X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=index%2Frecindex.c;h=3e562f47ff6350c7110b75d56ab19afa1f145058;hb=af7bc27d33843a8e667022f24589450c39f907f2;hp=6f06cd561fb56d2f80308a3d4684c761746bd7e4;hpb=226fb73f42a86cc30ff4f27eb452ab3f6c19ae01;p=idzebra-moved-to-github.git diff --git a/index/recindex.c b/index/recindex.c index 6f06cd5..3e562f4 100644 --- a/index/recindex.c +++ b/index/recindex.c @@ -1,10 +1,35 @@ /* - * Copyright (C) 1994-1995, Index Data I/S + * Copyright (C) 1994-1996, Index Data I/S * All rights reserved. * Sebastian Hammer, Adam Dickmeiss * * $Log: recindex.c,v $ - * Revision 1.10 1995-12-06 12:41:24 adam + * Revision 1.16 1996-06-04 10:19:00 adam + * Minor changes - removed include of ctype.h. + * + * Revision 1.15 1996/05/13 14:23:06 adam + * Work on compaction of set/use bytes in dictionary. + * + * Revision 1.14 1996/02/01 20:48:15 adam + * The total size of records are always checked in rec_cache_insert to + * reduce memory usage. + * + * Revision 1.13 1995/12/11 09:12:49 adam + * The rec_get function returns NULL if record doesn't exist - will + * happen in the server if the result set records have been deleted since + * the creation of the set (i.e. the search). + * The server saves a result temporarily if it is 'volatile', i.e. the + * set is register dependent. + * + * Revision 1.12 1995/12/07 17:38:47 adam + * Work locking mechanisms for concurrent updates/commit. + * + * Revision 1.11 1995/12/06 13:58:26 adam + * Improved flushing of records - all flushes except the last one + * don't write the last accessed. Also flush takes place if record + * info occupy more than about 256k. + * + * Revision 1.10 1995/12/06 12:41:24 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 @@ -48,7 +73,6 @@ #include #include #include -#include #include "recindxp.h" @@ -110,8 +134,8 @@ static void rec_release_blocks (Records p, int sysno) if (read_indx (p, sysno, &entry, sizeof(entry), 1) != 1) return ; - p->head.total_bytes -= entry.u.used.size; - freeblock = entry.u.used.next; + p->head.total_bytes -= entry.size; + freeblock = entry.next; assert (freeblock > 0); dst_type = freeblock & 7; assert (dst_type < REC_BLOCK_TYPES); @@ -142,7 +166,8 @@ static void rec_delete_single (Records p, Record rec) rec_release_blocks (p, rec->sysno); - entry.u.free.next = p->head.index_free; + entry.next = p->head.index_free; + entry.size = 0; p->head.index_free = rec->sysno; write_indx (p, rec->sysno, &entry, sizeof(entry)); } @@ -198,8 +223,8 @@ static void rec_write_single (Records p, Record rec) block_free = p->head.block_last[dst_type]++; if (block_prev == -1) { - entry.u.used.next = block_free*8 + dst_type; - entry.u.used.size = size; + entry.next = block_free*8 + dst_type; + entry.size = size; p->head.total_bytes += size; write_indx (p, rec->sysno, &entry, sizeof(entry)); } @@ -300,10 +325,13 @@ Records rec_open (int rw) return p; } -static void rec_cache_flush (Records p) +static void rec_cache_flush (Records p, int saveCount) { - int i; - for (i = 0; icache_cur; i++) + int i, j; + + if (saveCount >= p->cache_cur) + saveCount = 0; + for (i = 0; icache_cur - saveCount; i++) { struct record_cache_entry *e = p->record_cache + i; switch (e->flag) @@ -322,7 +350,10 @@ static void rec_cache_flush (Records p) } rec_rm (&e->rec); } - p->cache_cur = 0; + for (j = 0; jrecord_cache+j, p->record_cache+i, + sizeof(*p->record_cache)); + p->cache_cur = saveCount; } static Record *rec_cache_lookup (Records p, int sysno, @@ -347,7 +378,20 @@ static void rec_cache_insert (Records p, Record rec, enum recordCacheFlag flag) struct record_cache_entry *e; if (p->cache_cur == p->cache_max) - rec_cache_flush (p); + rec_cache_flush (p, 1); + else if (p->cache_cur > 0) + { + int i, j; + int used = 0; + for (i = 0; icache_cur; i++) + { + Record r = (p->record_cache + i)->rec; + for (j = 0; jsize[j]; + } + if (used > 256000) + rec_cache_flush (p, 1); + } assert (p->cache_cur < p->cache_max); e = p->record_cache + (p->cache_cur)++; @@ -362,7 +406,7 @@ void rec_close (Records *pp) assert (p); - rec_cache_flush (p); + rec_cache_flush (p, 0); xfree (p->record_cache); if (p->rw) @@ -397,16 +441,20 @@ Record rec_get (Records p, int sysno) if ((recp = rec_cache_lookup (p, sysno, recordFlagNop))) return rec_cp (*recp); - read_indx (p, sysno, &entry, sizeof(entry), 0); + if (!read_indx (p, sysno, &entry, sizeof(entry), 1)) + return NULL; /* record is not there! */ + + if (!entry.size) + return NULL; /* record is deleted */ - dst_type = entry.u.used.next & 7; + dst_type = entry.next & 7; assert (dst_type < REC_BLOCK_TYPES); - freeblock = entry.u.used.next / 8; + freeblock = entry.next / 8; assert (freeblock > 0); rec = xmalloc (sizeof(*rec)); - rec_tmp_expand (p, entry.u.used.size, dst_type); + rec_tmp_expand (p, entry.size, dst_type); cptr = p->tmp_buf; bf_read (p->data_BFile[dst_type], freeblock, 0, 0, cptr); @@ -458,7 +506,7 @@ Record rec_new (Records p) read_indx (p, p->head.index_free, &entry, sizeof(entry), 0); sysno = p->head.index_free; - p->head.index_free = entry.u.free.next; + p->head.index_free = entry.next; } (p->head.no_records)++; rec->sysno = sysno; @@ -475,6 +523,7 @@ void rec_del (Records p, Record *recpp) { Record *recp; + (p->head.no_records)--; if ((recp = rec_cache_lookup (p, (*recpp)->sysno, recordFlagDelete))) { rec_rm (recp); @@ -508,6 +557,9 @@ void rec_put (Records p, Record *recpp) void rec_rm (Record *recpp) { int i; + + if (!*recpp) + return ; for (i = 0; i < REC_NO_INFO; i++) xfree ((*recpp)->info[i]); xfree (*recpp);