-/* $Id: mandocdb.c,v 1.9 2011/11/20 12:39:08 kristaps Exp $ */
+/* $Id: mandocdb.c,v 1.55 2012/06/09 14:11:16 kristaps Exp $ */
/*
- * Copyright (c) 2011 Kristaps Dzonsons <kristaps@bsd.lv>
+ * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
+ * Copyright (c) 2011, 2012 Ingo Schwarze <schwarze@openbsd.org>
*
* Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
#endif
#include <sys/param.h>
+#include <sys/stat.h>
#include <assert.h>
-#include <dirent.h>
+#include <ctype.h>
+#include <errno.h>
#include <fcntl.h>
+#include <fts.h>
#include <getopt.h>
-#include <stdio.h>
+#include <stddef.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
+#include <unistd.h>
-#ifdef __linux__
-# include <db_185.h>
+#ifdef HAVE_OHASH
+#include <ohash.h>
#else
-# include <db.h>
+#include "compat_ohash.h"
#endif
+#include <sqlite3.h>
-#include "man.h"
#include "mdoc.h"
+#include "man.h"
#include "mandoc.h"
-#include "mandocdb.h"
+#include "manpath.h"
+#include "mansearch.h"
+
+/* Post a warning to stderr. */
+#define WARNING(_f, _b, _fmt, _args...) \
+ do if (warnings) { \
+ fprintf(stderr, "%s: ", (_b)); \
+ fprintf(stderr, (_fmt), ##_args); \
+ if ('\0' != *(_f)) \
+ fprintf(stderr, ": %s", (_f)); \
+ fprintf(stderr, "\n"); \
+ } while (/* CONSTCOND */ 0)
+/* Post a "verbose" message to stderr. */
+#define DEBUG(_f, _b, _fmt, _args...) \
+ do if (verb) { \
+ fprintf(stderr, "%s: ", (_b)); \
+ fprintf(stderr, (_fmt), ##_args); \
+ fprintf(stderr, ": %s\n", (_f)); \
+ } while (/* CONSTCOND */ 0)
+
+#define SQL_EXEC(_v) \
+ if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
+ fprintf(stderr, "%s\n", sqlite3_errmsg(db))
+#define SQL_BIND_TEXT(_s, _i, _v) \
+ if (SQLITE_OK != sqlite3_bind_text \
+ ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
+ fprintf(stderr, "%s\n", sqlite3_errmsg(db))
+#define SQL_BIND_INT(_s, _i, _v) \
+ if (SQLITE_OK != sqlite3_bind_int \
+ ((_s), (_i)++, (_v))) \
+ fprintf(stderr, "%s\n", sqlite3_errmsg(db))
+#define SQL_BIND_INT64(_s, _i, _v) \
+ if (SQLITE_OK != sqlite3_bind_int64 \
+ ((_s), (_i)++, (_v))) \
+ fprintf(stderr, "%s\n", sqlite3_errmsg(db))
+#define SQL_STEP(_s) \
+ if (SQLITE_DONE != sqlite3_step((_s))) \
+ fprintf(stderr, "%s\n", sqlite3_errmsg(db))
-#define MANDOC_BUFSZ BUFSIZ
-#define MANDOC_SLOP 1024
-
-/* Tiny list for files. No need to bring in QUEUE. */
+enum op {
+ OP_DEFAULT = 0, /* new dbs from dir list or default config */
+ OP_CONFFILE, /* new databases from custom config file */
+ OP_UPDATE, /* delete/add entries in existing database */
+ OP_DELETE, /* delete entries from existing database */
+ OP_TEST /* change no databases, report potential problems */
+};
-struct of {
- char *fname; /* heap-allocated */
- struct of *next; /* NULL for last one */
- struct of *first; /* first in list */
+enum form {
+ FORM_SRC, /* format is -man or -mdoc */
+ FORM_CAT, /* format is cat */
+ FORM_NONE /* format is unknown */
};
-/* Buffer for storing growable data. */
+struct str {
+ char *utf8; /* key in UTF-8 form */
+ const struct of *of; /* if set, the owning parse */
+ struct str *next; /* next in owning parse sequence */
+ uint64_t mask; /* bitmask in sequence */
+ char key[]; /* the string itself */
+};
-struct buf {
- char *cp;
- size_t len; /* current length */
- size_t size; /* total buffer size */
+struct id {
+ ino_t ino;
+ dev_t dev;
};
-/* Operation we're going to perform. */
+struct of {
+ struct id id; /* used for hashing routine */
+ struct of *next; /* next in ofs */
+ enum form dform; /* path-cued form */
+ enum form sform; /* suffix-cued form */
+ char file[MAXPATHLEN]; /* filename rel. to manpath */
+ const char *desc; /* parsed description */
+ const char *sec; /* suffix-cued section (or empty) */
+ const char *dsec; /* path-cued section (or empty) */
+ const char *arch; /* path-cued arch. (or empty) */
+ const char *name; /* name (from filename) (not empty) */
+};
-enum op {
- OP_NEW = 0, /* new database */
- OP_UPDATE, /* delete/add entries in existing database */
- OP_DELETE /* delete entries from existing database */
+enum stmt {
+ STMT_DELETE = 0, /* delete manpage */
+ STMT_INSERT_DOC, /* insert manpage */
+ STMT_INSERT_KEY, /* insert parsed key */
+ STMT__MAX
};
-#define MAN_ARGS DB *hash, \
- struct buf *buf, \
- struct buf *dbuf, \
- const struct man_node *n
-#define MDOC_ARGS DB *hash, \
- struct buf *buf, \
- struct buf *dbuf, \
- const struct mdoc_node *n, \
- const struct mdoc_meta *m
-
-static void buf_appendmdoc(struct buf *,
- const struct mdoc_node *, int);
-static void buf_append(struct buf *, const char *);
-static void buf_appendb(struct buf *,
- const void *, size_t);
-static void dbt_put(DB *, const char *, DBT *, DBT *);
-static void hash_put(DB *, const struct buf *, uint64_t);
-static void hash_reset(DB **);
-static void index_merge(const struct of *, struct mparse *,
- struct buf *, struct buf *,
- DB *, DB *, const char *,
- DB *, const char *, int,
- recno_t, const recno_t *, size_t);
-static void index_prune(const struct of *, DB *,
- const char *, DB *, const char *,
- int, recno_t *, recno_t **, size_t *);
-static void ofile_argbuild(char *[], int, int, struct of **);
-static int ofile_dirbuild(const char *, int, struct of **);
-static void ofile_free(struct of *);
-static int pman_node(MAN_ARGS);
-static void pmdoc_node(MDOC_ARGS);
-static void pmdoc_An(MDOC_ARGS);
-static void pmdoc_Cd(MDOC_ARGS);
-static void pmdoc_Er(MDOC_ARGS);
-static void pmdoc_Ev(MDOC_ARGS);
-static void pmdoc_Fd(MDOC_ARGS);
-static void pmdoc_In(MDOC_ARGS);
-static void pmdoc_Fn(MDOC_ARGS);
-static void pmdoc_Fo(MDOC_ARGS);
-static void pmdoc_Nd(MDOC_ARGS);
-static void pmdoc_Nm(MDOC_ARGS);
-static void pmdoc_Pa(MDOC_ARGS);
-static void pmdoc_St(MDOC_ARGS);
-static void pmdoc_Vt(MDOC_ARGS);
-static void pmdoc_Xr(MDOC_ARGS);
-static void usage(void);
-
-typedef void (*pmdoc_nf)(MDOC_ARGS);
-
-static const pmdoc_nf mdocs[MDOC_MAX] = {
- NULL, /* Ap */
- NULL, /* Dd */
- NULL, /* Dt */
- NULL, /* Os */
- NULL, /* Sh */
- NULL, /* Ss */
- NULL, /* Pp */
- NULL, /* D1 */
- NULL, /* Dl */
- NULL, /* Bd */
- NULL, /* Ed */
- NULL, /* Bl */
- NULL, /* El */
- NULL, /* It */
- NULL, /* Ad */
- pmdoc_An, /* An */
- NULL, /* Ar */
- pmdoc_Cd, /* Cd */
- NULL, /* Cm */
- NULL, /* Dv */
- pmdoc_Er, /* Er */
- pmdoc_Ev, /* Ev */
- NULL, /* Ex */
- NULL, /* Fa */
- pmdoc_Fd, /* Fd */
- NULL, /* Fl */
- pmdoc_Fn, /* Fn */
- NULL, /* Ft */
- NULL, /* Ic */
- pmdoc_In, /* In */
- NULL, /* Li */
- pmdoc_Nd, /* Nd */
- pmdoc_Nm, /* Nm */
- NULL, /* Op */
- NULL, /* Ot */
- pmdoc_Pa, /* Pa */
- NULL, /* Rv */
- pmdoc_St, /* St */
- pmdoc_Vt, /* Va */
- pmdoc_Vt, /* Vt */
- pmdoc_Xr, /* Xr */
- NULL, /* %A */
- NULL, /* %B */
- NULL, /* %D */
- NULL, /* %I */
- NULL, /* %J */
- NULL, /* %N */
- NULL, /* %O */
- NULL, /* %P */
- NULL, /* %R */
- NULL, /* %T */
- NULL, /* %V */
- NULL, /* Ac */
- NULL, /* Ao */
- NULL, /* Aq */
- NULL, /* At */
- NULL, /* Bc */
- NULL, /* Bf */
- NULL, /* Bo */
- NULL, /* Bq */
- NULL, /* Bsx */
- NULL, /* Bx */
- NULL, /* Db */
- NULL, /* Dc */
- NULL, /* Do */
- NULL, /* Dq */
- NULL, /* Ec */
- NULL, /* Ef */
- NULL, /* Em */
- NULL, /* Eo */
- NULL, /* Fx */
- NULL, /* Ms */
- NULL, /* No */
- NULL, /* Ns */
- NULL, /* Nx */
- NULL, /* Ox */
- NULL, /* Pc */
- NULL, /* Pf */
- NULL, /* Po */
- NULL, /* Pq */
- NULL, /* Qc */
- NULL, /* Ql */
- NULL, /* Qo */
- NULL, /* Qq */
- NULL, /* Re */
- NULL, /* Rs */
- NULL, /* Sc */
- NULL, /* So */
- NULL, /* Sq */
- NULL, /* Sm */
- NULL, /* Sx */
- NULL, /* Sy */
- NULL, /* Tn */
- NULL, /* Ux */
- NULL, /* Xc */
- NULL, /* Xo */
- pmdoc_Fo, /* Fo */
- NULL, /* Fc */
- NULL, /* Oo */
- NULL, /* Oc */
- NULL, /* Bk */
- NULL, /* Ek */
- NULL, /* Bt */
- NULL, /* Hf */
- NULL, /* Fr */
- NULL, /* Ud */
- NULL, /* Lb */
- NULL, /* Lp */
- NULL, /* Lk */
- NULL, /* Mt */
- NULL, /* Brq */
- NULL, /* Bro */
- NULL, /* Brc */
- NULL, /* %C */
- NULL, /* Es */
- NULL, /* En */
- NULL, /* Dx */
- NULL, /* %Q */
- NULL, /* br */
- NULL, /* sp */
- NULL, /* %U */
- NULL, /* Ta */
+typedef int (*mdoc_fp)(struct of *, const struct mdoc_node *);
+
+struct mdoc_handler {
+ mdoc_fp fp; /* optional handler */
+ uint64_t mask; /* set unless handler returns 0 */
+ int flags; /* for use by pmdoc_node */
+#define MDOCF_CHILD 0x01 /* automatically index child nodes */
};
-static const char *progname;
+static void dbclose(const char *, int);
+static void dbindex(struct mchars *, int,
+ const struct of *, const char *);
+static int dbopen(const char *, int);
+static void dbprune(const char *);
+static void fileadd(struct of *);
+static int filecheck(const char *);
+static void filescan(const char *, const char *);
+static struct str *hashget(const char *, size_t);
+static void *hash_alloc(size_t, void *);
+static void hash_free(void *, size_t, void *);
+static void *hash_halloc(size_t, void *);
+static void inoadd(const struct stat *, struct of *);
+static int inocheck(const struct stat *);
+static void ofadd(const char *, int, const char *,
+ const char *, const char *, const char *,
+ const char *, const struct stat *);
+static void offree(void);
+static int ofmerge(struct mchars *, struct mparse *, const char *);
+static void parse_catpage(struct of *, const char *);
+static int parse_man(struct of *,
+ const struct man_node *);
+static void parse_mdoc(struct of *, const struct mdoc_node *);
+static int parse_mdoc_body(struct of *, const struct mdoc_node *);
+static int parse_mdoc_head(struct of *, const struct mdoc_node *);
+static int parse_mdoc_Fd(struct of *, const struct mdoc_node *);
+static int parse_mdoc_Fn(struct of *, const struct mdoc_node *);
+static int parse_mdoc_In(struct of *, const struct mdoc_node *);
+static int parse_mdoc_Nd(struct of *, const struct mdoc_node *);
+static int parse_mdoc_Nm(struct of *, const struct mdoc_node *);
+static int parse_mdoc_Sh(struct of *, const struct mdoc_node *);
+static int parse_mdoc_St(struct of *, const struct mdoc_node *);
+static int parse_mdoc_Xr(struct of *, const struct mdoc_node *);
+static int path_reset(const char *, int, const char *);
+static void putkey(const struct of *,
+ const char *, uint64_t);
+static void putkeys(const struct of *,
+ const char *, int, uint64_t);
+static void putmdockey(const struct of *,
+ const struct mdoc_node *, uint64_t);
+static char *stradd(const char *);
+static char *straddbuf(const char *, size_t);
+static int treescan(const char *);
+static size_t utf8(unsigned int, char [7]);
+static void utf8key(struct mchars *, struct str *);
+static void wordaddbuf(const struct of *,
+ const char *, size_t, uint64_t);
+
+static char *progname;
+static int use_all; /* use all found files */
+static int nodb; /* no database changes */
+static int verb; /* print what we're doing */
+static int warnings; /* warn about crap */
+static enum op op; /* operational mode */
+static struct ohash inos; /* table of inodes/devices */
+static struct ohash filenames; /* table of filenames */
+static struct ohash strings; /* table of all strings */
+static struct of *ofs = NULL; /* vector of files to parse */
+static struct str *words = NULL; /* word list in current parse */
+static sqlite3 *db = NULL; /* current database */
+static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
+
+static const struct mdoc_handler mdocs[MDOC_MAX] = {
+ { NULL, 0, 0 }, /* Ap */
+ { NULL, 0, 0 }, /* Dd */
+ { NULL, 0, 0 }, /* Dt */
+ { NULL, 0, 0 }, /* Os */
+ { parse_mdoc_Sh, TYPE_Sh, MDOCF_CHILD }, /* Sh */
+ { parse_mdoc_head, TYPE_Ss, MDOCF_CHILD }, /* Ss */
+ { NULL, 0, 0 }, /* Pp */
+ { NULL, 0, 0 }, /* D1 */
+ { NULL, 0, 0 }, /* Dl */
+ { NULL, 0, 0 }, /* Bd */
+ { NULL, 0, 0 }, /* Ed */
+ { NULL, 0, 0 }, /* Bl */
+ { NULL, 0, 0 }, /* El */
+ { NULL, 0, 0 }, /* It */
+ { NULL, 0, 0 }, /* Ad */
+ { NULL, TYPE_An, MDOCF_CHILD }, /* An */
+ { NULL, TYPE_Ar, MDOCF_CHILD }, /* Ar */
+ { NULL, TYPE_Cd, MDOCF_CHILD }, /* Cd */
+ { NULL, TYPE_Cm, MDOCF_CHILD }, /* Cm */
+ { NULL, TYPE_Dv, MDOCF_CHILD }, /* Dv */
+ { NULL, TYPE_Er, MDOCF_CHILD }, /* Er */
+ { NULL, TYPE_Ev, MDOCF_CHILD }, /* Ev */
+ { NULL, 0, 0 }, /* Ex */
+ { NULL, TYPE_Fa, MDOCF_CHILD }, /* Fa */
+ { parse_mdoc_Fd, TYPE_In, 0 }, /* Fd */
+ { NULL, TYPE_Fl, MDOCF_CHILD }, /* Fl */
+ { parse_mdoc_Fn, 0, 0 }, /* Fn */
+ { NULL, TYPE_Ft, MDOCF_CHILD }, /* Ft */
+ { NULL, TYPE_Ic, MDOCF_CHILD }, /* Ic */
+ { parse_mdoc_In, TYPE_In, MDOCF_CHILD }, /* In */
+ { NULL, TYPE_Li, MDOCF_CHILD }, /* Li */
+ { parse_mdoc_Nd, TYPE_Nd, MDOCF_CHILD }, /* Nd */
+ { parse_mdoc_Nm, TYPE_Nm, MDOCF_CHILD }, /* Nm */
+ { NULL, 0, 0 }, /* Op */
+ { NULL, 0, 0 }, /* Ot */
+ { NULL, TYPE_Pa, MDOCF_CHILD }, /* Pa */
+ { NULL, 0, 0 }, /* Rv */
+ { parse_mdoc_St, TYPE_St, 0 }, /* St */
+ { NULL, TYPE_Va, MDOCF_CHILD }, /* Va */
+ { parse_mdoc_body, TYPE_Va, MDOCF_CHILD }, /* Vt */
+ { parse_mdoc_Xr, TYPE_Xr, 0 }, /* Xr */
+ { NULL, 0, 0 }, /* %A */
+ { NULL, 0, 0 }, /* %B */
+ { NULL, 0, 0 }, /* %D */
+ { NULL, 0, 0 }, /* %I */
+ { NULL, 0, 0 }, /* %J */
+ { NULL, 0, 0 }, /* %N */
+ { NULL, 0, 0 }, /* %O */
+ { NULL, 0, 0 }, /* %P */
+ { NULL, 0, 0 }, /* %R */
+ { NULL, 0, 0 }, /* %T */
+ { NULL, 0, 0 }, /* %V */
+ { NULL, 0, 0 }, /* Ac */
+ { NULL, 0, 0 }, /* Ao */
+ { NULL, 0, 0 }, /* Aq */
+ { NULL, TYPE_At, MDOCF_CHILD }, /* At */
+ { NULL, 0, 0 }, /* Bc */
+ { NULL, 0, 0 }, /* Bf */
+ { NULL, 0, 0 }, /* Bo */
+ { NULL, 0, 0 }, /* Bq */
+ { NULL, TYPE_Bsx, MDOCF_CHILD }, /* Bsx */
+ { NULL, TYPE_Bx, MDOCF_CHILD }, /* Bx */
+ { NULL, 0, 0 }, /* Db */
+ { NULL, 0, 0 }, /* Dc */
+ { NULL, 0, 0 }, /* Do */
+ { NULL, 0, 0 }, /* Dq */
+ { NULL, 0, 0 }, /* Ec */
+ { NULL, 0, 0 }, /* Ef */
+ { NULL, TYPE_Em, MDOCF_CHILD }, /* Em */
+ { NULL, 0, 0 }, /* Eo */
+ { NULL, TYPE_Fx, MDOCF_CHILD }, /* Fx */
+ { NULL, TYPE_Ms, MDOCF_CHILD }, /* Ms */
+ { NULL, 0, 0 }, /* No */
+ { NULL, 0, 0 }, /* Ns */
+ { NULL, TYPE_Nx, MDOCF_CHILD }, /* Nx */
+ { NULL, TYPE_Ox, MDOCF_CHILD }, /* Ox */
+ { NULL, 0, 0 }, /* Pc */
+ { NULL, 0, 0 }, /* Pf */
+ { NULL, 0, 0 }, /* Po */
+ { NULL, 0, 0 }, /* Pq */
+ { NULL, 0, 0 }, /* Qc */
+ { NULL, 0, 0 }, /* Ql */
+ { NULL, 0, 0 }, /* Qo */
+ { NULL, 0, 0 }, /* Qq */
+ { NULL, 0, 0 }, /* Re */
+ { NULL, 0, 0 }, /* Rs */
+ { NULL, 0, 0 }, /* Sc */
+ { NULL, 0, 0 }, /* So */
+ { NULL, 0, 0 }, /* Sq */
+ { NULL, 0, 0 }, /* Sm */
+ { NULL, 0, 0 }, /* Sx */
+ { NULL, TYPE_Sy, MDOCF_CHILD }, /* Sy */
+ { NULL, TYPE_Tn, MDOCF_CHILD }, /* Tn */
+ { NULL, 0, 0 }, /* Ux */
+ { NULL, 0, 0 }, /* Xc */
+ { NULL, 0, 0 }, /* Xo */
+ { parse_mdoc_head, TYPE_Fn, 0 }, /* Fo */
+ { NULL, 0, 0 }, /* Fc */
+ { NULL, 0, 0 }, /* Oo */
+ { NULL, 0, 0 }, /* Oc */
+ { NULL, 0, 0 }, /* Bk */
+ { NULL, 0, 0 }, /* Ek */
+ { NULL, 0, 0 }, /* Bt */
+ { NULL, 0, 0 }, /* Hf */
+ { NULL, 0, 0 }, /* Fr */
+ { NULL, 0, 0 }, /* Ud */
+ { NULL, TYPE_Lb, MDOCF_CHILD }, /* Lb */
+ { NULL, 0, 0 }, /* Lp */
+ { NULL, TYPE_Lk, MDOCF_CHILD }, /* Lk */
+ { NULL, TYPE_Mt, MDOCF_CHILD }, /* Mt */
+ { NULL, 0, 0 }, /* Brq */
+ { NULL, 0, 0 }, /* Bro */
+ { NULL, 0, 0 }, /* Brc */
+ { NULL, 0, 0 }, /* %C */
+ { NULL, 0, 0 }, /* Es */
+ { NULL, 0, 0 }, /* En */
+ { NULL, TYPE_Dx, MDOCF_CHILD }, /* Dx */
+ { NULL, 0, 0 }, /* %Q */
+ { NULL, 0, 0 }, /* br */
+ { NULL, 0, 0 }, /* sp */
+ { NULL, 0, 0 }, /* %U */
+ { NULL, 0, 0 }, /* Ta */
+};
int
main(int argc, char *argv[])
{
- struct mparse *mp; /* parse sequence */
- enum op op; /* current operation */
- const char *dir;
- char ibuf[MAXPATHLEN], /* index fname */
- fbuf[MAXPATHLEN]; /* btree fname */
- int verb, /* output verbosity */
- ch, i, flags;
- DB *idx, /* index database */
- *db, /* keyword database */
- *hash; /* temporary keyword hashtable */
- BTREEINFO info; /* btree configuration */
- recno_t maxrec; /* supremum of all records */
- recno_t *recs; /* buffer of empty records */
- size_t sz1, sz2,
- recsz, /* buffer size of recs */
- reccur; /* valid number of recs */
- struct buf buf, /* keyword buffer */
- dbuf; /* description buffer */
- struct of *of; /* list of files for processing */
- extern int optind;
- extern char *optarg;
+ char cwd[MAXPATHLEN];
+ int ch, rc, fd, i;
+ unsigned int index;
+ size_t j, sz;
+ const char *dir;
+ struct str *s;
+ struct mchars *mc;
+ struct manpaths dirs;
+ struct mparse *mp;
+ struct ohash_info ino_info, filename_info, str_info;
+
+ memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
+ memset(&dirs, 0, sizeof(struct manpaths));
+
+ ino_info.halloc = filename_info.halloc =
+ str_info.halloc = hash_halloc;
+ ino_info.hfree = filename_info.hfree =
+ str_info.hfree = hash_free;
+ ino_info.alloc = filename_info.alloc =
+ str_info.alloc = hash_alloc;
+
+ ino_info.key_offset = offsetof(struct of, id);
+ filename_info.key_offset = offsetof(struct of, file);
+ str_info.key_offset = offsetof(struct str, key);
progname = strrchr(argv[0], '/');
if (progname == NULL)
else
++progname;
- verb = 0;
- of = NULL;
- db = idx = NULL;
- mp = NULL;
- hash = NULL;
- recs = NULL;
- recsz = reccur = 0;
- maxrec = 0;
- op = OP_NEW;
+ /*
+ * Remember where we started by keeping a fd open to the origin
+ * path component: throughout this utility, we chdir() a lot to
+ * handle relative paths, and by doing this, we can return to
+ * the starting point.
+ */
+ if (NULL == getcwd(cwd, MAXPATHLEN)) {
+ perror(NULL);
+ return(EXIT_FAILURE);
+ } else if (-1 == (fd = open(cwd, O_RDONLY, 0))) {
+ perror(cwd);
+ return(EXIT_FAILURE);
+ }
+
+ /*
+ * We accept a few different invocations.
+ * The CHECKOP macro makes sure that invocation styles don't
+ * clobber each other.
+ */
+#define CHECKOP(_op, _ch) do \
+ if (OP_DEFAULT != (_op)) { \
+ fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \
+ goto usage; \
+ } while (/*CONSTCOND*/0)
+
dir = NULL;
+ op = OP_DEFAULT;
- while (-1 != (ch = getopt(argc, argv, "d:u:v")))
+ while (-1 != (ch = getopt(argc, argv, "aC:d:ntu:vW")))
switch (ch) {
+ case ('a'):
+ use_all = 1;
+ break;
+ case ('C'):
+ CHECKOP(op, ch);
+ dir = optarg;
+ op = OP_CONFFILE;
+ break;
case ('d'):
+ CHECKOP(op, ch);
dir = optarg;
op = OP_UPDATE;
break;
+ case ('n'):
+ nodb = 1;
+ break;
+ case ('t'):
+ CHECKOP(op, ch);
+ dup2(STDOUT_FILENO, STDERR_FILENO);
+ op = OP_TEST;
+ nodb = warnings = 1;
+ break;
case ('u'):
+ CHECKOP(op, ch);
dir = optarg;
op = OP_DELETE;
break;
case ('v'):
verb++;
break;
+ case ('W'):
+ warnings = 1;
+ break;
default:
- usage();
- return((int)MANDOCLEVEL_BADARG);
+ goto usage;
}
argc -= optind;
argv += optind;
- memset(&info, 0, sizeof(BTREEINFO));
- info.flags = R_DUP;
-
- mp = mparse_alloc(MPARSE_AUTO, MANDOCLEVEL_FATAL, NULL, NULL);
-
- memset(&buf, 0, sizeof(struct buf));
- memset(&dbuf, 0, sizeof(struct buf));
-
- buf.size = dbuf.size = MANDOC_BUFSZ;
-
- buf.cp = mandoc_malloc(buf.size);
- dbuf.cp = mandoc_malloc(dbuf.size);
-
- flags = OP_NEW == op ? O_CREAT|O_TRUNC|O_RDWR : O_CREAT|O_RDWR;
-
- if (OP_UPDATE == op || OP_DELETE == op) {
- ibuf[0] = fbuf[0] = '\0';
-
- strlcat(fbuf, dir, MAXPATHLEN);
- strlcat(fbuf, "/", MAXPATHLEN);
- sz1 = strlcat(fbuf, MANDOC_DB, MAXPATHLEN);
-
- strlcat(ibuf, dir, MAXPATHLEN);
- strlcat(ibuf, "/", MAXPATHLEN);
- sz2 = strlcat(ibuf, MANDOC_IDX, MAXPATHLEN);
-
- if (sz1 >= MAXPATHLEN || sz2 >= MAXPATHLEN) {
- fprintf(stderr, "%s: Path too long\n", dir);
- exit((int)MANDOCLEVEL_BADARG);
- }
-
- db = dbopen(fbuf, flags, 0644, DB_BTREE, &info);
- idx = dbopen(ibuf, flags, 0644, DB_RECNO, NULL);
-
- if (NULL == db) {
- perror(fbuf);
- exit((int)MANDOCLEVEL_SYSERR);
- } else if (NULL == db) {
- perror(ibuf);
- exit((int)MANDOCLEVEL_SYSERR);
- }
-
- if (verb > 2) {
- printf("%s: Opened\n", fbuf);
- printf("%s: Opened\n", ibuf);
- }
-
- ofile_argbuild(argv, argc, verb, &of);
- if (NULL == of)
- goto out;
-
- of = of->first;
-
- index_prune(of, db, fbuf, idx, ibuf, verb,
- &maxrec, &recs, &recsz);
-
- if (OP_UPDATE == op)
- index_merge(of, mp, &dbuf, &buf, hash,
- db, fbuf, idx, ibuf, verb,
- maxrec, recs, reccur);
-
- goto out;
+ if (OP_CONFFILE == op && argc > 0) {
+ fprintf(stderr, "-C: Too many arguments\n");
+ goto usage;
}
- for (i = 0; i < argc; i++) {
- ibuf[0] = fbuf[0] = '\0';
+ rc = 1;
+ mp = mparse_alloc(MPARSE_AUTO,
+ MANDOCLEVEL_FATAL, NULL, NULL, NULL);
+ mc = mchars_alloc();
- strlcat(fbuf, argv[i], MAXPATHLEN);
- strlcat(fbuf, "/", MAXPATHLEN);
- sz1 = strlcat(fbuf, MANDOC_DB, MAXPATHLEN);
+ ohash_init(&strings, 6, &str_info);
+ ohash_init(&inos, 6, &ino_info);
+ ohash_init(&filenames, 6, &filename_info);
- strlcat(ibuf, argv[i], MAXPATHLEN);
- strlcat(ibuf, "/", MAXPATHLEN);
- sz2 = strlcat(ibuf, MANDOC_IDX, MAXPATHLEN);
-
- if (sz1 >= MAXPATHLEN || sz2 >= MAXPATHLEN) {
- fprintf(stderr, "%s: Path too long\n", argv[i]);
- exit((int)MANDOCLEVEL_BADARG);
- }
-
- db = dbopen(fbuf, flags, 0644, DB_BTREE, &info);
- idx = dbopen(ibuf, flags, 0644, DB_RECNO, NULL);
+ if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
+ /*
+ * Force processing all files.
+ */
+ use_all = 1;
+ if (NULL == dir)
+ dir = cwd;
+ /*
+ * All of these deal with a specific directory.
+ * Jump into that directory then collect files specified
+ * on the command-line.
+ */
+ if (0 == path_reset(cwd, fd, dir))
+ goto out;
+ for (i = 0; i < argc; i++)
+ filescan(argv[i], dir);
+ if (0 == dbopen(dir, 1))
+ goto out;
+ if (OP_TEST != op)
+ dbprune(dir);
+ if (OP_DELETE != op)
+ rc = ofmerge(mc, mp, dir);
+ dbclose(dir, 1);
+ } else {
+ /*
+ * If we have arguments, use them as our manpaths.
+ * If we don't, grok from manpath(1) or however else
+ * manpath_parse() wants to do it.
+ */
+ if (argc > 0) {
+ dirs.paths = mandoc_calloc
+ (argc, sizeof(char *));
+ dirs.sz = (size_t)argc;
+ for (i = 0; i < argc; i++)
+ dirs.paths[i] = mandoc_strdup(argv[i]);
+ } else
+ manpath_parse(&dirs, dir, NULL, NULL);
- if (NULL == db) {
- perror(fbuf);
- exit((int)MANDOCLEVEL_SYSERR);
- } else if (NULL == db) {
- perror(ibuf);
- exit((int)MANDOCLEVEL_SYSERR);
- }
+ /*
+ * First scan the tree rooted at a base directory.
+ * Then whak its database (if one exists), parse, and
+ * build up the database.
+ * Ignore zero-length directories and strip trailing
+ * slashes.
+ */
+ for (j = 0; j < dirs.sz; j++) {
+ sz = strlen(dirs.paths[j]);
+ if (sz && '/' == dirs.paths[j][sz - 1])
+ dirs.paths[j][--sz] = '\0';
+ if (0 == sz)
+ continue;
+ if (0 == path_reset(cwd, fd, dirs.paths[j]))
+ goto out;
+ if (0 == treescan(dirs.paths[j]))
+ goto out;
+ if (0 == path_reset(cwd, fd, dirs.paths[j]))
+ goto out;
+ if (0 == dbopen(dirs.paths[j], 0))
+ goto out;
+
+ /*
+ * Since we're opening up a new database, we can
+ * turn off synchronous mode for much better
+ * performance.
+ */
+#ifndef __APPLE__
+ SQL_EXEC("PRAGMA synchronous = OFF");
+#endif
- if (verb > 2) {
- printf("%s: Truncated\n", fbuf);
- printf("%s: Truncated\n", ibuf);
+ if (0 == ofmerge(mc, mp, dirs.paths[j]))
+ goto out;
+ dbclose(dirs.paths[j], 0);
+ offree();
+ ohash_delete(&inos);
+ ohash_init(&inos, 6, &ino_info);
+ ohash_delete(&filenames);
+ ohash_init(&filenames, 6, &filename_info);
}
-
- ofile_free(of);
- of = NULL;
-
- if ( ! ofile_dirbuild(argv[i], verb, &of))
- exit((int)MANDOCLEVEL_SYSERR);
-
- if (NULL == of)
- continue;
-
- of = of->first;
-
- index_merge(of, mp, &dbuf, &buf, hash, db, fbuf,
- idx, ibuf, verb, maxrec, recs, reccur);
}
-
out:
- if (db)
- (*db->close)(db);
- if (idx)
- (*idx->close)(idx);
- if (hash)
- (*hash->close)(hash);
- if (mp)
- mparse_free(mp);
-
- ofile_free(of);
- free(buf.cp);
- free(dbuf.cp);
- free(recs);
-
- return(MANDOCLEVEL_OK);
+ close(fd);
+ manpath_free(&dirs);
+ mchars_free(mc);
+ mparse_free(mp);
+ for (s = ohash_first(&strings, &index);
+ NULL != s; s = ohash_next(&strings, &index)) {
+ if (s->utf8 != s->key)
+ free(s->utf8);
+ free(s);
+ }
+ ohash_delete(&strings);
+ ohash_delete(&inos);
+ ohash_delete(&filenames);
+ offree();
+ return(rc ? EXIT_SUCCESS : EXIT_FAILURE);
+usage:
+ fprintf(stderr, "usage: %s [-anvW] [-C file]\n"
+ " %s [-anvW] dir ...\n"
+ " %s [-nvW] -d dir [file ...]\n"
+ " %s [-nvW] -u dir [file ...]\n"
+ " %s -t file ...\n",
+ progname, progname, progname,
+ progname, progname);
+
+ return(EXIT_FAILURE);
}
-void
-index_merge(const struct of *of, struct mparse *mp,
- struct buf *dbuf, struct buf *buf,
- DB *hash, DB *db, const char *dbf,
- DB *idx, const char *idxf, int verb,
- recno_t maxrec, const recno_t *recs, size_t reccur)
+/*
+ * Scan a directory tree rooted at "base" for manpages.
+ * We use fts(), scanning directory parts along the way for clues to our
+ * section and architecture.
+ *
+ * If use_all has been specified, grok all files.
+ * If not, sanitise paths to the following:
+ *
+ * [./]man*[/<arch>]/<name>.<section>
+ * or
+ * [./]cat<section>[/<arch>]/<name>.0
+ *
+ * TODO: accomodate for multi-language directories.
+ */
+static int
+treescan(const char *base)
{
- recno_t rec;
- int ch;
- DBT key, val;
- struct mdoc *mdoc;
- struct man *man;
- const char *fn, *msec, *mtitle, *arch;
- size_t sv;
- unsigned seq;
- struct db_val vbuf;
-
- for (rec = 0; of; of = of->next) {
- fn = of->fname;
- if (reccur > 0) {
- --reccur;
- rec = recs[(int)reccur];
- } else if (maxrec > 0) {
- rec = maxrec;
- maxrec = 0;
- } else
- rec++;
-
- mparse_reset(mp);
- hash_reset(&hash);
-
- if (mparse_readfd(mp, -1, fn) >= MANDOCLEVEL_FATAL) {
- fprintf(stderr, "%s: Parse failure\n", fn);
- continue;
- }
-
- mparse_result(mp, &mdoc, &man);
- if (NULL == mdoc && NULL == man)
- continue;
-
- msec = NULL != mdoc ?
- mdoc_meta(mdoc)->msec : man_meta(man)->msec;
- mtitle = NULL != mdoc ?
- mdoc_meta(mdoc)->title : man_meta(man)->title;
- arch = NULL != mdoc ?
- mdoc_meta(mdoc)->arch : NULL;
-
- if (NULL == arch)
- arch = "";
-
- /*
- * The index record value consists of a nil-terminated
- * filename, a nil-terminated manual section, and a
- * nil-terminated description. Since the description
- * may not be set, we set a sentinel to see if we're
- * going to write a nil byte in its place.
- */
-
- dbuf->len = 0;
- buf_appendb(dbuf, fn, strlen(fn) + 1);
- buf_appendb(dbuf, msec, strlen(msec) + 1);
- buf_appendb(dbuf, mtitle, strlen(mtitle) + 1);
- buf_appendb(dbuf, arch, strlen(arch) + 1);
+ FTS *f;
+ FTSENT *ff;
+ int dform;
+ char *sec;
+ const char *dsec, *arch, *cp, *name, *path;
+ const char *argv[2];
- sv = dbuf->len;
+ argv[0] = ".";
+ argv[1] = (char *)NULL;
- /* Fix the record number in the btree value. */
+ /*
+ * Walk through all components under the directory, using the
+ * logical descent of files.
+ */
+ f = fts_open((char * const *)argv, FTS_LOGICAL, NULL);
+ if (NULL == f) {
+ perror(base);
+ return(0);
+ }
- if (mdoc)
- pmdoc_node(hash, buf, dbuf,
- mdoc_node(mdoc), mdoc_meta(mdoc));
- else
- pman_node(hash, buf, dbuf, man_node(man));
+ dsec = arch = NULL;
+ dform = FORM_NONE;
+ while (NULL != (ff = fts_read(f))) {
+ path = ff->fts_path + 2;
/*
- * Copy from the in-memory hashtable of pending keywords
- * into the database.
+ * If we're a regular file, add an "of" by using the
+ * stored directory data and handling the filename.
+ * Disallow duplicate (hard-linked) files.
*/
+ if (FTS_F == ff->fts_info) {
+ if ( ! use_all && ff->fts_level < 2) {
+ WARNING(path, base, "Extraneous file");
+ continue;
+ } else if (inocheck(ff->fts_statp)) {
+ WARNING(path, base, "Duplicate file");
+ continue;
+ }
- vbuf.rec = rec;
- seq = R_FIRST;
- while (0 == (ch = (*hash->seq)(hash, &key, &val, seq))) {
- seq = R_NEXT;
+ cp = ff->fts_name;
- vbuf.mask = *(uint64_t *)val.data;
- val.size = sizeof(struct db_val);
- val.data = &vbuf;
+ if (0 == strcmp(cp, "mandocdb.db")) {
+ WARNING(path, base, "Skip database");
+ continue;
+ } else if (NULL != (cp = strrchr(cp, '.'))) {
+ if (0 == strcmp(cp + 1, "html")) {
+ WARNING(path, base, "Skip html");
+ continue;
+ } else if (0 == strcmp(cp + 1, "gz")) {
+ WARNING(path, base, "Skip gz");
+ continue;
+ } else if (0 == strcmp(cp + 1, "ps")) {
+ WARNING(path, base, "Skip ps");
+ continue;
+ } else if (0 == strcmp(cp + 1, "pdf")) {
+ WARNING(path, base, "Skip pdf");
+ continue;
+ }
+ }
- if (verb > 1)
- printf("%s: Added keyword: %s\n",
- fn, (char *)key.data);
- dbt_put(db, dbf, &key, &val);
- }
- if (ch < 0) {
- perror("hash");
- exit((int)MANDOCLEVEL_SYSERR);
- }
-
- /*
- * Apply to the index. If we haven't had a description
- * set, put an empty one in now.
- */
+ if (NULL != (sec = strrchr(ff->fts_name, '.'))) {
+ *sec = '\0';
+ sec = stradd(sec + 1);
+ }
+ name = stradd(ff->fts_name);
+ ofadd(base, dform, path,
+ name, dsec, sec, arch, ff->fts_statp);
+ continue;
+ } else if (FTS_D != ff->fts_info &&
+ FTS_DP != ff->fts_info)
+ continue;
- if (dbuf->len == sv)
- buf_appendb(dbuf, "", 1);
+ switch (ff->fts_level) {
+ case (0):
+ /* Ignore the root directory. */
+ break;
+ case (1):
+ /*
+ * This might contain manX/ or catX/.
+ * Try to infer this from the name.
+ * If we're not in use_all, enforce it.
+ */
+ dsec = NULL;
+ dform = FORM_NONE;
+ cp = ff->fts_name;
+ if (FTS_DP == ff->fts_info)
+ break;
- key.data = &rec;
- key.size = sizeof(recno_t);
+ if (0 == strncmp(cp, "man", 3)) {
+ dform = FORM_SRC;
+ dsec = stradd(cp + 3);
+ } else if (0 == strncmp(cp, "cat", 3)) {
+ dform = FORM_CAT;
+ dsec = stradd(cp + 3);
+ }
- val.data = dbuf->cp;
- val.size = dbuf->len;
+ if (NULL != dsec || use_all)
+ break;
- if (verb)
- printf("%s: Added index\n", fn);
- dbt_put(idx, idxf, &key, &val);
+ WARNING(path, base, "Unknown directory part");
+ fts_set(f, ff, FTS_SKIP);
+ break;
+ case (2):
+ /*
+ * Possibly our architecture.
+ * If we're descending, keep tabs on it.
+ */
+ arch = NULL;
+ if (FTS_DP != ff->fts_info && NULL != dsec)
+ arch = stradd(ff->fts_name);
+ break;
+ default:
+ if (FTS_DP == ff->fts_info || use_all)
+ break;
+ WARNING(path, base, "Extraneous directory part");
+ fts_set(f, ff, FTS_SKIP);
+ break;
+ }
}
+
+ fts_close(f);
+ return(1);
}
/*
- * Scan through all entries in the index file `idx' and prune those
- * entries in `ofile'.
- * Pruning consists of removing from `db', then invalidating the entry
- * in `idx' (zeroing its value size).
+ * Add a file to the file vector.
+ * Do not verify that it's a "valid" looking manpage (we'll do that
+ * later).
+ *
+ * Try to infer the manual section, architecture, and page name from the
+ * path, assuming it looks like
+ *
+ * [./]man*[/<arch>]/<name>.<section>
+ * or
+ * [./]cat<section>[/<arch>]/<name>.0
+ *
+ * Stuff this information directly into the "of" vector.
+ * See treescan() for the fts(3) version of this.
*/
static void
-index_prune(const struct of *ofile, DB *db, const char *dbf,
- DB *idx, const char *idxf, int verb,
- recno_t *maxrec, recno_t **recs, size_t *recsz)
+filescan(const char *file, const char *base)
{
- const struct of *of;
- const char *fn;
- struct db_val *vbuf;
- unsigned seq, sseq;
- DBT key, val;
- size_t reccur;
- int ch;
-
- reccur = 0;
- seq = R_FIRST;
- while (0 == (ch = (*idx->seq)(idx, &key, &val, seq))) {
- seq = R_NEXT;
- *maxrec = *(recno_t *)key.data;
- if (0 == val.size) {
- if (reccur >= *recsz) {
- *recsz += MANDOC_SLOP;
- *recs = mandoc_realloc(*recs,
- *recsz * sizeof(recno_t));
- }
- (*recs)[(int)reccur] = *maxrec;
- reccur++;
- continue;
- }
+ const char *sec, *arch, *name, *dsec;
+ char *p, *start, *buf;
+ int dform;
+ struct stat st;
- fn = (char *)val.data;
- for (of = ofile; of; of = of->next)
- if (0 == strcmp(fn, of->fname))
- break;
+ assert(use_all);
- if (NULL == of)
- continue;
+ if (0 == strncmp(file, "./", 2))
+ file += 2;
- sseq = R_FIRST;
- while (0 == (ch = (*db->seq)(db, &key, &val, sseq))) {
- sseq = R_NEXT;
- assert(sizeof(struct db_val) == val.size);
- vbuf = val.data;
- if (*maxrec != vbuf->rec)
- continue;
- if (verb)
- printf("%s: Deleted keyword: %s\n",
- fn, (char *)key.data);
- ch = (*db->del)(db, &key, R_CURSOR);
- if (ch < 0)
- break;
- }
- if (ch < 0) {
- perror(dbf);
- exit((int)MANDOCLEVEL_SYSERR);
- }
+ if (-1 == stat(file, &st)) {
+ WARNING(file, base, "%s", strerror(errno));
+ return;
+ } else if ( ! (S_IFREG & st.st_mode)) {
+ WARNING(file, base, "Not a regular file");
+ return;
+ } else if (inocheck(&st)) {
+ WARNING(file, base, "Duplicate file");
+ return;
+ }
- if (verb)
- printf("%s: Deleted index\n", fn);
+ buf = mandoc_strdup(file);
+ start = buf;
+ sec = arch = name = dsec = NULL;
+ dform = FORM_NONE;
- val.size = 0;
- ch = (*idx->put)(idx, &key, &val, R_CURSOR);
- if (ch < 0) {
- perror(idxf);
- exit((int)MANDOCLEVEL_SYSERR);
+ /*
+ * First try to guess our directory structure.
+ * If we find a separator, try to look for man* or cat*.
+ * If we find one of these and what's underneath is a directory,
+ * assume it's an architecture.
+ */
+ if (NULL != (p = strchr(start, '/'))) {
+ *p++ = '\0';
+ if (0 == strncmp(start, "man", 3)) {
+ dform = FORM_SRC;
+ dsec = start + 3;
+ } else if (0 == strncmp(start, "cat", 3)) {
+ dform = FORM_CAT;
+ dsec = start + 3;
}
- if (reccur >= *recsz) {
- *recsz += MANDOC_SLOP;
- *recs = mandoc_realloc
- (*recs, *recsz * sizeof(recno_t));
- }
+ start = p;
+ if (NULL != dsec && NULL != (p = strchr(start, '/'))) {
+ *p++ = '\0';
+ arch = start;
+ start = p;
+ }
+ }
+
+ /*
+ * Now check the file suffix.
+ * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
+ */
+ p = strrchr(start, '\0');
+ while (p-- > start && '/' != *p && '.' != *p)
+ /* Loop. */ ;
- (*recs)[(int)reccur] = *maxrec;
- reccur++;
+ if ('.' == *p) {
+ *p++ = '\0';
+ sec = p;
}
- (*maxrec)++;
+
+ /*
+ * Now try to parse the name.
+ * Use the filename portion of the path.
+ */
+ name = start;
+ if (NULL != (p = strrchr(start, '/'))) {
+ name = p + 1;
+ *p = '\0';
+ }
+
+ ofadd(base, dform, file, name, dsec, sec, arch, &st);
+ free(buf);
}
/*
- * Grow the buffer (if necessary) and copy in a binary string.
+ * See fileadd().
*/
-static void
-buf_appendb(struct buf *buf, const void *cp, size_t sz)
+static int
+filecheck(const char *name)
{
+ unsigned int index;
- /* Overshoot by MANDOC_BUFSZ. */
-
- while (buf->len + sz >= buf->size) {
- buf->size = buf->len + sz + MANDOC_BUFSZ;
- buf->cp = mandoc_realloc(buf->cp, buf->size);
- }
-
- memcpy(buf->cp + (int)buf->len, cp, sz);
- buf->len += sz;
+ index = ohash_qlookup(&filenames, name);
+ return(NULL != ohash_find(&filenames, index));
}
/*
- * Append a nil-terminated string to the buffer.
- * This can be invoked multiple times.
- * The buffer string will be nil-terminated.
- * If invoked multiple times, a space is put between strings.
+ * Use the standard hashing mechanism (K&R) to see if the given filename
+ * already exists.
*/
static void
-buf_append(struct buf *buf, const char *cp)
+fileadd(struct of *of)
{
- size_t sz;
+ unsigned int index;
- if (0 == (sz = strlen(cp)))
- return;
+ index = ohash_qlookup(&filenames, of->file);
+ assert(NULL == ohash_find(&filenames, index));
+ ohash_insert(&filenames, index, of);
+}
+
+/*
+ * See inoadd().
+ */
+static int
+inocheck(const struct stat *st)
+{
+ struct id id;
+ uint32_t hash;
+ unsigned int index;
- if (buf->len)
- buf->cp[(int)buf->len - 1] = ' ';
+ memset(&id, 0, sizeof(id));
+ id.ino = hash = st->st_ino;
+ id.dev = st->st_dev;
+ index = ohash_lookup_memory
+ (&inos, (char *)&id, sizeof(id), hash);
- buf_appendb(buf, cp, sz + 1);
+ return(NULL != ohash_find(&inos, index));
}
/*
- * Recursively add all text from a given node.
- * This is optimised for general mdoc nodes in this context, which do
- * not consist of subexpressions and having a recursive call for n->next
- * would be wasteful.
- * The "f" variable should be 0 unless called from pmdoc_Nd for the
- * description buffer, which does not start at the beginning of the
- * buffer.
+ * The hashing function used here is quite simple: simply take the inode
+ * and use uint32_t of its bits.
+ * Then when we do the lookup, use both the inode and device identifier.
*/
static void
-buf_appendmdoc(struct buf *buf, const struct mdoc_node *n, int f)
+inoadd(const struct stat *st, struct of *of)
{
+ uint32_t hash;
+ unsigned int index;
- for ( ; n; n = n->next) {
- if (n->child)
- buf_appendmdoc(buf, n->child, f);
+ of->id.ino = hash = st->st_ino;
+ of->id.dev = st->st_dev;
+ index = ohash_lookup_memory
+ (&inos, (char *)&of->id, sizeof(of->id), hash);
- if (MDOC_TEXT == n->type && f) {
- f = 0;
- buf_appendb(buf, n->string,
- strlen(n->string) + 1);
- } else if (MDOC_TEXT == n->type)
- buf_append(buf, n->string);
+ assert(NULL == ohash_find(&inos, index));
+ ohash_insert(&inos, index, of);
+}
+static void
+ofadd(const char *base, int dform, const char *file,
+ const char *name, const char *dsec, const char *sec,
+ const char *arch, const struct stat *st)
+{
+ struct of *of;
+ int sform;
+
+ assert(NULL != file);
+
+ if (NULL == name)
+ name = "";
+ if (NULL == sec)
+ sec = "";
+ if (NULL == dsec)
+ dsec = "";
+ if (NULL == arch)
+ arch = "";
+
+ sform = FORM_NONE;
+ if (NULL != sec && *sec <= '9' && *sec >= '1')
+ sform = FORM_SRC;
+ else if (NULL != sec && *sec == '0') {
+ sec = dsec;
+ sform = FORM_CAT;
}
+
+ of = mandoc_calloc(1, sizeof(struct of));
+ strlcpy(of->file, file, MAXPATHLEN);
+ of->name = name;
+ of->sec = sec;
+ of->dsec = dsec;
+ of->arch = arch;
+ of->sform = sform;
+ of->dform = dform;
+ of->next = ofs;
+ ofs = of;
+
+ /*
+ * Add to unique identifier hash.
+ * Then if it's a source manual and we're going to use source in
+ * favour of catpages, add it to that hash.
+ */
+ inoadd(st, of);
+ fileadd(of);
}
-/* ARGSUSED */
static void
-pmdoc_An(MDOC_ARGS)
+offree(void)
{
-
- if (SEC_AUTHORS != n->sec)
- return;
+ struct of *of;
- buf_appendmdoc(buf, n->child, 0);
- hash_put(hash, buf, TYPE_An);
+ while (NULL != (of = ofs)) {
+ ofs = of->next;
+ free(of);
+ }
}
-static void
-hash_reset(DB **db)
+/*
+ * Run through the files in the global vector "ofs" and add them to the
+ * database specified in "base".
+ *
+ * This handles the parsing scheme itself, using the cues of directory
+ * and filename to determine whether the file is parsable or not.
+ */
+static int
+ofmerge(struct mchars *mc, struct mparse *mp, const char *base)
{
- DB *hash;
+ int form;
+ size_t sz;
+ struct mdoc *mdoc;
+ struct man *man;
+ char buf[MAXPATHLEN];
+ char *bufp;
+ const char *msec, *march, *mtitle, *cp;
+ struct of *of;
+ enum mandoclevel lvl;
+
+ for (of = ofs; NULL != of; of = of->next) {
+ /*
+ * If we're a catpage (as defined by our path), then see
+ * if a manpage exists by the same name (ignoring the
+ * suffix).
+ * If it does, then we want to use it instead of our
+ * own.
+ */
+ if ( ! use_all && FORM_CAT == of->dform) {
+ sz = strlcpy(buf, of->file, MAXPATHLEN);
+ if (sz >= MAXPATHLEN) {
+ WARNING(of->file, base,
+ "Filename too long");
+ continue;
+ }
+ bufp = strstr(buf, "cat");
+ assert(NULL != bufp);
+ memcpy(bufp, "man", 3);
+ if (NULL != (bufp = strrchr(buf, '.')))
+ *++bufp = '\0';
+ strlcat(buf, of->dsec, MAXPATHLEN);
+ if (filecheck(buf)) {
+ WARNING(of->file, base, "Man "
+ "source exists: %s", buf);
+ continue;
+ }
+ }
+
+ words = NULL;
+ mparse_reset(mp);
+ mdoc = NULL;
+ man = NULL;
+ form = 0;
+ msec = of->dsec;
+ march = of->arch;
+ mtitle = of->name;
+
+ /*
+ * Try interpreting the file as mdoc(7) or man(7)
+ * source code, unless it is already known to be
+ * formatted. Fall back to formatted mode.
+ */
+ if (FORM_SRC == of->dform || FORM_SRC == of->sform) {
+ lvl = mparse_readfd(mp, -1, of->file);
+ if (lvl < MANDOCLEVEL_FATAL)
+ mparse_result(mp, &mdoc, &man);
+ }
+
+ if (NULL != mdoc) {
+ form = 1;
+ msec = mdoc_meta(mdoc)->msec;
+ march = mdoc_meta(mdoc)->arch;
+ mtitle = mdoc_meta(mdoc)->title;
+ } else if (NULL != man) {
+ form = 1;
+ msec = man_meta(man)->msec;
+ march = "";
+ mtitle = man_meta(man)->title;
+ }
+
+ if (NULL == msec)
+ msec = "";
+ if (NULL == march)
+ march = "";
+ if (NULL == mtitle)
+ mtitle = "";
- if (NULL != (hash = *db))
- (*hash->close)(hash);
+ /*
+ * Check whether the manual section given in a file
+ * agrees with the directory where the file is located.
+ * Some manuals have suffixes like (3p) on their
+ * section number either inside the file or in the
+ * directory name, some are linked into more than one
+ * section, like encrypt(1) = makekey(8). Do not skip
+ * manuals for such reasons.
+ */
+ if ( ! use_all && form && strcasecmp(msec, of->dsec))
+ WARNING(of->file, base, "Section \"%s\" "
+ "manual in %s directory",
+ msec, of->dsec);
- *db = dbopen(NULL, O_CREAT|O_RDWR, 0644, DB_HASH, NULL);
- if (NULL == *db) {
- perror("hash");
- exit((int)MANDOCLEVEL_SYSERR);
+ /*
+ * Manual page directories exist for each kernel
+ * architecture as returned by machine(1).
+ * However, many manuals only depend on the
+ * application architecture as returned by arch(1).
+ * For example, some (2/ARM) manuals are shared
+ * across the "armish" and "zaurus" kernel
+ * architectures.
+ * A few manuals are even shared across completely
+ * different architectures, for example fdformat(1)
+ * on amd64, i386, sparc, and sparc64.
+ * Thus, warn about architecture mismatches,
+ * but don't skip manuals for this reason.
+ */
+ if ( ! use_all && strcasecmp(march, of->arch))
+ WARNING(of->file, base, "Architecture \"%s\" "
+ "manual in \"%s\" directory",
+ march, of->arch);
+
+ putkey(of, of->name, TYPE_Nm);
+
+ if (NULL != mdoc) {
+ if (NULL != (cp = mdoc_meta(mdoc)->name))
+ putkey(of, cp, TYPE_Nm);
+ parse_mdoc(of, mdoc_node(mdoc));
+ } else if (NULL != man)
+ parse_man(of, man_node(man));
+ else
+ parse_catpage(of, base);
+
+ dbindex(mc, form, of, base);
}
+
+ return(1);
}
-/* ARGSUSED */
static void
-pmdoc_Fd(MDOC_ARGS)
+parse_catpage(struct of *of, const char *base)
{
- const char *start, *end;
- size_t sz;
-
- if (SEC_SYNOPSIS != n->sec)
- return;
- if (NULL == (n = n->child) || MDOC_TEXT != n->type)
+ FILE *stream;
+ char *line, *p, *title;
+ size_t len, plen, titlesz;
+
+ if (NULL == (stream = fopen(of->file, "r"))) {
+ WARNING(of->file, base, "%s", strerror(errno));
return;
+ }
+
+ /* Skip to first blank line. */
+
+ while (NULL != (line = fgetln(stream, &len)))
+ if ('\n' == *line)
+ break;
/*
- * Only consider those `Fd' macro fields that begin with an
- * "inclusion" token (versus, e.g., #define).
+ * Assume the first line that is not indented
+ * is the first section header. Skip to it.
*/
- if (strcmp("#include", n->string))
- return;
-
- if (NULL == (n = n->next) || MDOC_TEXT != n->type)
- return;
+ while (NULL != (line = fgetln(stream, &len)))
+ if ('\n' != *line && ' ' != *line)
+ break;
+
/*
- * Strip away the enclosing angle brackets and make sure we're
- * not zero-length.
+ * Read up until the next section into a buffer.
+ * Strip the leading and trailing newline from each read line,
+ * appending a trailing space.
+ * Ignore empty (whitespace-only) lines.
*/
- start = n->string;
- if ('<' == *start || '"' == *start)
- start++;
+ titlesz = 0;
+ title = NULL;
- if (0 == (sz = strlen(start)))
+ while (NULL != (line = fgetln(stream, &len))) {
+ if (' ' != *line || '\n' != line[len - 1])
+ break;
+ while (len > 0 && isspace((unsigned char)*line)) {
+ line++;
+ len--;
+ }
+ if (1 == len)
+ continue;
+ title = mandoc_realloc(title, titlesz + len);
+ memcpy(title + titlesz, line, len);
+ titlesz += len;
+ title[titlesz - 1] = ' ';
+ }
+
+ /*
+ * If no page content can be found, or the input line
+ * is already the next section header, or there is no
+ * trailing newline, reuse the page title as the page
+ * description.
+ */
+
+ if (NULL == title || '\0' == *title) {
+ WARNING(of->file, base, "Cannot find NAME section");
+ fclose(stream);
+ free(title);
return;
+ }
- end = &start[(int)sz - 1];
- if ('>' == *end || '"' == *end)
- end--;
+ title = mandoc_realloc(title, titlesz + 1);
+ title[titlesz] = '\0';
+
+ /*
+ * Skip to the first dash.
+ * Use the remaining line as the description (no more than 70
+ * bytes).
+ */
+
+ if (NULL != (p = strstr(title, "- "))) {
+ for (p += 2; ' ' == *p || '\b' == *p; p++)
+ /* Skip to next word. */ ;
+ } else {
+ WARNING(of->file, base, "No dash in title line");
+ p = title;
+ }
- assert(end >= start);
+ plen = strlen(p);
- buf_appendb(buf, start, (size_t)(end - start + 1));
- buf_appendb(buf, "", 1);
+ /* Strip backspace-encoding from line. */
- hash_put(hash, buf, TYPE_In);
+ while (NULL != (line = memchr(p, '\b', plen))) {
+ len = line - p;
+ if (0 == len) {
+ memmove(line, line + 1, plen--);
+ continue;
+ }
+ memmove(line - 1, line + 1, plen - len);
+ plen -= 2;
+ }
+
+ of->desc = stradd(p);
+ putkey(of, p, TYPE_Nd);
+ fclose(stream);
+ free(title);
}
-/* ARGSUSED */
+/*
+ * Put a type/word pair into the word database for this particular file.
+ */
static void
-pmdoc_Cd(MDOC_ARGS)
+putkey(const struct of *of, const char *value, uint64_t type)
{
-
- if (SEC_SYNOPSIS != n->sec)
- return;
- buf_appendmdoc(buf, n->child, 0);
- hash_put(hash, buf, TYPE_Cd);
+ assert(NULL != value);
+ wordaddbuf(of, value, strlen(value), type);
}
-/* ARGSUSED */
-static void
-pmdoc_In(MDOC_ARGS)
+/*
+ * Like putkey() but for unterminated strings.
+ */
+static void
+putkeys(const struct of *of, const char *value, int sz, uint64_t type)
{
-
- if (SEC_SYNOPSIS != n->sec)
- return;
- if (NULL == n->child || MDOC_TEXT != n->child->type)
- return;
- buf_append(buf, n->child->string);
- hash_put(hash, buf, TYPE_In);
+ wordaddbuf(of, value, sz, type);
}
-/* ARGSUSED */
+/*
+ * Grok all nodes at or below a certain mdoc node into putkey().
+ */
static void
-pmdoc_Fn(MDOC_ARGS)
+putmdockey(const struct of *of, const struct mdoc_node *n, uint64_t m)
{
- const char *cp;
-
- if (SEC_SYNOPSIS != n->sec)
- return;
- if (NULL == n->child || MDOC_TEXT != n->child->type)
- return;
- /* .Fn "struct type *arg" "foo" */
+ for ( ; NULL != n; n = n->next) {
+ if (NULL != n->child)
+ putmdockey(of, n->child, m);
+ if (MDOC_TEXT == n->type)
+ putkey(of, n->string, m);
+ }
+}
- cp = strrchr(n->child->string, ' ');
- if (NULL == cp)
- cp = n->child->string;
+static int
+parse_man(struct of *of, const struct man_node *n)
+{
+ const struct man_node *head, *body;
+ char *start, *sv, *title;
+ char byte;
+ size_t sz, titlesz;
- /* Strip away pointer symbol. */
+ if (NULL == n)
+ return(0);
- while ('*' == *cp)
- cp++;
+ /*
+ * We're only searching for one thing: the first text child in
+ * the BODY of a NAME section. Since we don't keep track of
+ * sections in -man, run some hoops to find out whether we're in
+ * the correct section or not.
+ */
- buf_append(buf, cp);
- hash_put(hash, buf, TYPE_Fn);
-}
+ if (MAN_BODY == n->type && MAN_SH == n->tok) {
+ body = n;
+ assert(body->parent);
+ if (NULL != (head = body->parent->head) &&
+ 1 == head->nchild &&
+ NULL != (head = (head->child)) &&
+ MAN_TEXT == head->type &&
+ 0 == strcmp(head->string, "NAME") &&
+ NULL != (body = body->child) &&
+ MAN_TEXT == body->type) {
-/* ARGSUSED */
-static void
-pmdoc_St(MDOC_ARGS)
-{
-
- if (SEC_STANDARDS != n->sec)
- return;
- if (NULL == n->child || MDOC_TEXT != n->child->type)
- return;
+ title = NULL;
+ titlesz = 0;
- buf_append(buf, n->child->string);
- hash_put(hash, buf, TYPE_St);
-}
+ /*
+ * Suck the entire NAME section into memory.
+ * Yes, we might run away.
+ * But too many manuals have big, spread-out
+ * NAME sections over many lines.
+ */
-/* ARGSUSED */
-static void
-pmdoc_Xr(MDOC_ARGS)
-{
+ for ( ; NULL != body; body = body->next) {
+ if (MAN_TEXT != body->type)
+ break;
+ if (0 == (sz = strlen(body->string)))
+ continue;
+ title = mandoc_realloc
+ (title, titlesz + sz + 1);
+ memcpy(title + titlesz, body->string, sz);
+ titlesz += sz + 1;
+ title[titlesz - 1] = ' ';
+ }
+ if (NULL == title)
+ return(1);
- if (NULL == (n = n->child))
- return;
+ title = mandoc_realloc(title, titlesz + 1);
+ title[titlesz] = '\0';
- buf_appendb(buf, n->string, strlen(n->string));
+ /* Skip leading space. */
- if (NULL != (n = n->next)) {
- buf_appendb(buf, ".", 1);
- buf_appendb(buf, n->string, strlen(n->string) + 1);
- } else
- buf_appendb(buf, ".", 2);
+ sv = title;
+ while (isspace((unsigned char)*sv))
+ sv++;
+
+ if (0 == (sz = strlen(sv))) {
+ free(title);
+ return(1);
+ }
+
+ /* Erase trailing space. */
+
+ start = &sv[sz - 1];
+ while (start > sv && isspace((unsigned char)*start))
+ *start-- = '\0';
+
+ if (start == sv) {
+ free(title);
+ return(1);
+ }
+
+ start = sv;
+
+ /*
+ * Go through a special heuristic dance here.
+ * Conventionally, one or more manual names are
+ * comma-specified prior to a whitespace, then a
+ * dash, then a description. Try to puzzle out
+ * the name parts here.
+ */
+
+ for ( ;; ) {
+ sz = strcspn(start, " ,");
+ if ('\0' == start[sz])
+ break;
+
+ byte = start[sz];
+ start[sz] = '\0';
+
+ putkey(of, start, TYPE_Nm);
+
+ if (' ' == byte) {
+ start += sz + 1;
+ break;
+ }
+
+ assert(',' == byte);
+ start += sz + 1;
+ while (' ' == *start)
+ start++;
+ }
+
+ if (sv == start) {
+ putkey(of, start, TYPE_Nm);
+ free(title);
+ return(1);
+ }
+
+ while (isspace((unsigned char)*start))
+ start++;
+
+ if (0 == strncmp(start, "-", 1))
+ start += 1;
+ else if (0 == strncmp(start, "\\-\\-", 4))
+ start += 4;
+ else if (0 == strncmp(start, "\\-", 2))
+ start += 2;
+ else if (0 == strncmp(start, "\\(en", 4))
+ start += 4;
+ else if (0 == strncmp(start, "\\(em", 4))
+ start += 4;
- hash_put(hash, buf, TYPE_Xr);
+ while (' ' == *start)
+ start++;
+
+ assert(NULL == of->desc);
+ of->desc = stradd(start);
+ putkey(of, start, TYPE_Nd);
+ free(title);
+ return(1);
+ }
+ }
+
+ for (n = n->child; n; n = n->next)
+ if (parse_man(of, n))
+ return(1);
+
+ return(0);
}
-/* ARGSUSED */
static void
-pmdoc_Vt(MDOC_ARGS)
+parse_mdoc(struct of *of, const struct mdoc_node *n)
{
- const char *start;
+
+ assert(NULL != n);
+ for (n = n->child; NULL != n; n = n->next) {
+ switch (n->type) {
+ case (MDOC_ELEM):
+ /* FALLTHROUGH */
+ case (MDOC_BLOCK):
+ /* FALLTHROUGH */
+ case (MDOC_HEAD):
+ /* FALLTHROUGH */
+ case (MDOC_BODY):
+ /* FALLTHROUGH */
+ case (MDOC_TAIL):
+ if (NULL != mdocs[n->tok].fp)
+ if (0 == (*mdocs[n->tok].fp)(of, n))
+ break;
+
+ if (MDOCF_CHILD & mdocs[n->tok].flags)
+ putmdockey(of, n->child, mdocs[n->tok].mask);
+ break;
+ default:
+ assert(MDOC_ROOT != n->type);
+ continue;
+ }
+ if (NULL != n->child)
+ parse_mdoc(of, n);
+ }
+}
+
+static int
+parse_mdoc_Fd(struct of *of, const struct mdoc_node *n)
+{
+ const char *start, *end;
size_t sz;
-
- if (SEC_SYNOPSIS != n->sec)
- return;
- if (MDOC_Vt == n->tok && MDOC_BODY != n->type)
- return;
- if (NULL == n->last || MDOC_TEXT != n->last->type)
- return;
+
+ if (SEC_SYNOPSIS != n->sec ||
+ NULL == (n = n->child) ||
+ MDOC_TEXT != n->type)
+ return(0);
/*
- * Strip away leading pointer symbol '*' and trailing ';'.
+ * Only consider those `Fd' macro fields that begin with an
+ * "inclusion" token (versus, e.g., #define).
*/
- start = n->last->string;
+ if (strcmp("#include", n->string))
+ return(0);
+
+ if (NULL == (n = n->next) || MDOC_TEXT != n->type)
+ return(0);
+
+ /*
+ * Strip away the enclosing angle brackets and make sure we're
+ * not zero-length.
+ */
- while ('*' == *start)
+ start = n->string;
+ if ('<' == *start || '"' == *start)
start++;
if (0 == (sz = strlen(start)))
- return;
+ return(0);
- if (';' == start[(int)sz - 1])
- sz--;
+ end = &start[(int)sz - 1];
+ if ('>' == *end || '"' == *end)
+ end--;
- if (0 == sz)
- return;
+ if (end > start)
+ putkeys(of, start, end - start + 1, TYPE_In);
+ return(1);
+}
+
+static int
+parse_mdoc_In(struct of *of, const struct mdoc_node *n)
+{
+
+ if (NULL != n->child && MDOC_TEXT == n->child->type)
+ return(0);
- buf_appendb(buf, start, sz);
- buf_appendb(buf, "", 1);
- hash_put(hash, buf, TYPE_Va);
+ putkey(of, n->child->string, TYPE_In);
+ return(1);
}
-/* ARGSUSED */
-static void
-pmdoc_Fo(MDOC_ARGS)
+static int
+parse_mdoc_Fn(struct of *of, const struct mdoc_node *n)
{
-
- if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type)
- return;
+ const char *cp;
+
+ if (NULL == (n = n->child) || MDOC_TEXT != n->type)
+ return(0);
+
+ /*
+ * Parse: .Fn "struct type *name" "char *arg".
+ * First strip away pointer symbol.
+ * Then store the function name, then type.
+ * Finally, store the arguments.
+ */
+
+ if (NULL == (cp = strrchr(n->string, ' ')))
+ cp = n->string;
+
+ while ('*' == *cp)
+ cp++;
+
+ putkey(of, cp, TYPE_Fn);
+
+ if (n->string < cp)
+ putkeys(of, n->string, cp - n->string, TYPE_Ft);
+
+ for (n = n->next; NULL != n; n = n->next)
+ if (MDOC_TEXT == n->type)
+ putkey(of, n->string, TYPE_Fa);
+
+ return(0);
+}
+
+static int
+parse_mdoc_St(struct of *of, const struct mdoc_node *n)
+{
+
if (NULL == n->child || MDOC_TEXT != n->child->type)
- return;
+ return(0);
- buf_append(buf, n->child->string);
- hash_put(hash, buf, TYPE_Fn);
+ putkey(of, n->child->string, TYPE_St);
+ return(1);
}
+static int
+parse_mdoc_Xr(struct of *of, const struct mdoc_node *n)
+{
+
+ if (NULL == (n = n->child))
+ return(0);
-/* ARGSUSED */
-static void
-pmdoc_Nd(MDOC_ARGS)
+ putkey(of, n->string, TYPE_Xr);
+ return(1);
+}
+
+static int
+parse_mdoc_Nd(struct of *of, const struct mdoc_node *n)
{
+ size_t sz;
+ char *sv, *desc;
if (MDOC_BODY != n->type)
- return;
+ return(0);
- buf_appendmdoc(dbuf, n->child, 1);
- buf_appendmdoc(buf, n->child, 0);
+ /*
+ * Special-case the `Nd' because we need to put the description
+ * into the document table.
+ */
- hash_put(hash, buf, TYPE_Nd);
+ desc = NULL;
+ for (n = n->child; NULL != n; n = n->next) {
+ if (MDOC_TEXT == n->type) {
+ sz = strlen(n->string) + 1;
+ if (NULL != (sv = desc))
+ sz += strlen(desc) + 1;
+ desc = mandoc_realloc(desc, sz);
+ if (NULL != sv)
+ strlcat(desc, " ", sz);
+ else
+ *desc = '\0';
+ strlcat(desc, n->string, sz);
+ }
+ if (NULL != n->child)
+ parse_mdoc_Nd(of, n);
+ }
+
+ of->desc = NULL != desc ? stradd(desc) : NULL;
+ free(desc);
+ return(1);
}
-/* ARGSUSED */
-static void
-pmdoc_Er(MDOC_ARGS)
+static int
+parse_mdoc_Nm(struct of *of, const struct mdoc_node *n)
{
- if (SEC_ERRORS != n->sec)
- return;
-
- buf_appendmdoc(buf, n->child, 0);
- hash_put(hash, buf, TYPE_Er);
+ if (SEC_NAME == n->sec)
+ return(1);
+ else if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type)
+ return(0);
+
+ return(1);
}
-/* ARGSUSED */
-static void
-pmdoc_Ev(MDOC_ARGS)
+static int
+parse_mdoc_Sh(struct of *of, const struct mdoc_node *n)
{
- if (SEC_ENVIRONMENT != n->sec)
- return;
-
- buf_appendmdoc(buf, n->child, 0);
- hash_put(hash, buf, TYPE_Ev);
+ return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
}
-/* ARGSUSED */
-static void
-pmdoc_Pa(MDOC_ARGS)
+static int
+parse_mdoc_head(struct of *of, const struct mdoc_node *n)
{
- if (SEC_FILES != n->sec)
- return;
-
- buf_appendmdoc(buf, n->child, 0);
- hash_put(hash, buf, TYPE_Pa);
+ return(MDOC_HEAD == n->type);
}
-/* ARGSUSED */
-static void
-pmdoc_Nm(MDOC_ARGS)
+static int
+parse_mdoc_body(struct of *of, const struct mdoc_node *n)
{
-
- if (SEC_NAME == n->sec) {
- buf_appendmdoc(buf, n->child, 0);
- hash_put(hash, buf, TYPE_Nm);
- return;
- } else if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type)
- return;
- if (NULL == n->child)
- buf_append(buf, m->name);
-
- buf_appendmdoc(buf, n->child, 0);
- hash_put(hash, buf, TYPE_Nm);
+ return(MDOC_BODY == n->type);
}
-static void
-hash_put(DB *db, const struct buf *buf, uint64_t mask)
+/*
+ * See straddbuf().
+ */
+static char *
+stradd(const char *cp)
{
- DBT key, val;
- int rc;
- if (buf->len < 2)
- return;
+ return(straddbuf(cp, strlen(cp)));
+}
- key.data = buf->cp;
- key.size = buf->len;
+/*
+ * This looks up or adds a string to the string table.
+ * The string table is a table of all strings encountered during parse
+ * or file scan.
+ * In using it, we avoid having thousands of (e.g.) "cat1" string
+ * allocations for the "of" table.
+ * We also have a layer atop the string table for keeping track of words
+ * in a parse sequence (see wordaddbuf()).
+ */
+static char *
+straddbuf(const char *cp, size_t sz)
+{
+ struct str *s;
+ unsigned int index;
+ const char *end;
- if ((rc = (*db->get)(db, &key, &val, 0)) < 0) {
- perror("hash");
- exit((int)MANDOCLEVEL_SYSERR);
- } else if (0 == rc)
- mask |= *(uint64_t *)val.data;
+ if (NULL != (s = hashget(cp, sz)))
+ return(s->key);
- val.data = &mask;
- val.size = sizeof(uint64_t);
+ s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
+ memcpy(s->key, cp, sz);
- if ((rc = (*db->put)(db, &key, &val, 0)) < 0) {
- perror("hash");
- exit((int)MANDOCLEVEL_SYSERR);
- }
+ end = cp + sz;
+ index = ohash_qlookupi(&strings, cp, &end);
+ assert(NULL == ohash_find(&strings, index));
+ ohash_insert(&strings, index, s);
+ return(s->key);
}
-static void
-dbt_put(DB *db, const char *dbn, DBT *key, DBT *val)
+static struct str *
+hashget(const char *cp, size_t sz)
{
+ unsigned int index;
+ const char *end;
- assert(key->size);
- assert(val->size);
-
- if (0 == (*db->put)(db, key, val, 0))
- return;
-
- perror(dbn);
- exit((int)MANDOCLEVEL_SYSERR);
- /* NOTREACHED */
+ end = cp + sz;
+ index = ohash_qlookupi(&strings, cp, &end);
+ return(ohash_find(&strings, index));
}
/*
- * Call out to per-macro handlers after clearing the persistent database
- * key. If the macro sets the database key, flush it to the database.
+ * Add a word to the current parse sequence.
+ * Within the hashtable of strings, we maintain a list of strings that
+ * are currently indexed.
+ * Each of these ("words") has a bitmask modified within the parse.
+ * When we finish a parse, we'll dump the list, then remove the head
+ * entry -- since the next parse will have a new "of", it can keep track
+ * of its entries without conflict.
*/
static void
-pmdoc_node(MDOC_ARGS)
+wordaddbuf(const struct of *of,
+ const char *cp, size_t sz, uint64_t v)
{
+ struct str *s;
+ unsigned int index;
+ const char *end;
- if (NULL == n)
+ if (0 == sz)
return;
- switch (n->type) {
- case (MDOC_HEAD):
- /* FALLTHROUGH */
- case (MDOC_BODY):
- /* FALLTHROUGH */
- case (MDOC_TAIL):
- /* FALLTHROUGH */
- case (MDOC_BLOCK):
- /* FALLTHROUGH */
- case (MDOC_ELEM):
- if (NULL == mdocs[n->tok])
- break;
+ s = hashget(cp, sz);
- buf->len = 0;
- (*mdocs[n->tok])(hash, buf, dbuf, n, m);
- break;
- default:
- break;
+ if (NULL != s && of == s->of) {
+ s->mask |= v;
+ return;
+ } else if (NULL == s) {
+ s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
+ memcpy(s->key, cp, sz);
+ end = cp + sz;
+ index = ohash_qlookupi(&strings, cp, &end);
+ assert(NULL == ohash_find(&strings, index));
+ ohash_insert(&strings, index, s);
}
- pmdoc_node(hash, buf, dbuf, n->child, m);
- pmdoc_node(hash, buf, dbuf, n->next, m);
+ s->next = words;
+ s->of = of;
+ s->mask = v;
+ words = s;
}
-static int
-pman_node(MAN_ARGS)
+/*
+ * Take a Unicode codepoint and produce its UTF-8 encoding.
+ * This isn't the best way to do this, but it works.
+ * The magic numbers are from the UTF-8 packaging.
+ * They're not as scary as they seem: read the UTF-8 spec for details.
+ */
+static size_t
+utf8(unsigned int cp, char out[7])
{
- const struct man_node *head, *body;
- const char *start, *sv;
- size_t sz;
-
- if (NULL == n)
+ size_t rc;
+
+ rc = 0;
+ if (cp <= 0x0000007F) {
+ rc = 1;
+ out[0] = (char)cp;
+ } else if (cp <= 0x000007FF) {
+ rc = 2;
+ out[0] = (cp >> 6 & 31) | 192;
+ out[1] = (cp & 63) | 128;
+ } else if (cp <= 0x0000FFFF) {
+ rc = 3;
+ out[0] = (cp >> 12 & 15) | 224;
+ out[1] = (cp >> 6 & 63) | 128;
+ out[2] = (cp & 63) | 128;
+ } else if (cp <= 0x001FFFFF) {
+ rc = 4;
+ out[0] = (cp >> 18 & 7) | 240;
+ out[1] = (cp >> 12 & 63) | 128;
+ out[2] = (cp >> 6 & 63) | 128;
+ out[3] = (cp & 63) | 128;
+ } else if (cp <= 0x03FFFFFF) {
+ rc = 5;
+ out[0] = (cp >> 24 & 3) | 248;
+ out[1] = (cp >> 18 & 63) | 128;
+ out[2] = (cp >> 12 & 63) | 128;
+ out[3] = (cp >> 6 & 63) | 128;
+ out[4] = (cp & 63) | 128;
+ } else if (cp <= 0x7FFFFFFF) {
+ rc = 6;
+ out[0] = (cp >> 30 & 1) | 252;
+ out[1] = (cp >> 24 & 63) | 128;
+ out[2] = (cp >> 18 & 63) | 128;
+ out[3] = (cp >> 12 & 63) | 128;
+ out[4] = (cp >> 6 & 63) | 128;
+ out[5] = (cp & 63) | 128;
+ } else
return(0);
+ out[rc] = '\0';
+ return(rc);
+}
+
+/*
+ * Store the UTF-8 version of a key, or alias the pointer if the key has
+ * no UTF-8 transcription marks in it.
+ */
+static void
+utf8key(struct mchars *mc, struct str *key)
+{
+ size_t sz, bsz, pos;
+ char utfbuf[7], res[5];
+ char *buf;
+ const char *seq, *cpp, *val;
+ int len, u;
+ enum mandoc_esc esc;
+
+ assert(NULL == key->utf8);
+
+ res[0] = '\\';
+ res[1] = '\t';
+ res[2] = ASCII_NBRSP;
+ res[3] = ASCII_HYPH;
+ res[4] = '\0';
+
+ val = key->key;
+ bsz = strlen(val);
+
/*
- * We're only searching for one thing: the first text child in
- * the BODY of a NAME section. Since we don't keep track of
- * sections in -man, run some hoops to find out whether we're in
- * the correct section or not.
+ * Pre-check: if we have no stop-characters, then set the
+ * pointer as ourselvse and get out of here.
*/
+ if (strcspn(val, res) == bsz) {
+ key->utf8 = key->key;
+ return;
+ }
- if (MAN_BODY == n->type && MAN_SH == n->tok) {
- body = n;
- assert(body->parent);
- if (NULL != (head = body->parent->head) &&
- 1 == head->nchild &&
- NULL != (head = (head->child)) &&
- MAN_TEXT == head->type &&
- 0 == strcmp(head->string, "NAME") &&
- NULL != (body = body->child) &&
- MAN_TEXT == body->type) {
+ /* Pre-allocate by the length of the input */
- assert(body->string);
- start = sv = body->string;
+ buf = mandoc_malloc(++bsz);
+ pos = 0;
- /*
- * Go through a special heuristic dance here.
- * This is why -man manuals are great!
- * (I'm being sarcastic: my eyes are bleeding.)
- * Conventionally, one or more manual names are
- * comma-specified prior to a whitespace, then a
- * dash, then a description. Try to puzzle out
- * the name parts here.
- */
+ while ('\0' != *val) {
+ /*
+ * Halt on the first escape sequence.
+ * This also halts on the end of string, in which case
+ * we just copy, fallthrough, and exit the loop.
+ */
+ if ((sz = strcspn(val, res)) > 0) {
+ memcpy(&buf[pos], val, sz);
+ pos += sz;
+ val += sz;
+ }
- for ( ;; ) {
- sz = strcspn(start, " ,");
- if ('\0' == start[(int)sz])
- break;
+ if (ASCII_HYPH == *val) {
+ buf[pos++] = '-';
+ val++;
+ continue;
+ } else if ('\t' == *val || ASCII_NBRSP == *val) {
+ buf[pos++] = ' ';
+ val++;
+ continue;
+ } else if ('\\' != *val)
+ break;
- buf->len = 0;
- buf_appendb(buf, start, sz);
- buf_appendb(buf, "", 1);
+ /* Read past the slash. */
- hash_put(hash, buf, TYPE_Nm);
+ val++;
+ u = 0;
- if (' ' == start[(int)sz]) {
- start += (int)sz + 1;
- break;
- }
+ /*
+ * Parse the escape sequence and see if it's a
+ * predefined character or special character.
+ */
+ esc = mandoc_escape
+ ((const char **)&val, &seq, &len);
+ if (ESCAPE_ERROR == esc)
+ break;
- assert(',' == start[(int)sz]);
- start += (int)sz + 1;
- while (' ' == *start)
- start++;
- }
+ if (ESCAPE_SPECIAL != esc)
+ continue;
+ if (0 == (u = mchars_spec2cp(mc, seq, len)))
+ continue;
- buf->len = 0;
+ /*
+ * If we have a Unicode codepoint, try to convert that
+ * to a UTF-8 byte string.
+ */
+ cpp = utfbuf;
+ if (0 == (sz = utf8(u, utfbuf)))
+ continue;
- if (sv == start) {
- buf_append(buf, start);
- return(1);
- }
+ /* Copy the rendered glyph into the stream. */
- while (' ' == *start)
- start++;
+ sz = strlen(cpp);
+ bsz += sz;
- if (0 == strncmp(start, "-", 1))
- start += 1;
- else if (0 == strncmp(start, "\\-", 2))
- start += 2;
- else if (0 == strncmp(start, "\\(en", 4))
- start += 4;
- else if (0 == strncmp(start, "\\(em", 4))
- start += 4;
+ buf = mandoc_realloc(buf, bsz);
- while (' ' == *start)
- start++;
+ memcpy(&buf[pos], cpp, sz);
+ pos += sz;
+ }
- sz = strlen(start) + 1;
- buf_appendb(dbuf, start, sz);
- buf_appendb(buf, start, sz);
+ buf[pos] = '\0';
+ key->utf8 = buf;
+}
- hash_put(hash, buf, TYPE_Nd);
- }
+/*
+ * Flush the current page's terms (and their bits) into the database.
+ * Wrap the entire set of additions in a transaction to make sqlite be a
+ * little faster.
+ * Also, UTF-8-encode the description at the last possible moment.
+ */
+static void
+dbindex(struct mchars *mc, int form,
+ const struct of *of, const char *base)
+{
+ struct str *key;
+ const char *desc;
+ int64_t recno;
+ size_t i;
+
+ DEBUG(of->file, base, "Adding to index");
+
+ if (nodb)
+ return;
+
+ desc = "";
+ if (NULL != of->desc) {
+ key = hashget(of->desc, strlen(of->desc));
+ assert(NULL != key);
+ if (NULL == key->utf8)
+ utf8key(mc, key);
+ desc = key->utf8;
}
- for (n = n->child; n; n = n->next)
- if (pman_node(hash, buf, dbuf, n))
- return(1);
+ SQL_EXEC("BEGIN TRANSACTION");
+
+ i = 1;
+ SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, of->file);
+ SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, of->sec);
+ SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, of->arch);
+ SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, desc);
+ SQL_BIND_INT(stmts[STMT_INSERT_DOC], i, form);
+ SQL_STEP(stmts[STMT_INSERT_DOC]);
+ recno = sqlite3_last_insert_rowid(db);
+ sqlite3_reset(stmts[STMT_INSERT_DOC]);
+
+ for (key = words; NULL != key; key = key->next) {
+ assert(key->of == of);
+ if (NULL == key->utf8)
+ utf8key(mc, key);
+ i = 1;
+ SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
+ SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->utf8);
+ SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, recno);
+ SQL_STEP(stmts[STMT_INSERT_KEY]);
+ sqlite3_reset(stmts[STMT_INSERT_KEY]);
+ }
- return(0);
+ SQL_EXEC("END TRANSACTION");
}
static void
-ofile_argbuild(char *argv[], int argc, int verb, struct of **of)
+dbprune(const char *base)
{
- int i;
- struct of *nof;
-
- for (i = 0; i < argc; i++) {
- nof = mandoc_calloc(1, sizeof(struct of));
- nof->fname = strdup(argv[i]);
- if (verb > 2)
- printf("%s: Scheduling\n", argv[i]);
- if (NULL == *of) {
- *of = nof;
- (*of)->first = nof;
- } else {
- nof->first = (*of)->first;
- (*of)->next = nof;
- *of = nof;
- }
+ struct of *of;
+ size_t i;
+
+ if (nodb)
+ return;
+
+ for (of = ofs; NULL != of; of = of->next) {
+ i = 1;
+ SQL_BIND_TEXT(stmts[STMT_DELETE], i, of->file);
+ SQL_STEP(stmts[STMT_DELETE]);
+ sqlite3_reset(stmts[STMT_DELETE]);
+ DEBUG(of->file, base, "Deleted from index");
}
}
/*
- * Recursively build up a list of files to parse.
- * We use this instead of ftw() and so on because I don't want global
- * variables hanging around.
- * This ignores the mandoc.db and mandoc.index files, but assumes that
- * everything else is a manual.
- * Pass in a pointer to a NULL structure for the first invocation.
+ * Close an existing database and its prepared statements.
+ * If "real" is not set, rename the temporary file into the real one.
+ */
+static void
+dbclose(const char *base, int real)
+{
+ size_t i;
+ char file[MAXPATHLEN];
+
+ if (nodb)
+ return;
+
+ for (i = 0; i < STMT__MAX; i++) {
+ sqlite3_finalize(stmts[i]);
+ stmts[i] = NULL;
+ }
+
+ sqlite3_close(db);
+ db = NULL;
+
+ if (real)
+ return;
+
+ strlcpy(file, MANDOC_DB, MAXPATHLEN);
+ strlcat(file, "~", MAXPATHLEN);
+ if (-1 == rename(file, MANDOC_DB))
+ perror(MANDOC_DB);
+}
+
+/*
+ * This is straightforward stuff.
+ * Open a database connection to a "temporary" database, then open a set
+ * of prepared statements we'll use over and over again.
+ * If "real" is set, we use the existing database; if not, we truncate a
+ * temporary one.
+ * Must be matched by dbclose().
*/
static int
-ofile_dirbuild(const char *dir, int verb, struct of **of)
+dbopen(const char *base, int real)
{
- char buf[MAXPATHLEN];
+ char file[MAXPATHLEN];
+ const char *sql;
+ int rc, ofl;
size_t sz;
- DIR *d;
- const char *fn;
- struct of *nof;
- struct dirent *dp;
- if (NULL == (d = opendir(dir))) {
- perror(dir);
+ if (nodb)
+ return(1);
+
+ sz = strlcpy(file, MANDOC_DB, MAXPATHLEN);
+ if ( ! real)
+ sz = strlcat(file, "~", MAXPATHLEN);
+
+ if (sz >= MAXPATHLEN) {
+ fprintf(stderr, "%s: Path too long\n", file);
return(0);
}
- while (NULL != (dp = readdir(d))) {
- fn = dp->d_name;
- if (DT_DIR == dp->d_type) {
- if (0 == strcmp(".", fn))
- continue;
- if (0 == strcmp("..", fn))
- continue;
+ if ( ! real)
+ remove(file);
- buf[0] = '\0';
- strlcat(buf, dir, MAXPATHLEN);
- strlcat(buf, "/", MAXPATHLEN);
- sz = strlcat(buf, fn, MAXPATHLEN);
+ ofl = SQLITE_OPEN_READWRITE |
+ (0 == real ? SQLITE_OPEN_EXCLUSIVE : 0);
- if (sz < MAXPATHLEN) {
- if ( ! ofile_dirbuild(buf, verb, of))
- return(0);
- continue;
- } else if (sz < MAXPATHLEN)
- continue;
+ rc = sqlite3_open_v2(file, &db, ofl, NULL);
+ if (SQLITE_OK == rc)
+ return(1);
+ if (SQLITE_CANTOPEN != rc) {
+ perror(file);
+ return(0);
+ }
- fprintf(stderr, "%s: Path too long\n", dir);
- return(0);
- }
- if (DT_REG != dp->d_type)
- continue;
+ sqlite3_close(db);
+ db = NULL;
- if (0 == strcmp(MANDOC_DB, fn) ||
- 0 == strcmp(MANDOC_IDX, fn))
- continue;
+ if (SQLITE_OK != (rc = sqlite3_open(file, &db))) {
+ perror(file);
+ return(0);
+ }
- buf[0] = '\0';
- strlcat(buf, dir, MAXPATHLEN);
- strlcat(buf, "/", MAXPATHLEN);
- sz = strlcat(buf, fn, MAXPATHLEN);
- if (sz >= MAXPATHLEN) {
- fprintf(stderr, "%s: Path too long\n", dir);
- return(0);
- }
+ sql = "CREATE TABLE \"docs\" (\n"
+ " \"file\" TEXT NOT NULL,\n"
+ " \"sec\" TEXT NOT NULL,\n"
+ " \"arch\" TEXT NOT NULL,\n"
+ " \"desc\" TEXT NOT NULL,\n"
+ " \"form\" INTEGER NOT NULL,\n"
+ " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
+ ");\n"
+ "\n"
+ "CREATE TABLE \"keys\" (\n"
+ " \"bits\" INTEGER NOT NULL,\n"
+ " \"key\" TEXT NOT NULL,\n"
+ " \"docid\" INTEGER NOT NULL REFERENCES docs(id) "
+ "ON DELETE CASCADE,\n"
+ " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
+ ");\n"
+ "\n"
+ "CREATE INDEX \"key_index\" ON keys (key);\n";
+
+ if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
+ perror(sqlite3_errmsg(db));
+ return(0);
+ }
- nof = mandoc_calloc(1, sizeof(struct of));
- nof->fname = mandoc_strdup(buf);
+ sql = "DELETE FROM docs where file=?";
+ sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE], NULL);
+ sql = "INSERT INTO docs "
+ "(file,sec,arch,desc,form) VALUES (?,?,?,?,?)";
+ sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_DOC], NULL);
+ sql = "INSERT INTO keys "
+ "(bits,key,docid) VALUES (?,?,?)";
+ sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
+ return(1);
+}
- if (verb > 2)
- printf("%s: Scheduling\n", buf);
+static void *
+hash_halloc(size_t sz, void *arg)
+{
- if (NULL == *of) {
- *of = nof;
- (*of)->first = nof;
- } else {
- nof->first = (*of)->first;
- (*of)->next = nof;
- *of = nof;
- }
- }
+ return(mandoc_calloc(sz, 1));
+}
- closedir(d);
- return(1);
+static void *
+hash_alloc(size_t sz, void *arg)
+{
+
+ return(mandoc_malloc(sz));
}
static void
-ofile_free(struct of *of)
+hash_free(void *p, size_t sz, void *arg)
{
- struct of *nof;
- while (of) {
- nof = of->next;
- free(of->fname);
- free(of);
- of = nof;
- }
+ free(p);
}
-static void
-usage(void)
+static int
+path_reset(const char *cwd, int fd, const char *base)
{
- fprintf(stderr, "usage: %s [-v] "
- "[-d dir [files...] |"
- " -u dir [files...] |"
- " dir...]\n", progname);
+ if (-1 == fchdir(fd)) {
+ perror(cwd);
+ return(0);
+ } else if (-1 == chdir(base)) {
+ perror(base);
+ return(0);
+ }
+ return(1);
}
git.ckatri.com / mandoc.git / blobdiff