Annotation of mandoc/mandocdb.c, Revision 1.121
1.121 ! schwarze 1: /* $Id: mandocdb.c,v 1.120 2014/03/23 12:11:18 schwarze Exp $ */
1.1 kristaps 2: /*
1.48 schwarze 3: * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.95 schwarze 4: * Copyright (c) 2011, 2012, 2013, 2014 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>
1.115 schwarze 23: #include <sys/wait.h>
1.1 kristaps 24:
25: #include <assert.h>
1.43 kristaps 26: #include <ctype.h>
1.63 schwarze 27: #include <errno.h>
1.1 kristaps 28: #include <fcntl.h>
1.50 kristaps 29: #include <fts.h>
1.1 kristaps 30: #include <getopt.h>
1.58 schwarze 31: #include <limits.h>
1.50 kristaps 32: #include <stddef.h>
1.59 schwarze 33: #include <stdio.h>
1.1 kristaps 34: #include <stdint.h>
35: #include <stdlib.h>
36: #include <string.h>
1.17 schwarze 37: #include <unistd.h>
1.1 kristaps 38:
1.53 kristaps 39: #ifdef HAVE_OHASH
1.50 kristaps 40: #include <ohash.h>
1.53 kristaps 41: #else
42: #include "compat_ohash.h"
43: #endif
1.50 kristaps 44: #include <sqlite3.h>
1.1 kristaps 45:
1.50 kristaps 46: #include "mdoc.h"
1.1 kristaps 47: #include "man.h"
48: #include "mandoc.h"
1.119 schwarze 49: #include "mandoc_aux.h"
1.10 kristaps 50: #include "manpath.h"
1.55 kristaps 51: #include "mansearch.h"
1.1 kristaps 52:
1.111 schwarze 53: extern int mansearch_keymax;
54: extern const char *const mansearch_keynames[];
55:
1.52 kristaps 56: #define SQL_EXEC(_v) \
57: if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
58: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
59: #define SQL_BIND_TEXT(_s, _i, _v) \
60: if (SQLITE_OK != sqlite3_bind_text \
61: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
62: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
63: #define SQL_BIND_INT(_s, _i, _v) \
64: if (SQLITE_OK != sqlite3_bind_int \
65: ((_s), (_i)++, (_v))) \
66: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
67: #define SQL_BIND_INT64(_s, _i, _v) \
68: if (SQLITE_OK != sqlite3_bind_int64 \
69: ((_s), (_i)++, (_v))) \
70: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
71: #define SQL_STEP(_s) \
72: if (SQLITE_DONE != sqlite3_step((_s))) \
73: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
74:
1.50 kristaps 75: enum op {
76: OP_DEFAULT = 0, /* new dbs from dir list or default config */
77: OP_CONFFILE, /* new databases from custom config file */
78: OP_UPDATE, /* delete/add entries in existing database */
79: OP_DELETE, /* delete entries from existing database */
80: OP_TEST /* change no databases, report potential problems */
1.38 schwarze 81: };
82:
1.50 kristaps 83: enum form {
1.84 schwarze 84: FORM_NONE, /* format is unknown */
85: FORM_SRC, /* format is -man or -mdoc */
86: FORM_CAT /* format is cat */
1.50 kristaps 87: };
1.38 schwarze 88:
1.50 kristaps 89: struct str {
1.96 schwarze 90: char *rendered; /* key in UTF-8 or ASCII form */
1.78 schwarze 91: const struct mpage *mpage; /* if set, the owning parse */
1.50 kristaps 92: uint64_t mask; /* bitmask in sequence */
1.96 schwarze 93: char key[]; /* may contain escape sequences */
1.38 schwarze 94: };
95:
1.79 schwarze 96: struct inodev {
97: ino_t st_ino;
98: dev_t st_dev;
1.50 kristaps 99: };
1.5 kristaps 100:
1.78 schwarze 101: struct mpage {
1.79 schwarze 102: struct inodev inodev; /* used for hashing routine */
1.118 schwarze 103: int64_t recno; /* id in mpages SQL table */
1.82 schwarze 104: enum form form; /* format from file content */
105: char *sec; /* section from file content */
106: char *arch; /* architecture from file content */
107: char *title; /* title from file content */
108: char *desc; /* description from file content */
109: struct mlink *mlinks; /* singly linked list */
110: };
111:
112: struct mlink {
1.58 schwarze 113: char file[PATH_MAX]; /* filename rel. to manpath */
1.82 schwarze 114: enum form dform; /* format from directory */
1.84 schwarze 115: enum form fform; /* format from file name suffix */
1.82 schwarze 116: char *dsec; /* section from directory */
117: char *arch; /* architecture from directory */
118: char *name; /* name from file name (not empty) */
119: char *fsec; /* section from file name suffix */
1.85 schwarze 120: struct mlink *next; /* singly linked list */
1.118 schwarze 121: struct mpage *mpage; /* parent */
1.3 kristaps 122: };
123:
1.50 kristaps 124: enum stmt {
1.81 schwarze 125: STMT_DELETE_PAGE = 0, /* delete mpage */
126: STMT_INSERT_PAGE, /* insert mpage */
127: STMT_INSERT_LINK, /* insert mlink */
128: STMT_INSERT_KEY, /* insert parsed key */
1.50 kristaps 129: STMT__MAX
1.1 kristaps 130: };
131:
1.78 schwarze 132: typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_node *);
1.1 kristaps 133:
1.50 kristaps 134: struct mdoc_handler {
135: mdoc_fp fp; /* optional handler */
136: uint64_t mask; /* set unless handler returns 0 */
1.1 kristaps 137: };
138:
1.59 schwarze 139: static void dbclose(int);
1.118 schwarze 140: static void dbadd(struct mpage *, struct mchars *);
141: static void dbadd_mlink(const struct mlink *mlink);
1.59 schwarze 142: static int dbopen(int);
143: static void dbprune(void);
144: static void filescan(const char *);
1.50 kristaps 145: static void *hash_alloc(size_t, void *);
146: static void hash_free(void *, size_t, void *);
147: static void *hash_halloc(size_t, void *);
1.84 schwarze 148: static void mlink_add(struct mlink *, const struct stat *);
1.93 schwarze 149: static int mlink_check(struct mpage *, struct mlink *);
1.82 schwarze 150: static void mlink_free(struct mlink *);
1.89 schwarze 151: static void mlinks_undupe(struct mpage *);
1.78 schwarze 152: static void mpages_free(void);
1.101 schwarze 153: static void mpages_merge(struct mchars *, struct mparse *);
1.78 schwarze 154: static void parse_cat(struct mpage *);
155: static void parse_man(struct mpage *, const struct man_node *);
156: static void parse_mdoc(struct mpage *, const struct mdoc_node *);
157: static int parse_mdoc_body(struct mpage *, const struct mdoc_node *);
158: static int parse_mdoc_head(struct mpage *, const struct mdoc_node *);
159: static int parse_mdoc_Fd(struct mpage *, const struct mdoc_node *);
160: static int parse_mdoc_Fn(struct mpage *, const struct mdoc_node *);
161: static int parse_mdoc_Nd(struct mpage *, const struct mdoc_node *);
162: static int parse_mdoc_Nm(struct mpage *, const struct mdoc_node *);
163: static int parse_mdoc_Sh(struct mpage *, const struct mdoc_node *);
164: static int parse_mdoc_Xr(struct mpage *, const struct mdoc_node *);
1.112 schwarze 165: static void putkey(const struct mpage *, char *, uint64_t);
1.78 schwarze 166: static void putkeys(const struct mpage *,
1.64 schwarze 167: const char *, size_t, uint64_t);
1.78 schwarze 168: static void putmdockey(const struct mpage *,
1.50 kristaps 169: const struct mdoc_node *, uint64_t);
1.96 schwarze 170: static void render_key(struct mchars *, struct str *);
1.59 schwarze 171: static void say(const char *, const char *, ...);
1.80 schwarze 172: static int set_basedir(const char *);
1.59 schwarze 173: static int treescan(void);
1.50 kristaps 174: static size_t utf8(unsigned int, char [7]);
175:
1.115 schwarze 176: static char tempfilename[32];
1.50 kristaps 177: static char *progname;
1.102 schwarze 178: static int nodb; /* no database changes */
1.116 schwarze 179: static int mparse_options; /* abort the parse early */
1.50 kristaps 180: static int use_all; /* use all found files */
181: static int verb; /* print what we're doing */
182: static int warnings; /* warn about crap */
1.95 schwarze 183: static int write_utf8; /* write UTF-8 output; else ASCII */
1.59 schwarze 184: static int exitcode; /* to be returned by main */
1.50 kristaps 185: static enum op op; /* operational mode */
1.59 schwarze 186: static char basedir[PATH_MAX]; /* current base directory */
1.78 schwarze 187: static struct ohash mpages; /* table of distinct manual pages */
1.83 schwarze 188: static struct ohash mlinks; /* table of directory entries */
1.50 kristaps 189: static struct ohash strings; /* table of all strings */
190: static sqlite3 *db = NULL; /* current database */
191: static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
1.25 schwarze 192:
193: static const struct mdoc_handler mdocs[MDOC_MAX] = {
1.68 schwarze 194: { NULL, 0 }, /* Ap */
195: { NULL, 0 }, /* Dd */
196: { NULL, 0 }, /* Dt */
197: { NULL, 0 }, /* Os */
198: { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
199: { parse_mdoc_head, TYPE_Ss }, /* Ss */
200: { NULL, 0 }, /* Pp */
201: { NULL, 0 }, /* D1 */
202: { NULL, 0 }, /* Dl */
203: { NULL, 0 }, /* Bd */
204: { NULL, 0 }, /* Ed */
205: { NULL, 0 }, /* Bl */
206: { NULL, 0 }, /* El */
207: { NULL, 0 }, /* It */
208: { NULL, 0 }, /* Ad */
209: { NULL, TYPE_An }, /* An */
210: { NULL, TYPE_Ar }, /* Ar */
211: { NULL, TYPE_Cd }, /* Cd */
212: { NULL, TYPE_Cm }, /* Cm */
213: { NULL, TYPE_Dv }, /* Dv */
214: { NULL, TYPE_Er }, /* Er */
215: { NULL, TYPE_Ev }, /* Ev */
216: { NULL, 0 }, /* Ex */
217: { NULL, TYPE_Fa }, /* Fa */
218: { parse_mdoc_Fd, 0 }, /* Fd */
219: { NULL, TYPE_Fl }, /* Fl */
220: { parse_mdoc_Fn, 0 }, /* Fn */
221: { NULL, TYPE_Ft }, /* Ft */
222: { NULL, TYPE_Ic }, /* Ic */
1.92 schwarze 223: { NULL, TYPE_In }, /* In */
1.68 schwarze 224: { NULL, TYPE_Li }, /* Li */
225: { parse_mdoc_Nd, TYPE_Nd }, /* Nd */
226: { parse_mdoc_Nm, TYPE_Nm }, /* Nm */
227: { NULL, 0 }, /* Op */
228: { NULL, 0 }, /* Ot */
229: { NULL, TYPE_Pa }, /* Pa */
230: { NULL, 0 }, /* Rv */
1.92 schwarze 231: { NULL, TYPE_St }, /* St */
1.68 schwarze 232: { NULL, TYPE_Va }, /* Va */
233: { parse_mdoc_body, TYPE_Va }, /* Vt */
234: { parse_mdoc_Xr, 0 }, /* Xr */
235: { NULL, 0 }, /* %A */
236: { NULL, 0 }, /* %B */
237: { NULL, 0 }, /* %D */
238: { NULL, 0 }, /* %I */
239: { NULL, 0 }, /* %J */
240: { NULL, 0 }, /* %N */
241: { NULL, 0 }, /* %O */
242: { NULL, 0 }, /* %P */
243: { NULL, 0 }, /* %R */
244: { NULL, 0 }, /* %T */
245: { NULL, 0 }, /* %V */
246: { NULL, 0 }, /* Ac */
247: { NULL, 0 }, /* Ao */
248: { NULL, 0 }, /* Aq */
249: { NULL, TYPE_At }, /* At */
250: { NULL, 0 }, /* Bc */
251: { NULL, 0 }, /* Bf */
252: { NULL, 0 }, /* Bo */
253: { NULL, 0 }, /* Bq */
254: { NULL, TYPE_Bsx }, /* Bsx */
255: { NULL, TYPE_Bx }, /* Bx */
256: { NULL, 0 }, /* Db */
257: { NULL, 0 }, /* Dc */
258: { NULL, 0 }, /* Do */
259: { NULL, 0 }, /* Dq */
260: { NULL, 0 }, /* Ec */
261: { NULL, 0 }, /* Ef */
262: { NULL, TYPE_Em }, /* Em */
263: { NULL, 0 }, /* Eo */
264: { NULL, TYPE_Fx }, /* Fx */
265: { NULL, TYPE_Ms }, /* Ms */
266: { NULL, 0 }, /* No */
267: { NULL, 0 }, /* Ns */
268: { NULL, TYPE_Nx }, /* Nx */
269: { NULL, TYPE_Ox }, /* Ox */
270: { NULL, 0 }, /* Pc */
271: { NULL, 0 }, /* Pf */
272: { NULL, 0 }, /* Po */
273: { NULL, 0 }, /* Pq */
274: { NULL, 0 }, /* Qc */
275: { NULL, 0 }, /* Ql */
276: { NULL, 0 }, /* Qo */
277: { NULL, 0 }, /* Qq */
278: { NULL, 0 }, /* Re */
279: { NULL, 0 }, /* Rs */
280: { NULL, 0 }, /* Sc */
281: { NULL, 0 }, /* So */
282: { NULL, 0 }, /* Sq */
283: { NULL, 0 }, /* Sm */
284: { NULL, 0 }, /* Sx */
285: { NULL, TYPE_Sy }, /* Sy */
286: { NULL, TYPE_Tn }, /* Tn */
287: { NULL, 0 }, /* Ux */
288: { NULL, 0 }, /* Xc */
289: { NULL, 0 }, /* Xo */
290: { parse_mdoc_head, 0 }, /* Fo */
291: { NULL, 0 }, /* Fc */
292: { NULL, 0 }, /* Oo */
293: { NULL, 0 }, /* Oc */
294: { NULL, 0 }, /* Bk */
295: { NULL, 0 }, /* Ek */
296: { NULL, 0 }, /* Bt */
297: { NULL, 0 }, /* Hf */
298: { NULL, 0 }, /* Fr */
299: { NULL, 0 }, /* Ud */
300: { NULL, TYPE_Lb }, /* Lb */
301: { NULL, 0 }, /* Lp */
302: { NULL, TYPE_Lk }, /* Lk */
303: { NULL, TYPE_Mt }, /* Mt */
304: { NULL, 0 }, /* Brq */
305: { NULL, 0 }, /* Bro */
306: { NULL, 0 }, /* Brc */
307: { NULL, 0 }, /* %C */
308: { NULL, 0 }, /* Es */
309: { NULL, 0 }, /* En */
310: { NULL, TYPE_Dx }, /* Dx */
311: { NULL, 0 }, /* %Q */
312: { NULL, 0 }, /* br */
313: { NULL, 0 }, /* sp */
314: { NULL, 0 }, /* %U */
315: { NULL, 0 }, /* Ta */
1.1 kristaps 316: };
317:
318: int
319: main(int argc, char *argv[])
320: {
1.59 schwarze 321: int ch, i;
1.50 kristaps 322: size_t j, sz;
1.59 schwarze 323: const char *path_arg;
1.50 kristaps 324: struct mchars *mc;
325: struct manpaths dirs;
326: struct mparse *mp;
1.83 schwarze 327: struct ohash_info mpages_info, mlinks_info;
1.50 kristaps 328:
329: memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
330: memset(&dirs, 0, sizeof(struct manpaths));
331:
1.83 schwarze 332: mpages_info.alloc = mlinks_info.alloc = hash_alloc;
333: mpages_info.halloc = mlinks_info.halloc = hash_halloc;
334: mpages_info.hfree = mlinks_info.hfree = hash_free;
1.50 kristaps 335:
1.79 schwarze 336: mpages_info.key_offset = offsetof(struct mpage, inodev);
1.83 schwarze 337: mlinks_info.key_offset = offsetof(struct mlink, file);
1.1 kristaps 338:
339: progname = strrchr(argv[0], '/');
340: if (progname == NULL)
341: progname = argv[0];
342: else
343: ++progname;
344:
1.50 kristaps 345: /*
346: * We accept a few different invocations.
347: * The CHECKOP macro makes sure that invocation styles don't
348: * clobber each other.
349: */
350: #define CHECKOP(_op, _ch) do \
351: if (OP_DEFAULT != (_op)) { \
352: fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \
353: goto usage; \
354: } while (/*CONSTCOND*/0)
1.10 kristaps 355:
1.59 schwarze 356: path_arg = NULL;
1.38 schwarze 357: op = OP_DEFAULT;
1.1 kristaps 358:
1.102 schwarze 359: while (-1 != (ch = getopt(argc, argv, "aC:d:nQT:tu:vW")))
1.1 kristaps 360: switch (ch) {
1.12 schwarze 361: case ('a'):
362: use_all = 1;
363: break;
1.34 schwarze 364: case ('C'):
1.50 kristaps 365: CHECKOP(op, ch);
1.59 schwarze 366: path_arg = optarg;
1.38 schwarze 367: op = OP_CONFFILE;
1.34 schwarze 368: break;
1.5 kristaps 369: case ('d'):
1.50 kristaps 370: CHECKOP(op, ch);
1.59 schwarze 371: path_arg = optarg;
1.5 kristaps 372: op = OP_UPDATE;
373: break;
1.50 kristaps 374: case ('n'):
375: nodb = 1;
376: break;
1.102 schwarze 377: case ('Q'):
1.116 schwarze 378: mparse_options |= MPARSE_QUICK;
1.102 schwarze 379: break;
1.95 schwarze 380: case ('T'):
381: if (strcmp(optarg, "utf8")) {
382: fprintf(stderr, "-T%s: Unsupported "
383: "output format\n", optarg);
384: goto usage;
385: }
386: write_utf8 = 1;
387: break;
1.38 schwarze 388: case ('t'):
1.50 kristaps 389: CHECKOP(op, ch);
1.38 schwarze 390: dup2(STDOUT_FILENO, STDERR_FILENO);
391: op = OP_TEST;
1.50 kristaps 392: nodb = warnings = 1;
1.38 schwarze 393: break;
1.5 kristaps 394: case ('u'):
1.50 kristaps 395: CHECKOP(op, ch);
1.59 schwarze 396: path_arg = optarg;
1.5 kristaps 397: op = OP_DELETE;
398: break;
399: case ('v'):
400: verb++;
401: break;
1.38 schwarze 402: case ('W'):
403: warnings = 1;
404: break;
1.1 kristaps 405: default:
1.38 schwarze 406: goto usage;
1.1 kristaps 407: }
408:
409: argc -= optind;
410: argv += optind;
411:
1.38 schwarze 412: if (OP_CONFFILE == op && argc > 0) {
1.50 kristaps 413: fprintf(stderr, "-C: Too many arguments\n");
1.38 schwarze 414: goto usage;
415: }
416:
1.59 schwarze 417: exitcode = (int)MANDOCLEVEL_OK;
1.116 schwarze 418: mp = mparse_alloc(mparse_options, MANDOCLEVEL_FATAL, NULL, NULL);
1.50 kristaps 419: mc = mchars_alloc();
420:
1.78 schwarze 421: ohash_init(&mpages, 6, &mpages_info);
1.83 schwarze 422: ohash_init(&mlinks, 6, &mlinks_info);
1.50 kristaps 423:
424: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
425: /*
426: * Force processing all files.
427: */
428: use_all = 1;
1.59 schwarze 429:
1.50 kristaps 430: /*
431: * All of these deal with a specific directory.
432: * Jump into that directory then collect files specified
433: * on the command-line.
434: */
1.59 schwarze 435: if (0 == set_basedir(path_arg))
1.50 kristaps 436: goto out;
437: for (i = 0; i < argc; i++)
1.59 schwarze 438: filescan(argv[i]);
439: if (0 == dbopen(1))
1.50 kristaps 440: goto out;
441: if (OP_TEST != op)
1.59 schwarze 442: dbprune();
1.50 kristaps 443: if (OP_DELETE != op)
1.101 schwarze 444: mpages_merge(mc, mp);
1.59 schwarze 445: dbclose(1);
1.50 kristaps 446: } else {
447: /*
448: * If we have arguments, use them as our manpaths.
449: * If we don't, grok from manpath(1) or however else
450: * manpath_parse() wants to do it.
451: */
452: if (argc > 0) {
453: dirs.paths = mandoc_calloc
454: (argc, sizeof(char *));
455: dirs.sz = (size_t)argc;
456: for (i = 0; i < argc; i++)
457: dirs.paths[i] = mandoc_strdup(argv[i]);
458: } else
1.59 schwarze 459: manpath_parse(&dirs, path_arg, NULL, NULL);
1.50 kristaps 460:
461: /*
1.69 schwarze 462: * First scan the tree rooted at a base directory, then
463: * build a new database and finally move it into place.
1.50 kristaps 464: * Ignore zero-length directories and strip trailing
465: * slashes.
466: */
467: for (j = 0; j < dirs.sz; j++) {
468: sz = strlen(dirs.paths[j]);
469: if (sz && '/' == dirs.paths[j][sz - 1])
470: dirs.paths[j][--sz] = '\0';
471: if (0 == sz)
472: continue;
1.66 schwarze 473:
474: if (j) {
1.78 schwarze 475: ohash_init(&mpages, 6, &mpages_info);
1.83 schwarze 476: ohash_init(&mlinks, 6, &mlinks_info);
1.66 schwarze 477: }
478:
1.59 schwarze 479: if (0 == set_basedir(dirs.paths[j]))
1.50 kristaps 480: goto out;
1.59 schwarze 481: if (0 == treescan())
1.50 kristaps 482: goto out;
1.59 schwarze 483: if (0 == set_basedir(dirs.paths[j]))
1.50 kristaps 484: goto out;
1.59 schwarze 485: if (0 == dbopen(0))
1.50 kristaps 486: goto out;
1.52 kristaps 487:
1.101 schwarze 488: mpages_merge(mc, mp);
1.59 schwarze 489: dbclose(0);
1.66 schwarze 490:
491: if (j + 1 < dirs.sz) {
1.80 schwarze 492: mpages_free();
1.78 schwarze 493: ohash_delete(&mpages);
1.83 schwarze 494: ohash_delete(&mlinks);
1.66 schwarze 495: }
1.50 kristaps 496: }
497: }
498: out:
1.59 schwarze 499: set_basedir(NULL);
1.50 kristaps 500: manpath_free(&dirs);
501: mchars_free(mc);
502: mparse_free(mp);
1.80 schwarze 503: mpages_free();
1.78 schwarze 504: ohash_delete(&mpages);
1.83 schwarze 505: ohash_delete(&mlinks);
1.59 schwarze 506: return(exitcode);
1.50 kristaps 507: usage:
1.102 schwarze 508: fprintf(stderr, "usage: %s [-anQvW] [-C file] [-Tutf8]\n"
509: " %s [-anQvW] [-Tutf8] dir ...\n"
510: " %s [-nQvW] [-Tutf8] -d dir [file ...]\n"
1.50 kristaps 511: " %s [-nvW] -u dir [file ...]\n"
1.102 schwarze 512: " %s [-Q] -t file ...\n",
1.50 kristaps 513: progname, progname, progname,
514: progname, progname);
515:
1.59 schwarze 516: return((int)MANDOCLEVEL_BADARG);
1.50 kristaps 517: }
518:
519: /*
1.59 schwarze 520: * Scan a directory tree rooted at "basedir" for manpages.
1.50 kristaps 521: * We use fts(), scanning directory parts along the way for clues to our
522: * section and architecture.
523: *
524: * If use_all has been specified, grok all files.
525: * If not, sanitise paths to the following:
526: *
527: * [./]man*[/<arch>]/<name>.<section>
528: * or
529: * [./]cat<section>[/<arch>]/<name>.0
530: *
531: * TODO: accomodate for multi-language directories.
532: */
533: static int
1.59 schwarze 534: treescan(void)
1.50 kristaps 535: {
536: FTS *f;
537: FTSENT *ff;
1.84 schwarze 538: struct mlink *mlink;
1.50 kristaps 539: int dform;
1.94 schwarze 540: char *dsec, *arch, *fsec, *cp;
541: const char *path;
1.50 kristaps 542: const char *argv[2];
543:
544: argv[0] = ".";
545: argv[1] = (char *)NULL;
546:
547: /*
548: * Walk through all components under the directory, using the
549: * logical descent of files.
550: */
551: f = fts_open((char * const *)argv, FTS_LOGICAL, NULL);
552: if (NULL == f) {
1.59 schwarze 553: exitcode = (int)MANDOCLEVEL_SYSERR;
554: say("", NULL);
1.50 kristaps 555: return(0);
556: }
557:
558: dsec = arch = NULL;
559: dform = FORM_NONE;
560:
561: while (NULL != (ff = fts_read(f))) {
562: path = ff->fts_path + 2;
563: /*
1.83 schwarze 564: * If we're a regular file, add an mlink by using the
1.50 kristaps 565: * stored directory data and handling the filename.
566: */
567: if (FTS_F == ff->fts_info) {
1.60 schwarze 568: if (0 == strcmp(path, MANDOC_DB))
569: continue;
1.50 kristaps 570: if ( ! use_all && ff->fts_level < 2) {
1.56 schwarze 571: if (warnings)
1.59 schwarze 572: say(path, "Extraneous file");
1.50 kristaps 573: continue;
1.84 schwarze 574: } else if (NULL == (fsec =
1.60 schwarze 575: strrchr(ff->fts_name, '.'))) {
576: if ( ! use_all) {
1.56 schwarze 577: if (warnings)
1.60 schwarze 578: say(path,
579: "No filename suffix");
1.50 kristaps 580: continue;
581: }
1.84 schwarze 582: } else if (0 == strcmp(++fsec, "html")) {
1.60 schwarze 583: if (warnings)
584: say(path, "Skip html");
585: continue;
1.84 schwarze 586: } else if (0 == strcmp(fsec, "gz")) {
1.60 schwarze 587: if (warnings)
588: say(path, "Skip gz");
589: continue;
1.84 schwarze 590: } else if (0 == strcmp(fsec, "ps")) {
1.60 schwarze 591: if (warnings)
592: say(path, "Skip ps");
593: continue;
1.84 schwarze 594: } else if (0 == strcmp(fsec, "pdf")) {
1.60 schwarze 595: if (warnings)
596: say(path, "Skip pdf");
597: continue;
598: } else if ( ! use_all &&
1.84 schwarze 599: ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
600: (FORM_CAT == dform && strcmp(fsec, "0")))) {
1.60 schwarze 601: if (warnings)
602: say(path, "Wrong filename suffix");
603: continue;
1.66 schwarze 604: } else
1.84 schwarze 605: fsec[-1] = '\0';
1.94 schwarze 606:
1.84 schwarze 607: mlink = mandoc_calloc(1, sizeof(struct mlink));
608: strlcpy(mlink->file, path, sizeof(mlink->file));
609: mlink->dform = dform;
1.94 schwarze 610: mlink->dsec = dsec;
611: mlink->arch = arch;
612: mlink->name = ff->fts_name;
613: mlink->fsec = fsec;
1.84 schwarze 614: mlink_add(mlink, ff->fts_statp);
1.50 kristaps 615: continue;
1.84 schwarze 616: } else if (FTS_D != ff->fts_info &&
1.60 schwarze 617: FTS_DP != ff->fts_info) {
618: if (warnings)
619: say(path, "Not a regular file");
1.50 kristaps 620: continue;
1.60 schwarze 621: }
1.1 kristaps 622:
1.50 kristaps 623: switch (ff->fts_level) {
624: case (0):
625: /* Ignore the root directory. */
626: break;
627: case (1):
628: /*
629: * This might contain manX/ or catX/.
630: * Try to infer this from the name.
631: * If we're not in use_all, enforce it.
632: */
633: cp = ff->fts_name;
634: if (FTS_DP == ff->fts_info)
635: break;
1.1 kristaps 636:
1.50 kristaps 637: if (0 == strncmp(cp, "man", 3)) {
638: dform = FORM_SRC;
1.66 schwarze 639: dsec = cp + 3;
1.50 kristaps 640: } else if (0 == strncmp(cp, "cat", 3)) {
641: dform = FORM_CAT;
1.66 schwarze 642: dsec = cp + 3;
1.94 schwarze 643: } else {
644: dform = FORM_NONE;
645: dsec = NULL;
1.50 kristaps 646: }
1.1 kristaps 647:
1.50 kristaps 648: if (NULL != dsec || use_all)
649: break;
1.1 kristaps 650:
1.56 schwarze 651: if (warnings)
1.59 schwarze 652: say(path, "Unknown directory part");
1.50 kristaps 653: fts_set(f, ff, FTS_SKIP);
654: break;
655: case (2):
656: /*
657: * Possibly our architecture.
658: * If we're descending, keep tabs on it.
659: */
660: if (FTS_DP != ff->fts_info && NULL != dsec)
1.66 schwarze 661: arch = ff->fts_name;
1.94 schwarze 662: else
663: arch = NULL;
1.50 kristaps 664: break;
665: default:
666: if (FTS_DP == ff->fts_info || use_all)
667: break;
1.56 schwarze 668: if (warnings)
1.59 schwarze 669: say(path, "Extraneous directory part");
1.50 kristaps 670: fts_set(f, ff, FTS_SKIP);
671: break;
1.5 kristaps 672: }
1.50 kristaps 673: }
674:
675: fts_close(f);
676: return(1);
677: }
1.5 kristaps 678:
1.50 kristaps 679: /*
1.89 schwarze 680: * Add a file to the mlinks table.
1.50 kristaps 681: * Do not verify that it's a "valid" looking manpage (we'll do that
682: * later).
683: *
684: * Try to infer the manual section, architecture, and page name from the
685: * path, assuming it looks like
686: *
687: * [./]man*[/<arch>]/<name>.<section>
688: * or
689: * [./]cat<section>[/<arch>]/<name>.0
690: *
691: * See treescan() for the fts(3) version of this.
692: */
693: static void
1.59 schwarze 694: filescan(const char *file)
1.50 kristaps 695: {
1.59 schwarze 696: char buf[PATH_MAX];
1.84 schwarze 697: struct stat st;
698: struct mlink *mlink;
1.59 schwarze 699: char *p, *start;
1.5 kristaps 700:
1.50 kristaps 701: assert(use_all);
1.5 kristaps 702:
1.50 kristaps 703: if (0 == strncmp(file, "./", 2))
704: file += 2;
1.5 kristaps 705:
1.59 schwarze 706: if (NULL == realpath(file, buf)) {
707: exitcode = (int)MANDOCLEVEL_BADARG;
708: say(file, NULL);
709: return;
1.106 schwarze 710: }
711:
712: if (strstr(buf, basedir) == buf)
713: start = buf + strlen(basedir) + 1;
714: else if (OP_TEST == op)
715: start = buf;
716: else {
1.59 schwarze 717: exitcode = (int)MANDOCLEVEL_BADARG;
718: say("", "%s: outside base directory", buf);
719: return;
1.106 schwarze 720: }
721:
722: if (-1 == stat(buf, &st)) {
1.59 schwarze 723: exitcode = (int)MANDOCLEVEL_BADARG;
724: say(file, NULL);
1.50 kristaps 725: return;
726: } else if ( ! (S_IFREG & st.st_mode)) {
1.59 schwarze 727: exitcode = (int)MANDOCLEVEL_BADARG;
728: say(file, "Not a regular file");
1.50 kristaps 729: return;
730: }
1.106 schwarze 731:
1.84 schwarze 732: mlink = mandoc_calloc(1, sizeof(struct mlink));
733: strlcpy(mlink->file, start, sizeof(mlink->file));
1.17 schwarze 734:
1.50 kristaps 735: /*
736: * First try to guess our directory structure.
737: * If we find a separator, try to look for man* or cat*.
738: * If we find one of these and what's underneath is a directory,
739: * assume it's an architecture.
740: */
741: if (NULL != (p = strchr(start, '/'))) {
742: *p++ = '\0';
743: if (0 == strncmp(start, "man", 3)) {
1.84 schwarze 744: mlink->dform = FORM_SRC;
1.94 schwarze 745: mlink->dsec = start + 3;
1.50 kristaps 746: } else if (0 == strncmp(start, "cat", 3)) {
1.84 schwarze 747: mlink->dform = FORM_CAT;
1.94 schwarze 748: mlink->dsec = start + 3;
1.17 schwarze 749: }
1.5 kristaps 750:
1.50 kristaps 751: start = p;
1.84 schwarze 752: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
1.50 kristaps 753: *p++ = '\0';
1.94 schwarze 754: mlink->arch = start;
1.50 kristaps 755: start = p;
1.84 schwarze 756: }
1.50 kristaps 757: }
758:
759: /*
760: * Now check the file suffix.
761: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
762: */
763: p = strrchr(start, '\0');
764: while (p-- > start && '/' != *p && '.' != *p)
765: /* Loop. */ ;
766:
767: if ('.' == *p) {
768: *p++ = '\0';
1.94 schwarze 769: mlink->fsec = p;
1.5 kristaps 770: }
771:
1.10 kristaps 772: /*
1.50 kristaps 773: * Now try to parse the name.
774: * Use the filename portion of the path.
1.10 kristaps 775: */
1.84 schwarze 776: mlink->name = start;
1.50 kristaps 777: if (NULL != (p = strrchr(start, '/'))) {
1.84 schwarze 778: mlink->name = p + 1;
1.50 kristaps 779: *p = '\0';
1.84 schwarze 780: }
781: mlink_add(mlink, &st);
1.50 kristaps 782: }
1.5 kristaps 783:
1.50 kristaps 784: static void
1.84 schwarze 785: mlink_add(struct mlink *mlink, const struct stat *st)
1.50 kristaps 786: {
1.83 schwarze 787: struct inodev inodev;
1.78 schwarze 788: struct mpage *mpage;
1.83 schwarze 789: unsigned int slot;
1.50 kristaps 790:
1.84 schwarze 791: assert(NULL != mlink->file);
1.50 kristaps 792:
1.94 schwarze 793: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
794: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
795: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
796: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.84 schwarze 797:
798: if ('0' == *mlink->fsec) {
799: free(mlink->fsec);
800: mlink->fsec = mandoc_strdup(mlink->dsec);
801: mlink->fform = FORM_CAT;
802: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
803: mlink->fform = FORM_SRC;
1.74 schwarze 804: else
1.84 schwarze 805: mlink->fform = FORM_NONE;
1.82 schwarze 806:
1.83 schwarze 807: slot = ohash_qlookup(&mlinks, mlink->file);
808: assert(NULL == ohash_find(&mlinks, slot));
809: ohash_insert(&mlinks, slot, mlink);
810:
811: inodev.st_ino = st->st_ino;
812: inodev.st_dev = st->st_dev;
813: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
814: sizeof(struct inodev), inodev.st_ino);
815: mpage = ohash_find(&mpages, slot);
816: if (NULL == mpage) {
817: mpage = mandoc_calloc(1, sizeof(struct mpage));
818: mpage->inodev.st_ino = inodev.st_ino;
819: mpage->inodev.st_dev = inodev.st_dev;
820: ohash_insert(&mpages, slot, mpage);
821: } else
1.85 schwarze 822: mlink->next = mpage->mlinks;
1.82 schwarze 823: mpage->mlinks = mlink;
1.118 schwarze 824: mlink->mpage = mpage;
1.82 schwarze 825: }
826:
827: static void
828: mlink_free(struct mlink *mlink)
829: {
830:
831: free(mlink->dsec);
832: free(mlink->arch);
833: free(mlink->name);
834: free(mlink->fsec);
835: free(mlink);
1.50 kristaps 836: }
1.3 kristaps 837:
1.50 kristaps 838: static void
1.78 schwarze 839: mpages_free(void)
1.50 kristaps 840: {
1.78 schwarze 841: struct mpage *mpage;
1.82 schwarze 842: struct mlink *mlink;
1.80 schwarze 843: unsigned int slot;
1.3 kristaps 844:
1.80 schwarze 845: mpage = ohash_first(&mpages, &slot);
846: while (NULL != mpage) {
1.82 schwarze 847: while (NULL != (mlink = mpage->mlinks)) {
1.85 schwarze 848: mpage->mlinks = mlink->next;
1.82 schwarze 849: mlink_free(mlink);
850: }
1.78 schwarze 851: free(mpage->sec);
852: free(mpage->arch);
1.82 schwarze 853: free(mpage->title);
854: free(mpage->desc);
1.78 schwarze 855: free(mpage);
1.80 schwarze 856: mpage = ohash_next(&mpages, &slot);
1.50 kristaps 857: }
858: }
1.38 schwarze 859:
1.50 kristaps 860: /*
1.89 schwarze 861: * For each mlink to the mpage, check whether the path looks like
862: * it is formatted, and if it does, check whether a source manual
863: * exists by the same name, ignoring the suffix.
864: * If both conditions hold, drop the mlink.
865: */
866: static void
867: mlinks_undupe(struct mpage *mpage)
868: {
869: char buf[PATH_MAX];
870: struct mlink **prev;
871: struct mlink *mlink;
872: char *bufp;
873:
874: mpage->form = FORM_CAT;
1.90 schwarze 875: prev = &mpage->mlinks;
876: while (NULL != (mlink = *prev)) {
1.89 schwarze 877: if (FORM_CAT != mlink->dform) {
878: mpage->form = FORM_NONE;
1.90 schwarze 879: goto nextlink;
1.89 schwarze 880: }
881: if (strlcpy(buf, mlink->file, PATH_MAX) >= PATH_MAX) {
882: if (warnings)
883: say(mlink->file, "Filename too long");
1.90 schwarze 884: goto nextlink;
1.89 schwarze 885: }
886: bufp = strstr(buf, "cat");
887: assert(NULL != bufp);
888: memcpy(bufp, "man", 3);
889: if (NULL != (bufp = strrchr(buf, '.')))
890: *++bufp = '\0';
891: strlcat(buf, mlink->dsec, PATH_MAX);
892: if (NULL == ohash_find(&mlinks,
893: ohash_qlookup(&mlinks, buf)))
1.90 schwarze 894: goto nextlink;
1.89 schwarze 895: if (warnings)
896: say(mlink->file, "Man source exists: %s", buf);
897: if (use_all)
1.90 schwarze 898: goto nextlink;
1.89 schwarze 899: *prev = mlink->next;
900: mlink_free(mlink);
1.90 schwarze 901: continue;
902: nextlink:
903: prev = &(*prev)->next;
1.89 schwarze 904: }
905: }
906:
1.93 schwarze 907: static int
908: mlink_check(struct mpage *mpage, struct mlink *mlink)
909: {
910: int match;
911:
912: match = 1;
913:
914: /*
915: * Check whether the manual section given in a file
916: * agrees with the directory where the file is located.
917: * Some manuals have suffixes like (3p) on their
918: * section number either inside the file or in the
919: * directory name, some are linked into more than one
920: * section, like encrypt(1) = makekey(8).
921: */
922:
923: if (FORM_SRC == mpage->form &&
924: strcasecmp(mpage->sec, mlink->dsec)) {
925: match = 0;
926: say(mlink->file, "Section \"%s\" manual in %s directory",
927: mpage->sec, mlink->dsec);
928: }
929:
930: /*
931: * Manual page directories exist for each kernel
932: * architecture as returned by machine(1).
933: * However, many manuals only depend on the
934: * application architecture as returned by arch(1).
935: * For example, some (2/ARM) manuals are shared
936: * across the "armish" and "zaurus" kernel
937: * architectures.
938: * A few manuals are even shared across completely
939: * different architectures, for example fdformat(1)
940: * on amd64, i386, sparc, and sparc64.
941: */
942:
943: if (strcasecmp(mpage->arch, mlink->arch)) {
944: match = 0;
945: say(mlink->file, "Architecture \"%s\" manual in "
946: "\"%s\" directory", mpage->arch, mlink->arch);
947: }
948:
949: if (strcasecmp(mpage->title, mlink->name))
950: match = 0;
951:
952: return(match);
953: }
954:
1.89 schwarze 955: /*
1.80 schwarze 956: * Run through the files in the global vector "mpages"
957: * and add them to the database specified in "basedir".
1.50 kristaps 958: *
959: * This handles the parsing scheme itself, using the cues of directory
960: * and filename to determine whether the file is parsable or not.
961: */
1.59 schwarze 962: static void
1.101 schwarze 963: mpages_merge(struct mchars *mc, struct mparse *mp)
1.50 kristaps 964: {
1.113 schwarze 965: char any[] = "any";
1.101 schwarze 966: struct ohash_info str_info;
1.118 schwarze 967: struct mpage *mpage, *mpage_dest;
968: struct mlink *mlink, *mlink_dest;
1.69 schwarze 969: struct mdoc *mdoc;
970: struct man *man;
1.118 schwarze 971: char *sodest;
1.112 schwarze 972: char *cp;
1.69 schwarze 973: int match;
1.101 schwarze 974: unsigned int pslot;
1.69 schwarze 975: enum mandoclevel lvl;
976:
1.71 schwarze 977: str_info.alloc = hash_alloc;
978: str_info.halloc = hash_halloc;
979: str_info.hfree = hash_free;
980: str_info.key_offset = offsetof(struct str, key);
981:
1.107 schwarze 982: if (0 == nodb)
983: SQL_EXEC("BEGIN TRANSACTION");
984:
1.80 schwarze 985: mpage = ohash_first(&mpages, &pslot);
986: while (NULL != mpage) {
1.89 schwarze 987: mlinks_undupe(mpage);
988: if (NULL == mpage->mlinks) {
989: mpage = ohash_next(&mpages, &pslot);
990: continue;
1.50 kristaps 991: }
992:
1.71 schwarze 993: ohash_init(&strings, 6, &str_info);
1.50 kristaps 994: mparse_reset(mp);
995: mdoc = NULL;
996: man = NULL;
1.14 schwarze 997:
998: /*
1.33 schwarze 999: * Try interpreting the file as mdoc(7) or man(7)
1000: * source code, unless it is already known to be
1001: * formatted. Fall back to formatted mode.
1.14 schwarze 1002: */
1.82 schwarze 1003: if (FORM_CAT != mpage->mlinks->dform ||
1.84 schwarze 1004: FORM_CAT != mpage->mlinks->fform) {
1.82 schwarze 1005: lvl = mparse_readfd(mp, -1, mpage->mlinks->file);
1.50 kristaps 1006: if (lvl < MANDOCLEVEL_FATAL)
1.118 schwarze 1007: mparse_result(mp, &mdoc, &man, &sodest);
1.76 schwarze 1008: }
1.14 schwarze 1009:
1.118 schwarze 1010: if (NULL != sodest) {
1011: mlink_dest = ohash_find(&mlinks,
1012: ohash_qlookup(&mlinks, sodest));
1013: if (NULL != mlink_dest) {
1014:
1015: /* The .so target exists. */
1016:
1017: mpage_dest = mlink_dest->mpage;
1018: mlink = mpage->mlinks;
1019: while (1) {
1020: mlink->mpage = mpage_dest;
1021:
1022: /*
1023: * If the target was already
1024: * processed, add the links
1025: * to the database now.
1026: * Otherwise, this will
1027: * happen when we come
1028: * to the target.
1029: */
1030:
1031: if (mpage_dest->recno)
1032: dbadd_mlink(mlink);
1033:
1034: if (NULL == mlink->next)
1035: break;
1036: mlink = mlink->next;
1037: }
1038:
1039: /* Move all links to the target. */
1040:
1041: mlink->next = mlink_dest->next;
1042: mlink_dest->next = mpage->mlinks;
1043: mpage->mlinks = NULL;
1044: }
1045: ohash_delete(&strings);
1046: mpage = ohash_next(&mpages, &pslot);
1047: continue;
1048: } else if (NULL != mdoc) {
1.82 schwarze 1049: mpage->form = FORM_SRC;
1050: mpage->sec =
1051: mandoc_strdup(mdoc_meta(mdoc)->msec);
1052: mpage->arch = mdoc_meta(mdoc)->arch;
1.84 schwarze 1053: mpage->arch = mandoc_strdup(
1054: NULL == mpage->arch ? "" : mpage->arch);
1.82 schwarze 1055: mpage->title =
1056: mandoc_strdup(mdoc_meta(mdoc)->title);
1.14 schwarze 1057: } else if (NULL != man) {
1.82 schwarze 1058: mpage->form = FORM_SRC;
1059: mpage->sec =
1060: mandoc_strdup(man_meta(man)->msec);
1061: mpage->arch =
1062: mandoc_strdup(mpage->mlinks->arch);
1063: mpage->title =
1064: mandoc_strdup(man_meta(man)->title);
1.76 schwarze 1065: } else {
1.82 schwarze 1066: mpage->form = FORM_CAT;
1067: mpage->sec =
1068: mandoc_strdup(mpage->mlinks->dsec);
1069: mpage->arch =
1070: mandoc_strdup(mpage->mlinks->arch);
1071: mpage->title =
1072: mandoc_strdup(mpage->mlinks->name);
1.76 schwarze 1073: }
1.97 schwarze 1074: putkey(mpage, mpage->sec, TYPE_sec);
1.98 schwarze 1075: putkey(mpage, '\0' == *mpage->arch ?
1.113 schwarze 1076: any : mpage->arch, TYPE_arch);
1.50 kristaps 1077:
1.97 schwarze 1078: for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1079: if ('\0' != *mlink->dsec)
1080: putkey(mpage, mlink->dsec, TYPE_sec);
1081: if ('\0' != *mlink->fsec)
1082: putkey(mpage, mlink->fsec, TYPE_sec);
1.98 schwarze 1083: putkey(mpage, '\0' == *mlink->arch ?
1.113 schwarze 1084: any : mlink->arch, TYPE_arch);
1.93 schwarze 1085: putkey(mpage, mlink->name, TYPE_Nm);
1.97 schwarze 1086: }
1.12 schwarze 1087:
1.93 schwarze 1088: if (warnings && !use_all) {
1.69 schwarze 1089: match = 0;
1.93 schwarze 1090: for (mlink = mpage->mlinks; mlink;
1091: mlink = mlink->next)
1092: if (mlink_check(mpage, mlink))
1093: match = 1;
1094: } else
1095: match = 1;
1.12 schwarze 1096:
1.50 kristaps 1097: if (NULL != mdoc) {
1098: if (NULL != (cp = mdoc_meta(mdoc)->name))
1.78 schwarze 1099: putkey(mpage, cp, TYPE_Nm);
1100: assert(NULL == mpage->desc);
1101: parse_mdoc(mpage, mdoc_node(mdoc));
1102: putkey(mpage, NULL != mpage->desc ?
1.82 schwarze 1103: mpage->desc : mpage->mlinks->name, TYPE_Nd);
1.50 kristaps 1104: } else if (NULL != man)
1.78 schwarze 1105: parse_man(mpage, man_node(man));
1.50 kristaps 1106: else
1.78 schwarze 1107: parse_cat(mpage);
1.44 kristaps 1108:
1.105 schwarze 1109: dbadd(mpage, mc);
1.66 schwarze 1110: ohash_delete(&strings);
1.80 schwarze 1111: mpage = ohash_next(&mpages, &pslot);
1.50 kristaps 1112: }
1.107 schwarze 1113:
1114: if (0 == nodb)
1115: SQL_EXEC("END TRANSACTION");
1.50 kristaps 1116: }
1.12 schwarze 1117:
1.50 kristaps 1118: static void
1.78 schwarze 1119: parse_cat(struct mpage *mpage)
1.50 kristaps 1120: {
1121: FILE *stream;
1122: char *line, *p, *title;
1123: size_t len, plen, titlesz;
1.12 schwarze 1124:
1.82 schwarze 1125: if (NULL == (stream = fopen(mpage->mlinks->file, "r"))) {
1.56 schwarze 1126: if (warnings)
1.82 schwarze 1127: say(mpage->mlinks->file, NULL);
1.50 kristaps 1128: return;
1129: }
1.1 kristaps 1130:
1.50 kristaps 1131: /* Skip to first blank line. */
1.1 kristaps 1132:
1.50 kristaps 1133: while (NULL != (line = fgetln(stream, &len)))
1134: if ('\n' == *line)
1135: break;
1.1 kristaps 1136:
1.50 kristaps 1137: /*
1138: * Assume the first line that is not indented
1139: * is the first section header. Skip to it.
1140: */
1.1 kristaps 1141:
1.50 kristaps 1142: while (NULL != (line = fgetln(stream, &len)))
1143: if ('\n' != *line && ' ' != *line)
1144: break;
1145:
1146: /*
1147: * Read up until the next section into a buffer.
1148: * Strip the leading and trailing newline from each read line,
1149: * appending a trailing space.
1150: * Ignore empty (whitespace-only) lines.
1151: */
1.1 kristaps 1152:
1.50 kristaps 1153: titlesz = 0;
1154: title = NULL;
1.38 schwarze 1155:
1.50 kristaps 1156: while (NULL != (line = fgetln(stream, &len))) {
1157: if (' ' != *line || '\n' != line[len - 1])
1158: break;
1159: while (len > 0 && isspace((unsigned char)*line)) {
1160: line++;
1161: len--;
1162: }
1163: if (1 == len)
1.38 schwarze 1164: continue;
1.50 kristaps 1165: title = mandoc_realloc(title, titlesz + len);
1166: memcpy(title + titlesz, line, len);
1167: titlesz += len;
1168: title[titlesz - 1] = ' ';
1169: }
1.38 schwarze 1170:
1.50 kristaps 1171: /*
1172: * If no page content can be found, or the input line
1173: * is already the next section header, or there is no
1174: * trailing newline, reuse the page title as the page
1175: * description.
1176: */
1.44 kristaps 1177:
1.50 kristaps 1178: if (NULL == title || '\0' == *title) {
1.56 schwarze 1179: if (warnings)
1.82 schwarze 1180: say(mpage->mlinks->file,
1181: "Cannot find NAME section");
1.78 schwarze 1182: assert(NULL == mpage->desc);
1.82 schwarze 1183: mpage->desc = mandoc_strdup(mpage->mlinks->name);
1184: putkey(mpage, mpage->mlinks->name, TYPE_Nd);
1.50 kristaps 1185: fclose(stream);
1186: free(title);
1187: return;
1188: }
1.1 kristaps 1189:
1.50 kristaps 1190: title = mandoc_realloc(title, titlesz + 1);
1191: title[titlesz] = '\0';
1.33 schwarze 1192:
1.50 kristaps 1193: /*
1194: * Skip to the first dash.
1195: * Use the remaining line as the description (no more than 70
1196: * bytes).
1197: */
1.33 schwarze 1198:
1.50 kristaps 1199: if (NULL != (p = strstr(title, "- "))) {
1200: for (p += 2; ' ' == *p || '\b' == *p; p++)
1201: /* Skip to next word. */ ;
1202: } else {
1.56 schwarze 1203: if (warnings)
1.82 schwarze 1204: say(mpage->mlinks->file,
1205: "No dash in title line");
1.50 kristaps 1206: p = title;
1207: }
1.38 schwarze 1208:
1.50 kristaps 1209: plen = strlen(p);
1.1 kristaps 1210:
1.50 kristaps 1211: /* Strip backspace-encoding from line. */
1.1 kristaps 1212:
1.50 kristaps 1213: while (NULL != (line = memchr(p, '\b', plen))) {
1214: len = line - p;
1215: if (0 == len) {
1216: memmove(line, line + 1, plen--);
1217: continue;
1218: }
1219: memmove(line - 1, line + 1, plen - len);
1220: plen -= 2;
1221: }
1.1 kristaps 1222:
1.78 schwarze 1223: assert(NULL == mpage->desc);
1224: mpage->desc = mandoc_strdup(p);
1225: putkey(mpage, mpage->desc, TYPE_Nd);
1.50 kristaps 1226: fclose(stream);
1227: free(title);
1228: }
1.1 kristaps 1229:
1.50 kristaps 1230: /*
1231: * Put a type/word pair into the word database for this particular file.
1232: */
1233: static void
1.112 schwarze 1234: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.50 kristaps 1235: {
1.112 schwarze 1236: char *cp;
1.18 kristaps 1237:
1.50 kristaps 1238: assert(NULL != value);
1.112 schwarze 1239: if (TYPE_arch == type)
1240: for (cp = value; *cp; cp++)
1241: if (isupper((unsigned char)*cp))
1242: *cp = _tolower((unsigned char)*cp);
1.78 schwarze 1243: putkeys(mpage, value, strlen(value), type);
1.3 kristaps 1244: }
1245:
1246: /*
1.50 kristaps 1247: * Grok all nodes at or below a certain mdoc node into putkey().
1.3 kristaps 1248: */
1249: static void
1.78 schwarze 1250: putmdockey(const struct mpage *mpage,
1251: const struct mdoc_node *n, uint64_t m)
1.3 kristaps 1252: {
1.18 kristaps 1253:
1.50 kristaps 1254: for ( ; NULL != n; n = n->next) {
1255: if (NULL != n->child)
1.78 schwarze 1256: putmdockey(mpage, n->child, m);
1.50 kristaps 1257: if (MDOC_TEXT == n->type)
1.78 schwarze 1258: putkey(mpage, n->string, m);
1.50 kristaps 1259: }
1260: }
1.18 kristaps 1261:
1.61 schwarze 1262: static void
1.78 schwarze 1263: parse_man(struct mpage *mpage, const struct man_node *n)
1.50 kristaps 1264: {
1265: const struct man_node *head, *body;
1.121 ! schwarze 1266: char *start, *title;
1.50 kristaps 1267: char byte;
1.121 ! schwarze 1268: size_t sz;
1.18 kristaps 1269:
1.50 kristaps 1270: if (NULL == n)
1.61 schwarze 1271: return;
1.18 kristaps 1272:
1.50 kristaps 1273: /*
1274: * We're only searching for one thing: the first text child in
1275: * the BODY of a NAME section. Since we don't keep track of
1276: * sections in -man, run some hoops to find out whether we're in
1277: * the correct section or not.
1278: */
1.18 kristaps 1279:
1.50 kristaps 1280: if (MAN_BODY == n->type && MAN_SH == n->tok) {
1281: body = n;
1282: assert(body->parent);
1283: if (NULL != (head = body->parent->head) &&
1284: 1 == head->nchild &&
1285: NULL != (head = (head->child)) &&
1286: MAN_TEXT == head->type &&
1287: 0 == strcmp(head->string, "NAME") &&
1.121 ! schwarze 1288: NULL != body->child) {
1.3 kristaps 1289:
1.50 kristaps 1290: /*
1291: * Suck the entire NAME section into memory.
1292: * Yes, we might run away.
1293: * But too many manuals have big, spread-out
1294: * NAME sections over many lines.
1295: */
1.3 kristaps 1296:
1.121 ! schwarze 1297: title = NULL;
! 1298: man_deroff(&title, body);
1.50 kristaps 1299: if (NULL == title)
1.61 schwarze 1300: return;
1.18 kristaps 1301:
1.50 kristaps 1302: /*
1303: * Go through a special heuristic dance here.
1304: * Conventionally, one or more manual names are
1305: * comma-specified prior to a whitespace, then a
1306: * dash, then a description. Try to puzzle out
1307: * the name parts here.
1308: */
1.18 kristaps 1309:
1.121 ! schwarze 1310: start = title;
1.50 kristaps 1311: for ( ;; ) {
1312: sz = strcspn(start, " ,");
1313: if ('\0' == start[sz])
1314: break;
1.1 kristaps 1315:
1.50 kristaps 1316: byte = start[sz];
1317: start[sz] = '\0';
1.110 schwarze 1318:
1319: /*
1320: * Assume a stray trailing comma in the
1321: * name list if a name begins with a dash.
1322: */
1323:
1324: if ('-' == start[0] ||
1325: ('\\' == start[0] && '-' == start[1]))
1326: break;
1.1 kristaps 1327:
1.78 schwarze 1328: putkey(mpage, start, TYPE_Nm);
1.1 kristaps 1329:
1.50 kristaps 1330: if (' ' == byte) {
1331: start += sz + 1;
1332: break;
1333: }
1.1 kristaps 1334:
1.50 kristaps 1335: assert(',' == byte);
1336: start += sz + 1;
1337: while (' ' == *start)
1338: start++;
1339: }
1.1 kristaps 1340:
1.121 ! schwarze 1341: if (start == title) {
1.78 schwarze 1342: putkey(mpage, start, TYPE_Nm);
1.50 kristaps 1343: free(title);
1.61 schwarze 1344: return;
1.50 kristaps 1345: }
1.1 kristaps 1346:
1.50 kristaps 1347: while (isspace((unsigned char)*start))
1348: start++;
1.1 kristaps 1349:
1.50 kristaps 1350: if (0 == strncmp(start, "-", 1))
1351: start += 1;
1352: else if (0 == strncmp(start, "\\-\\-", 4))
1353: start += 4;
1354: else if (0 == strncmp(start, "\\-", 2))
1355: start += 2;
1356: else if (0 == strncmp(start, "\\(en", 4))
1357: start += 4;
1358: else if (0 == strncmp(start, "\\(em", 4))
1359: start += 4;
1.1 kristaps 1360:
1.50 kristaps 1361: while (' ' == *start)
1362: start++;
1.1 kristaps 1363:
1.78 schwarze 1364: assert(NULL == mpage->desc);
1365: mpage->desc = mandoc_strdup(start);
1366: putkey(mpage, mpage->desc, TYPE_Nd);
1.50 kristaps 1367: free(title);
1.61 schwarze 1368: return;
1.50 kristaps 1369: }
1370: }
1.1 kristaps 1371:
1.77 schwarze 1372: for (n = n->child; n; n = n->next) {
1.78 schwarze 1373: if (NULL != mpage->desc)
1.77 schwarze 1374: break;
1.78 schwarze 1375: parse_man(mpage, n);
1.77 schwarze 1376: }
1.1 kristaps 1377: }
1378:
1379: static void
1.78 schwarze 1380: parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1381: {
1382:
1.50 kristaps 1383: assert(NULL != n);
1384: for (n = n->child; NULL != n; n = n->next) {
1385: switch (n->type) {
1386: case (MDOC_ELEM):
1387: /* FALLTHROUGH */
1388: case (MDOC_BLOCK):
1389: /* FALLTHROUGH */
1390: case (MDOC_HEAD):
1391: /* FALLTHROUGH */
1392: case (MDOC_BODY):
1393: /* FALLTHROUGH */
1394: case (MDOC_TAIL):
1395: if (NULL != mdocs[n->tok].fp)
1.78 schwarze 1396: if (0 == (*mdocs[n->tok].fp)(mpage, n))
1.50 kristaps 1397: break;
1.68 schwarze 1398: if (mdocs[n->tok].mask)
1.78 schwarze 1399: putmdockey(mpage, n->child,
1400: mdocs[n->tok].mask);
1.50 kristaps 1401: break;
1402: default:
1403: assert(MDOC_ROOT != n->type);
1404: continue;
1405: }
1406: if (NULL != n->child)
1.78 schwarze 1407: parse_mdoc(mpage, n);
1.1 kristaps 1408: }
1409: }
1410:
1.25 schwarze 1411: static int
1.78 schwarze 1412: parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1413: {
1414: const char *start, *end;
1415: size_t sz;
1.25 schwarze 1416:
1.50 kristaps 1417: if (SEC_SYNOPSIS != n->sec ||
1418: NULL == (n = n->child) ||
1419: MDOC_TEXT != n->type)
1.25 schwarze 1420: return(0);
1.1 kristaps 1421:
1422: /*
1423: * Only consider those `Fd' macro fields that begin with an
1424: * "inclusion" token (versus, e.g., #define).
1425: */
1.50 kristaps 1426:
1.1 kristaps 1427: if (strcmp("#include", n->string))
1.25 schwarze 1428: return(0);
1.1 kristaps 1429:
1430: if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1.25 schwarze 1431: return(0);
1.1 kristaps 1432:
1433: /*
1434: * Strip away the enclosing angle brackets and make sure we're
1435: * not zero-length.
1436: */
1437:
1438: start = n->string;
1439: if ('<' == *start || '"' == *start)
1440: start++;
1441:
1442: if (0 == (sz = strlen(start)))
1.25 schwarze 1443: return(0);
1.1 kristaps 1444:
1445: end = &start[(int)sz - 1];
1446: if ('>' == *end || '"' == *end)
1447: end--;
1448:
1.50 kristaps 1449: if (end > start)
1.78 schwarze 1450: putkeys(mpage, start, end - start + 1, TYPE_In);
1.92 schwarze 1451: return(0);
1.1 kristaps 1452: }
1453:
1.25 schwarze 1454: static int
1.78 schwarze 1455: parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1456: {
1.112 schwarze 1457: char *cp;
1.1 kristaps 1458:
1.50 kristaps 1459: if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1.25 schwarze 1460: return(0);
1461:
1.50 kristaps 1462: /*
1463: * Parse: .Fn "struct type *name" "char *arg".
1464: * First strip away pointer symbol.
1465: * Then store the function name, then type.
1466: * Finally, store the arguments.
1467: */
1.1 kristaps 1468:
1.50 kristaps 1469: if (NULL == (cp = strrchr(n->string, ' ')))
1470: cp = n->string;
1.1 kristaps 1471:
1472: while ('*' == *cp)
1473: cp++;
1474:
1.78 schwarze 1475: putkey(mpage, cp, TYPE_Fn);
1.25 schwarze 1476:
1.50 kristaps 1477: if (n->string < cp)
1.78 schwarze 1478: putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1.25 schwarze 1479:
1.50 kristaps 1480: for (n = n->next; NULL != n; n = n->next)
1481: if (MDOC_TEXT == n->type)
1.78 schwarze 1482: putkey(mpage, n->string, TYPE_Fa);
1.25 schwarze 1483:
1484: return(0);
1.1 kristaps 1485: }
1486:
1.25 schwarze 1487: static int
1.78 schwarze 1488: parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1489: {
1.67 schwarze 1490: char *cp;
1.1 kristaps 1491:
1492: if (NULL == (n = n->child))
1.25 schwarze 1493: return(0);
1.1 kristaps 1494:
1.67 schwarze 1495: if (NULL == n->next) {
1.78 schwarze 1496: putkey(mpage, n->string, TYPE_Xr);
1.67 schwarze 1497: return(0);
1498: }
1499:
1.120 schwarze 1500: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.78 schwarze 1501: putkey(mpage, cp, TYPE_Xr);
1.67 schwarze 1502: free(cp);
1503: return(0);
1.1 kristaps 1504: }
1505:
1.25 schwarze 1506: static int
1.78 schwarze 1507: parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1508: {
1.50 kristaps 1509: size_t sz;
1.1 kristaps 1510:
1511: if (MDOC_BODY != n->type)
1.25 schwarze 1512: return(0);
1.1 kristaps 1513:
1.50 kristaps 1514: /*
1515: * Special-case the `Nd' because we need to put the description
1516: * into the document table.
1517: */
1518:
1519: for (n = n->child; NULL != n; n = n->next) {
1520: if (MDOC_TEXT == n->type) {
1.78 schwarze 1521: if (NULL != mpage->desc) {
1522: sz = strlen(mpage->desc) +
1.66 schwarze 1523: strlen(n->string) + 2;
1.78 schwarze 1524: mpage->desc = mandoc_realloc(
1525: mpage->desc, sz);
1526: strlcat(mpage->desc, " ", sz);
1527: strlcat(mpage->desc, n->string, sz);
1.66 schwarze 1528: } else
1.78 schwarze 1529: mpage->desc = mandoc_strdup(n->string);
1.50 kristaps 1530: }
1531: if (NULL != n->child)
1.78 schwarze 1532: parse_mdoc_Nd(mpage, n);
1.50 kristaps 1533: }
1.25 schwarze 1534: return(1);
1.1 kristaps 1535: }
1536:
1.25 schwarze 1537: static int
1.78 schwarze 1538: parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1539: {
1540:
1.92 schwarze 1541: return(SEC_NAME == n->sec ||
1542: (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type));
1.1 kristaps 1543: }
1544:
1.25 schwarze 1545: static int
1.78 schwarze 1546: parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1547: {
1548:
1.25 schwarze 1549: return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1.1 kristaps 1550: }
1551:
1.50 kristaps 1552: static int
1.78 schwarze 1553: parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1554: {
1555:
1.50 kristaps 1556: return(MDOC_HEAD == n->type);
1557: }
1.1 kristaps 1558:
1.50 kristaps 1559: static int
1.78 schwarze 1560: parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1.50 kristaps 1561: {
1.1 kristaps 1562:
1.50 kristaps 1563: return(MDOC_BODY == n->type);
1.1 kristaps 1564: }
1565:
1.50 kristaps 1566: /*
1.66 schwarze 1567: * Add a string to the hash table for the current manual.
1568: * Each string has a bitmask telling which macros it belongs to.
1569: * When we finish the manual, we'll dump the table.
1.50 kristaps 1570: */
1571: static void
1.78 schwarze 1572: putkeys(const struct mpage *mpage,
1573: const char *cp, size_t sz, uint64_t v)
1.50 kristaps 1574: {
1575: struct str *s;
1.111 schwarze 1576: const char *end;
1577: uint64_t mask;
1.65 schwarze 1578: unsigned int slot;
1.111 schwarze 1579: int i;
1.25 schwarze 1580:
1.50 kristaps 1581: if (0 == sz)
1582: return;
1.111 schwarze 1583:
1584: if (verb > 1) {
1585: for (i = 0, mask = 1;
1586: i < mansearch_keymax;
1587: i++, mask <<= 1)
1588: if (mask & v)
1589: break;
1590: say(mpage->mlinks->file, "Adding key %s=%*s",
1591: mansearch_keynames[i], sz, cp);
1592: }
1.25 schwarze 1593:
1.65 schwarze 1594: end = cp + sz;
1595: slot = ohash_qlookupi(&strings, cp, &end);
1596: s = ohash_find(&strings, slot);
1.25 schwarze 1597:
1.78 schwarze 1598: if (NULL != s && mpage == s->mpage) {
1.50 kristaps 1599: s->mask |= v;
1600: return;
1601: } else if (NULL == s) {
1.51 kristaps 1602: s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
1.50 kristaps 1603: memcpy(s->key, cp, sz);
1.65 schwarze 1604: ohash_insert(&strings, slot, s);
1.1 kristaps 1605: }
1.78 schwarze 1606: s->mpage = mpage;
1.50 kristaps 1607: s->mask = v;
1.1 kristaps 1608: }
1609:
1.50 kristaps 1610: /*
1611: * Take a Unicode codepoint and produce its UTF-8 encoding.
1612: * This isn't the best way to do this, but it works.
1613: * The magic numbers are from the UTF-8 packaging.
1614: * They're not as scary as they seem: read the UTF-8 spec for details.
1615: */
1616: static size_t
1617: utf8(unsigned int cp, char out[7])
1.1 kristaps 1618: {
1.50 kristaps 1619: size_t rc;
1.1 kristaps 1620:
1.50 kristaps 1621: rc = 0;
1622: if (cp <= 0x0000007F) {
1623: rc = 1;
1624: out[0] = (char)cp;
1625: } else if (cp <= 0x000007FF) {
1626: rc = 2;
1627: out[0] = (cp >> 6 & 31) | 192;
1628: out[1] = (cp & 63) | 128;
1629: } else if (cp <= 0x0000FFFF) {
1630: rc = 3;
1631: out[0] = (cp >> 12 & 15) | 224;
1632: out[1] = (cp >> 6 & 63) | 128;
1633: out[2] = (cp & 63) | 128;
1634: } else if (cp <= 0x001FFFFF) {
1635: rc = 4;
1636: out[0] = (cp >> 18 & 7) | 240;
1637: out[1] = (cp >> 12 & 63) | 128;
1638: out[2] = (cp >> 6 & 63) | 128;
1639: out[3] = (cp & 63) | 128;
1640: } else if (cp <= 0x03FFFFFF) {
1641: rc = 5;
1642: out[0] = (cp >> 24 & 3) | 248;
1643: out[1] = (cp >> 18 & 63) | 128;
1644: out[2] = (cp >> 12 & 63) | 128;
1645: out[3] = (cp >> 6 & 63) | 128;
1646: out[4] = (cp & 63) | 128;
1647: } else if (cp <= 0x7FFFFFFF) {
1648: rc = 6;
1649: out[0] = (cp >> 30 & 1) | 252;
1650: out[1] = (cp >> 24 & 63) | 128;
1651: out[2] = (cp >> 18 & 63) | 128;
1652: out[3] = (cp >> 12 & 63) | 128;
1653: out[4] = (cp >> 6 & 63) | 128;
1654: out[5] = (cp & 63) | 128;
1655: } else
1.1 kristaps 1656: return(0);
1657:
1.50 kristaps 1658: out[rc] = '\0';
1659: return(rc);
1660: }
1.1 kristaps 1661:
1.50 kristaps 1662: /*
1.96 schwarze 1663: * Store the rendered version of a key, or alias the pointer
1664: * if the key contains no escape sequences.
1.50 kristaps 1665: */
1666: static void
1.96 schwarze 1667: render_key(struct mchars *mc, struct str *key)
1.50 kristaps 1668: {
1669: size_t sz, bsz, pos;
1.114 schwarze 1670: char utfbuf[7], res[6];
1.50 kristaps 1671: char *buf;
1672: const char *seq, *cpp, *val;
1673: int len, u;
1674: enum mandoc_esc esc;
1675:
1.96 schwarze 1676: assert(NULL == key->rendered);
1.50 kristaps 1677:
1678: res[0] = '\\';
1679: res[1] = '\t';
1680: res[2] = ASCII_NBRSP;
1681: res[3] = ASCII_HYPH;
1.114 schwarze 1682: res[4] = ASCII_BREAK;
1683: res[5] = '\0';
1.1 kristaps 1684:
1.50 kristaps 1685: val = key->key;
1686: bsz = strlen(val);
1.46 kristaps 1687:
1.50 kristaps 1688: /*
1689: * Pre-check: if we have no stop-characters, then set the
1690: * pointer as ourselvse and get out of here.
1691: */
1692: if (strcspn(val, res) == bsz) {
1.96 schwarze 1693: key->rendered = key->key;
1.50 kristaps 1694: return;
1695: }
1.46 kristaps 1696:
1.50 kristaps 1697: /* Pre-allocate by the length of the input */
1.46 kristaps 1698:
1.50 kristaps 1699: buf = mandoc_malloc(++bsz);
1700: pos = 0;
1.46 kristaps 1701:
1.50 kristaps 1702: while ('\0' != *val) {
1703: /*
1704: * Halt on the first escape sequence.
1705: * This also halts on the end of string, in which case
1706: * we just copy, fallthrough, and exit the loop.
1707: */
1708: if ((sz = strcspn(val, res)) > 0) {
1709: memcpy(&buf[pos], val, sz);
1710: pos += sz;
1711: val += sz;
1712: }
1.46 kristaps 1713:
1.114 schwarze 1714: switch (*val) {
1715: case (ASCII_HYPH):
1.50 kristaps 1716: buf[pos++] = '-';
1717: val++;
1718: continue;
1.114 schwarze 1719: case ('\t'):
1720: /* FALLTHROUGH */
1721: case (ASCII_NBRSP):
1.50 kristaps 1722: buf[pos++] = ' ';
1723: val++;
1.114 schwarze 1724: /* FALLTHROUGH */
1725: case (ASCII_BREAK):
1.50 kristaps 1726: continue;
1.114 schwarze 1727: default:
1728: break;
1729: }
1730: if ('\\' != *val)
1.50 kristaps 1731: break;
1.46 kristaps 1732:
1.50 kristaps 1733: /* Read past the slash. */
1.46 kristaps 1734:
1.50 kristaps 1735: val++;
1.46 kristaps 1736:
1.50 kristaps 1737: /*
1738: * Parse the escape sequence and see if it's a
1739: * predefined character or special character.
1740: */
1.95 schwarze 1741:
1.50 kristaps 1742: esc = mandoc_escape
1743: ((const char **)&val, &seq, &len);
1744: if (ESCAPE_ERROR == esc)
1745: break;
1746: if (ESCAPE_SPECIAL != esc)
1747: continue;
1.1 kristaps 1748:
1.50 kristaps 1749: /*
1.95 schwarze 1750: * Render the special character
1751: * as either UTF-8 or ASCII.
1.50 kristaps 1752: */
1.95 schwarze 1753:
1754: if (write_utf8) {
1755: if (0 == (u = mchars_spec2cp(mc, seq, len)))
1756: continue;
1757: cpp = utfbuf;
1758: if (0 == (sz = utf8(u, utfbuf)))
1759: continue;
1760: sz = strlen(cpp);
1761: } else {
1762: cpp = mchars_spec2str(mc, seq, len, &sz);
1763: if (NULL == cpp)
1764: continue;
1765: if (ASCII_NBRSP == *cpp) {
1766: cpp = " ";
1767: sz = 1;
1768: }
1769: }
1.1 kristaps 1770:
1.50 kristaps 1771: /* Copy the rendered glyph into the stream. */
1.1 kristaps 1772:
1.50 kristaps 1773: bsz += sz;
1774: buf = mandoc_realloc(buf, bsz);
1775: memcpy(&buf[pos], cpp, sz);
1776: pos += sz;
1.1 kristaps 1777: }
1778:
1.50 kristaps 1779: buf[pos] = '\0';
1.96 schwarze 1780: key->rendered = buf;
1.1 kristaps 1781: }
1782:
1.118 schwarze 1783: static void
1784: dbadd_mlink(const struct mlink *mlink)
1785: {
1786: size_t i;
1787:
1788: i = 1;
1789: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1790: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1791: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1792: SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->recno);
1793: SQL_STEP(stmts[STMT_INSERT_LINK]);
1794: sqlite3_reset(stmts[STMT_INSERT_LINK]);
1795: }
1796:
1.14 schwarze 1797: /*
1.50 kristaps 1798: * Flush the current page's terms (and their bits) into the database.
1799: * Wrap the entire set of additions in a transaction to make sqlite be a
1800: * little faster.
1.96 schwarze 1801: * Also, handle escape sequences at the last possible moment.
1.14 schwarze 1802: */
1803: static void
1.118 schwarze 1804: dbadd(struct mpage *mpage, struct mchars *mc)
1.14 schwarze 1805: {
1.87 schwarze 1806: struct mlink *mlink;
1.50 kristaps 1807: struct str *key;
1.52 kristaps 1808: size_t i;
1.66 schwarze 1809: unsigned int slot;
1.14 schwarze 1810:
1.56 schwarze 1811: if (verb)
1.105 schwarze 1812: say(mpage->mlinks->file, "Adding to database");
1.43 kristaps 1813:
1.50 kristaps 1814: if (nodb)
1.14 schwarze 1815: return;
1.28 kristaps 1816:
1.52 kristaps 1817: i = 1;
1.82 schwarze 1818: SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form);
1.81 schwarze 1819: SQL_STEP(stmts[STMT_INSERT_PAGE]);
1.118 schwarze 1820: mpage->recno = sqlite3_last_insert_rowid(db);
1.81 schwarze 1821: sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1822:
1.118 schwarze 1823: for (mlink = mpage->mlinks; mlink; mlink = mlink->next)
1824: dbadd_mlink(mlink);
1.50 kristaps 1825:
1.66 schwarze 1826: for (key = ohash_first(&strings, &slot); NULL != key;
1827: key = ohash_next(&strings, &slot)) {
1.78 schwarze 1828: assert(key->mpage == mpage);
1.96 schwarze 1829: if (NULL == key->rendered)
1830: render_key(mc, key);
1.52 kristaps 1831: i = 1;
1832: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1.96 schwarze 1833: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered);
1.118 schwarze 1834: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->recno);
1.52 kristaps 1835: SQL_STEP(stmts[STMT_INSERT_KEY]);
1.50 kristaps 1836: sqlite3_reset(stmts[STMT_INSERT_KEY]);
1.96 schwarze 1837: if (key->rendered != key->key)
1838: free(key->rendered);
1.66 schwarze 1839: free(key);
1.38 schwarze 1840: }
1.14 schwarze 1841: }
1842:
1.5 kristaps 1843: static void
1.59 schwarze 1844: dbprune(void)
1.5 kristaps 1845: {
1.78 schwarze 1846: struct mpage *mpage;
1.82 schwarze 1847: struct mlink *mlink;
1.52 kristaps 1848: size_t i;
1.80 schwarze 1849: unsigned int slot;
1.5 kristaps 1850:
1.106 schwarze 1851: if (0 == nodb)
1852: SQL_EXEC("BEGIN TRANSACTION");
1.12 schwarze 1853:
1.106 schwarze 1854: for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
1855: mpage = ohash_next(&mpages, &slot)) {
1.82 schwarze 1856: mlink = mpage->mlinks;
1.56 schwarze 1857: if (verb)
1.106 schwarze 1858: say(mlink->file, "Deleting from database");
1859: if (nodb)
1860: continue;
1861: for ( ; NULL != mlink; mlink = mlink->next) {
1862: i = 1;
1863: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
1864: i, mlink->dsec);
1865: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
1866: i, mlink->arch);
1867: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
1868: i, mlink->name);
1869: SQL_STEP(stmts[STMT_DELETE_PAGE]);
1870: sqlite3_reset(stmts[STMT_DELETE_PAGE]);
1871: }
1.5 kristaps 1872: }
1.106 schwarze 1873:
1874: if (0 == nodb)
1875: SQL_EXEC("END TRANSACTION");
1.5 kristaps 1876: }
1877:
1.4 kristaps 1878: /*
1.50 kristaps 1879: * Close an existing database and its prepared statements.
1880: * If "real" is not set, rename the temporary file into the real one.
1.4 kristaps 1881: */
1.35 kristaps 1882: static void
1.59 schwarze 1883: dbclose(int real)
1.4 kristaps 1884: {
1.50 kristaps 1885: size_t i;
1.115 schwarze 1886: int status;
1887: pid_t child;
1.4 kristaps 1888:
1.50 kristaps 1889: if (nodb)
1.38 schwarze 1890: return;
1.50 kristaps 1891:
1892: for (i = 0; i < STMT__MAX; i++) {
1893: sqlite3_finalize(stmts[i]);
1894: stmts[i] = NULL;
1.4 kristaps 1895: }
1896:
1.50 kristaps 1897: sqlite3_close(db);
1898: db = NULL;
1.12 schwarze 1899:
1.50 kristaps 1900: if (real)
1901: return;
1.12 schwarze 1902:
1.115 schwarze 1903: if ('\0' == *tempfilename) {
1904: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
1905: exitcode = (int)MANDOCLEVEL_SYSERR;
1906: say(MANDOC_DB, "%s", strerror(errno));
1907: }
1908: return;
1909: }
1910:
1911: switch (child = fork()) {
1912: case (-1):
1913: exitcode = (int)MANDOCLEVEL_SYSERR;
1914: say("fork cmp", "%s", strerror(errno));
1915: return;
1916: case (0):
1917: execlp("cmp", "cmp", "-s",
1918: tempfilename, MANDOC_DB, NULL);
1919: say("exec cmp", "%s", strerror(errno));
1920: exit(0);
1921: default:
1922: break;
1923: }
1924: if (-1 == waitpid(child, &status, 0)) {
1925: exitcode = (int)MANDOCLEVEL_SYSERR;
1926: say("wait cmp", "%s", strerror(errno));
1927: } else if (WIFSIGNALED(status)) {
1928: exitcode = (int)MANDOCLEVEL_SYSERR;
1929: say("cmp", "Died from a signal");
1930: } else if (WEXITSTATUS(status)) {
1.59 schwarze 1931: exitcode = (int)MANDOCLEVEL_SYSERR;
1.115 schwarze 1932: say(MANDOC_DB,
1933: "Data changed, but cannot replace database");
1934: }
1935:
1936: *strrchr(tempfilename, '/') = '\0';
1937: switch (child = fork()) {
1938: case (-1):
1939: exitcode = (int)MANDOCLEVEL_SYSERR;
1940: say("fork rm", "%s", strerror(errno));
1941: return;
1942: case (0):
1943: execlp("rm", "rm", "-rf", tempfilename, NULL);
1944: say("exec rm", "%s", strerror(errno));
1945: exit((int)MANDOCLEVEL_SYSERR);
1946: default:
1947: break;
1948: }
1949: if (-1 == waitpid(child, &status, 0)) {
1950: exitcode = (int)MANDOCLEVEL_SYSERR;
1951: say("wait rm", "%s", strerror(errno));
1952: } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
1953: exitcode = (int)MANDOCLEVEL_SYSERR;
1954: say(tempfilename,
1955: "Cannot remove temporary directory");
1.59 schwarze 1956: }
1.50 kristaps 1957: }
1.14 schwarze 1958:
1.50 kristaps 1959: /*
1960: * This is straightforward stuff.
1961: * Open a database connection to a "temporary" database, then open a set
1962: * of prepared statements we'll use over and over again.
1963: * If "real" is set, we use the existing database; if not, we truncate a
1964: * temporary one.
1965: * Must be matched by dbclose().
1966: */
1967: static int
1.59 schwarze 1968: dbopen(int real)
1.50 kristaps 1969: {
1.115 schwarze 1970: const char *sql;
1.50 kristaps 1971: int rc, ofl;
1.12 schwarze 1972:
1.50 kristaps 1973: if (nodb)
1974: return(1);
1.12 schwarze 1975:
1.115 schwarze 1976: *tempfilename = '\0';
1.63 schwarze 1977: ofl = SQLITE_OPEN_READWRITE;
1.115 schwarze 1978:
1979: if (real) {
1980: rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
1981: if (SQLITE_OK != rc) {
1.63 schwarze 1982: exitcode = (int)MANDOCLEVEL_SYSERR;
1.115 schwarze 1983: say(MANDOC_DB, "%s", sqlite3_errmsg(db));
1.63 schwarze 1984: return(0);
1985: }
1.115 schwarze 1986: goto prepare_statements;
1987: }
1988:
1989: ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
1.45 kristaps 1990:
1.115 schwarze 1991: remove(MANDOC_DB "~");
1992: rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
1.50 kristaps 1993: if (SQLITE_OK == rc)
1.115 schwarze 1994: goto create_tables;
1.116 schwarze 1995: if (MPARSE_QUICK & mparse_options) {
1.59 schwarze 1996: exitcode = (int)MANDOCLEVEL_SYSERR;
1.115 schwarze 1997: say(MANDOC_DB "~", "%s", sqlite3_errmsg(db));
1.50 kristaps 1998: return(0);
1999: }
1.12 schwarze 2000:
1.115 schwarze 2001: if (strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2002: sizeof(tempfilename)) >= sizeof(tempfilename)) {
2003: exitcode = (int)MANDOCLEVEL_SYSERR;
2004: say("/tmp/mandocdb.XXXXXX", "Filename too long");
2005: return(0);
2006: }
2007: if (NULL == mkdtemp(tempfilename)) {
2008: exitcode = (int)MANDOCLEVEL_SYSERR;
2009: say(tempfilename, "%s", strerror(errno));
2010: return(0);
2011: }
2012: if (strlcat(tempfilename, "/" MANDOC_DB,
2013: sizeof(tempfilename)) >= sizeof(tempfilename)) {
2014: exitcode = (int)MANDOCLEVEL_SYSERR;
2015: say(tempfilename, "Filename too long");
2016: return(0);
2017: }
2018: rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2019: if (SQLITE_OK != rc) {
1.59 schwarze 2020: exitcode = (int)MANDOCLEVEL_SYSERR;
1.115 schwarze 2021: say(tempfilename, "%s", sqlite3_errmsg(db));
1.50 kristaps 2022: return(0);
2023: }
1.12 schwarze 2024:
1.115 schwarze 2025: create_tables:
1.81 schwarze 2026: sql = "CREATE TABLE \"mpages\" (\n"
1.50 kristaps 2027: " \"form\" INTEGER NOT NULL,\n"
2028: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2029: ");\n"
2030: "\n"
1.81 schwarze 2031: "CREATE TABLE \"mlinks\" (\n"
2032: " \"sec\" TEXT NOT NULL,\n"
2033: " \"arch\" TEXT NOT NULL,\n"
2034: " \"name\" TEXT NOT NULL,\n"
2035: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1.109 schwarze 2036: "ON DELETE CASCADE\n"
1.81 schwarze 2037: ");\n"
2038: "\n"
1.50 kristaps 2039: "CREATE TABLE \"keys\" (\n"
2040: " \"bits\" INTEGER NOT NULL,\n"
2041: " \"key\" TEXT NOT NULL,\n"
1.81 schwarze 2042: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1.109 schwarze 2043: "ON DELETE CASCADE\n"
1.108 schwarze 2044: ");\n";
1.14 schwarze 2045:
1.50 kristaps 2046: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
1.59 schwarze 2047: exitcode = (int)MANDOCLEVEL_SYSERR;
1.115 schwarze 2048: say(MANDOC_DB, "%s", sqlite3_errmsg(db));
1.50 kristaps 2049: return(0);
2050: }
1.4 kristaps 2051:
1.57 schwarze 2052: prepare_statements:
2053: SQL_EXEC("PRAGMA foreign_keys = ON");
1.106 schwarze 2054: sql = "DELETE FROM mpages WHERE id IN "
2055: "(SELECT pageid FROM mlinks WHERE "
2056: "sec=? AND arch=? AND name=?)";
1.81 schwarze 2057: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2058: sql = "INSERT INTO mpages "
1.103 schwarze 2059: "(form) VALUES (?)";
1.81 schwarze 2060: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2061: sql = "INSERT INTO mlinks "
1.104 schwarze 2062: "(sec,arch,name,pageid) VALUES (?,?,?,?)";
1.81 schwarze 2063: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1.50 kristaps 2064: sql = "INSERT INTO keys "
1.81 schwarze 2065: "(bits,key,pageid) VALUES (?,?,?)";
1.50 kristaps 2066: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.70 schwarze 2067:
2068: #ifndef __APPLE__
2069: /*
2070: * When opening a new database, we can turn off
2071: * synchronous mode for much better performance.
2072: */
2073:
2074: if (real)
2075: SQL_EXEC("PRAGMA synchronous = OFF");
2076: #endif
2077:
1.50 kristaps 2078: return(1);
2079: }
1.5 kristaps 2080:
1.50 kristaps 2081: static void *
2082: hash_halloc(size_t sz, void *arg)
2083: {
1.12 schwarze 2084:
1.50 kristaps 2085: return(mandoc_calloc(sz, 1));
2086: }
1.12 schwarze 2087:
1.50 kristaps 2088: static void *
2089: hash_alloc(size_t sz, void *arg)
2090: {
1.5 kristaps 2091:
1.50 kristaps 2092: return(mandoc_malloc(sz));
2093: }
1.41 kristaps 2094:
1.50 kristaps 2095: static void
2096: hash_free(void *p, size_t sz, void *arg)
2097: {
1.4 kristaps 2098:
1.50 kristaps 2099: free(p);
1.4 kristaps 2100: }
2101:
1.50 kristaps 2102: static int
1.59 schwarze 2103: set_basedir(const char *targetdir)
1.4 kristaps 2104: {
1.59 schwarze 2105: static char startdir[PATH_MAX];
2106: static int fd;
1.4 kristaps 2107:
1.59 schwarze 2108: /*
2109: * Remember where we started by keeping a fd open to the origin
2110: * path component: throughout this utility, we chdir() a lot to
2111: * handle relative paths, and by doing this, we can return to
2112: * the starting point.
2113: */
2114: if ('\0' == *startdir) {
2115: if (NULL == getcwd(startdir, PATH_MAX)) {
2116: exitcode = (int)MANDOCLEVEL_SYSERR;
2117: if (NULL != targetdir)
2118: say(".", NULL);
2119: return(0);
2120: }
2121: if (-1 == (fd = open(startdir, O_RDONLY, 0))) {
2122: exitcode = (int)MANDOCLEVEL_SYSERR;
2123: say(startdir, NULL);
2124: return(0);
2125: }
2126: if (NULL == targetdir)
2127: targetdir = startdir;
2128: } else {
2129: if (-1 == fd)
2130: return(0);
2131: if (-1 == fchdir(fd)) {
2132: close(fd);
2133: basedir[0] = '\0';
2134: exitcode = (int)MANDOCLEVEL_SYSERR;
2135: say(startdir, NULL);
2136: return(0);
2137: }
2138: if (NULL == targetdir) {
2139: close(fd);
2140: return(1);
2141: }
2142: }
2143: if (NULL == realpath(targetdir, basedir)) {
2144: basedir[0] = '\0';
2145: exitcode = (int)MANDOCLEVEL_BADARG;
2146: say(targetdir, NULL);
1.50 kristaps 2147: return(0);
1.59 schwarze 2148: } else if (-1 == chdir(basedir)) {
2149: exitcode = (int)MANDOCLEVEL_BADARG;
2150: say("", NULL);
1.50 kristaps 2151: return(0);
1.4 kristaps 2152: }
1.50 kristaps 2153: return(1);
1.56 schwarze 2154: }
2155:
2156: static void
1.59 schwarze 2157: say(const char *file, const char *format, ...)
1.56 schwarze 2158: {
2159: va_list ap;
2160:
1.59 schwarze 2161: if ('\0' != *basedir)
2162: fprintf(stderr, "%s", basedir);
2163: if ('\0' != *basedir && '\0' != *file)
2164: fputs("//", stderr);
1.56 schwarze 2165: if ('\0' != *file)
1.59 schwarze 2166: fprintf(stderr, "%s", file);
1.56 schwarze 2167: fputs(": ", stderr);
1.59 schwarze 2168:
2169: if (NULL == format) {
2170: perror(NULL);
2171: return;
2172: }
1.56 schwarze 2173:
2174: va_start(ap, format);
2175: vfprintf(stderr, format, ap);
2176: va_end(ap);
2177:
2178: fputc('\n', stderr);
1.1 kristaps 2179: }
CVSweb