.50 kristaps 706: sec = arch = name = dsec = NULL;
707: dform = FORM_NONE;
1.17 schwarze 708:
1.50 kristaps 709: /*
710: * First try to guess our directory structure.
711: * If we find a separator, try to look for man* or cat*.
712: * If we find one of these and what's underneath is a directory,
713: * assume it's an architecture.
714: */
715: if (NULL != (p = strchr(start, '/'))) {
716: *p++ = '\0';
717: if (0 == strncmp(start, "man", 3)) {
718: dform = FORM_SRC;
719: dsec = start + 3;
720: } else if (0 == strncmp(start, "cat", 3)) {
721: dform = FORM_CAT;
722: dsec = start + 3;
1.17 schwarze 723: }
1.5 kristaps 724:
1.50 kristaps 725: start = p;
726: if (NULL != dsec && NULL != (p = strchr(start, '/'))) {
727: *p++ = '\0';
728: arch = start;
729: start = p;
730: }
731: }
732:
733: /*
734: * Now check the file suffix.
735: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
736: */
737: p = strrchr(start, '\0');
738: while (p-- > start && '/' != *p && '.' != *p)
739: /* Loop. */ ;
740:
741: if ('.' == *p) {
742: *p++ = '\0';
743: sec = p;
1.5 kristaps 744: }
745:
1.10 kristaps 746: /*
1.50 kristaps 747: * Now try to parse the name.
748: * Use the filename portion of the path.
1.10 kristaps 749: */
1.50 kristaps 750: name = start;
751: if (NULL != (p = strrchr(start, '/'))) {
752: name = p + 1;
753: *p = '\0';
754: }
755:
1.59 schwarze 756: ofadd(dform, file, name, dsec, sec, arch, &st);
1.50 kristaps 757: }
758:
759: /*
760: * See fileadd().
761: */
762: static int
763: filecheck(const char *name)
764: {
1.10 kristaps 765:
1.65 schwarze 766: return(NULL != ohash_find(&filenames,
767: ohash_qlookup(&filenames, name)));
1.50 kristaps 768: }
1.10 kristaps 769:
1.50 kristaps 770: /*
771: * Use the standard hashing mechanism (K&R) to see if the given filename
772: * already exists.
773: */
774: static void
775: fileadd(struct of *of)
776: {
1.65 schwarze 777: unsigned int slot;
1.1 kristaps 778:
1.65 schwarze 779: slot = ohash_qlookup(&filenames, of->file);
780: assert(NULL == ohash_find(&filenames, slot));
781: ohash_insert(&filenames, slot, of);
1.50 kristaps 782: }
1.3 kristaps 783:
1.50 kristaps 784: /*
785: * See inoadd().
786: */
787: static int
788: inocheck(const struct stat *st)
789: {
790: struct id id;
791: uint32_t hash;
792:
793: memset(&id, 0, sizeof(id));
794: id.ino = hash = st->st_ino;
795: id.dev = st->st_dev;
1.13 schwarze 796:
1.65 schwarze 797: return(NULL != ohash_find(&inos, ohash_lookup_memory(
798: &inos, (char *)&id, sizeof(id), hash)));
1.50 kristaps 799: }
1.5 kristaps 800:
1.50 kristaps 801: /*
802: * The hashing function used here is quite simple: simply take the inode
803: * and use uint32_t of its bits.
804: * Then when we do the lookup, use both the inode and device identifier.
805: */
806: static void
807: inoadd(const struct stat *st, struct of *of)
808: {
809: uint32_t hash;
1.65 schwarze 810: unsigned int slot;
1.1 kristaps 811:
1.50 kristaps 812: of->id.ino = hash = st->st_ino;
813: of->id.dev = st->st_dev;
1.65 schwarze 814: slot = ohash_lookup_memory
1.50 kristaps 815: (&inos, (char *)&of->id, sizeof(of->id), hash);
1.1 kristaps 816:
1.65 schwarze 817: assert(NULL == ohash_find(&inos, slot));
818: ohash_insert(&inos, slot, of);
1.50 kristaps 819: }
1.35 kristaps 820:
1.50 kristaps 821: static void
1.59 schwarze 822: ofadd(int dform, const char *file, const char *name, const char *dsec,
823: const char *sec, const char *arch, const struct stat *st)
1.50 kristaps 824: {
825: struct of *of;
826: int sform;
827:
828: assert(NULL != file);
829:
830: if (NULL == name)
831: name = "";
832: if (NULL == sec)
833: sec = "";
834: if (NULL == dsec)
835: dsec = "";
836: if (NULL == arch)
837: arch = "";
838:
1.74 schwarze 839: if ('0' == *sec) {
1.50 kristaps 840: sec = dsec;
841: sform = FORM_CAT;
1.74 schwarze 842: } else if ('1' <= *sec && '9' >= *sec)
843: sform = FORM_SRC;
844: else
845: sform = FORM_NONE;
1.50 kristaps 846:
847: of = mandoc_calloc(1, sizeof(struct of));
1.58 schwarze 848: strlcpy(of->file, file, PATH_MAX);
1.66 schwarze 849: of->name = mandoc_strdup(name);
850: of->sec = mandoc_strdup(sec);
851: of->dsec = mandoc_strdup(dsec);
852: of->arch = mandoc_strdup(arch);
1.50 kristaps 853: of->sform = sform;
854: of->dform = dform;
855: of->next = ofs;
856: ofs = of;
1.3 kristaps 857:
1.50 kristaps 858: /*
859: * Add to unique identifier hash.
860: * Then if it's a source manual and we're going to use source in
861: * favour of catpages, add it to that hash.
862: */
863: inoadd(st, of);
864: fileadd(of);
865: }
1.3 kristaps 866:
1.50 kristaps 867: static void
868: offree(void)
869: {
870: struct of *of;
1.3 kristaps 871:
1.50 kristaps 872: while (NULL != (of = ofs)) {
873: ofs = of->next;
1.66 schwarze 874: free(of->name);
875: free(of->sec);
876: free(of->dsec);
877: free(of->arch);
1.50 kristaps 878: free(of);
879: }
880: }
1.38 schwarze 881:
1.50 kristaps 882: /*
883: * Run through the files in the global vector "ofs" and add them to the
1.59 schwarze 884: * database specified in "basedir".
1.50 kristaps 885: *
886: * This handles the parsing scheme itself, using the cues of directory
887: * and filename to determine whether the file is parsable or not.
888: */
1.59 schwarze 889: static void
1.71 schwarze 890: ofmerge(struct mchars *mc, struct mparse *mp, int check_reachable)
1.50 kristaps 891: {
1.69 schwarze 892: struct ohash title_table;
1.71 schwarze 893: struct ohash_info title_info, str_info;
1.69 schwarze 894: char buf[PATH_MAX];
895: struct of *of;
896: struct mdoc *mdoc;
897: struct man *man;
898: struct title *title_entry;
899: char *bufp, *title_str;
900: const char *msec, *march, *mtitle, *cp;
901: size_t sz;
902: int form;
903: int match;
904: unsigned int slot;
905: enum mandoclevel lvl;
906:
1.71 schwarze 907: str_info.alloc = hash_alloc;
908: str_info.halloc = hash_halloc;
909: str_info.hfree = hash_free;
910: str_info.key_offset = offsetof(struct str, key);
911:
1.69 schwarze 912: if (check_reachable) {
913: title_info.alloc = hash_alloc;
914: title_info.halloc = hash_halloc;
915: title_info.hfree = hash_free;
916: title_info.key_offset = offsetof(struct title, title);
917: ohash_init(&title_table, 6, &title_info);
918: }
1.50 kristaps 919:
920: for (of = ofs; NULL != of; of = of->next) {
921: /*
922: * If we're a catpage (as defined by our path), then see
923: * if a manpage exists by the same name (ignoring the
924: * suffix).
925: * If it does, then we want to use it instead of our
926: * own.
927: */
928: if ( ! use_all && FORM_CAT == of->dform) {
1.58 schwarze 929: sz = strlcpy(buf, of->file, PATH_MAX);
930: if (sz >= PATH_MAX) {
1.56 schwarze 931: if (warnings)
1.59 schwarze 932: say(of->file, "Filename too long");
1.50 kristaps 933: continue;
934: }
935: bufp = strstr(buf, "cat");
936: assert(NULL != bufp);
937: memcpy(bufp, "man", 3);
938: if (NULL != (bufp = strrchr(buf, '.')))
939: *++bufp = '\0';
1.58 schwarze 940: strlcat(buf, of->dsec, PATH_MAX);
1.50 kristaps 941: if (filecheck(buf)) {
1.56 schwarze 942: if (warnings)
1.59 schwarze 943: say(of->file, "Man "
1.56 schwarze 944: "source exists: %s", buf);
1.50 kristaps 945: continue;
946: }
947: }
948:
1.71 schwarze 949: ohash_init(&strings, 6, &str_info);
1.50 kristaps 950: mparse_reset(mp);
951: mdoc = NULL;
952: man = NULL;
953: form = 0;
954: msec = of->dsec;
955: march = of->arch;
956: mtitle = of->name;
1.69 schwarze 957: match = 1;
1.14 schwarze 958:
959: /*
1.33 schwarze 960: * Try interpreting the file as mdoc(7) or man(7)
961: * source code, unless it is already known to be
962: * formatted. Fall back to formatted mode.
1.14 schwarze 963: */
1.75 ! schwarze 964: if (FORM_CAT != of->dform || FORM_CAT != of->sform) {
1.50 kristaps 965: lvl = mparse_readfd(mp, -1, of->file);
966: if (lvl < MANDOCLEVEL_FATAL)
967: mparse_result(mp, &mdoc, &man);
968: }
1.14 schwarze 969:
970: if (NULL != mdoc) {
1.50 kristaps 971: form = 1;
1.14 schwarze 972: msec = mdoc_meta(mdoc)->msec;
1.38 schwarze 973: march = mdoc_meta(mdoc)->arch;
1.14 schwarze 974: mtitle = mdoc_meta(mdoc)->title;
975: } else if (NULL != man) {
1.50 kristaps 976: form = 1;
1.14 schwarze 977: msec = man_meta(man)->msec;
1.38 schwarze 978: march = "";
1.14 schwarze 979: mtitle = man_meta(man)->title;
1.50 kristaps 980: }
981:
982: if (NULL == msec)
983: msec = "";
984: if (NULL == march)
985: march = "";
986: if (NULL == mtitle)
987: mtitle = "";
1.1 kristaps 988:
1.12 schwarze 989: /*
1.44 kristaps 990: * Check whether the manual section given in a file
991: * agrees with the directory where the file is located.
992: * Some manuals have suffixes like (3p) on their
993: * section number either inside the file or in the
994: * directory name, some are linked into more than one
995: * section, like encrypt(1) = makekey(8). Do not skip
996: * manuals for such reasons.
1.12 schwarze 997: */
1.56 schwarze 998: if (warnings && !use_all && form &&
1.69 schwarze 999: strcasecmp(msec, of->dsec)) {
1000: match = 0;
1.59 schwarze 1001: say(of->file, "Section \"%s\" "
1.50 kristaps 1002: "manual in %s directory",
1003: msec, of->dsec);
1.69 schwarze 1004: }
1.12 schwarze 1005:
1.42 schwarze 1006: /*
1007: * Manual page directories exist for each kernel
1008: * architecture as returned by machine(1).
1009: * However, many manuals only depend on the
1010: * application architecture as returned by arch(1).
1011: * For example, some (2/ARM) manuals are shared
1012: * across the "armish" and "zaurus" kernel
1013: * architectures.
1014: * A few manuals are even shared across completely
1015: * different architectures, for example fdformat(1)
1016: * on amd64, i386, sparc, and sparc64.
1017: * Thus, warn about architecture mismatches,
1018: * but don't skip manuals for this reason.
1019: */
1.69 schwarze 1020: if (warnings && !use_all && strcasecmp(march, of->arch)) {
1021: match = 0;
1.59 schwarze 1022: say(of->file, "Architecture \"%s\" "
1.47 schwarze 1023: "manual in \"%s\" directory",
1.45 kristaps 1024: march, of->arch);
1.69 schwarze 1025: }
1026: if (warnings && !use_all && strcasecmp(mtitle, of->name))
1027: match = 0;
1.12 schwarze 1028:
1.50 kristaps 1029: putkey(of, of->name, TYPE_Nm);
1.12 schwarze 1030:
1.50 kristaps 1031: if (NULL != mdoc) {
1032: if (NULL != (cp = mdoc_meta(mdoc)->name))
1033: putkey(of, cp, TYPE_Nm);
1.66 schwarze 1034: assert(NULL == of->desc);
1.50 kristaps 1035: parse_mdoc(of, mdoc_node(mdoc));
1.66 schwarze 1036: putkey(of, NULL != of->desc ?
1037: of->desc : of->name, TYPE_Nd);
1.50 kristaps 1038: } else if (NULL != man)
1039: parse_man(of, man_node(man));
1040: else
1.59 schwarze 1041: parse_catpage(of);
1.44 kristaps 1042:
1.69 schwarze 1043: /*
1044: * Build a title string for the file. If it matches
1045: * the location of the file, remember the title as
1046: * found; else, remember it as missing.
1047: */
1048:
1049: if (check_reachable) {
1050: if (-1 == asprintf(&title_str, "%s(%s%s%s)", mtitle,
1051: msec, '\0' == *march ? "" : "/", march)) {
1052: perror(NULL);
1053: exit((int)MANDOCLEVEL_SYSERR);
1054: }
1055: slot = ohash_qlookup(&title_table, title_str);
1056: title_entry = ohash_find(&title_table, slot);
1057: if (NULL == title_entry) {
1058: title_entry = mandoc_malloc(
1059: sizeof(struct title));
1060: title_entry->title = title_str;
1061: title_entry->file = mandoc_strdup(
1062: match ? "" : of->file);
1063: ohash_insert(&title_table, slot,
1064: title_entry);
1065: } else {
1066: if (match)
1067: *title_entry->file = '\0';
1068: free(title_str);
1069: }
1070: }
1071:
1.59 schwarze 1072: dbindex(mc, form, of);
1.66 schwarze 1073: ohash_delete(&strings);
1.69 schwarze 1074: }
1075:
1076: if (check_reachable) {
1077: title_entry = ohash_first(&title_table, &slot);
1078: while (NULL != title_entry) {
1079: if ('\0' != *title_entry->file)
1080: say(title_entry->file,
1081: "Probably unreachable, title is %s",
1082: title_entry->title);
1083: free(title_entry->title);
1084: free(title_entry->file);
1085: free(title_entry);
1086: title_entry = ohash_next(&title_table, &slot);
1087: }
1088: ohash_delete(&title_table);
1.50 kristaps 1089: }
1090: }
1.12 schwarze 1091:
1.50 kristaps 1092: static void
1.59 schwarze 1093: parse_catpage(struct of *of)
1.50 kristaps 1094: {
1095: FILE *stream;
1096: char *line, *p, *title;
1097: size_t len, plen, titlesz;
1.12 schwarze 1098:
1.50 kristaps 1099: if (NULL == (stream = fopen(of->file, "r"))) {
1.56 schwarze 1100: if (warnings)
1.59 schwarze 1101: say(of->file, NULL);
1.50 kristaps 1102: return;
1103: }
1.1 kristaps 1104:
1.50 kristaps 1105: /* Skip to first blank line. */
1.1 kristaps 1106:
1.50 kristaps 1107: while (NULL != (line = fgetln(stream, &len)))
1108: if ('\n' == *line)
1109: break;
1.1 kristaps 1110:
1.50 kristaps 1111: /*
1112: * Assume the first line that is not indented
1113: * is the first section header. Skip to it.
1114: */
1.1 kristaps 1115:
1.50 kristaps 1116: while (NULL != (line = fgetln(stream, &len)))
1117: if ('\n' != *line && ' ' != *line)
1118: break;
1119:
1120: /*
1121: * Read up until the next section into a buffer.
1122: * Strip the leading and trailing newline from each read line,
1123: * appending a trailing space.
1124: * Ignore empty (whitespace-only) lines.
1125: */
1.1 kristaps 1126:
1.50 kristaps 1127: titlesz = 0;
1128: title = NULL;
1.38 schwarze 1129:
1.50 kristaps 1130: while (NULL != (line = fgetln(stream, &len))) {
1131: if (' ' != *line || '\n' != line[len - 1])
1132: break;
1133: while (len > 0 && isspace((unsigned char)*line)) {
1134: line++;
1135: len--;
1136: }
1137: if (1 == len)
1.38 schwarze 1138: continue;
1.50 kristaps 1139: title = mandoc_realloc(title, titlesz + len);
1140: memcpy(title + titlesz, line, len);
1141: titlesz += len;
1142: title[titlesz - 1] = ' ';
1143: }
1.38 schwarze 1144:
1.50 kristaps 1145: /*
1146: * If no page content can be found, or the input line
1147: * is already the next section header, or there is no
1148: * trailing newline, reuse the page title as the page
1149: * description.
1150: */
1.44 kristaps 1151:
1.50 kristaps 1152: if (NULL == title || '\0' == *title) {
1.56 schwarze 1153: if (warnings)
1.59 schwarze 1154: say(of->file, "Cannot find NAME section");
1.66 schwarze 1155: assert(NULL == of->desc);
1156: of->desc = mandoc_strdup(of->name);
1.62 schwarze 1157: putkey(of, of->name, TYPE_Nd);
1.50 kristaps 1158: fclose(stream);
1159: free(title);
1160: return;
1161: }
1.1 kristaps 1162:
1.50 kristaps 1163: title = mandoc_realloc(title, titlesz + 1);
1164: title[titlesz] = '\0';
1.33 schwarze 1165:
1.50 kristaps 1166: /*
1167: * Skip to the first dash.
1168: * Use the remaining line as the description (no more than 70
1169: * bytes).
1170: */
1.33 schwarze 1171:
1.50 kristaps 1172: if (NULL != (p = strstr(title, "- "))) {
1173: for (p += 2; ' ' == *p || '\b' == *p; p++)
1174: /* Skip to next word. */ ;
1175: } else {
1.56 schwarze 1176: if (warnings)
1.59 schwarze 1177: say(of->file, "No dash in title line");
1.50 kristaps 1178: p = title;
1179: }
1.38 schwarze 1180:
1.50 kristaps 1181: plen = strlen(p);
1.1 kristaps 1182:
1.50 kristaps 1183: /* Strip backspace-encoding from line. */
1.1 kristaps 1184:
1.50 kristaps 1185: while (NULL != (line = memchr(p, '\b', plen))) {
1186: len = line - p;
1187: if (0 == len) {
1188: memmove(line, line + 1, plen--);
1189: continue;
1190: }
1191: memmove(line - 1, line + 1, plen - len);
1192: plen -= 2;
1193: }
1.1 kristaps 1194:
1.66 schwarze 1195: assert(NULL == of->desc);
1196: of->desc = mandoc_strdup(p);
1197: putkey(of, of->desc, TYPE_Nd);
1.50 kristaps 1198: fclose(stream);
1199: free(title);
1200: }
1.1 kristaps 1201:
1.50 kristaps 1202: /*
1203: * Put a type/word pair into the word database for this particular file.
1204: */
1205: static void
1206: putkey(const struct of *of, const char *value, uint64_t type)
1207: {
1.18 kristaps 1208:
1.50 kristaps 1209: assert(NULL != value);
1.64 schwarze 1210: putkeys(of, value, strlen(value), type);
1.3 kristaps 1211: }
1212:
1213: /*
1.50 kristaps 1214: * Grok all nodes at or below a certain mdoc node into putkey().
1.3 kristaps 1215: */
1216: static void
1.50 kristaps 1217: putmdockey(const struct of *of, const struct mdoc_node *n, uint64_t m)
1.3 kristaps 1218: {
1.18 kristaps 1219:
1.50 kristaps 1220: for ( ; NULL != n; n = n->next) {
1221: if (NULL != n->child)
1222: putmdockey(of, n->child, m);
1223: if (MDOC_TEXT == n->type)
1224: putkey(of, n->string, m);
1225: }
1226: }
1.18 kristaps 1227:
1.61 schwarze 1228: static void
1.50 kristaps 1229: parse_man(struct of *of, const struct man_node *n)
1230: {
1231: const struct man_node *head, *body;
1232: char *start, *sv, *title;
1233: char byte;
1234: size_t sz, titlesz;
1.18 kristaps 1235:
1.50 kristaps 1236: if (NULL == n)
1.61 schwarze 1237: return;
1.18 kristaps 1238:
1.50 kristaps 1239: /*
1240: * We're only searching for one thing: the first text child in
1241: * the BODY of a NAME section. Since we don't keep track of
1242: * sections in -man, run some hoops to find out whether we're in
1243: * the correct section or not.
1244: */
1.18 kristaps 1245:
1.50 kristaps 1246: if (MAN_BODY == n->type && MAN_SH == n->tok) {
1247: body = n;
1248: assert(body->parent);
1249: if (NULL != (head = body->parent->head) &&
1250: 1 == head->nchild &&
1251: NULL != (head = (head->child)) &&
1252: MAN_TEXT == head->type &&
1253: 0 == strcmp(head->string, "NAME") &&
1254: NULL != (body = body->child) &&
1255: MAN_TEXT == body->type) {
1.3 kristaps 1256:
1.50 kristaps 1257: title = NULL;
1258: titlesz = 0;
1.3 kristaps 1259:
1.50 kristaps 1260: /*
1261: * Suck the entire NAME section into memory.
1262: * Yes, we might run away.
1263: * But too many manuals have big, spread-out
1264: * NAME sections over many lines.
1265: */
1.3 kristaps 1266:
1.50 kristaps 1267: for ( ; NULL != body; body = body->next) {
1268: if (MAN_TEXT != body->type)
1269: break;
1270: if (0 == (sz = strlen(body->string)))
1271: continue;
1272: title = mandoc_realloc
1273: (title, titlesz + sz + 1);
1274: memcpy(title + titlesz, body->string, sz);
1275: titlesz += sz + 1;
1276: title[titlesz - 1] = ' ';
1277: }
1278: if (NULL == title)
1.61 schwarze 1279: return;
1.18 kristaps 1280:
1.50 kristaps 1281: title = mandoc_realloc(title, titlesz + 1);
1282: title[titlesz] = '\0';
1.18 kristaps 1283:
1.50 kristaps 1284: /* Skip leading space. */
1.18 kristaps 1285:
1.50 kristaps 1286: sv = title;
1287: while (isspace((unsigned char)*sv))
1288: sv++;
1.18 kristaps 1289:
1.50 kristaps 1290: if (0 == (sz = strlen(sv))) {
1291: free(title);
1.61 schwarze 1292: return;
1.50 kristaps 1293: }
1.1 kristaps 1294:
1.50 kristaps 1295: /* Erase trailing space. */
1.1 kristaps 1296:
1.50 kristaps 1297: start = &sv[sz - 1];
1298: while (start > sv && isspace((unsigned char)*start))
1299: *start-- = '\0';
1.1 kristaps 1300:
1.50 kristaps 1301: if (start == sv) {
1302: free(title);
1.61 schwarze 1303: return;
1.50 kristaps 1304: }
1.1 kristaps 1305:
1.50 kristaps 1306: start = sv;
1.18 kristaps 1307:
1.50 kristaps 1308: /*
1309: * Go through a special heuristic dance here.
1310: * Conventionally, one or more manual names are
1311: * comma-specified prior to a whitespace, then a
1312: * dash, then a description. Try to puzzle out
1313: * the name parts here.
1314: */
1.18 kristaps 1315:
1.50 kristaps 1316: for ( ;; ) {
1317: sz = strcspn(start, " ,");
1318: if ('\0' == start[sz])
1319: break;
1.1 kristaps 1320:
1.50 kristaps 1321: byte = start[sz];
1322: start[sz] = '\0';
1.1 kristaps 1323:
1.50 kristaps 1324: putkey(of, start, TYPE_Nm);
1.1 kristaps 1325:
1.50 kristaps 1326: if (' ' == byte) {
1327: start += sz + 1;
1328: break;
1329: }
1.1 kristaps 1330:
1.50 kristaps 1331: assert(',' == byte);
1332: start += sz + 1;
1333: while (' ' == *start)
1334: start++;
1335: }
1.1 kristaps 1336:
1.50 kristaps 1337: if (sv == start) {
1338: putkey(of, start, TYPE_Nm);
1339: free(title);
1.61 schwarze 1340: return;
1.50 kristaps 1341: }
1.1 kristaps 1342:
1.50 kristaps 1343: while (isspace((unsigned char)*start))
1344: start++;
1.1 kristaps 1345:
1.50 kristaps 1346: if (0 == strncmp(start, "-", 1))
1347: start += 1;
1348: else if (0 == strncmp(start, "\\-\\-", 4))
1349: start += 4;
1350: else if (0 == strncmp(start, "\\-", 2))
1351: start += 2;
1352: else if (0 == strncmp(start, "\\(en", 4))
1353: start += 4;
1354: else if (0 == strncmp(start, "\\(em", 4))
1355: start += 4;
1.1 kristaps 1356:
1.50 kristaps 1357: while (' ' == *start)
1358: start++;
1.1 kristaps 1359:
1.50 kristaps 1360: assert(NULL == of->desc);
1.66 schwarze 1361: of->desc = mandoc_strdup(start);
1362: putkey(of, of->desc, TYPE_Nd);
1.50 kristaps 1363: free(title);
1.61 schwarze 1364: return;
1.50 kristaps 1365: }
1366: }
1.1 kristaps 1367:
1.50 kristaps 1368: for (n = n->child; n; n = n->next)
1.61 schwarze 1369: parse_man(of, n);
1.1 kristaps 1370: }
1371:
1372: static void
1.50 kristaps 1373: parse_mdoc(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1374: {
1375:
1.50 kristaps 1376: assert(NULL != n);
1377: for (n = n->child; NULL != n; n = n->next) {
1378: switch (n->type) {
1379: case (MDOC_ELEM):
1380: /* FALLTHROUGH */
1381: case (MDOC_BLOCK):
1382: /* FALLTHROUGH */
1383: case (MDOC_HEAD):
1384: /* FALLTHROUGH */
1385: case (MDOC_BODY):
1386: /* FALLTHROUGH */
1387: case (MDOC_TAIL):
1388: if (NULL != mdocs[n->tok].fp)
1389: if (0 == (*mdocs[n->tok].fp)(of, n))
1390: break;
1.68 schwarze 1391: if (mdocs[n->tok].mask)
1.50 kristaps 1392: putmdockey(of, n->child, mdocs[n->tok].mask);
1393: break;
1394: default:
1395: assert(MDOC_ROOT != n->type);
1396: continue;
1397: }
1398: if (NULL != n->child)
1399: parse_mdoc(of, n);
1.1 kristaps 1400: }
1401: }
1402:
1.25 schwarze 1403: static int
1.50 kristaps 1404: parse_mdoc_Fd(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1405: {
1406: const char *start, *end;
1407: size_t sz;
1.25 schwarze 1408:
1.50 kristaps 1409: if (SEC_SYNOPSIS != n->sec ||
1410: NULL == (n = n->child) ||
1411: MDOC_TEXT != n->type)
1.25 schwarze 1412: return(0);
1.1 kristaps 1413:
1414: /*
1415: * Only consider those `Fd' macro fields that begin with an
1416: * "inclusion" token (versus, e.g., #define).
1417: */
1.50 kristaps 1418:
1.1 kristaps 1419: if (strcmp("#include", n->string))
1.25 schwarze 1420: return(0);
1.1 kristaps 1421:
1422: if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1.25 schwarze 1423: return(0);
1.1 kristaps 1424:
1425: /*
1426: * Strip away the enclosing angle brackets and make sure we're
1427: * not zero-length.
1428: */
1429:
1430: start = n->string;
1431: if ('<' == *start || '"' == *start)
1432: start++;
1433:
1434: if (0 == (sz = strlen(start)))
1.25 schwarze 1435: return(0);
1.1 kristaps 1436:
1437: end = &start[(int)sz - 1];
1438: if ('>' == *end || '"' == *end)
1439: end--;
1440:
1.50 kristaps 1441: if (end > start)
1442: putkeys(of, start, end - start + 1, TYPE_In);
1.25 schwarze 1443: return(1);
1.1 kristaps 1444: }
1445:
1.25 schwarze 1446: static int
1.50 kristaps 1447: parse_mdoc_In(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1448: {
1449:
1.50 kristaps 1450: if (NULL != n->child && MDOC_TEXT == n->child->type)
1.25 schwarze 1451: return(0);
1.1 kristaps 1452:
1.50 kristaps 1453: putkey(of, n->child->string, TYPE_In);
1.25 schwarze 1454: return(1);
1.1 kristaps 1455: }
1456:
1.25 schwarze 1457: static int
1.50 kristaps 1458: parse_mdoc_Fn(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1459: {
1460: const char *cp;
1461:
1.50 kristaps 1462: if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1.25 schwarze 1463: return(0);
1464:
1.50 kristaps 1465: /*
1466: * Parse: .Fn "struct type *name" "char *arg".
1467: * First strip away pointer symbol.
1468: * Then store the function name, then type.
1469: * Finally, store the arguments.
1470: */
1.1 kristaps 1471:
1.50 kristaps 1472: if (NULL == (cp = strrchr(n->string, ' ')))
1473: cp = n->string;
1.1 kristaps 1474:
1475: while ('*' == *cp)
1476: cp++;
1477:
1.50 kristaps 1478: putkey(of, cp, TYPE_Fn);
1.25 schwarze 1479:
1.50 kristaps 1480: if (n->string < cp)
1481: putkeys(of, n->string, cp - n->string, TYPE_Ft);
1.25 schwarze 1482:
1.50 kristaps 1483: for (n = n->next; NULL != n; n = n->next)
1484: if (MDOC_TEXT == n->type)
1485: putkey(of, n->string, TYPE_Fa);
1.25 schwarze 1486:
1487: return(0);
1.1 kristaps 1488: }
1489:
1.25 schwarze 1490: static int
1.50 kristaps 1491: parse_mdoc_St(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1492: {
1.25 schwarze 1493:
1.1 kristaps 1494: if (NULL == n->child || MDOC_TEXT != n->child->type)
1.25 schwarze 1495: return(0);
1.1 kristaps 1496:
1.50 kristaps 1497: putkey(of, n->child->string, TYPE_St);
1.25 schwarze 1498: return(1);
1.1 kristaps 1499: }
1500:
1.25 schwarze 1501: static int
1.50 kristaps 1502: parse_mdoc_Xr(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1503: {
1.67 schwarze 1504: char *cp;
1.1 kristaps 1505:
1506: if (NULL == (n = n->child))
1.25 schwarze 1507: return(0);
1.1 kristaps 1508:
1.67 schwarze 1509: if (NULL == n->next) {
1510: putkey(of, n->string, TYPE_Xr);
1511: return(0);
1512: }
1513:
1514: if (-1 == asprintf(&cp, "%s(%s)", n->string, n->next->string)) {
1515: perror(NULL);
1516: exit((int)MANDOCLEVEL_SYSERR);
1517: }
1518: putkey(of, cp, TYPE_Xr);
1519: free(cp);
1520: return(0);
1.1 kristaps 1521: }
1522:
1.25 schwarze 1523: static int
1.50 kristaps 1524: parse_mdoc_Nd(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1525: {
1.50 kristaps 1526: size_t sz;
1.1 kristaps 1527:
1528: if (MDOC_BODY != n->type)
1.25 schwarze 1529: return(0);
1.1 kristaps 1530:
1.50 kristaps 1531: /*
1532: * Special-case the `Nd' because we need to put the description
1533: * into the document table.
1534: */
1535:
1536: for (n = n->child; NULL != n; n = n->next) {
1537: if (MDOC_TEXT == n->type) {
1.66 schwarze 1538: if (NULL != of->desc) {
1539: sz = strlen(of->desc) +
1540: strlen(n->string) + 2;
1541: of->desc = mandoc_realloc(of->desc, sz);
1542: strlcat(of->desc, " ", sz);
1543: strlcat(of->desc, n->string, sz);
1544: } else
1545: of->desc = mandoc_strdup(n->string);
1.50 kristaps 1546: }
1547: if (NULL != n->child)
1548: parse_mdoc_Nd(of, n);
1549: }
1.25 schwarze 1550: return(1);
1.1 kristaps 1551: }
1552:
1.25 schwarze 1553: static int
1.50 kristaps 1554: parse_mdoc_Nm(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1555: {
1556:
1.25 schwarze 1557: if (SEC_NAME == n->sec)
1558: return(1);
1559: else if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type)
1560: return(0);
1.1 kristaps 1561:
1.25 schwarze 1562: return(1);
1.1 kristaps 1563: }
1564:
1.25 schwarze 1565: static int
1.50 kristaps 1566: parse_mdoc_Sh(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1567: {
1568:
1.25 schwarze 1569: return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1.1 kristaps 1570: }
1571:
1.50 kristaps 1572: static int
1573: parse_mdoc_head(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1574: {
1575:
1.50 kristaps 1576: return(MDOC_HEAD == n->type);
1577: }
1.1 kristaps 1578:
1.50 kristaps 1579: static int
1580: parse_mdoc_body(struct of *of, const struct mdoc_node *n)
1581: {
1.1 kristaps 1582:
1.50 kristaps 1583: return(MDOC_BODY == n->type);
1.1 kristaps 1584: }
1585:
1.50 kristaps 1586: /*
1.66 schwarze 1587: * Add a string to the hash table for the current manual.
1588: * Each string has a bitmask telling which macros it belongs to.
1589: * When we finish the manual, we'll dump the table.
1.50 kristaps 1590: */
1591: static void
1.64 schwarze 1592: putkeys(const struct of *of, const char *cp, size_t sz, uint64_t v)
1.50 kristaps 1593: {
1594: struct str *s;
1.65 schwarze 1595: unsigned int slot;
1.50 kristaps 1596: const char *end;
1.25 schwarze 1597:
1.50 kristaps 1598: if (0 == sz)
1599: return;
1.25 schwarze 1600:
1.65 schwarze 1601: end = cp + sz;
1602: slot = ohash_qlookupi(&strings, cp, &end);
1603: s = ohash_find(&strings, slot);
1.25 schwarze 1604:
1.50 kristaps 1605: if (NULL != s && of == s->of) {
1606: s->mask |= v;
1607: return;
1608: } else if (NULL == s) {
1.51 kristaps 1609: s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
1.50 kristaps 1610: memcpy(s->key, cp, sz);
1.65 schwarze 1611: ohash_insert(&strings, slot, s);
1.1 kristaps 1612: }
1.50 kristaps 1613: s->of = of;
1614: s->mask = v;
1.1 kristaps 1615: }
1616:
1.50 kristaps 1617: /*
1618: * Take a Unicode codepoint and produce its UTF-8 encoding.
1619: * This isn't the best way to do this, but it works.
1620: * The magic numbers are from the UTF-8 packaging.
1621: * They're not as scary as they seem: read the UTF-8 spec for details.
1622: */
1623: static size_t
1624: utf8(unsigned int cp, char out[7])
1.1 kristaps 1625: {
1.50 kristaps 1626: size_t rc;
1.1 kristaps 1627:
1.50 kristaps 1628: rc = 0;
1629: if (cp <= 0x0000007F) {
1630: rc = 1;
1631: out[0] = (char)cp;
1632: } else if (cp <= 0x000007FF) {
1633: rc = 2;
1634: out[0] = (cp >> 6 & 31) | 192;
1635: out[1] = (cp & 63) | 128;
1636: } else if (cp <= 0x0000FFFF) {
1637: rc = 3;
1638: out[0] = (cp >> 12 & 15) | 224;
1639: out[1] = (cp >> 6 & 63) | 128;
1640: out[2] = (cp & 63) | 128;
1641: } else if (cp <= 0x001FFFFF) {
1642: rc = 4;
1643: out[0] = (cp >> 18 & 7) | 240;
1644: out[1] = (cp >> 12 & 63) | 128;
1645: out[2] = (cp >> 6 & 63) | 128;
1646: out[3] = (cp & 63) | 128;
1647: } else if (cp <= 0x03FFFFFF) {
1648: rc = 5;
1649: out[0] = (cp >> 24 & 3) | 248;
1650: out[1] = (cp >> 18 & 63) | 128;
1651: out[2] = (cp >> 12 & 63) | 128;
1652: out[3] = (cp >> 6 & 63) | 128;
1653: out[4] = (cp & 63) | 128;
1654: } else if (cp <= 0x7FFFFFFF) {
1655: rc = 6;
1656: out[0] = (cp >> 30 & 1) | 252;
1657: out[1] = (cp >> 24 & 63) | 128;
1658: out[2] = (cp >> 18 & 63) | 128;
1659: out[3] = (cp >> 12 & 63) | 128;
1660: out[4] = (cp >> 6 & 63) | 128;
1661: out[5] = (cp & 63) | 128;
1662: } else
1.1 kristaps 1663: return(0);
1664:
1.50 kristaps 1665: out[rc] = '\0';
1666: return(rc);
1667: }
1.1 kristaps 1668:
1.50 kristaps 1669: /*
1670: * Store the UTF-8 version of a key, or alias the pointer if the key has
1671: * no UTF-8 transcription marks in it.
1672: */
1673: static void
1674: utf8key(struct mchars *mc, struct str *key)
1675: {
1676: size_t sz, bsz, pos;
1677: char utfbuf[7], res[5];
1678: char *buf;
1679: const char *seq, *cpp, *val;
1680: int len, u;
1681: enum mandoc_esc esc;
1682:
1683: assert(NULL == key->utf8);
1684:
1685: res[0] = '\\';
1686: res[1] = '\t';
1687: res[2] = ASCII_NBRSP;
1688: res[3] = ASCII_HYPH;
1689: res[4] = '\0';
1.1 kristaps 1690:
1.50 kristaps 1691: val = key->key;
1692: bsz = strlen(val);
1.46 kristaps 1693:
1.50 kristaps 1694: /*
1695: * Pre-check: if we have no stop-characters, then set the
1696: * pointer as ourselvse and get out of here.
1697: */
1698: if (strcspn(val, res) == bsz) {
1699: key->utf8 = key->key;
1700: return;
1701: }
1.46 kristaps 1702:
1.50 kristaps 1703: /* Pre-allocate by the length of the input */
1.46 kristaps 1704:
1.50 kristaps 1705: buf = mandoc_malloc(++bsz);
1706: pos = 0;
1.46 kristaps 1707:
1.50 kristaps 1708: while ('\0' != *val) {
1709: /*
1710: * Halt on the first escape sequence.
1711: * This also halts on the end of string, in which case
1712: * we just copy, fallthrough, and exit the loop.
1713: */
1714: if ((sz = strcspn(val, res)) > 0) {
1715: memcpy(&buf[pos], val, sz);
1716: pos += sz;
1717: val += sz;
1718: }
1.46 kristaps 1719:
1.50 kristaps 1720: if (ASCII_HYPH == *val) {
1721: buf[pos++] = '-';
1722: val++;
1723: continue;
1724: } else if ('\t' == *val || ASCII_NBRSP == *val) {
1725: buf[pos++] = ' ';
1726: val++;
1727: continue;
1728: } else if ('\\' != *val)
1729: break;
1.46 kristaps 1730:
1.50 kristaps 1731: /* Read past the slash. */
1.46 kristaps 1732:
1.50 kristaps 1733: val++;
1734: u = 0;
1.46 kristaps 1735:
1.50 kristaps 1736: /*
1737: * Parse the escape sequence and see if it's a
1738: * predefined character or special character.
1739: */
1740: esc = mandoc_escape
1741: ((const char **)&val, &seq, &len);
1742: if (ESCAPE_ERROR == esc)
1743: break;
1.1 kristaps 1744:
1.50 kristaps 1745: if (ESCAPE_SPECIAL != esc)
1746: continue;
1747: if (0 == (u = mchars_spec2cp(mc, seq, len)))
1748: continue;
1.1 kristaps 1749:
1.50 kristaps 1750: /*
1751: * If we have a Unicode codepoint, try to convert that
1752: * to a UTF-8 byte string.
1753: */
1754: cpp = utfbuf;
1755: if (0 == (sz = utf8(u, utfbuf)))
1756: continue;
1.1 kristaps 1757:
1.50 kristaps 1758: /* Copy the rendered glyph into the stream. */
1.1 kristaps 1759:
1.50 kristaps 1760: sz = strlen(cpp);
1761: bsz += sz;
1.1 kristaps 1762:
1.50 kristaps 1763: buf = mandoc_realloc(buf, bsz);
1.1 kristaps 1764:
1.50 kristaps 1765: memcpy(&buf[pos], cpp, sz);
1766: pos += sz;
1.1 kristaps 1767: }
1768:
1.50 kristaps 1769: buf[pos] = '\0';
1770: key->utf8 = buf;
1.1 kristaps 1771: }
1772:
1.14 schwarze 1773: /*
1.50 kristaps 1774: * Flush the current page's terms (and their bits) into the database.
1775: * Wrap the entire set of additions in a transaction to make sqlite be a
1776: * little faster.
1777: * Also, UTF-8-encode the description at the last possible moment.
1.14 schwarze 1778: */
1779: static void
1.59 schwarze 1780: dbindex(struct mchars *mc, int form, const struct of *of)
1.14 schwarze 1781: {
1.50 kristaps 1782: struct str *key;
1783: const char *desc;
1784: int64_t recno;
1.52 kristaps 1785: size_t i;
1.66 schwarze 1786: unsigned int slot;
1.14 schwarze 1787:
1.56 schwarze 1788: if (verb)
1.59 schwarze 1789: say(of->file, "Adding to index");
1.43 kristaps 1790:
1.50 kristaps 1791: if (nodb)
1.14 schwarze 1792: return;
1.28 kristaps 1793:
1.50 kristaps 1794: desc = "";
1.73 schwarze 1795: if (NULL != of->desc && '\0' != *of->desc) {
1.65 schwarze 1796: key = ohash_find(&strings,
1797: ohash_qlookup(&strings, of->desc));
1.50 kristaps 1798: assert(NULL != key);
1799: if (NULL == key->utf8)
1800: utf8key(mc, key);
1801: desc = key->utf8;
1802: }
1803:
1.52 kristaps 1804: SQL_EXEC("BEGIN TRANSACTION");
1.50 kristaps 1805:
1.52 kristaps 1806: i = 1;
1807: SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, of->file);
1808: SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, of->sec);
1809: SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, of->arch);
1810: SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, desc);
1811: SQL_BIND_INT(stmts[STMT_INSERT_DOC], i, form);
1812: SQL_STEP(stmts[STMT_INSERT_DOC]);
1.50 kristaps 1813: recno = sqlite3_last_insert_rowid(db);
1814: sqlite3_reset(stmts[STMT_INSERT_DOC]);
1815:
1.66 schwarze 1816: for (key = ohash_first(&strings, &slot); NULL != key;
1817: key = ohash_next(&strings, &slot)) {
1.50 kristaps 1818: assert(key->of == of);
1819: if (NULL == key->utf8)
1820: utf8key(mc, key);
1.52 kristaps 1821: i = 1;
1822: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1823: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->utf8);
1824: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, recno);
1825: SQL_STEP(stmts[STMT_INSERT_KEY]);
1.50 kristaps 1826: sqlite3_reset(stmts[STMT_INSERT_KEY]);
1.66 schwarze 1827: if (key->utf8 != key->key)
1828: free(key->utf8);
1829: free(key);
1.38 schwarze 1830: }
1.28 kristaps 1831:
1.52 kristaps 1832: SQL_EXEC("END TRANSACTION");
1.14 schwarze 1833: }
1834:
1.5 kristaps 1835: static void
1.59 schwarze 1836: dbprune(void)
1.5 kristaps 1837: {
1.50 kristaps 1838: struct of *of;
1.52 kristaps 1839: size_t i;
1.5 kristaps 1840:
1.50 kristaps 1841: if (nodb)
1842: return;
1.12 schwarze 1843:
1.50 kristaps 1844: for (of = ofs; NULL != of; of = of->next) {
1.52 kristaps 1845: i = 1;
1846: SQL_BIND_TEXT(stmts[STMT_DELETE], i, of->file);
1847: SQL_STEP(stmts[STMT_DELETE]);
1.50 kristaps 1848: sqlite3_reset(stmts[STMT_DELETE]);
1.56 schwarze 1849: if (verb)
1.59 schwarze 1850: say(of->file, "Deleted from index");
1.5 kristaps 1851: }
1852: }
1853:
1.4 kristaps 1854: /*
1.50 kristaps 1855: * Close an existing database and its prepared statements.
1856: * If "real" is not set, rename the temporary file into the real one.
1.4 kristaps 1857: */
1.35 kristaps 1858: static void
1.59 schwarze 1859: dbclose(int real)
1.4 kristaps 1860: {
1.50 kristaps 1861: size_t i;
1.4 kristaps 1862:
1.50 kristaps 1863: if (nodb)
1.38 schwarze 1864: return;
1.50 kristaps 1865:
1866: for (i = 0; i < STMT__MAX; i++) {
1867: sqlite3_finalize(stmts[i]);
1868: stmts[i] = NULL;
1.4 kristaps 1869: }
1870:
1.50 kristaps 1871: sqlite3_close(db);
1872: db = NULL;
1.12 schwarze 1873:
1.50 kristaps 1874: if (real)
1875: return;
1.12 schwarze 1876:
1.63 schwarze 1877: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
1.59 schwarze 1878: exitcode = (int)MANDOCLEVEL_SYSERR;
1879: say(MANDOC_DB, NULL);
1880: }
1.50 kristaps 1881: }
1.14 schwarze 1882:
1.50 kristaps 1883: /*
1884: * This is straightforward stuff.
1885: * Open a database connection to a "temporary" database, then open a set
1886: * of prepared statements we'll use over and over again.
1887: * If "real" is set, we use the existing database; if not, we truncate a
1888: * temporary one.
1889: * Must be matched by dbclose().
1890: */
1891: static int
1.59 schwarze 1892: dbopen(int real)
1.50 kristaps 1893: {
1.63 schwarze 1894: const char *file, *sql;
1.50 kristaps 1895: int rc, ofl;
1.12 schwarze 1896:
1.50 kristaps 1897: if (nodb)
1898: return(1);
1.12 schwarze 1899:
1.63 schwarze 1900: ofl = SQLITE_OPEN_READWRITE;
1901: if (0 == real) {
1902: file = MANDOC_DB "~";
1903: if (-1 == remove(file) && ENOENT != errno) {
1904: exitcode = (int)MANDOCLEVEL_SYSERR;
1905: say(file, NULL);
1906: return(0);
1907: }
1908: ofl |= SQLITE_OPEN_EXCLUSIVE;
1909: } else
1910: file = MANDOC_DB;
1.45 kristaps 1911:
1.50 kristaps 1912: rc = sqlite3_open_v2(file, &db, ofl, NULL);
1913: if (SQLITE_OK == rc)
1.57 schwarze 1914: goto prepare_statements;
1.50 kristaps 1915: if (SQLITE_CANTOPEN != rc) {
1.59 schwarze 1916: exitcode = (int)MANDOCLEVEL_SYSERR;
1917: say(file, NULL);
1.50 kristaps 1918: return(0);
1919: }
1.12 schwarze 1920:
1.50 kristaps 1921: sqlite3_close(db);
1922: db = NULL;
1.12 schwarze 1923:
1.50 kristaps 1924: if (SQLITE_OK != (rc = sqlite3_open(file, &db))) {
1.59 schwarze 1925: exitcode = (int)MANDOCLEVEL_SYSERR;
1926: say(file, NULL);
1.50 kristaps 1927: return(0);
1928: }
1.12 schwarze 1929:
1.50 kristaps 1930: sql = "CREATE TABLE \"docs\" (\n"
1931: " \"file\" TEXT NOT NULL,\n"
1932: " \"sec\" TEXT NOT NULL,\n"
1933: " \"arch\" TEXT NOT NULL,\n"
1934: " \"desc\" TEXT NOT NULL,\n"
1935: " \"form\" INTEGER NOT NULL,\n"
1936: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1937: ");\n"
1938: "\n"
1939: "CREATE TABLE \"keys\" (\n"
1940: " \"bits\" INTEGER NOT NULL,\n"
1941: " \"key\" TEXT NOT NULL,\n"
1942: " \"docid\" INTEGER NOT NULL REFERENCES docs(id) "
1943: "ON DELETE CASCADE,\n"
1944: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1945: ");\n"
1946: "\n"
1947: "CREATE INDEX \"key_index\" ON keys (key);\n";
1.14 schwarze 1948:
1.50 kristaps 1949: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
1.59 schwarze 1950: exitcode = (int)MANDOCLEVEL_SYSERR;
1951: say(file, "%s", sqlite3_errmsg(db));
1.50 kristaps 1952: return(0);
1953: }
1.4 kristaps 1954:
1.57 schwarze 1955: prepare_statements:
1956: SQL_EXEC("PRAGMA foreign_keys = ON");
1.50 kristaps 1957: sql = "DELETE FROM docs where file=?";
1958: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE], NULL);
1959: sql = "INSERT INTO docs "
1960: "(file,sec,arch,desc,form) VALUES (?,?,?,?,?)";
1961: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_DOC], NULL);
1962: sql = "INSERT INTO keys "
1963: "(bits,key,docid) VALUES (?,?,?)";
1964: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.70 schwarze 1965:
1966: #ifndef __APPLE__
1967: /*
1968: * When opening a new database, we can turn off
1969: * synchronous mode for much better performance.
1970: */
1971:
1972: if (real)
1973: SQL_EXEC("PRAGMA synchronous = OFF");
1974: #endif
1975:
1.50 kristaps 1976: return(1);
1977: }
1.5 kristaps 1978:
1.50 kristaps 1979: static void *
1980: hash_halloc(size_t sz, void *arg)
1981: {
1.12 schwarze 1982:
1.50 kristaps 1983: return(mandoc_calloc(sz, 1));
1984: }
1.12 schwarze 1985:
1.50 kristaps 1986: static void *
1987: hash_alloc(size_t sz, void *arg)
1988: {
1.5 kristaps 1989:
1.50 kristaps 1990: return(mandoc_malloc(sz));
1991: }
1.41 kristaps 1992:
1.50 kristaps 1993: static void
1994: hash_free(void *p, size_t sz, void *arg)
1995: {
1.4 kristaps 1996:
1.50 kristaps 1997: free(p);
1.4 kristaps 1998: }
1999:
1.50 kristaps 2000: static int
1.59 schwarze 2001: set_basedir(const char *targetdir)
1.4 kristaps 2002: {
1.59 schwarze 2003: static char startdir[PATH_MAX];
2004: static int fd;
1.4 kristaps 2005:
1.59 schwarze 2006: /*
2007: * Remember where we started by keeping a fd open to the origin
2008: * path component: throughout this utility, we chdir() a lot to
2009: * handle relative paths, and by doing this, we can return to
2010: * the starting point.
2011: */
2012: if ('\0' == *startdir) {
2013: if (NULL == getcwd(startdir, PATH_MAX)) {
2014: exitcode = (int)MANDOCLEVEL_SYSERR;
2015: if (NULL != targetdir)
2016: say(".", NULL);
2017: return(0);
2018: }
2019: if (-1 == (fd = open(startdir, O_RDONLY, 0))) {
2020: exitcode = (int)MANDOCLEVEL_SYSERR;
2021: say(startdir, NULL);
2022: return(0);
2023: }
2024: if (NULL == targetdir)
2025: targetdir = startdir;
2026: } else {
2027: if (-1 == fd)
2028: return(0);
2029: if (-1 == fchdir(fd)) {
2030: close(fd);
2031: basedir[0] = '\0';
2032: exitcode = (int)MANDOCLEVEL_SYSERR;
2033: say(startdir, NULL);
2034: return(0);
2035: }
2036: if (NULL == targetdir) {
2037: close(fd);
2038: return(1);
2039: }
2040: }
2041: if (NULL == realpath(targetdir, basedir)) {
2042: basedir[0] = '\0';
2043: exitcode = (int)MANDOCLEVEL_BADARG;
2044: say(targetdir, NULL);
1.50 kristaps 2045: return(0);
1.59 schwarze 2046: } else if (-1 == chdir(basedir)) {
2047: exitcode = (int)MANDOCLEVEL_BADARG;
2048: say("", NULL);
1.50 kristaps 2049: return(0);
1.4 kristaps 2050: }
1.50 kristaps 2051: return(1);
1.56 schwarze 2052: }
2053:
2054: static void
1.59 schwarze 2055: say(const char *file, const char *format, ...)
1.56 schwarze 2056: {
2057: va_list ap;
2058:
1.59 schwarze 2059: if ('\0' != *basedir)
2060: fprintf(stderr, "%s", basedir);
2061: if ('\0' != *basedir && '\0' != *file)
2062: fputs("//", stderr);
1.56 schwarze 2063: if ('\0' != *file)
1.59 schwarze 2064: fprintf(stderr, "%s", file);
1.56 schwarze 2065: fputs(": ", stderr);
1.59 schwarze 2066:
2067: if (NULL == format) {
2068: perror(NULL);
2069: return;
2070: }
1.56 schwarze 2071:
2072: va_start(ap, format);
2073: vfprintf(stderr, format, ap);
2074: va_end(ap);
2075:
2076: fputc('\n', stderr);
1.1 kristaps 2077: }
CVSweb