X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=index%2Fextract.c;h=1830850ce6a7aad046c90b413ef2c5db4ede0daf;hb=fc7107844c9ec8ea23e680ca4f3231923db4e9c5;hp=8464bb48ca44ef34942a064a35a5a6fc8f1a67c6;hpb=5bdf6c8ddbafa8515a82079f631ef172d98e055b;p=idzebra-moved-to-github.git diff --git a/index/extract.c b/index/extract.c index 8464bb4..1830850 100644 --- a/index/extract.c +++ b/index/extract.c @@ -4,7 +4,25 @@ * Sebastian Hammer, Adam Dickmeiss * * $Log: extract.c,v $ - * Revision 1.15 1995-10-02 15:42:53 adam + * Revision 1.21 1995-10-10 12:24:38 adam + * Temporary sort files are compressed. + * + * Revision 1.20 1995/10/06 13:52:05 adam + * Bug fixes. Handler may abort further scanning. + * + * Revision 1.19 1995/10/04 12:55:16 adam + * Bug fix in ranked search. Use=Any keys inserted. + * + * Revision 1.18 1995/10/04 09:37:08 quinn + * Fixed bug. + * + * Revision 1.17 1995/10/03 14:28:57 adam + * Buffered read in extract works. + * + * Revision 1.16 1995/10/03 14:28:45 adam + * Work on more effecient read handler in extract. + * + * Revision 1.15 1995/10/02 15:42:53 adam * Extract uses file descriptors instead of FILE pointers. * * Revision 1.14 1995/10/02 15:29:13 adam @@ -104,12 +122,73 @@ void key_open (int mem) exit (1); } } + +struct encode_info { + int sysno; + int seqno; + char buf[512]; +}; + +void encode_key_init (struct encode_info *i) +{ + i->sysno = 0; + i->seqno = 0; +} + +char *encode_key_int (int d, char *bp) +{ + if (d <= 63) + *bp++ = d; + else if (d <= 16383) + { + *bp++ = 64 + (d>>8); + *bp++ = d & 255; + } + else if (d <= 4194303) + { + *bp++ = 128 + (d>>16); + *bp++ = (d>>8) & 255; + *bp++ = d & 255; + } + else + { + *bp++ = 192 + (d>>24); + *bp++ = (d>>16) & 255; + *bp++ = (d>>8) & 255; + *bp++ = d & 255; + } + return bp; +} + +void encode_key_write (char *k, struct encode_info *i, FILE *outf) +{ + struct it_key key; + char *bp = i->buf; + + while ((*bp++ = *k++)) + ; + memcpy (&key, k+1, sizeof(struct it_key)); + bp = encode_key_int ( (key.sysno - i->sysno) * 2 + *k, bp); + if (i->sysno != key.sysno) + { + i->sysno = key.sysno; + i->seqno = 0; + } + bp = encode_key_int (key.seqno - i->seqno, bp); + i->seqno = key.seqno; + if (fwrite (i->buf, bp - i->buf, 1, outf) != 1) + { + logf (LOG_FATAL|LOG_ERRNO, "fwrite"); + exit (1); + } +} void key_flush (void) { FILE *outf; char out_fname[200]; char *prevcp, *cp; + struct encode_info encode_info; if (ptr_i <= 0) return; @@ -127,33 +206,19 @@ void key_flush (void) logf (LOG_LOG, "writing section %d", key_file_no); prevcp = cp = key_buf[ptr_top-ptr_i]; - if (fwrite (cp, strlen (cp)+2+sizeof(struct it_key), 1, outf) != 1) - { - logf (LOG_FATAL|LOG_ERRNO, "fwrite %s", out_fname); - exit (1); - } + encode_key_init (&encode_info); + encode_key_write (cp, &encode_info, outf); while (--ptr_i > 0) { cp = key_buf[ptr_top-ptr_i]; if (strcmp (cp, prevcp)) { - if (fwrite (cp, strlen (cp)+2+sizeof(struct it_key), 1, - outf) != 1) - { - logf (LOG_FATAL|LOG_ERRNO, "fwrite %s", out_fname); - exit (1); - } + encode_key_init (&encode_info); + encode_key_write (cp, &encode_info, outf); prevcp = cp; } else - { - cp = strlen (cp) + cp; - if (fwrite (cp, 2+sizeof(struct it_key), 1, outf) != 1) - { - logf (LOG_FATAL|LOG_ERRNO, "fwrite %s", out_fname); - exit (1); - } - } + encode_key_write (cp + strlen(cp), &encode_info, outf); } if (fclose (outf)) { @@ -210,11 +275,89 @@ static void wordAdd (const RecWord *p) kused += sizeof(key); } +static void wordAddAny (const RecWord *p) +{ + if (p->attrSet != 1 || p->attrUse != 1016) + { + RecWord w; + + memcpy (&w, p, sizeof(w)); + w.attrSet = 1; + w.attrUse = 1016; + wordAdd (&w); + } + wordAdd (p); +} + + +#define FILE_READ_BUF 1 +#if FILE_READ_BUF +static char *file_buf; +static int file_offset; +static int file_bufsize; + +static void file_read_start (int fd) +{ + file_offset = 0; + file_buf = xmalloc (4096); + file_bufsize = read (fd, file_buf, 4096); +} + +static void file_read_stop (int fd) +{ + xfree (file_buf); +} + +static int file_read (int fd, char *buf, size_t count) +{ + int l = file_bufsize - file_offset; + + if (count > l) + { + int r; + if (l > 0) + memcpy (buf, file_buf + file_offset, l); + count = count-l; + if (count > file_bufsize) + { + if ((r = read (fd, buf + l, count)) == -1) + { + logf (LOG_FATAL|LOG_ERRNO, "read"); + exit (1); + } + file_bufsize = 0; + file_offset = 0; + return r; + } + file_bufsize = r = read (fd, file_buf, 4096); + if (r == -1) + { + logf (LOG_FATAL|LOG_ERRNO, "read"); + exit (1); + } + else if (r <= count) + { + file_offset = r; + memcpy (buf + l, file_buf, r); + return l + r; + } + else + { + file_offset = count; + memcpy (buf + l, file_buf, count - l); + return count; + } + } + memcpy (buf, file_buf + file_offset, count); + file_offset += count; + return count; +} +#else static int file_read (int fd, char *buf, size_t count) { return read (fd, buf, count); } - +#endif void file_extract (int cmd, const char *fname, const char *kname) { int i, r; @@ -262,11 +405,17 @@ void file_extract (int cmd, const char *fname, const char *kname) } extractCtrl.subType = ""; extractCtrl.init = wordInit; - extractCtrl.add = wordAdd; + extractCtrl.add = wordAddAny; +#if FILE_READ_BUF + file_read_start (extractCtrl.fd); +#endif extractCtrl.readf = file_read; key_sysno = sysno; key_cmd = cmd; r = (*rt->extract)(&extractCtrl); +#if FILE_READ_BUF + file_read_stop (extractCtrl.fd); +#endif close (extractCtrl.fd); if (r) logf (LOG_WARN, "Couldn't extract file %s, code %d", fname, r);