Annotation of mandoc/mandocdb.c, Revision 1.59
1.59 ! schwarze 1: /* $Id: mandocdb.c,v 1.58 2013/06/05 02:00:26 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>
1.1 kristaps 26: #include <fcntl.h>
1.50 kristaps 27: #include <fts.h>
1.1 kristaps 28: #include <getopt.h>
1.58 schwarze 29: #include <limits.h>
1.50 kristaps 30: #include <stddef.h>
1.59 ! schwarze 31: #include <stdio.h>
1.1 kristaps 32: #include <stdint.h>
33: #include <stdlib.h>
34: #include <string.h>
1.17 schwarze 35: #include <unistd.h>
1.1 kristaps 36:
1.53 kristaps 37: #ifdef HAVE_OHASH
1.50 kristaps 38: #include <ohash.h>
1.53 kristaps 39: #else
40: #include "compat_ohash.h"
41: #endif
1.50 kristaps 42: #include <sqlite3.h>
1.1 kristaps 43:
1.50 kristaps 44: #include "mdoc.h"
1.1 kristaps 45: #include "man.h"
46: #include "mandoc.h"
1.10 kristaps 47: #include "manpath.h"
1.55 kristaps 48: #include "mansearch.h"
1.1 kristaps 49:
1.52 kristaps 50: #define SQL_EXEC(_v) \
51: if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
52: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
53: #define SQL_BIND_TEXT(_s, _i, _v) \
54: if (SQLITE_OK != sqlite3_bind_text \
55: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
56: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
57: #define SQL_BIND_INT(_s, _i, _v) \
58: if (SQLITE_OK != sqlite3_bind_int \
59: ((_s), (_i)++, (_v))) \
60: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
61: #define SQL_BIND_INT64(_s, _i, _v) \
62: if (SQLITE_OK != sqlite3_bind_int64 \
63: ((_s), (_i)++, (_v))) \
64: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
65: #define SQL_STEP(_s) \
66: if (SQLITE_DONE != sqlite3_step((_s))) \
67: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
68:
1.50 kristaps 69: enum op {
70: OP_DEFAULT = 0, /* new dbs from dir list or default config */
71: OP_CONFFILE, /* new databases from custom config file */
72: OP_UPDATE, /* delete/add entries in existing database */
73: OP_DELETE, /* delete entries from existing database */
74: OP_TEST /* change no databases, report potential problems */
1.38 schwarze 75: };
76:
1.50 kristaps 77: enum form {
78: FORM_SRC, /* format is -man or -mdoc */
79: FORM_CAT, /* format is cat */
80: FORM_NONE /* format is unknown */
81: };
1.38 schwarze 82:
1.50 kristaps 83: struct str {
84: char *utf8; /* key in UTF-8 form */
85: const struct of *of; /* if set, the owning parse */
86: struct str *next; /* next in owning parse sequence */
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 */
1.50 kristaps 102: const char *desc; /* parsed description */
103: const char *sec; /* suffix-cued section (or empty) */
104: const char *dsec; /* path-cued section (or empty) */
105: const char *arch; /* path-cued arch. (or empty) */
106: const char *name; /* name (from filename) (not empty) */
1.3 kristaps 107: };
108:
1.50 kristaps 109: enum stmt {
110: STMT_DELETE = 0, /* delete manpage */
111: STMT_INSERT_DOC, /* insert manpage */
112: STMT_INSERT_KEY, /* insert parsed key */
113: STMT__MAX
1.1 kristaps 114: };
115:
1.50 kristaps 116: typedef int (*mdoc_fp)(struct of *, const struct mdoc_node *);
1.1 kristaps 117:
1.50 kristaps 118: struct mdoc_handler {
119: mdoc_fp fp; /* optional handler */
120: uint64_t mask; /* set unless handler returns 0 */
121: int flags; /* for use by pmdoc_node */
122: #define MDOCF_CHILD 0x01 /* automatically index child nodes */
1.1 kristaps 123: };
124:
1.59 ! schwarze 125: static void dbclose(int);
! 126: static void dbindex(struct mchars *, int, const struct of *);
! 127: static int dbopen(int);
! 128: static void dbprune(void);
1.50 kristaps 129: static void fileadd(struct of *);
130: static int filecheck(const char *);
1.59 ! schwarze 131: static void filescan(const char *);
1.50 kristaps 132: static struct str *hashget(const char *, size_t);
133: static void *hash_alloc(size_t, void *);
134: static void hash_free(void *, size_t, void *);
135: static void *hash_halloc(size_t, void *);
136: static void inoadd(const struct stat *, struct of *);
137: static int inocheck(const struct stat *);
1.59 ! schwarze 138: static void ofadd(int, const char *, const char *, const char *,
! 139: const char *, const char *, const struct stat *);
1.50 kristaps 140: static void offree(void);
1.59 ! schwarze 141: static void ofmerge(struct mchars *, struct mparse *);
! 142: static void parse_catpage(struct of *);
1.50 kristaps 143: static int parse_man(struct of *,
144: const struct man_node *);
145: static void parse_mdoc(struct of *, const struct mdoc_node *);
146: static int parse_mdoc_body(struct of *, const struct mdoc_node *);
147: static int parse_mdoc_head(struct of *, const struct mdoc_node *);
148: static int parse_mdoc_Fd(struct of *, const struct mdoc_node *);
149: static int parse_mdoc_Fn(struct of *, const struct mdoc_node *);
150: static int parse_mdoc_In(struct of *, const struct mdoc_node *);
151: static int parse_mdoc_Nd(struct of *, const struct mdoc_node *);
152: static int parse_mdoc_Nm(struct of *, const struct mdoc_node *);
153: static int parse_mdoc_Sh(struct of *, const struct mdoc_node *);
154: static int parse_mdoc_St(struct of *, const struct mdoc_node *);
155: static int parse_mdoc_Xr(struct of *, const struct mdoc_node *);
1.59 ! schwarze 156: static int set_basedir(const char *);
1.50 kristaps 157: static void putkey(const struct of *,
158: const char *, uint64_t);
159: static void putkeys(const struct of *,
160: const char *, int, uint64_t);
161: static void putmdockey(const struct of *,
162: const struct mdoc_node *, uint64_t);
1.59 ! schwarze 163: static void say(const char *, const char *, ...);
1.50 kristaps 164: static char *stradd(const char *);
165: static char *straddbuf(const char *, size_t);
1.59 ! schwarze 166: static int treescan(void);
1.50 kristaps 167: static size_t utf8(unsigned int, char [7]);
168: static void utf8key(struct mchars *, struct str *);
169: static void wordaddbuf(const struct of *,
170: const char *, size_t, uint64_t);
171:
172: static char *progname;
173: static int use_all; /* use all found files */
174: static int nodb; /* no database changes */
175: static int verb; /* print what we're doing */
176: static int warnings; /* warn about crap */
1.59 ! schwarze 177: static int exitcode; /* to be returned by main */
1.50 kristaps 178: static enum op op; /* operational mode */
1.59 ! schwarze 179: static char basedir[PATH_MAX]; /* current base directory */
1.50 kristaps 180: static struct ohash inos; /* table of inodes/devices */
181: static struct ohash filenames; /* table of filenames */
182: static struct ohash strings; /* table of all strings */
183: static struct of *ofs = NULL; /* vector of files to parse */
184: static struct str *words = NULL; /* word list in current parse */
185: static sqlite3 *db = NULL; /* current database */
186: static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
1.25 schwarze 187:
188: static const struct mdoc_handler mdocs[MDOC_MAX] = {
189: { NULL, 0, 0 }, /* Ap */
190: { NULL, 0, 0 }, /* Dd */
191: { NULL, 0, 0 }, /* Dt */
192: { NULL, 0, 0 }, /* Os */
1.50 kristaps 193: { parse_mdoc_Sh, TYPE_Sh, MDOCF_CHILD }, /* Sh */
194: { parse_mdoc_head, TYPE_Ss, MDOCF_CHILD }, /* Ss */
1.25 schwarze 195: { NULL, 0, 0 }, /* Pp */
196: { NULL, 0, 0 }, /* D1 */
197: { NULL, 0, 0 }, /* Dl */
198: { NULL, 0, 0 }, /* Bd */
199: { NULL, 0, 0 }, /* Ed */
200: { NULL, 0, 0 }, /* Bl */
201: { NULL, 0, 0 }, /* El */
202: { NULL, 0, 0 }, /* It */
203: { NULL, 0, 0 }, /* Ad */
204: { NULL, TYPE_An, MDOCF_CHILD }, /* An */
205: { NULL, TYPE_Ar, MDOCF_CHILD }, /* Ar */
206: { NULL, TYPE_Cd, MDOCF_CHILD }, /* Cd */
207: { NULL, TYPE_Cm, MDOCF_CHILD }, /* Cm */
208: { NULL, TYPE_Dv, MDOCF_CHILD }, /* Dv */
209: { NULL, TYPE_Er, MDOCF_CHILD }, /* Er */
210: { NULL, TYPE_Ev, MDOCF_CHILD }, /* Ev */
211: { NULL, 0, 0 }, /* Ex */
212: { NULL, TYPE_Fa, MDOCF_CHILD }, /* Fa */
1.50 kristaps 213: { parse_mdoc_Fd, TYPE_In, 0 }, /* Fd */
1.25 schwarze 214: { NULL, TYPE_Fl, MDOCF_CHILD }, /* Fl */
1.50 kristaps 215: { parse_mdoc_Fn, 0, 0 }, /* Fn */
1.25 schwarze 216: { NULL, TYPE_Ft, MDOCF_CHILD }, /* Ft */
217: { NULL, TYPE_Ic, MDOCF_CHILD }, /* Ic */
1.50 kristaps 218: { parse_mdoc_In, TYPE_In, MDOCF_CHILD }, /* In */
1.25 schwarze 219: { NULL, TYPE_Li, MDOCF_CHILD }, /* Li */
1.50 kristaps 220: { parse_mdoc_Nd, TYPE_Nd, MDOCF_CHILD }, /* Nd */
221: { parse_mdoc_Nm, TYPE_Nm, MDOCF_CHILD }, /* Nm */
1.25 schwarze 222: { NULL, 0, 0 }, /* Op */
223: { NULL, 0, 0 }, /* Ot */
224: { NULL, TYPE_Pa, MDOCF_CHILD }, /* Pa */
225: { NULL, 0, 0 }, /* Rv */
1.50 kristaps 226: { parse_mdoc_St, TYPE_St, 0 }, /* St */
1.25 schwarze 227: { NULL, TYPE_Va, MDOCF_CHILD }, /* Va */
1.50 kristaps 228: { parse_mdoc_body, TYPE_Va, MDOCF_CHILD }, /* Vt */
229: { parse_mdoc_Xr, TYPE_Xr, 0 }, /* Xr */
1.25 schwarze 230: { NULL, 0, 0 }, /* %A */
231: { NULL, 0, 0 }, /* %B */
232: { NULL, 0, 0 }, /* %D */
233: { NULL, 0, 0 }, /* %I */
234: { NULL, 0, 0 }, /* %J */
235: { NULL, 0, 0 }, /* %N */
236: { NULL, 0, 0 }, /* %O */
237: { NULL, 0, 0 }, /* %P */
238: { NULL, 0, 0 }, /* %R */
239: { NULL, 0, 0 }, /* %T */
240: { NULL, 0, 0 }, /* %V */
241: { NULL, 0, 0 }, /* Ac */
242: { NULL, 0, 0 }, /* Ao */
243: { NULL, 0, 0 }, /* Aq */
244: { NULL, TYPE_At, MDOCF_CHILD }, /* At */
245: { NULL, 0, 0 }, /* Bc */
246: { NULL, 0, 0 }, /* Bf */
247: { NULL, 0, 0 }, /* Bo */
248: { NULL, 0, 0 }, /* Bq */
249: { NULL, TYPE_Bsx, MDOCF_CHILD }, /* Bsx */
250: { NULL, TYPE_Bx, MDOCF_CHILD }, /* Bx */
251: { NULL, 0, 0 }, /* Db */
252: { NULL, 0, 0 }, /* Dc */
253: { NULL, 0, 0 }, /* Do */
254: { NULL, 0, 0 }, /* Dq */
255: { NULL, 0, 0 }, /* Ec */
256: { NULL, 0, 0 }, /* Ef */
257: { NULL, TYPE_Em, MDOCF_CHILD }, /* Em */
258: { NULL, 0, 0 }, /* Eo */
259: { NULL, TYPE_Fx, MDOCF_CHILD }, /* Fx */
260: { NULL, TYPE_Ms, MDOCF_CHILD }, /* Ms */
261: { NULL, 0, 0 }, /* No */
262: { NULL, 0, 0 }, /* Ns */
263: { NULL, TYPE_Nx, MDOCF_CHILD }, /* Nx */
264: { NULL, TYPE_Ox, MDOCF_CHILD }, /* Ox */
265: { NULL, 0, 0 }, /* Pc */
266: { NULL, 0, 0 }, /* Pf */
267: { NULL, 0, 0 }, /* Po */
268: { NULL, 0, 0 }, /* Pq */
269: { NULL, 0, 0 }, /* Qc */
270: { NULL, 0, 0 }, /* Ql */
271: { NULL, 0, 0 }, /* Qo */
272: { NULL, 0, 0 }, /* Qq */
273: { NULL, 0, 0 }, /* Re */
274: { NULL, 0, 0 }, /* Rs */
275: { NULL, 0, 0 }, /* Sc */
276: { NULL, 0, 0 }, /* So */
277: { NULL, 0, 0 }, /* Sq */
278: { NULL, 0, 0 }, /* Sm */
279: { NULL, 0, 0 }, /* Sx */
280: { NULL, TYPE_Sy, MDOCF_CHILD }, /* Sy */
281: { NULL, TYPE_Tn, MDOCF_CHILD }, /* Tn */
282: { NULL, 0, 0 }, /* Ux */
283: { NULL, 0, 0 }, /* Xc */
284: { NULL, 0, 0 }, /* Xo */
1.50 kristaps 285: { parse_mdoc_head, TYPE_Fn, 0 }, /* Fo */
1.25 schwarze 286: { NULL, 0, 0 }, /* Fc */
287: { NULL, 0, 0 }, /* Oo */
288: { NULL, 0, 0 }, /* Oc */
289: { NULL, 0, 0 }, /* Bk */
290: { NULL, 0, 0 }, /* Ek */
291: { NULL, 0, 0 }, /* Bt */
292: { NULL, 0, 0 }, /* Hf */
293: { NULL, 0, 0 }, /* Fr */
294: { NULL, 0, 0 }, /* Ud */
295: { NULL, TYPE_Lb, MDOCF_CHILD }, /* Lb */
296: { NULL, 0, 0 }, /* Lp */
297: { NULL, TYPE_Lk, MDOCF_CHILD }, /* Lk */
298: { NULL, TYPE_Mt, MDOCF_CHILD }, /* Mt */
299: { NULL, 0, 0 }, /* Brq */
300: { NULL, 0, 0 }, /* Bro */
301: { NULL, 0, 0 }, /* Brc */
302: { NULL, 0, 0 }, /* %C */
303: { NULL, 0, 0 }, /* Es */
304: { NULL, 0, 0 }, /* En */
305: { NULL, TYPE_Dx, MDOCF_CHILD }, /* Dx */
306: { NULL, 0, 0 }, /* %Q */
307: { NULL, 0, 0 }, /* br */
308: { NULL, 0, 0 }, /* sp */
309: { NULL, 0, 0 }, /* %U */
310: { NULL, 0, 0 }, /* Ta */
1.1 kristaps 311: };
312:
313: int
314: main(int argc, char *argv[])
315: {
1.59 ! schwarze 316: int ch, i;
1.53 kristaps 317: unsigned int index;
1.50 kristaps 318: size_t j, sz;
1.59 ! schwarze 319: const char *path_arg;
1.50 kristaps 320: struct str *s;
321: struct mchars *mc;
322: struct manpaths dirs;
323: struct mparse *mp;
324: struct ohash_info ino_info, filename_info, str_info;
325:
326: memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
327: memset(&dirs, 0, sizeof(struct manpaths));
328:
329: ino_info.halloc = filename_info.halloc =
330: str_info.halloc = hash_halloc;
331: ino_info.hfree = filename_info.hfree =
332: str_info.hfree = hash_free;
333: ino_info.alloc = filename_info.alloc =
334: str_info.alloc = hash_alloc;
335:
336: ino_info.key_offset = offsetof(struct of, id);
337: filename_info.key_offset = offsetof(struct of, file);
338: str_info.key_offset = offsetof(struct str, key);
1.1 kristaps 339:
340: progname = strrchr(argv[0], '/');
341: if (progname == NULL)
342: progname = argv[0];
343: else
344: ++progname;
345:
1.50 kristaps 346: /*
347: * We accept a few different invocations.
348: * The CHECKOP macro makes sure that invocation styles don't
349: * clobber each other.
350: */
351: #define CHECKOP(_op, _ch) do \
352: if (OP_DEFAULT != (_op)) { \
353: fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \
354: goto usage; \
355: } while (/*CONSTCOND*/0)
1.10 kristaps 356:
1.59 ! schwarze 357: path_arg = NULL;
1.38 schwarze 358: op = OP_DEFAULT;
1.1 kristaps 359:
1.50 kristaps 360: while (-1 != (ch = getopt(argc, argv, "aC:d:ntu:vW")))
1.1 kristaps 361: switch (ch) {
1.12 schwarze 362: case ('a'):
363: use_all = 1;
364: break;
1.34 schwarze 365: case ('C'):
1.50 kristaps 366: CHECKOP(op, ch);
1.59 ! schwarze 367: path_arg = optarg;
1.38 schwarze 368: op = OP_CONFFILE;
1.34 schwarze 369: break;
1.5 kristaps 370: case ('d'):
1.50 kristaps 371: CHECKOP(op, ch);
1.59 ! schwarze 372: path_arg = optarg;
1.5 kristaps 373: op = OP_UPDATE;
374: break;
1.50 kristaps 375: case ('n'):
376: nodb = 1;
377: break;
1.38 schwarze 378: case ('t'):
1.50 kristaps 379: CHECKOP(op, ch);
1.38 schwarze 380: dup2(STDOUT_FILENO, STDERR_FILENO);
381: op = OP_TEST;
1.50 kristaps 382: nodb = warnings = 1;
1.38 schwarze 383: break;
1.5 kristaps 384: case ('u'):
1.50 kristaps 385: CHECKOP(op, ch);
1.59 ! schwarze 386: path_arg = optarg;
1.5 kristaps 387: op = OP_DELETE;
388: break;
389: case ('v'):
390: verb++;
391: break;
1.38 schwarze 392: case ('W'):
393: warnings = 1;
394: break;
1.1 kristaps 395: default:
1.38 schwarze 396: goto usage;
1.1 kristaps 397: }
398:
399: argc -= optind;
400: argv += optind;
401:
1.38 schwarze 402: if (OP_CONFFILE == op && argc > 0) {
1.50 kristaps 403: fprintf(stderr, "-C: Too many arguments\n");
1.38 schwarze 404: goto usage;
405: }
406:
1.59 ! schwarze 407: exitcode = (int)MANDOCLEVEL_OK;
1.50 kristaps 408: mp = mparse_alloc(MPARSE_AUTO,
409: MANDOCLEVEL_FATAL, NULL, NULL, NULL);
410: mc = mchars_alloc();
411:
412: ohash_init(&strings, 6, &str_info);
413: ohash_init(&inos, 6, &ino_info);
414: ohash_init(&filenames, 6, &filename_info);
415:
416: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
417: /*
418: * Force processing all files.
419: */
420: use_all = 1;
1.59 ! schwarze 421:
1.50 kristaps 422: /*
423: * All of these deal with a specific directory.
424: * Jump into that directory then collect files specified
425: * on the command-line.
426: */
1.59 ! schwarze 427: if (0 == set_basedir(path_arg))
1.50 kristaps 428: goto out;
429: for (i = 0; i < argc; i++)
1.59 ! schwarze 430: filescan(argv[i]);
! 431: if (0 == dbopen(1))
1.50 kristaps 432: goto out;
433: if (OP_TEST != op)
1.59 ! schwarze 434: dbprune();
1.50 kristaps 435: if (OP_DELETE != op)
1.59 ! schwarze 436: ofmerge(mc, mp);
! 437: dbclose(1);
1.50 kristaps 438: } else {
439: /*
440: * If we have arguments, use them as our manpaths.
441: * If we don't, grok from manpath(1) or however else
442: * manpath_parse() wants to do it.
443: */
444: if (argc > 0) {
445: dirs.paths = mandoc_calloc
446: (argc, sizeof(char *));
447: dirs.sz = (size_t)argc;
448: for (i = 0; i < argc; i++)
449: dirs.paths[i] = mandoc_strdup(argv[i]);
450: } else
1.59 ! schwarze 451: manpath_parse(&dirs, path_arg, NULL, NULL);
1.50 kristaps 452:
453: /*
454: * First scan the tree rooted at a base directory.
455: * Then whak its database (if one exists), parse, and
456: * build up the database.
457: * Ignore zero-length directories and strip trailing
458: * slashes.
459: */
460: for (j = 0; j < dirs.sz; j++) {
461: sz = strlen(dirs.paths[j]);
462: if (sz && '/' == dirs.paths[j][sz - 1])
463: dirs.paths[j][--sz] = '\0';
464: if (0 == sz)
465: continue;
1.59 ! schwarze 466: if (0 == set_basedir(dirs.paths[j]))
1.50 kristaps 467: goto out;
1.59 ! schwarze 468: if (0 == treescan())
1.50 kristaps 469: goto out;
1.59 ! schwarze 470: if (0 == set_basedir(dirs.paths[j]))
1.50 kristaps 471: goto out;
1.59 ! schwarze 472: if (0 == dbopen(0))
1.50 kristaps 473: goto out;
1.52 kristaps 474:
475: /*
476: * Since we're opening up a new database, we can
477: * turn off synchronous mode for much better
478: * performance.
479: */
1.54 kristaps 480: #ifndef __APPLE__
1.52 kristaps 481: SQL_EXEC("PRAGMA synchronous = OFF");
1.54 kristaps 482: #endif
1.52 kristaps 483:
1.59 ! schwarze 484: ofmerge(mc, mp);
! 485: dbclose(0);
1.50 kristaps 486: offree();
487: ohash_delete(&inos);
488: ohash_init(&inos, 6, &ino_info);
489: ohash_delete(&filenames);
490: ohash_init(&filenames, 6, &filename_info);
491: }
492: }
493: out:
1.59 ! schwarze 494: set_basedir(NULL);
1.50 kristaps 495: manpath_free(&dirs);
496: mchars_free(mc);
497: mparse_free(mp);
1.53 kristaps 498: for (s = ohash_first(&strings, &index);
499: NULL != s; s = ohash_next(&strings, &index)) {
1.50 kristaps 500: if (s->utf8 != s->key)
501: free(s->utf8);
502: free(s);
503: }
504: ohash_delete(&strings);
505: ohash_delete(&inos);
506: ohash_delete(&filenames);
507: offree();
1.59 ! schwarze 508: return(exitcode);
1.50 kristaps 509: usage:
510: fprintf(stderr, "usage: %s [-anvW] [-C file]\n"
511: " %s [-anvW] dir ...\n"
512: " %s [-nvW] -d dir [file ...]\n"
513: " %s [-nvW] -u dir [file ...]\n"
514: " %s -t file ...\n",
515: progname, progname, progname,
516: progname, progname);
517:
1.59 ! schwarze 518: return((int)MANDOCLEVEL_BADARG);
1.50 kristaps 519: }
520:
521: /*
1.59 ! schwarze 522: * Scan a directory tree rooted at "basedir" for manpages.
1.50 kristaps 523: * We use fts(), scanning directory parts along the way for clues to our
524: * section and architecture.
525: *
526: * If use_all has been specified, grok all files.
527: * If not, sanitise paths to the following:
528: *
529: * [./]man*[/<arch>]/<name>.<section>
530: * or
531: * [./]cat<section>[/<arch>]/<name>.0
532: *
533: * TODO: accomodate for multi-language directories.
534: */
535: static int
1.59 ! schwarze 536: treescan(void)
1.50 kristaps 537: {
538: FTS *f;
539: FTSENT *ff;
540: int dform;
541: char *sec;
542: const char *dsec, *arch, *cp, *name, *path;
543: const char *argv[2];
544:
545: argv[0] = ".";
546: argv[1] = (char *)NULL;
547:
548: /*
549: * Walk through all components under the directory, using the
550: * logical descent of files.
551: */
552: f = fts_open((char * const *)argv, FTS_LOGICAL, NULL);
553: if (NULL == f) {
1.59 ! schwarze 554: exitcode = (int)MANDOCLEVEL_SYSERR;
! 555: say("", NULL);
1.50 kristaps 556: return(0);
557: }
558:
559: dsec = arch = NULL;
560: dform = FORM_NONE;
561:
562: while (NULL != (ff = fts_read(f))) {
563: path = ff->fts_path + 2;
564: /*
565: * If we're a regular file, add an "of" by using the
566: * stored directory data and handling the filename.
567: * Disallow duplicate (hard-linked) files.
568: */
569: if (FTS_F == ff->fts_info) {
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;
574: } else if (inocheck(ff->fts_statp)) {
1.56 schwarze 575: if (warnings)
1.59 ! schwarze 576: say(path, "Duplicate file");
1.50 kristaps 577: continue;
578: }
579:
580: cp = ff->fts_name;
581:
582: if (0 == strcmp(cp, "mandocdb.db")) {
1.56 schwarze 583: if (warnings)
1.59 ! schwarze 584: say(path, "Skip database");
1.50 kristaps 585: continue;
586: } else if (NULL != (cp = strrchr(cp, '.'))) {
587: if (0 == strcmp(cp + 1, "html")) {
1.56 schwarze 588: if (warnings)
1.59 ! schwarze 589: say(path, "Skip html");
1.50 kristaps 590: continue;
591: } else if (0 == strcmp(cp + 1, "gz")) {
1.56 schwarze 592: if (warnings)
1.59 ! schwarze 593: say(path, "Skip gz");
1.50 kristaps 594: continue;
595: } else if (0 == strcmp(cp + 1, "ps")) {
1.56 schwarze 596: if (warnings)
1.59 ! schwarze 597: say(path, "Skip ps");
1.50 kristaps 598: continue;
599: } else if (0 == strcmp(cp + 1, "pdf")) {
1.56 schwarze 600: if (warnings)
1.59 ! schwarze 601: say(path, "Skip pdf");
1.50 kristaps 602: continue;
603: }
604: }
1.1 kristaps 605:
1.50 kristaps 606: if (NULL != (sec = strrchr(ff->fts_name, '.'))) {
607: *sec = '\0';
608: sec = stradd(sec + 1);
609: }
610: name = stradd(ff->fts_name);
1.59 ! schwarze 611: ofadd(dform, path,
1.50 kristaps 612: name, dsec, sec, arch, ff->fts_statp);
613: continue;
614: } else if (FTS_D != ff->fts_info &&
615: FTS_DP != ff->fts_info)
616: continue;
1.1 kristaps 617:
1.50 kristaps 618: switch (ff->fts_level) {
619: case (0):
620: /* Ignore the root directory. */
621: break;
622: case (1):
623: /*
624: * This might contain manX/ or catX/.
625: * Try to infer this from the name.
626: * If we're not in use_all, enforce it.
627: */
628: dsec = NULL;
629: dform = FORM_NONE;
630: cp = ff->fts_name;
631: if (FTS_DP == ff->fts_info)
632: break;
1.1 kristaps 633:
1.50 kristaps 634: if (0 == strncmp(cp, "man", 3)) {
635: dform = FORM_SRC;
636: dsec = stradd(cp + 3);
637: } else if (0 == strncmp(cp, "cat", 3)) {
638: dform = FORM_CAT;
639: dsec = stradd(cp + 3);
640: }
1.1 kristaps 641:
1.50 kristaps 642: if (NULL != dsec || use_all)
643: break;
1.1 kristaps 644:
1.56 schwarze 645: if (warnings)
1.59 ! schwarze 646: say(path, "Unknown directory part");
1.50 kristaps 647: fts_set(f, ff, FTS_SKIP);
648: break;
649: case (2):
650: /*
651: * Possibly our architecture.
652: * If we're descending, keep tabs on it.
653: */
654: arch = NULL;
655: if (FTS_DP != ff->fts_info && NULL != dsec)
656: arch = stradd(ff->fts_name);
657: break;
658: default:
659: if (FTS_DP == ff->fts_info || use_all)
660: break;
1.56 schwarze 661: if (warnings)
1.59 ! schwarze 662: say(path, "Extraneous directory part");
1.50 kristaps 663: fts_set(f, ff, FTS_SKIP);
664: break;
1.5 kristaps 665: }
1.50 kristaps 666: }
667:
668: fts_close(f);
669: return(1);
670: }
1.5 kristaps 671:
1.50 kristaps 672: /*
673: * Add a file to the file vector.
674: * Do not verify that it's a "valid" looking manpage (we'll do that
675: * later).
676: *
677: * Try to infer the manual section, architecture, and page name from the
678: * path, assuming it looks like
679: *
680: * [./]man*[/<arch>]/<name>.<section>
681: * or
682: * [./]cat<section>[/<arch>]/<name>.0
683: *
684: * Stuff this information directly into the "of" vector.
685: * See treescan() for the fts(3) version of this.
686: */
687: static void
1.59 ! schwarze 688: filescan(const char *file)
1.50 kristaps 689: {
1.59 ! schwarze 690: char buf[PATH_MAX];
1.50 kristaps 691: const char *sec, *arch, *name, *dsec;
1.59 ! schwarze 692: char *p, *start;
1.50 kristaps 693: int dform;
694: struct stat st;
1.5 kristaps 695:
1.50 kristaps 696: assert(use_all);
1.5 kristaps 697:
1.50 kristaps 698: if (0 == strncmp(file, "./", 2))
699: file += 2;
1.5 kristaps 700:
1.59 ! schwarze 701: if (NULL == realpath(file, buf)) {
! 702: exitcode = (int)MANDOCLEVEL_BADARG;
! 703: say(file, NULL);
! 704: return;
! 705: } else if (strstr(buf, basedir) != buf) {
! 706: exitcode = (int)MANDOCLEVEL_BADARG;
! 707: say("", "%s: outside base directory", buf);
! 708: return;
! 709: } else if (-1 == stat(buf, &st)) {
! 710: exitcode = (int)MANDOCLEVEL_BADARG;
! 711: say(file, NULL);
1.50 kristaps 712: return;
713: } else if ( ! (S_IFREG & st.st_mode)) {
1.59 ! schwarze 714: exitcode = (int)MANDOCLEVEL_BADARG;
! 715: say(file, "Not a regular file");
1.50 kristaps 716: return;
717: } else if (inocheck(&st)) {
1.56 schwarze 718: if (warnings)
1.59 ! schwarze 719: say(file, "Duplicate file");
1.50 kristaps 720: return;
721: }
1.59 ! schwarze 722: start = buf + strlen(basedir);
1.50 kristaps 723: sec = arch = name = dsec = NULL;
724: dform = FORM_NONE;
1.17 schwarze 725:
1.50 kristaps 726: /*
727: * First try to guess our directory structure.
728: * If we find a separator, try to look for man* or cat*.
729: * If we find one of these and what's underneath is a directory,
730: * assume it's an architecture.
731: */
732: if (NULL != (p = strchr(start, '/'))) {
733: *p++ = '\0';
734: if (0 == strncmp(start, "man", 3)) {
735: dform = FORM_SRC;
736: dsec = start + 3;
737: } else if (0 == strncmp(start, "cat", 3)) {
738: dform = FORM_CAT;
739: dsec = start + 3;
1.17 schwarze 740: }
1.5 kristaps 741:
1.50 kristaps 742: start = p;
743: if (NULL != dsec && NULL != (p = strchr(start, '/'))) {
744: *p++ = '\0';
745: arch = start;
746: start = p;
747: }
748: }
749:
750: /*
751: * Now check the file suffix.
752: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
753: */
754: p = strrchr(start, '\0');
755: while (p-- > start && '/' != *p && '.' != *p)
756: /* Loop. */ ;
757:
758: if ('.' == *p) {
759: *p++ = '\0';
760: sec = p;
1.5 kristaps 761: }
762:
1.10 kristaps 763: /*
1.50 kristaps 764: * Now try to parse the name.
765: * Use the filename portion of the path.
1.10 kristaps 766: */
1.50 kristaps 767: name = start;
768: if (NULL != (p = strrchr(start, '/'))) {
769: name = p + 1;
770: *p = '\0';
771: }
772:
1.59 ! schwarze 773: ofadd(dform, file, name, dsec, sec, arch, &st);
1.50 kristaps 774: }
775:
776: /*
777: * See fileadd().
778: */
779: static int
780: filecheck(const char *name)
781: {
782: unsigned int index;
1.10 kristaps 783:
1.50 kristaps 784: index = ohash_qlookup(&filenames, name);
785: return(NULL != ohash_find(&filenames, index));
786: }
1.10 kristaps 787:
1.50 kristaps 788: /*
789: * Use the standard hashing mechanism (K&R) to see if the given filename
790: * already exists.
791: */
792: static void
793: fileadd(struct of *of)
794: {
795: unsigned int index;
1.1 kristaps 796:
1.50 kristaps 797: index = ohash_qlookup(&filenames, of->file);
798: assert(NULL == ohash_find(&filenames, index));
799: ohash_insert(&filenames, index, of);
800: }
1.3 kristaps 801:
1.50 kristaps 802: /*
803: * See inoadd().
804: */
805: static int
806: inocheck(const struct stat *st)
807: {
808: struct id id;
809: uint32_t hash;
810: unsigned int index;
811:
812: memset(&id, 0, sizeof(id));
813: id.ino = hash = st->st_ino;
814: id.dev = st->st_dev;
815: index = ohash_lookup_memory
816: (&inos, (char *)&id, sizeof(id), hash);
1.13 schwarze 817:
1.50 kristaps 818: return(NULL != ohash_find(&inos, index));
819: }
1.5 kristaps 820:
1.50 kristaps 821: /*
822: * The hashing function used here is quite simple: simply take the inode
823: * and use uint32_t of its bits.
824: * Then when we do the lookup, use both the inode and device identifier.
825: */
826: static void
827: inoadd(const struct stat *st, struct of *of)
828: {
829: uint32_t hash;
830: unsigned int index;
1.1 kristaps 831:
1.50 kristaps 832: of->id.ino = hash = st->st_ino;
833: of->id.dev = st->st_dev;
834: index = ohash_lookup_memory
835: (&inos, (char *)&of->id, sizeof(of->id), hash);
1.1 kristaps 836:
1.50 kristaps 837: assert(NULL == ohash_find(&inos, index));
838: ohash_insert(&inos, index, of);
839: }
1.35 kristaps 840:
1.50 kristaps 841: static void
1.59 ! schwarze 842: ofadd(int dform, const char *file, const char *name, const char *dsec,
! 843: const char *sec, const char *arch, const struct stat *st)
1.50 kristaps 844: {
845: struct of *of;
846: int sform;
847:
848: assert(NULL != file);
849:
850: if (NULL == name)
851: name = "";
852: if (NULL == sec)
853: sec = "";
854: if (NULL == dsec)
855: dsec = "";
856: if (NULL == arch)
857: arch = "";
858:
859: sform = FORM_NONE;
860: if (NULL != sec && *sec <= '9' && *sec >= '1')
861: sform = FORM_SRC;
862: else if (NULL != sec && *sec == '0') {
863: sec = dsec;
864: sform = FORM_CAT;
865: }
866:
867: of = mandoc_calloc(1, sizeof(struct of));
1.58 schwarze 868: strlcpy(of->file, file, PATH_MAX);
1.50 kristaps 869: of->name = name;
870: of->sec = sec;
871: of->dsec = dsec;
872: of->arch = arch;
873: of->sform = sform;
874: of->dform = dform;
875: of->next = ofs;
876: ofs = of;
1.3 kristaps 877:
1.50 kristaps 878: /*
879: * Add to unique identifier hash.
880: * Then if it's a source manual and we're going to use source in
881: * favour of catpages, add it to that hash.
882: */
883: inoadd(st, of);
884: fileadd(of);
885: }
1.3 kristaps 886:
1.50 kristaps 887: static void
888: offree(void)
889: {
890: struct of *of;
1.3 kristaps 891:
1.50 kristaps 892: while (NULL != (of = ofs)) {
893: ofs = of->next;
894: free(of);
895: }
896: }
1.38 schwarze 897:
1.50 kristaps 898: /*
899: * Run through the files in the global vector "ofs" and add them to the
1.59 ! schwarze 900: * database specified in "basedir".
1.50 kristaps 901: *
902: * This handles the parsing scheme itself, using the cues of directory
903: * and filename to determine whether the file is parsable or not.
904: */
1.59 ! schwarze 905: static void
! 906: ofmerge(struct mchars *mc, struct mparse *mp)
1.50 kristaps 907: {
908: int form;
909: size_t sz;
1.3 kristaps 910: struct mdoc *mdoc;
911: struct man *man;
1.58 schwarze 912: char buf[PATH_MAX];
1.50 kristaps 913: char *bufp;
914: const char *msec, *march, *mtitle, *cp;
915: struct of *of;
916: enum mandoclevel lvl;
917:
918: for (of = ofs; NULL != of; of = of->next) {
919: /*
920: * If we're a catpage (as defined by our path), then see
921: * if a manpage exists by the same name (ignoring the
922: * suffix).
923: * If it does, then we want to use it instead of our
924: * own.
925: */
926: if ( ! use_all && FORM_CAT == of->dform) {
1.58 schwarze 927: sz = strlcpy(buf, of->file, PATH_MAX);
928: if (sz >= PATH_MAX) {
1.56 schwarze 929: if (warnings)
1.59 ! schwarze 930: say(of->file, "Filename too long");
1.50 kristaps 931: continue;
932: }
933: bufp = strstr(buf, "cat");
934: assert(NULL != bufp);
935: memcpy(bufp, "man", 3);
936: if (NULL != (bufp = strrchr(buf, '.')))
937: *++bufp = '\0';
1.58 schwarze 938: strlcat(buf, of->dsec, PATH_MAX);
1.50 kristaps 939: if (filecheck(buf)) {
1.56 schwarze 940: if (warnings)
1.59 ! schwarze 941: say(of->file, "Man "
1.56 schwarze 942: "source exists: %s", buf);
1.50 kristaps 943: continue;
944: }
945: }
946:
947: words = NULL;
948: mparse_reset(mp);
949: mdoc = NULL;
950: man = NULL;
951: form = 0;
952: msec = of->dsec;
953: march = of->arch;
954: mtitle = of->name;
1.14 schwarze 955:
956: /*
1.33 schwarze 957: * Try interpreting the file as mdoc(7) or man(7)
958: * source code, unless it is already known to be
959: * formatted. Fall back to formatted mode.
1.14 schwarze 960: */
1.50 kristaps 961: if (FORM_SRC == of->dform || FORM_SRC == of->sform) {
962: lvl = mparse_readfd(mp, -1, of->file);
963: if (lvl < MANDOCLEVEL_FATAL)
964: mparse_result(mp, &mdoc, &man);
965: }
1.14 schwarze 966:
967: if (NULL != mdoc) {
1.50 kristaps 968: form = 1;
1.14 schwarze 969: msec = mdoc_meta(mdoc)->msec;
1.38 schwarze 970: march = mdoc_meta(mdoc)->arch;
1.14 schwarze 971: mtitle = mdoc_meta(mdoc)->title;
972: } else if (NULL != man) {
1.50 kristaps 973: form = 1;
1.14 schwarze 974: msec = man_meta(man)->msec;
1.38 schwarze 975: march = "";
1.14 schwarze 976: mtitle = man_meta(man)->title;
1.50 kristaps 977: }
978:
979: if (NULL == msec)
980: msec = "";
981: if (NULL == march)
982: march = "";
983: if (NULL == mtitle)
984: mtitle = "";
1.1 kristaps 985:
1.12 schwarze 986: /*
1.44 kristaps 987: * Check whether the manual section given in a file
988: * agrees with the directory where the file is located.
989: * Some manuals have suffixes like (3p) on their
990: * section number either inside the file or in the
991: * directory name, some are linked into more than one
992: * section, like encrypt(1) = makekey(8). Do not skip
993: * manuals for such reasons.
1.12 schwarze 994: */
1.56 schwarze 995: if (warnings && !use_all && form &&
996: strcasecmp(msec, of->dsec))
1.59 ! schwarze 997: say(of->file, "Section \"%s\" "
1.50 kristaps 998: "manual in %s directory",
999: msec, of->dsec);
1.12 schwarze 1000:
1.42 schwarze 1001: /*
1002: * Manual page directories exist for each kernel
1003: * architecture as returned by machine(1).
1004: * However, many manuals only depend on the
1005: * application architecture as returned by arch(1).
1006: * For example, some (2/ARM) manuals are shared
1007: * across the "armish" and "zaurus" kernel
1008: * architectures.
1009: * A few manuals are even shared across completely
1010: * different architectures, for example fdformat(1)
1011: * on amd64, i386, sparc, and sparc64.
1012: * Thus, warn about architecture mismatches,
1013: * but don't skip manuals for this reason.
1014: */
1.56 schwarze 1015: if (warnings && !use_all && strcasecmp(march, of->arch))
1.59 ! schwarze 1016: say(of->file, "Architecture \"%s\" "
1.47 schwarze 1017: "manual in \"%s\" directory",
1.45 kristaps 1018: march, of->arch);
1.12 schwarze 1019:
1.50 kristaps 1020: putkey(of, of->name, TYPE_Nm);
1.12 schwarze 1021:
1.50 kristaps 1022: if (NULL != mdoc) {
1023: if (NULL != (cp = mdoc_meta(mdoc)->name))
1024: putkey(of, cp, TYPE_Nm);
1025: parse_mdoc(of, mdoc_node(mdoc));
1026: } else if (NULL != man)
1027: parse_man(of, man_node(man));
1028: else
1.59 ! schwarze 1029: parse_catpage(of);
1.44 kristaps 1030:
1.59 ! schwarze 1031: dbindex(mc, form, of);
1.50 kristaps 1032: }
1033: }
1.12 schwarze 1034:
1.50 kristaps 1035: static void
1.59 ! schwarze 1036: parse_catpage(struct of *of)
1.50 kristaps 1037: {
1038: FILE *stream;
1039: char *line, *p, *title;
1040: size_t len, plen, titlesz;
1.12 schwarze 1041:
1.50 kristaps 1042: if (NULL == (stream = fopen(of->file, "r"))) {
1.56 schwarze 1043: if (warnings)
1.59 ! schwarze 1044: say(of->file, NULL);
1.50 kristaps 1045: return;
1046: }
1.1 kristaps 1047:
1.50 kristaps 1048: /* Skip to first blank line. */
1.1 kristaps 1049:
1.50 kristaps 1050: while (NULL != (line = fgetln(stream, &len)))
1051: if ('\n' == *line)
1052: break;
1.1 kristaps 1053:
1.50 kristaps 1054: /*
1055: * Assume the first line that is not indented
1056: * is the first section header. Skip to it.
1057: */
1.1 kristaps 1058:
1.50 kristaps 1059: while (NULL != (line = fgetln(stream, &len)))
1060: if ('\n' != *line && ' ' != *line)
1061: break;
1062:
1063: /*
1064: * Read up until the next section into a buffer.
1065: * Strip the leading and trailing newline from each read line,
1066: * appending a trailing space.
1067: * Ignore empty (whitespace-only) lines.
1068: */
1.1 kristaps 1069:
1.50 kristaps 1070: titlesz = 0;
1071: title = NULL;
1.38 schwarze 1072:
1.50 kristaps 1073: while (NULL != (line = fgetln(stream, &len))) {
1074: if (' ' != *line || '\n' != line[len - 1])
1075: break;
1076: while (len > 0 && isspace((unsigned char)*line)) {
1077: line++;
1078: len--;
1079: }
1080: if (1 == len)
1.38 schwarze 1081: continue;
1.50 kristaps 1082: title = mandoc_realloc(title, titlesz + len);
1083: memcpy(title + titlesz, line, len);
1084: titlesz += len;
1085: title[titlesz - 1] = ' ';
1086: }
1.38 schwarze 1087:
1.50 kristaps 1088: /*
1089: * If no page content can be found, or the input line
1090: * is already the next section header, or there is no
1091: * trailing newline, reuse the page title as the page
1092: * description.
1093: */
1.44 kristaps 1094:
1.50 kristaps 1095: if (NULL == title || '\0' == *title) {
1.56 schwarze 1096: if (warnings)
1.59 ! schwarze 1097: say(of->file, "Cannot find NAME section");
1.50 kristaps 1098: fclose(stream);
1099: free(title);
1100: return;
1101: }
1.1 kristaps 1102:
1.50 kristaps 1103: title = mandoc_realloc(title, titlesz + 1);
1104: title[titlesz] = '\0';
1.33 schwarze 1105:
1.50 kristaps 1106: /*
1107: * Skip to the first dash.
1108: * Use the remaining line as the description (no more than 70
1109: * bytes).
1110: */
1.33 schwarze 1111:
1.50 kristaps 1112: if (NULL != (p = strstr(title, "- "))) {
1113: for (p += 2; ' ' == *p || '\b' == *p; p++)
1114: /* Skip to next word. */ ;
1115: } else {
1.56 schwarze 1116: if (warnings)
1.59 ! schwarze 1117: say(of->file, "No dash in title line");
1.50 kristaps 1118: p = title;
1119: }
1.38 schwarze 1120:
1.50 kristaps 1121: plen = strlen(p);
1.1 kristaps 1122:
1.50 kristaps 1123: /* Strip backspace-encoding from line. */
1.1 kristaps 1124:
1.50 kristaps 1125: while (NULL != (line = memchr(p, '\b', plen))) {
1126: len = line - p;
1127: if (0 == len) {
1128: memmove(line, line + 1, plen--);
1129: continue;
1130: }
1131: memmove(line - 1, line + 1, plen - len);
1132: plen -= 2;
1133: }
1.1 kristaps 1134:
1.50 kristaps 1135: of->desc = stradd(p);
1136: putkey(of, p, TYPE_Nd);
1137: fclose(stream);
1138: free(title);
1139: }
1.1 kristaps 1140:
1.50 kristaps 1141: /*
1142: * Put a type/word pair into the word database for this particular file.
1143: */
1144: static void
1145: putkey(const struct of *of, const char *value, uint64_t type)
1146: {
1.18 kristaps 1147:
1.50 kristaps 1148: assert(NULL != value);
1149: wordaddbuf(of, value, strlen(value), type);
1150: }
1.44 kristaps 1151:
1.50 kristaps 1152: /*
1153: * Like putkey() but for unterminated strings.
1154: */
1155: static void
1156: putkeys(const struct of *of, const char *value, int sz, uint64_t type)
1157: {
1.44 kristaps 1158:
1.50 kristaps 1159: wordaddbuf(of, value, sz, type);
1.3 kristaps 1160: }
1161:
1162: /*
1.50 kristaps 1163: * Grok all nodes at or below a certain mdoc node into putkey().
1.3 kristaps 1164: */
1165: static void
1.50 kristaps 1166: putmdockey(const struct of *of, const struct mdoc_node *n, uint64_t m)
1.3 kristaps 1167: {
1.18 kristaps 1168:
1.50 kristaps 1169: for ( ; NULL != n; n = n->next) {
1170: if (NULL != n->child)
1171: putmdockey(of, n->child, m);
1172: if (MDOC_TEXT == n->type)
1173: putkey(of, n->string, m);
1174: }
1175: }
1.18 kristaps 1176:
1.50 kristaps 1177: static int
1178: parse_man(struct of *of, const struct man_node *n)
1179: {
1180: const struct man_node *head, *body;
1181: char *start, *sv, *title;
1182: char byte;
1183: size_t sz, titlesz;
1.18 kristaps 1184:
1.50 kristaps 1185: if (NULL == n)
1186: return(0);
1.18 kristaps 1187:
1.50 kristaps 1188: /*
1189: * We're only searching for one thing: the first text child in
1190: * the BODY of a NAME section. Since we don't keep track of
1191: * sections in -man, run some hoops to find out whether we're in
1192: * the correct section or not.
1193: */
1.18 kristaps 1194:
1.50 kristaps 1195: if (MAN_BODY == n->type && MAN_SH == n->tok) {
1196: body = n;
1197: assert(body->parent);
1198: if (NULL != (head = body->parent->head) &&
1199: 1 == head->nchild &&
1200: NULL != (head = (head->child)) &&
1201: MAN_TEXT == head->type &&
1202: 0 == strcmp(head->string, "NAME") &&
1203: NULL != (body = body->child) &&
1204: MAN_TEXT == body->type) {
1.3 kristaps 1205:
1.50 kristaps 1206: title = NULL;
1207: titlesz = 0;
1.3 kristaps 1208:
1.50 kristaps 1209: /*
1210: * Suck the entire NAME section into memory.
1211: * Yes, we might run away.
1212: * But too many manuals have big, spread-out
1213: * NAME sections over many lines.
1214: */
1.3 kristaps 1215:
1.50 kristaps 1216: for ( ; NULL != body; body = body->next) {
1217: if (MAN_TEXT != body->type)
1218: break;
1219: if (0 == (sz = strlen(body->string)))
1220: continue;
1221: title = mandoc_realloc
1222: (title, titlesz + sz + 1);
1223: memcpy(title + titlesz, body->string, sz);
1224: titlesz += sz + 1;
1225: title[titlesz - 1] = ' ';
1226: }
1227: if (NULL == title)
1228: return(1);
1.18 kristaps 1229:
1.50 kristaps 1230: title = mandoc_realloc(title, titlesz + 1);
1231: title[titlesz] = '\0';
1.18 kristaps 1232:
1.50 kristaps 1233: /* Skip leading space. */
1.18 kristaps 1234:
1.50 kristaps 1235: sv = title;
1236: while (isspace((unsigned char)*sv))
1237: sv++;
1.18 kristaps 1238:
1.50 kristaps 1239: if (0 == (sz = strlen(sv))) {
1240: free(title);
1241: return(1);
1242: }
1.1 kristaps 1243:
1.50 kristaps 1244: /* Erase trailing space. */
1.1 kristaps 1245:
1.50 kristaps 1246: start = &sv[sz - 1];
1247: while (start > sv && isspace((unsigned char)*start))
1248: *start-- = '\0';
1.1 kristaps 1249:
1.50 kristaps 1250: if (start == sv) {
1251: free(title);
1252: return(1);
1253: }
1.1 kristaps 1254:
1.50 kristaps 1255: start = sv;
1.18 kristaps 1256:
1.50 kristaps 1257: /*
1258: * Go through a special heuristic dance here.
1259: * Conventionally, one or more manual names are
1260: * comma-specified prior to a whitespace, then a
1261: * dash, then a description. Try to puzzle out
1262: * the name parts here.
1263: */
1.18 kristaps 1264:
1.50 kristaps 1265: for ( ;; ) {
1266: sz = strcspn(start, " ,");
1267: if ('\0' == start[sz])
1268: break;
1.1 kristaps 1269:
1.50 kristaps 1270: byte = start[sz];
1271: start[sz] = '\0';
1.1 kristaps 1272:
1.50 kristaps 1273: putkey(of, start, TYPE_Nm);
1.1 kristaps 1274:
1.50 kristaps 1275: if (' ' == byte) {
1276: start += sz + 1;
1277: break;
1278: }
1.1 kristaps 1279:
1.50 kristaps 1280: assert(',' == byte);
1281: start += sz + 1;
1282: while (' ' == *start)
1283: start++;
1284: }
1.1 kristaps 1285:
1.50 kristaps 1286: if (sv == start) {
1287: putkey(of, start, TYPE_Nm);
1288: free(title);
1289: return(1);
1290: }
1.1 kristaps 1291:
1.50 kristaps 1292: while (isspace((unsigned char)*start))
1293: start++;
1.1 kristaps 1294:
1.50 kristaps 1295: if (0 == strncmp(start, "-", 1))
1296: start += 1;
1297: else if (0 == strncmp(start, "\\-\\-", 4))
1298: start += 4;
1299: else if (0 == strncmp(start, "\\-", 2))
1300: start += 2;
1301: else if (0 == strncmp(start, "\\(en", 4))
1302: start += 4;
1303: else if (0 == strncmp(start, "\\(em", 4))
1304: start += 4;
1.1 kristaps 1305:
1.50 kristaps 1306: while (' ' == *start)
1307: start++;
1.1 kristaps 1308:
1.50 kristaps 1309: assert(NULL == of->desc);
1310: of->desc = stradd(start);
1311: putkey(of, start, TYPE_Nd);
1312: free(title);
1313: return(1);
1314: }
1315: }
1.1 kristaps 1316:
1.50 kristaps 1317: for (n = n->child; n; n = n->next)
1318: if (parse_man(of, n))
1319: return(1);
1.1 kristaps 1320:
1.50 kristaps 1321: return(0);
1.1 kristaps 1322: }
1323:
1324: static void
1.50 kristaps 1325: parse_mdoc(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1326: {
1327:
1.50 kristaps 1328: assert(NULL != n);
1329: for (n = n->child; NULL != n; n = n->next) {
1330: switch (n->type) {
1331: case (MDOC_ELEM):
1332: /* FALLTHROUGH */
1333: case (MDOC_BLOCK):
1334: /* FALLTHROUGH */
1335: case (MDOC_HEAD):
1336: /* FALLTHROUGH */
1337: case (MDOC_BODY):
1338: /* FALLTHROUGH */
1339: case (MDOC_TAIL):
1340: if (NULL != mdocs[n->tok].fp)
1341: if (0 == (*mdocs[n->tok].fp)(of, n))
1342: break;
1.1 kristaps 1343:
1.50 kristaps 1344: if (MDOCF_CHILD & mdocs[n->tok].flags)
1345: putmdockey(of, n->child, mdocs[n->tok].mask);
1346: break;
1347: default:
1348: assert(MDOC_ROOT != n->type);
1349: continue;
1350: }
1351: if (NULL != n->child)
1352: parse_mdoc(of, n);
1.1 kristaps 1353: }
1354: }
1355:
1.25 schwarze 1356: static int
1.50 kristaps 1357: parse_mdoc_Fd(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1358: {
1359: const char *start, *end;
1360: size_t sz;
1.25 schwarze 1361:
1.50 kristaps 1362: if (SEC_SYNOPSIS != n->sec ||
1363: NULL == (n = n->child) ||
1364: MDOC_TEXT != n->type)
1.25 schwarze 1365: return(0);
1.1 kristaps 1366:
1367: /*
1368: * Only consider those `Fd' macro fields that begin with an
1369: * "inclusion" token (versus, e.g., #define).
1370: */
1.50 kristaps 1371:
1.1 kristaps 1372: if (strcmp("#include", n->string))
1.25 schwarze 1373: return(0);
1.1 kristaps 1374:
1375: if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1.25 schwarze 1376: return(0);
1.1 kristaps 1377:
1378: /*
1379: * Strip away the enclosing angle brackets and make sure we're
1380: * not zero-length.
1381: */
1382:
1383: start = n->string;
1384: if ('<' == *start || '"' == *start)
1385: start++;
1386:
1387: if (0 == (sz = strlen(start)))
1.25 schwarze 1388: return(0);
1.1 kristaps 1389:
1390: end = &start[(int)sz - 1];
1391: if ('>' == *end || '"' == *end)
1392: end--;
1393:
1.50 kristaps 1394: if (end > start)
1395: putkeys(of, start, end - start + 1, TYPE_In);
1.25 schwarze 1396: return(1);
1.1 kristaps 1397: }
1398:
1.25 schwarze 1399: static int
1.50 kristaps 1400: parse_mdoc_In(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1401: {
1402:
1.50 kristaps 1403: if (NULL != n->child && MDOC_TEXT == n->child->type)
1.25 schwarze 1404: return(0);
1.1 kristaps 1405:
1.50 kristaps 1406: putkey(of, n->child->string, TYPE_In);
1.25 schwarze 1407: return(1);
1.1 kristaps 1408: }
1409:
1.25 schwarze 1410: static int
1.50 kristaps 1411: parse_mdoc_Fn(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1412: {
1413: const char *cp;
1414:
1.50 kristaps 1415: if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1.25 schwarze 1416: return(0);
1417:
1.50 kristaps 1418: /*
1419: * Parse: .Fn "struct type *name" "char *arg".
1420: * First strip away pointer symbol.
1421: * Then store the function name, then type.
1422: * Finally, store the arguments.
1423: */
1.1 kristaps 1424:
1.50 kristaps 1425: if (NULL == (cp = strrchr(n->string, ' ')))
1426: cp = n->string;
1.1 kristaps 1427:
1428: while ('*' == *cp)
1429: cp++;
1430:
1.50 kristaps 1431: putkey(of, cp, TYPE_Fn);
1.25 schwarze 1432:
1.50 kristaps 1433: if (n->string < cp)
1434: putkeys(of, n->string, cp - n->string, TYPE_Ft);
1.25 schwarze 1435:
1.50 kristaps 1436: for (n = n->next; NULL != n; n = n->next)
1437: if (MDOC_TEXT == n->type)
1438: putkey(of, n->string, TYPE_Fa);
1.25 schwarze 1439:
1440: return(0);
1.1 kristaps 1441: }
1442:
1.25 schwarze 1443: static int
1.50 kristaps 1444: parse_mdoc_St(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1445: {
1.25 schwarze 1446:
1.1 kristaps 1447: if (NULL == n->child || MDOC_TEXT != n->child->type)
1.25 schwarze 1448: return(0);
1.1 kristaps 1449:
1.50 kristaps 1450: putkey(of, n->child->string, TYPE_St);
1.25 schwarze 1451: return(1);
1.1 kristaps 1452: }
1453:
1.25 schwarze 1454: static int
1.50 kristaps 1455: parse_mdoc_Xr(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1456: {
1457:
1458: if (NULL == (n = n->child))
1.25 schwarze 1459: return(0);
1.1 kristaps 1460:
1.50 kristaps 1461: putkey(of, n->string, TYPE_Xr);
1.25 schwarze 1462: return(1);
1.1 kristaps 1463: }
1464:
1.25 schwarze 1465: static int
1.50 kristaps 1466: parse_mdoc_Nd(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1467: {
1.50 kristaps 1468: size_t sz;
1469: char *sv, *desc;
1.1 kristaps 1470:
1471: if (MDOC_BODY != n->type)
1.25 schwarze 1472: return(0);
1.1 kristaps 1473:
1.50 kristaps 1474: /*
1475: * Special-case the `Nd' because we need to put the description
1476: * into the document table.
1477: */
1478:
1479: desc = NULL;
1480: for (n = n->child; NULL != n; n = n->next) {
1481: if (MDOC_TEXT == n->type) {
1482: sz = strlen(n->string) + 1;
1483: if (NULL != (sv = desc))
1484: sz += strlen(desc) + 1;
1485: desc = mandoc_realloc(desc, sz);
1486: if (NULL != sv)
1487: strlcat(desc, " ", sz);
1488: else
1489: *desc = '\0';
1490: strlcat(desc, n->string, sz);
1491: }
1492: if (NULL != n->child)
1493: parse_mdoc_Nd(of, n);
1494: }
1495:
1496: of->desc = NULL != desc ? stradd(desc) : NULL;
1497: free(desc);
1.25 schwarze 1498: return(1);
1.1 kristaps 1499: }
1500:
1.25 schwarze 1501: static int
1.50 kristaps 1502: parse_mdoc_Nm(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1503: {
1504:
1.25 schwarze 1505: if (SEC_NAME == n->sec)
1506: return(1);
1507: else if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type)
1508: return(0);
1.1 kristaps 1509:
1.25 schwarze 1510: return(1);
1.1 kristaps 1511: }
1512:
1.25 schwarze 1513: static int
1.50 kristaps 1514: parse_mdoc_Sh(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1515: {
1516:
1.25 schwarze 1517: return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1.1 kristaps 1518: }
1519:
1.50 kristaps 1520: static int
1521: parse_mdoc_head(struct of *of, const struct mdoc_node *n)
1.1 kristaps 1522: {
1523:
1.50 kristaps 1524: return(MDOC_HEAD == n->type);
1525: }
1.1 kristaps 1526:
1.50 kristaps 1527: static int
1528: parse_mdoc_body(struct of *of, const struct mdoc_node *n)
1529: {
1.1 kristaps 1530:
1.50 kristaps 1531: return(MDOC_BODY == n->type);
1.1 kristaps 1532: }
1533:
1.50 kristaps 1534: /*
1535: * See straddbuf().
1536: */
1537: static char *
1538: stradd(const char *cp)
1.1 kristaps 1539: {
1540:
1.50 kristaps 1541: return(straddbuf(cp, strlen(cp)));
1.1 kristaps 1542: }
1543:
1544: /*
1.50 kristaps 1545: * This looks up or adds a string to the string table.
1546: * The string table is a table of all strings encountered during parse
1547: * or file scan.
1548: * In using it, we avoid having thousands of (e.g.) "cat1" string
1549: * allocations for the "of" table.
1550: * We also have a layer atop the string table for keeping track of words
1551: * in a parse sequence (see wordaddbuf()).
1.1 kristaps 1552: */
1.50 kristaps 1553: static char *
1554: straddbuf(const char *cp, size_t sz)
1.1 kristaps 1555: {
1.50 kristaps 1556: struct str *s;
1557: unsigned int index;
1558: const char *end;
1559:
1560: if (NULL != (s = hashget(cp, sz)))
1561: return(s->key);
1.1 kristaps 1562:
1.51 kristaps 1563: s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
1.50 kristaps 1564: memcpy(s->key, cp, sz);
1.1 kristaps 1565:
1.50 kristaps 1566: end = cp + sz;
1567: index = ohash_qlookupi(&strings, cp, &end);
1568: assert(NULL == ohash_find(&strings, index));
1569: ohash_insert(&strings, index, s);
1570: return(s->key);
1571: }
1.25 schwarze 1572:
1.50 kristaps 1573: static struct str *
1574: hashget(const char *cp, size_t sz)
1575: {
1576: unsigned int index;
1577: const char *end;
1.25 schwarze 1578:
1.50 kristaps 1579: end = cp + sz;
1580: index = ohash_qlookupi(&strings, cp, &end);
1581: return(ohash_find(&strings, index));
1582: }
1.1 kristaps 1583:
1.50 kristaps 1584: /*
1585: * Add a word to the current parse sequence.
1586: * Within the hashtable of strings, we maintain a list of strings that
1587: * are currently indexed.
1588: * Each of these ("words") has a bitmask modified within the parse.
1589: * When we finish a parse, we'll dump the list, then remove the head
1590: * entry -- since the next parse will have a new "of", it can keep track
1591: * of its entries without conflict.
1592: */
1593: static void
1594: wordaddbuf(const struct of *of,
1595: const char *cp, size_t sz, uint64_t v)
1596: {
1597: struct str *s;
1598: unsigned int index;
1599: const char *end;
1.25 schwarze 1600:
1.50 kristaps 1601: if (0 == sz)
1602: return;
1.25 schwarze 1603:
1.50 kristaps 1604: s = hashget(cp, sz);
1.25 schwarze 1605:
1.50 kristaps 1606: if (NULL != s && of == s->of) {
1607: s->mask |= v;
1608: return;
1609: } else if (NULL == s) {
1.51 kristaps 1610: s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
1.50 kristaps 1611: memcpy(s->key, cp, sz);
1612: end = cp + sz;
1613: index = ohash_qlookupi(&strings, cp, &end);
1614: assert(NULL == ohash_find(&strings, index));
1615: ohash_insert(&strings, index, s);
1.1 kristaps 1616: }
1617:
1.50 kristaps 1618: s->next = words;
1619: s->of = of;
1620: s->mask = v;
1621: words = s;
1.1 kristaps 1622: }
1623:
1.50 kristaps 1624: /*
1625: * Take a Unicode codepoint and produce its UTF-8 encoding.
1626: * This isn't the best way to do this, but it works.
1627: * The magic numbers are from the UTF-8 packaging.
1628: * They're not as scary as they seem: read the UTF-8 spec for details.
1629: */
1630: static size_t
1631: utf8(unsigned int cp, char out[7])
1.1 kristaps 1632: {
1.50 kristaps 1633: size_t rc;
1.1 kristaps 1634:
1.50 kristaps 1635: rc = 0;
1636: if (cp <= 0x0000007F) {
1637: rc = 1;
1638: out[0] = (char)cp;
1639: } else if (cp <= 0x000007FF) {
1640: rc = 2;
1641: out[0] = (cp >> 6 & 31) | 192;
1642: out[1] = (cp & 63) | 128;
1643: } else if (cp <= 0x0000FFFF) {
1644: rc = 3;
1645: out[0] = (cp >> 12 & 15) | 224;
1646: out[1] = (cp >> 6 & 63) | 128;
1647: out[2] = (cp & 63) | 128;
1648: } else if (cp <= 0x001FFFFF) {
1649: rc = 4;
1650: out[0] = (cp >> 18 & 7) | 240;
1651: out[1] = (cp >> 12 & 63) | 128;
1652: out[2] = (cp >> 6 & 63) | 128;
1653: out[3] = (cp & 63) | 128;
1654: } else if (cp <= 0x03FFFFFF) {
1655: rc = 5;
1656: out[0] = (cp >> 24 & 3) | 248;
1657: out[1] = (cp >> 18 & 63) | 128;
1658: out[2] = (cp >> 12 & 63) | 128;
1659: out[3] = (cp >> 6 & 63) | 128;
1660: out[4] = (cp & 63) | 128;
1661: } else if (cp <= 0x7FFFFFFF) {
1662: rc = 6;
1663: out[0] = (cp >> 30 & 1) | 252;
1664: out[1] = (cp >> 24 & 63) | 128;
1665: out[2] = (cp >> 18 & 63) | 128;
1666: out[3] = (cp >> 12 & 63) | 128;
1667: out[4] = (cp >> 6 & 63) | 128;
1668: out[5] = (cp & 63) | 128;
1669: } else
1.1 kristaps 1670: return(0);
1671:
1.50 kristaps 1672: out[rc] = '\0';
1673: return(rc);
1674: }
1.1 kristaps 1675:
1.50 kristaps 1676: /*
1677: * Store the UTF-8 version of a key, or alias the pointer if the key has
1678: * no UTF-8 transcription marks in it.
1679: */
1680: static void
1681: utf8key(struct mchars *mc, struct str *key)
1682: {
1683: size_t sz, bsz, pos;
1684: char utfbuf[7], res[5];
1685: char *buf;
1686: const char *seq, *cpp, *val;
1687: int len, u;
1688: enum mandoc_esc esc;
1689:
1690: assert(NULL == key->utf8);
1691:
1692: res[0] = '\\';
1693: res[1] = '\t';
1694: res[2] = ASCII_NBRSP;
1695: res[3] = ASCII_HYPH;
1696: res[4] = '\0';
1.1 kristaps 1697:
1.50 kristaps 1698: val = key->key;
1699: bsz = strlen(val);
1.46 kristaps 1700:
1.50 kristaps 1701: /*
1702: * Pre-check: if we have no stop-characters, then set the
1703: * pointer as ourselvse and get out of here.
1704: */
1705: if (strcspn(val, res) == bsz) {
1706: key->utf8 = key->key;
1707: return;
1708: }
1.46 kristaps 1709:
1.50 kristaps 1710: /* Pre-allocate by the length of the input */
1.46 kristaps 1711:
1.50 kristaps 1712: buf = mandoc_malloc(++bsz);
1713: pos = 0;
1.46 kristaps 1714:
1.50 kristaps 1715: while ('\0' != *val) {
1716: /*
1717: * Halt on the first escape sequence.
1718: * This also halts on the end of string, in which case
1719: * we just copy, fallthrough, and exit the loop.
1720: */
1721: if ((sz = strcspn(val, res)) > 0) {
1722: memcpy(&buf[pos], val, sz);
1723: pos += sz;
1724: val += sz;
1725: }
1.46 kristaps 1726:
1.50 kristaps 1727: if (ASCII_HYPH == *val) {
1728: buf[pos++] = '-';
1729: val++;
1730: continue;
1731: } else if ('\t' == *val || ASCII_NBRSP == *val) {
1732: buf[pos++] = ' ';
1733: val++;
1734: continue;
1735: } else if ('\\' != *val)
1736: break;
1.46 kristaps 1737:
1.50 kristaps 1738: /* Read past the slash. */
1.46 kristaps 1739:
1.50 kristaps 1740: val++;
1741: u = 0;
1.46 kristaps 1742:
1.50 kristaps 1743: /*
1744: * Parse the escape sequence and see if it's a
1745: * predefined character or special character.
1746: */
1747: esc = mandoc_escape
1748: ((const char **)&val, &seq, &len);
1749: if (ESCAPE_ERROR == esc)
1750: break;
1.1 kristaps 1751:
1.50 kristaps 1752: if (ESCAPE_SPECIAL != esc)
1753: continue;
1754: if (0 == (u = mchars_spec2cp(mc, seq, len)))
1755: continue;
1.1 kristaps 1756:
1.50 kristaps 1757: /*
1758: * If we have a Unicode codepoint, try to convert that
1759: * to a UTF-8 byte string.
1760: */
1761: cpp = utfbuf;
1762: if (0 == (sz = utf8(u, utfbuf)))
1763: continue;
1.1 kristaps 1764:
1.50 kristaps 1765: /* Copy the rendered glyph into the stream. */
1.1 kristaps 1766:
1.50 kristaps 1767: sz = strlen(cpp);
1768: bsz += sz;
1.1 kristaps 1769:
1.50 kristaps 1770: buf = mandoc_realloc(buf, bsz);
1.1 kristaps 1771:
1.50 kristaps 1772: memcpy(&buf[pos], cpp, sz);
1773: pos += sz;
1.1 kristaps 1774: }
1775:
1.50 kristaps 1776: buf[pos] = '\0';
1777: key->utf8 = buf;
1.1 kristaps 1778: }
1779:
1.14 schwarze 1780: /*
1.50 kristaps 1781: * Flush the current page's terms (and their bits) into the database.
1782: * Wrap the entire set of additions in a transaction to make sqlite be a
1783: * little faster.
1784: * Also, UTF-8-encode the description at the last possible moment.
1.14 schwarze 1785: */
1786: static void
1.59 ! schwarze 1787: dbindex(struct mchars *mc, int form, const struct of *of)
1.14 schwarze 1788: {
1.50 kristaps 1789: struct str *key;
1790: const char *desc;
1791: int64_t recno;
1.52 kristaps 1792: size_t i;
1.14 schwarze 1793:
1.56 schwarze 1794: if (verb)
1.59 ! schwarze 1795: say(of->file, "Adding to index");
1.43 kristaps 1796:
1.50 kristaps 1797: if (nodb)
1.14 schwarze 1798: return;
1.28 kristaps 1799:
1.50 kristaps 1800: desc = "";
1801: if (NULL != of->desc) {
1802: key = hashget(of->desc, strlen(of->desc));
1803: assert(NULL != key);
1804: if (NULL == key->utf8)
1805: utf8key(mc, key);
1806: desc = key->utf8;
1807: }
1808:
1.52 kristaps 1809: SQL_EXEC("BEGIN TRANSACTION");
1.50 kristaps 1810:
1.52 kristaps 1811: i = 1;
1812: SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, of->file);
1813: SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, of->sec);
1814: SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, of->arch);
1815: SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, desc);
1816: SQL_BIND_INT(stmts[STMT_INSERT_DOC], i, form);
1817: SQL_STEP(stmts[STMT_INSERT_DOC]);
1.50 kristaps 1818: recno = sqlite3_last_insert_rowid(db);
1819: sqlite3_reset(stmts[STMT_INSERT_DOC]);
1820:
1821: for (key = words; NULL != key; key = key->next) {
1822: assert(key->of == of);
1823: if (NULL == key->utf8)
1824: utf8key(mc, key);
1.52 kristaps 1825: i = 1;
1826: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1827: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->utf8);
1828: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, recno);
1829: SQL_STEP(stmts[STMT_INSERT_KEY]);
1.50 kristaps 1830: sqlite3_reset(stmts[STMT_INSERT_KEY]);
1.38 schwarze 1831: }
1.28 kristaps 1832:
1.52 kristaps 1833: SQL_EXEC("END TRANSACTION");
1.14 schwarze 1834: }
1835:
1.5 kristaps 1836: static void
1.59 ! schwarze 1837: dbprune(void)
1.5 kristaps 1838: {
1.50 kristaps 1839: struct of *of;
1.52 kristaps 1840: size_t i;
1.5 kristaps 1841:
1.50 kristaps 1842: if (nodb)
1843: return;
1.12 schwarze 1844:
1.50 kristaps 1845: for (of = ofs; NULL != of; of = of->next) {
1.52 kristaps 1846: i = 1;
1847: SQL_BIND_TEXT(stmts[STMT_DELETE], i, of->file);
1848: SQL_STEP(stmts[STMT_DELETE]);
1.50 kristaps 1849: sqlite3_reset(stmts[STMT_DELETE]);
1.56 schwarze 1850: if (verb)
1.59 ! schwarze 1851: say(of->file, "Deleted from index");
1.5 kristaps 1852: }
1853: }
1854:
1.4 kristaps 1855: /*
1.50 kristaps 1856: * Close an existing database and its prepared statements.
1857: * If "real" is not set, rename the temporary file into the real one.
1.4 kristaps 1858: */
1.35 kristaps 1859: static void
1.59 ! schwarze 1860: dbclose(int real)
1.4 kristaps 1861: {
1.50 kristaps 1862: size_t i;
1.58 schwarze 1863: char file[PATH_MAX];
1.4 kristaps 1864:
1.50 kristaps 1865: if (nodb)
1.38 schwarze 1866: return;
1.50 kristaps 1867:
1868: for (i = 0; i < STMT__MAX; i++) {
1869: sqlite3_finalize(stmts[i]);
1870: stmts[i] = NULL;
1.4 kristaps 1871: }
1872:
1.50 kristaps 1873: sqlite3_close(db);
1874: db = NULL;
1.12 schwarze 1875:
1.50 kristaps 1876: if (real)
1877: return;
1.12 schwarze 1878:
1.58 schwarze 1879: strlcpy(file, MANDOC_DB, PATH_MAX);
1880: strlcat(file, "~", PATH_MAX);
1.59 ! schwarze 1881: if (-1 == rename(file, MANDOC_DB)) {
! 1882: exitcode = (int)MANDOCLEVEL_SYSERR;
! 1883: say(MANDOC_DB, NULL);
! 1884: }
1.50 kristaps 1885: }
1.14 schwarze 1886:
1.50 kristaps 1887: /*
1888: * This is straightforward stuff.
1889: * Open a database connection to a "temporary" database, then open a set
1890: * of prepared statements we'll use over and over again.
1891: * If "real" is set, we use the existing database; if not, we truncate a
1892: * temporary one.
1893: * Must be matched by dbclose().
1894: */
1895: static int
1.59 ! schwarze 1896: dbopen(int real)
1.50 kristaps 1897: {
1.58 schwarze 1898: char file[PATH_MAX];
1.50 kristaps 1899: const char *sql;
1900: int rc, ofl;
1901: size_t sz;
1.12 schwarze 1902:
1.50 kristaps 1903: if (nodb)
1904: return(1);
1.12 schwarze 1905:
1.58 schwarze 1906: sz = strlcpy(file, MANDOC_DB, PATH_MAX);
1.50 kristaps 1907: if ( ! real)
1.58 schwarze 1908: sz = strlcat(file, "~", PATH_MAX);
1.5 kristaps 1909:
1.58 schwarze 1910: if (sz >= PATH_MAX) {
1.50 kristaps 1911: fprintf(stderr, "%s: Path too long\n", file);
1912: return(0);
1913: }
1.5 kristaps 1914:
1.50 kristaps 1915: if ( ! real)
1916: remove(file);
1.38 schwarze 1917:
1.52 kristaps 1918: ofl = SQLITE_OPEN_READWRITE |
1.50 kristaps 1919: (0 == real ? SQLITE_OPEN_EXCLUSIVE : 0);
1.45 kristaps 1920:
1.50 kristaps 1921: rc = sqlite3_open_v2(file, &db, ofl, NULL);
1922: if (SQLITE_OK == rc)
1.57 schwarze 1923: goto prepare_statements;
1.50 kristaps 1924: if (SQLITE_CANTOPEN != rc) {
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: sqlite3_close(db);
1931: db = NULL;
1.12 schwarze 1932:
1.50 kristaps 1933: if (SQLITE_OK != (rc = sqlite3_open(file, &db))) {
1.59 ! schwarze 1934: exitcode = (int)MANDOCLEVEL_SYSERR;
! 1935: say(file, NULL);
1.50 kristaps 1936: return(0);
1937: }
1.12 schwarze 1938:
1.50 kristaps 1939: sql = "CREATE TABLE \"docs\" (\n"
1940: " \"file\" TEXT NOT NULL,\n"
1941: " \"sec\" TEXT NOT NULL,\n"
1942: " \"arch\" TEXT NOT NULL,\n"
1943: " \"desc\" TEXT NOT NULL,\n"
1944: " \"form\" INTEGER NOT NULL,\n"
1945: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1946: ");\n"
1947: "\n"
1948: "CREATE TABLE \"keys\" (\n"
1949: " \"bits\" INTEGER NOT NULL,\n"
1950: " \"key\" TEXT NOT NULL,\n"
1951: " \"docid\" INTEGER NOT NULL REFERENCES docs(id) "
1952: "ON DELETE CASCADE,\n"
1953: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1954: ");\n"
1955: "\n"
1956: "CREATE INDEX \"key_index\" ON keys (key);\n";
1.14 schwarze 1957:
1.50 kristaps 1958: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
1.59 ! schwarze 1959: exitcode = (int)MANDOCLEVEL_SYSERR;
! 1960: say(file, "%s", sqlite3_errmsg(db));
1.50 kristaps 1961: return(0);
1962: }
1.4 kristaps 1963:
1.57 schwarze 1964: prepare_statements:
1965: SQL_EXEC("PRAGMA foreign_keys = ON");
1.50 kristaps 1966: sql = "DELETE FROM docs where file=?";
1967: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE], NULL);
1968: sql = "INSERT INTO docs "
1969: "(file,sec,arch,desc,form) VALUES (?,?,?,?,?)";
1970: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_DOC], NULL);
1971: sql = "INSERT INTO keys "
1972: "(bits,key,docid) VALUES (?,?,?)";
1973: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1974: return(1);
1975: }
1.5 kristaps 1976:
1.50 kristaps 1977: static void *
1978: hash_halloc(size_t sz, void *arg)
1979: {
1.12 schwarze 1980:
1.50 kristaps 1981: return(mandoc_calloc(sz, 1));
1982: }
1.12 schwarze 1983:
1.50 kristaps 1984: static void *
1985: hash_alloc(size_t sz, void *arg)
1986: {
1.5 kristaps 1987:
1.50 kristaps 1988: return(mandoc_malloc(sz));
1989: }
1.41 kristaps 1990:
1.50 kristaps 1991: static void
1992: hash_free(void *p, size_t sz, void *arg)
1993: {
1.4 kristaps 1994:
1.50 kristaps 1995: free(p);
1.4 kristaps 1996: }
1997:
1.50 kristaps 1998: static int
1.59 ! schwarze 1999: set_basedir(const char *targetdir)
1.4 kristaps 2000: {
1.59 ! schwarze 2001: static char startdir[PATH_MAX];
! 2002: static int fd;
1.4 kristaps 2003:
1.59 ! schwarze 2004: /*
! 2005: * Remember where we started by keeping a fd open to the origin
! 2006: * path component: throughout this utility, we chdir() a lot to
! 2007: * handle relative paths, and by doing this, we can return to
! 2008: * the starting point.
! 2009: */
! 2010: if ('\0' == *startdir) {
! 2011: if (NULL == getcwd(startdir, PATH_MAX)) {
! 2012: exitcode = (int)MANDOCLEVEL_SYSERR;
! 2013: if (NULL != targetdir)
! 2014: say(".", NULL);
! 2015: return(0);
! 2016: }
! 2017: if (-1 == (fd = open(startdir, O_RDONLY, 0))) {
! 2018: exitcode = (int)MANDOCLEVEL_SYSERR;
! 2019: say(startdir, NULL);
! 2020: return(0);
! 2021: }
! 2022: if (NULL == targetdir)
! 2023: targetdir = startdir;
! 2024: } else {
! 2025: if (-1 == fd)
! 2026: return(0);
! 2027: if (-1 == fchdir(fd)) {
! 2028: close(fd);
! 2029: basedir[0] = '\0';
! 2030: exitcode = (int)MANDOCLEVEL_SYSERR;
! 2031: say(startdir, NULL);
! 2032: return(0);
! 2033: }
! 2034: if (NULL == targetdir) {
! 2035: close(fd);
! 2036: return(1);
! 2037: }
! 2038: }
! 2039: if (NULL == realpath(targetdir, basedir)) {
! 2040: basedir[0] = '\0';
! 2041: exitcode = (int)MANDOCLEVEL_BADARG;
! 2042: say(targetdir, NULL);
1.50 kristaps 2043: return(0);
1.59 ! schwarze 2044: } else if (-1 == chdir(basedir)) {
! 2045: exitcode = (int)MANDOCLEVEL_BADARG;
! 2046: say("", NULL);
1.50 kristaps 2047: return(0);
1.4 kristaps 2048: }
1.50 kristaps 2049: return(1);
1.56 schwarze 2050: }
2051:
2052: static void
1.59 ! schwarze 2053: say(const char *file, const char *format, ...)
1.56 schwarze 2054: {
2055: va_list ap;
2056:
1.59 ! schwarze 2057: if ('\0' != *basedir)
! 2058: fprintf(stderr, "%s", basedir);
! 2059: if ('\0' != *basedir && '\0' != *file)
! 2060: fputs("//", stderr);
1.56 schwarze 2061: if ('\0' != *file)
1.59 ! schwarze 2062: fprintf(stderr, "%s", file);
1.56 schwarze 2063: fputs(": ", stderr);
1.59 ! schwarze 2064:
! 2065: if (NULL == format) {
! 2066: perror(NULL);
! 2067: return;
! 2068: }
1.56 schwarze 2069:
2070: va_start(ap, format);
2071: vfprintf(stderr, format, ap);
2072: va_end(ap);
2073:
2074: fputc('\n', stderr);
1.1 kristaps 2075: }
CVSweb