X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=index%2Fkinput.c;h=94291692bc01cd72e2f0927477a21a6aa75ad613;hb=cdc04db6a4bcbf0630b5da7cfdf7195a88f0a9f9;hp=e9d467a8a35736eb7e7cb46db208ee24b5203de9;hpb=fc7107844c9ec8ea23e680ca4f3231923db4e9c5;p=idzebra-moved-to-github.git diff --git a/index/kinput.c b/index/kinput.c index e9d467a..9429169 100644 --- a/index/kinput.c +++ b/index/kinput.c @@ -4,7 +4,30 @@ * Sebastian Hammer, Adam Dickmeiss * * $Log: kinput.c,v $ - * Revision 1.9 1995-10-10 12:24:39 adam + * Revision 1.15 1996-03-21 14:50:09 adam + * File update uses modify-time instead of change-time. + * + * 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 @@ -52,6 +75,7 @@ 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; @@ -65,20 +89,36 @@ struct key_file { 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); @@ -98,6 +138,8 @@ void key_file_chunk_read (struct key_file *f) } f->buf_size = nr; f->buf_ptr = 0; + if (f->readHandler) + (*f->readHandler)(f, f->readInfo); close (fd); } @@ -111,6 +153,8 @@ struct key_file *key_file_init (int no, int chunk) 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'; @@ -343,11 +387,20 @@ int heap_inp (Dict dict, ISAM isam, struct heap_info *hi) { 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 { @@ -363,8 +416,40 @@ int heap_inp (Dict dict, ISAM isam, struct heap_info *hi) return 0; } +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) + int nkeys, int cache) + { Dict dict; ISAM isam; @@ -372,6 +457,7 @@ void key_input (const char *dict_fname, const char *isam_fname, char rbuf[1024]; int i, r; struct heap_info *hi; + struct progressInfo progressInfo; dict = dict_open (dict_fname, cache, 1); if (!dict) @@ -387,9 +473,18 @@ void key_input (const char *dict_fname, const char *isam_fname, } 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))) @@ -398,10 +493,16 @@ void key_input (const char *dict_fname, const char *isam_fname, 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); }