2 * Copyright (C) 1994-1995, Index Data I/S
4 * Sebastian Hammer, Adam Dickmeiss
7 * Revision 1.14 1996-02-07 14:06:37 adam
8 * Better progress report during register merge.
9 * New command: clean - removes temporary shadow files.
11 * Revision 1.13 1996/02/05 12:30:00 adam
12 * Logging reduced a bit.
13 * The remaining running time is estimated during register merge.
15 * Revision 1.12 1995/12/06 17:49:19 adam
16 * Uses dict_delete now.
18 * Revision 1.11 1995/12/06 16:06:43 adam
19 * Better diagnostics. Work on 'real' dictionary deletion.
21 * Revision 1.10 1995/12/06 12:41:22 adam
22 * New command 'stat' for the index program.
23 * Filenames can be read from stdin by specifying '-'.
24 * Bug fix/enhancement of the transformation from terms to regular
25 * expressons in the search engine.
27 * Revision 1.9 1995/10/10 12:24:39 adam
28 * Temporary sort files are compressed.
30 * Revision 1.8 1995/10/04 16:57:19 adam
31 * Key input and merge sort in one pass.
33 * Revision 1.7 1995/10/02 15:18:52 adam
34 * New member in recRetrieveCtrl: diagnostic.
36 * Revision 1.6 1995/09/29 15:51:56 adam
37 * First work on multi-way read.
39 * Revision 1.5 1995/09/29 14:01:43 adam
42 * Revision 1.4 1995/09/28 14:22:57 adam
43 * Sort uses smaller temporary files.
45 * Revision 1.3 1995/09/06 16:11:17 adam
46 * Option: only one word key per file.
48 * Revision 1.2 1995/09/04 12:33:42 adam
49 * Various cleanup. YAZ util used instead.
51 * Revision 1.1 1995/09/04 09:10:37 adam
52 * More work on index add/del/update.
53 * Merge sort implemented.
54 * Initial work on z39 server.
69 #define KEY_SIZE (1+sizeof(struct it_key))
70 #define INP_NAME_MAX 8192
71 #define INP_BUF_START 60000
72 #define INP_BUF_ADD 400000
74 static int no_diffs = 0;
75 static int no_updates = 0;
76 static int no_deletions = 0;
77 static int no_insertions = 0;
78 static int no_iterations = 0;
82 off_t offset; /* file offset */
83 unsigned char *buf; /* buffer block */
84 size_t buf_size; /* number of read bytes in block */
85 size_t chunk; /* number of bytes allocated */
86 size_t buf_ptr; /* current position in buffer */
87 char *prev_name; /* last word read */
88 int sysno; /* last sysno */
89 int seqno; /* last seqno */
90 off_t length; /* length of file */
91 /* handler invoked in each read */
92 void (*readHandler)(struct key_file *keyp, void *rinfo);
96 void getFnameTmp (char *fname, int no)
98 sprintf (fname, TEMP_FNAME, no);
101 void key_file_chunk_read (struct key_file *f)
105 getFnameTmp (fname, f->no);
106 fd = open (fname, O_RDONLY);
109 logf (LOG_FATAL|LOG_ERRNO, "cannot open %s", fname);
114 if ((f->length = lseek (fd, 0L, SEEK_END)) == (off_t) -1)
116 logf (LOG_FATAL|LOG_ERRNO, "cannot seek %s", fname);
120 if (lseek (fd, f->offset, SEEK_SET) == -1)
122 logf (LOG_FATAL|LOG_ERRNO, "cannot seek %s", fname);
125 while (f->chunk - nr > 0)
127 r = read (fd, f->buf + nr, f->chunk - nr);
134 logf (LOG_FATAL|LOG_ERRNO, "read of %s", fname);
140 (*f->readHandler)(f, f->readInfo);
144 struct key_file *key_file_init (int no, int chunk)
148 f = xmalloc (sizeof(*f));
155 f->readHandler = NULL;
156 f->buf = xmalloc (f->chunk);
157 f->prev_name = xmalloc (256);
158 *f->prev_name = '\0';
159 key_file_chunk_read (f);
163 int key_file_getc (struct key_file *f)
165 if (f->buf_ptr < f->buf_size)
166 return f->buf[(f->buf_ptr)++];
167 if (f->buf_size < f->chunk)
169 f->offset += f->buf_size;
170 key_file_chunk_read (f);
171 if (f->buf_ptr < f->buf_size)
172 return f->buf[(f->buf_ptr)++];
177 int key_file_decode (struct key_file *f)
181 c = key_file_getc (f);
188 d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
191 d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
192 d = (d << 8) + (key_file_getc (f) & 0xff);
195 d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
196 d = (d << 8) + (key_file_getc (f) & 0xff);
197 d = (d << 8) + (key_file_getc (f) & 0xff);
203 int key_file_read (struct key_file *f, char *key)
208 c = key_file_getc (f);
211 strcpy (key, f->prev_name);
220 while ((key[i++] = key_file_getc (f)))
222 strcpy (f->prev_name, key);
225 d = key_file_decode (f);
228 itkey.sysno = d + f->sysno;
231 f->sysno = itkey.sysno;
234 d = key_file_decode (f);
235 itkey.seqno = d + f->seqno;
236 f->seqno = itkey.seqno;
237 memcpy (key + i, &itkey, sizeof(struct it_key));
238 return i + sizeof (struct it_key);
243 struct key_file **file;
248 int (*cmp)(const void *p1, const void *p2);
251 struct heap_info *key_heap_init (int nkeys,
252 int (*cmp)(const void *p1, const void *p2))
254 struct heap_info *hi;
257 hi = xmalloc (sizeof(*hi));
258 hi->info.file = xmalloc (sizeof(*hi->info.file) * (1+nkeys));
259 hi->info.buf = xmalloc (sizeof(*hi->info.buf) * (1+nkeys));
261 hi->ptr = xmalloc (sizeof(*hi->ptr) * (1+nkeys));
263 for (i = 0; i<= nkeys; i++)
266 hi->info.buf[i] = xmalloc (512);
271 static void key_heap_swap (struct heap_info *hi, int i1, int i2)
276 hi->ptr[i1] = hi->ptr[i2];
281 static void key_heap_delete (struct heap_info *hi)
283 int cur = 1, child = 2;
285 assert (hi->heapnum > 0);
287 key_heap_swap (hi, 1, hi->heapnum);
289 while (child <= hi->heapnum) {
290 if (child < hi->heapnum &&
291 (*hi->cmp)(&hi->info.buf[hi->ptr[child]],
292 &hi->info.buf[hi->ptr[child+1]]) > 0)
294 if ((*hi->cmp)(&hi->info.buf[hi->ptr[cur]],
295 &hi->info.buf[hi->ptr[child]]) > 0)
297 key_heap_swap (hi, cur, child);
306 static void key_heap_insert (struct heap_info *hi, const char *buf, int nbytes,
311 cur = ++(hi->heapnum);
312 memcpy (hi->info.buf[hi->ptr[cur]], buf, nbytes);
313 hi->info.file[hi->ptr[cur]] = kf;
316 while (parent && (*hi->cmp)(&hi->info.buf[hi->ptr[parent]],
317 &hi->info.buf[hi->ptr[cur]]) > 0)
319 key_heap_swap (hi, cur, parent);
325 static int heap_read_one (struct heap_info *hi, char *name, char *key)
334 strcpy (name, hi->info.buf[n]);
335 kf = hi->info.file[n];
337 memcpy (key, hi->info.buf[n] + r+1, KEY_SIZE);
338 key_heap_delete (hi);
339 if ((r = key_file_read (kf, rbuf)))
340 key_heap_insert (hi, rbuf, r, kf);
345 int heap_inp (Dict dict, ISAM isam, struct heap_info *hi)
348 char next_name[INP_NAME_MAX+1];
349 char cur_name[INP_NAME_MAX+1];
350 int key_buf_size = INP_BUF_START;
356 next_key = xmalloc (KEY_SIZE);
357 key_buf = xmalloc (key_buf_size * (KEY_SIZE));
358 more = heap_read_one (hi, cur_name, key_buf);
359 while (more) /* EOF ? */
362 key_buf_ptr = KEY_SIZE;
365 if (!(more = heap_read_one (hi, next_name, next_key)))
367 if (*next_name && strcmp (next_name, cur_name))
369 memcpy (key_buf + key_buf_ptr, next_key, KEY_SIZE);
370 key_buf_ptr += KEY_SIZE;
371 if (key_buf_ptr+KEY_SIZE >= key_buf_size)
374 new_key_buf = xmalloc (key_buf_size + INP_BUF_ADD);
375 memcpy (new_key_buf, key_buf, key_buf_size);
376 key_buf_size += INP_BUF_ADD;
378 key_buf = new_key_buf;
382 nmemb = key_buf_ptr / KEY_SIZE;
383 assert (nmemb*KEY_SIZE == key_buf_ptr);
384 if ((info = dict_lookup (dict, cur_name)))
386 ISAM_P isam_p, isam_p2;
387 logf (LOG_DEBUG, "updating %s", cur_name);
388 memcpy (&isam_p, info+1, sizeof(ISAM_P));
389 isam_p2 = is_merge (isam, isam_p, nmemb, key_buf);
393 if (!dict_delete (dict, cur_name))
399 if (isam_p2 != isam_p)
400 dict_insert (dict, cur_name, sizeof(ISAM_P), &isam_p2);
406 logf (LOG_DEBUG, "inserting %s", cur_name);
408 isam_p = is_merge (isam, 0, nmemb, key_buf);
409 dict_insert (dict, cur_name, sizeof(ISAM_P), &isam_p);
411 memcpy (key_buf, next_key, KEY_SIZE);
412 strcpy (cur_name, next_name);
417 struct progressInfo {
424 void progressFunc (struct key_file *keyp, void *info)
426 struct progressInfo *p = info;
427 time_t now, remaining;
429 if (keyp->buf_size <= 0 || p->totalBytes <= 0)
433 if (now >= p->lastTime+10)
436 remaining = (now - p->startTime)*
437 ((double) p->totalBytes/p->totalOffset - 1.0);
438 if (remaining <= 130)
439 logf (LOG_LOG, "Merge %2.1f%% completed; %ld seconds remaining",
440 (100.0*p->totalOffset) / p->totalBytes, (long) remaining);
442 logf (LOG_LOG, "Merge %2.1f%% completed; %ld minutes remaining",
443 (100.0*p->totalOffset) / p->totalBytes, (long) remaining/60);
445 p->totalOffset += keyp->buf_size;
448 void key_input (const char *dict_fname, const char *isam_fname,
449 int nkeys, int cache)
454 struct key_file **kf;
457 struct heap_info *hi;
458 struct progressInfo progressInfo;
460 dict = dict_open (dict_fname, cache, 1);
463 logf (LOG_FATAL, "dict_open fail of `%s'", dict_fname);
466 isam = is_open (isam_fname, key_compare, 1, sizeof(struct it_key));
469 logf (LOG_FATAL, "is_open fail of `%s'", isam_fname);
473 kf = xmalloc ((1+nkeys) * sizeof(*kf));
474 progressInfo.totalBytes = 0;
475 progressInfo.totalOffset = 0;
476 time (&progressInfo.startTime);
477 time (&progressInfo.lastTime);
478 for (i = 1; i<=nkeys; i++)
480 kf[i] = key_file_init (i, 32768);
481 kf[i]->readHandler = progressFunc;
482 kf[i]->readInfo = &progressInfo;
483 progressInfo.totalBytes += kf[i]->length;
484 progressInfo.totalOffset += kf[i]->buf_size;
486 hi = key_heap_init (nkeys, key_qsort_compare);
487 for (i = 1; i<=nkeys; i++)
488 if ((r = key_file_read (kf[i], rbuf)))
489 key_heap_insert (hi, rbuf, r, kf[i]);
490 heap_inp (dict, isam, hi);
495 for (i = 1; i<=nkeys; i++)
497 getFnameTmp (rbuf, i);
500 logf (LOG_LOG, "Iterations . . .%7d", no_iterations);
501 logf (LOG_LOG, "Distinct words .%7d", no_diffs);
502 logf (LOG_LOG, "Updates. . . . .%7d", no_updates);
503 logf (LOG_LOG, "Deletions. . . .%7d", no_deletions);
504 logf (LOG_LOG, "Insertions . . .%7d", no_insertions);