* Sebastian Hammer, Adam Dickmeiss
*
* $Log: kinput.c,v $
- * Revision 1.8 1995-10-04 16:57:19 adam
+ * Revision 1.14 1996-02-07 14:06:37 adam
+ * Better progress report during register merge.
+ * New command: clean - removes temporary shadow files.
+ *
+ * Revision 1.13 1996/02/05 12:30:00 adam
+ * Logging reduced a bit.
+ * The remaining running time is estimated during register merge.
+ *
+ * Revision 1.12 1995/12/06 17:49:19 adam
+ * Uses dict_delete now.
+ *
+ * Revision 1.11 1995/12/06 16:06:43 adam
+ * Better diagnostics. Work on 'real' dictionary deletion.
+ *
+ * Revision 1.10 1995/12/06 12:41:22 adam
+ * New command 'stat' for the index program.
+ * Filenames can be read from stdin by specifying '-'.
+ * Bug fix/enhancement of the transformation from terms to regular
+ * expressons in the search engine.
+ *
+ * Revision 1.9 1995/10/10 12:24:39 adam
+ * Temporary sort files are compressed.
+ *
+ * Revision 1.8 1995/10/04 16:57:19 adam
* Key input and merge sort in one pass.
*
* Revision 1.7 1995/10/02 15:18:52 adam
#include <fcntl.h>
#include <unistd.h>
#include <stdlib.h>
+#include <time.h>
#include <string.h>
#include <stdio.h>
#include <ctype.h>
static int no_diffs = 0;
static int no_updates = 0;
+static int no_deletions = 0;
static int no_insertions = 0;
static int no_iterations = 0;
-static int read_one (FILE *inf, char *name, char *key)
-{
- int c;
- int i = 0;
- do
- {
- if ((c=getc(inf)) == EOF)
- return 0;
- name[i++] = c;
- } while (c);
- for (i = 0; i<KEY_SIZE; i++)
- ((char *)key)[i] = getc (inf);
- ++no_iterations;
- return 1;
-}
-
-static int inp (Dict dict, ISAM isam, const char *name)
-{
- FILE *inf;
- char *info;
- char next_name[INP_NAME_MAX+1];
- char cur_name[INP_NAME_MAX+1];
- int key_buf_size = INP_BUF_START;
- int key_buf_ptr;
- char *next_key;
- char *key_buf;
- int more;
-
- next_key = xmalloc (KEY_SIZE);
- key_buf = xmalloc (key_buf_size * (KEY_SIZE));
- if (!(inf = fopen (name, "r")))
- {
- logf (LOG_FATAL|LOG_ERRNO, "cannot open `%s'", name);
- exit (1);
- }
- more = read_one (inf, cur_name, key_buf);
- while (more) /* EOF ? */
- {
- int nmemb;
- key_buf_ptr = KEY_SIZE;
- while (1)
- {
- if (!(more = read_one (inf, next_name, next_key)))
- break;
- if (*next_name && strcmp (next_name, cur_name))
- break;
- memcpy (key_buf + key_buf_ptr, next_key, KEY_SIZE);
- key_buf_ptr += KEY_SIZE;
- if (key_buf_ptr+KEY_SIZE >= key_buf_size)
- {
- char *new_key_buf;
- new_key_buf = xmalloc (key_buf_size + INP_BUF_ADD);
- memcpy (new_key_buf, key_buf, key_buf_size);
- key_buf_size += INP_BUF_ADD;
- xfree (key_buf);
- key_buf = new_key_buf;
- }
- }
- no_diffs++;
- nmemb = key_buf_ptr / KEY_SIZE;
- assert (nmemb*KEY_SIZE == key_buf_ptr);
- if ((info = dict_lookup (dict, cur_name)))
- {
- ISAM_P isam_p, isam_p2;
- logf (LOG_DEBUG, "updating %s", cur_name);
- no_updates++;
- memcpy (&isam_p, info+1, sizeof(ISAM_P));
- isam_p2 = is_merge (isam, isam_p, nmemb, key_buf);
- if (isam_p2 != isam_p)
- dict_insert (dict, cur_name, sizeof(ISAM_P), &isam_p2);
- }
- else
- {
- ISAM_P isam_p;
- logf (LOG_DEBUG, "inserting %s", cur_name);
- no_insertions++;
- isam_p = is_merge (isam, 0, nmemb, key_buf);
- dict_insert (dict, cur_name, sizeof(ISAM_P), &isam_p);
- }
- memcpy (key_buf, next_key, KEY_SIZE);
- strcpy (cur_name, next_name);
- }
- fclose (inf);
- return 0;
-}
-
-void key_input (const char *dict_fname, const char *isam_fname,
- const char *key_fname, int cache)
-{
- Dict dict;
- ISAM isam;
-
- dict = dict_open (dict_fname, cache, 1);
- if (!dict)
- {
- logf (LOG_FATAL, "dict_open fail of `%s'", dict_fname);
- exit (1);
- }
- isam = is_open (isam_fname, key_compare, 1, sizeof(struct it_key));
- if (!isam)
- {
- logf (LOG_FATAL, "is_open fail of `%s'", isam_fname);
- exit (1);
- }
- inp (dict, isam, key_fname);
- dict_close (dict);
- is_close (isam);
- logf (LOG_LOG, "Iterations . . .%7d", no_iterations);
- logf (LOG_LOG, "Distinct words .%7d", no_diffs);
- logf (LOG_LOG, "Updates. . . . .%7d", no_updates);
- logf (LOG_LOG, "Insertions . . .%7d", no_insertions);
-}
-
-
struct key_file {
int no; /* file no */
off_t offset; /* file offset */
size_t chunk; /* number of bytes allocated */
size_t buf_ptr; /* current position in buffer */
char *prev_name; /* last word read */
+ int sysno; /* last sysno */
+ int seqno; /* last seqno */
+ off_t length; /* length of file */
+ /* handler invoked in each read */
+ void (*readHandler)(struct key_file *keyp, void *rinfo);
+ void *readInfo;
};
+void getFnameTmp (char *fname, int no)
+{
+ sprintf (fname, TEMP_FNAME, no);
+}
+
void key_file_chunk_read (struct key_file *f)
{
int nr = 0, r, fd;
char fname[256];
- sprintf (fname, TEMP_FNAME, f->no);
+ getFnameTmp (fname, f->no);
fd = open (fname, O_RDONLY);
if (fd == -1)
{
logf (LOG_FATAL|LOG_ERRNO, "cannot open %s", fname);
exit (1);
}
- logf (LOG_LOG, "reading chunk from %s", fname);
+ if (!f->length)
+ {
+ if ((f->length = lseek (fd, 0L, SEEK_END)) == (off_t) -1)
+ {
+ logf (LOG_FATAL|LOG_ERRNO, "cannot seek %s", fname);
+ exit (1);
+ }
+ }
if (lseek (fd, f->offset, SEEK_SET) == -1)
{
logf (LOG_FATAL|LOG_ERRNO, "cannot seek %s", fname);
}
f->buf_size = nr;
f->buf_ptr = 0;
+ if (f->readHandler)
+ (*f->readHandler)(f, f->readInfo);
close (fd);
}
struct key_file *f;
f = xmalloc (sizeof(*f));
+ f->sysno = 0;
+ f->seqno = 0;
f->no = no;
f->chunk = chunk;
f->offset = 0;
+ f->length = 0;
+ f->readHandler = NULL;
f->buf = xmalloc (f->chunk);
f->prev_name = xmalloc (256);
*f->prev_name = '\0';
return EOF;
}
+int key_file_decode (struct key_file *f)
+{
+ int c, d;
+
+ c = key_file_getc (f);
+ switch (c & 192)
+ {
+ case 0:
+ d = c;
+ break;
+ case 64:
+ d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
+ break;
+ case 128:
+ d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
+ d = (d << 8) + (key_file_getc (f) & 0xff);
+ break;
+ case 192:
+ d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
+ d = (d << 8) + (key_file_getc (f) & 0xff);
+ d = (d << 8) + (key_file_getc (f) & 0xff);
+ break;
+ }
+ return d;
+}
+
int key_file_read (struct key_file *f, char *key)
{
- int i, j, c;
+ int i, d, c;
+ struct it_key itkey;
c = key_file_getc (f);
if (c == 0)
while ((key[i++] = key_file_getc (f)))
;
strcpy (f->prev_name, key);
+ f->sysno = 0;
}
- for (j = KEY_SIZE; --j >= 0; )
- key[i++] = key_file_getc (f);
- return i;
+ d = key_file_decode (f);
+ key[i++] = d & 1;
+ d = d >> 1;
+ itkey.sysno = d + f->sysno;
+ if (d)
+ {
+ f->sysno = itkey.sysno;
+ f->seqno = 0;
+ }
+ d = key_file_decode (f);
+ itkey.seqno = d + f->seqno;
+ f->seqno = itkey.seqno;
+ memcpy (key + i, &itkey, sizeof(struct it_key));
+ return i + sizeof (struct it_key);
}
struct heap_info {
assert (hi->heapnum > 0);
-#if 1
key_heap_swap (hi, 1, hi->heapnum);
hi->heapnum--;
-#else
- hi->ptr[1] = hi->ptr[hi->heapnum--];
-#endif
while (child <= hi->heapnum) {
if (child < hi->heapnum &&
(*hi->cmp)(&hi->info.buf[hi->ptr[child]],
{
ISAM_P isam_p, isam_p2;
logf (LOG_DEBUG, "updating %s", cur_name);
- no_updates++;
memcpy (&isam_p, info+1, sizeof(ISAM_P));
isam_p2 = is_merge (isam, isam_p, nmemb, key_buf);
- if (isam_p2 != isam_p)
- dict_insert (dict, cur_name, sizeof(ISAM_P), &isam_p2);
+ if (!isam_p2)
+ {
+ no_deletions++;
+ if (!dict_delete (dict, cur_name))
+ abort ();
+ }
+ else
+ {
+ no_updates++;
+ if (isam_p2 != isam_p)
+ dict_insert (dict, cur_name, sizeof(ISAM_P), &isam_p2);
+ }
}
else
{
return 0;
}
-void key_input2 (const char *dict_fname, const char *isam_fname,
- int nkeys, int cache)
+struct progressInfo {
+ time_t startTime;
+ time_t lastTime;
+ off_t totalBytes;
+ off_t totalOffset;
+};
+
+void progressFunc (struct key_file *keyp, void *info)
+{
+ struct progressInfo *p = info;
+ time_t now, remaining;
+
+ if (keyp->buf_size <= 0 || p->totalBytes <= 0)
+ return ;
+ time (&now);
+
+ if (now >= p->lastTime+10)
+ {
+ p->lastTime = now;
+ remaining = (now - p->startTime)*
+ ((double) p->totalBytes/p->totalOffset - 1.0);
+ if (remaining <= 130)
+ logf (LOG_LOG, "Merge %2.1f%% completed; %ld seconds remaining",
+ (100.0*p->totalOffset) / p->totalBytes, (long) remaining);
+ else
+ logf (LOG_LOG, "Merge %2.1f%% completed; %ld minutes remaining",
+ (100.0*p->totalOffset) / p->totalBytes, (long) remaining/60);
+ }
+ p->totalOffset += keyp->buf_size;
+}
+
+void key_input (const char *dict_fname, const char *isam_fname,
+ int nkeys, int cache)
+
{
Dict dict;
ISAM isam;
char rbuf[1024];
int i, r;
struct heap_info *hi;
+ struct progressInfo progressInfo;
dict = dict_open (dict_fname, cache, 1);
if (!dict)
}
kf = xmalloc ((1+nkeys) * sizeof(*kf));
+ progressInfo.totalBytes = 0;
+ progressInfo.totalOffset = 0;
+ time (&progressInfo.startTime);
+ time (&progressInfo.lastTime);
for (i = 1; i<=nkeys; i++)
+ {
kf[i] = key_file_init (i, 32768);
-
+ kf[i]->readHandler = progressFunc;
+ kf[i]->readInfo = &progressInfo;
+ progressInfo.totalBytes += kf[i]->length;
+ progressInfo.totalOffset += kf[i]->buf_size;
+ }
hi = key_heap_init (nkeys, key_qsort_compare);
for (i = 1; i<=nkeys; i++)
if ((r = key_file_read (kf[i], rbuf)))
dict_close (dict);
is_close (isam);
-
+
+ for (i = 1; i<=nkeys; i++)
+ {
+ getFnameTmp (rbuf, i);
+ unlink (rbuf);
+ }
logf (LOG_LOG, "Iterations . . .%7d", no_iterations);
logf (LOG_LOG, "Distinct words .%7d", no_diffs);
logf (LOG_LOG, "Updates. . . . .%7d", no_updates);
+ logf (LOG_LOG, "Deletions. . . .%7d", no_deletions);
logf (LOG_LOG, "Insertions . . .%7d", no_insertions);
}