Changeset 1976

Timestamp:

12/05/07 05:12:58 (8 months ago)

Author:

jmruiz

Message:

merge.c fix for BDB. Not tested. First work

Files:

• swish-e/branches/2.6/src/merge.c (modified) (5 diffs)

swish-e/branches/2.6/src/merge.c

@@ -91 +91 @@
-                 loc_count = 0,
-                 word_count = 0;
-
+, *tmp
-    int          metadata_length = 0;
-    unsigned char   *worddata;
@@ -217 +217 @@
-                    {
-
-
-
-
-
-
-
-
-
-
+
-                        {
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-                        }
-
-                        filenum = 0;
-                        start = s;
-
-                        while(1)
-                        {                   /* Read on all items */
-                            uncompress_location_values(&s,&flag,&tmpval,&frequency);
-                            filenum += tmpval;
-                            /* Use stack array when possible to avoid malloc/free overhead */
-                            if(frequency > MAX_STACK_POSITIONS)
-                                posdata = (unsigned int *) emalloc(frequency * sizeof(int));
-                            else
-                                posdata = local_posdata;
-
-                            /* Read the positions */
-                            uncompress_location_positions(&s,flag,frequency,posdata);
-
-
-                            /* now we have the word data */
-                            for (i = 0; i < frequency; i++, loc_count++)
-                                write_word_pos( cur_index, sw_output, cur_index->merge_file_num_map, filenum, e, metaID, posdata[i]);
-
-                            if(e->tfrequency)
-                            {
-                                /* 08/2002 jmruiz - We will call CompressCurrentLocEntry from time
-                                ** to time to help addentry.
-                                ** If we do not do this, addentry routine will have to run linked lists
-                                ** of positions with thousands of elements and makes the merge proccess
-                                ** very slow
-                                */
-                                if(!(loc_count % 100))
-                                    CompressCurrentLocEntry(sw_output, e);
-                            }
-
-
-                            if(posdata != local_posdata)
-                                efree(posdata);
-
-                            /* Check for enf of worddata */
-                            if ((s - worddata) == sz_worddata)
-                                break;   /* End of worddata */
-
-                            /* Check for end of current metaID data */
-                            if ( metadata_length == (s - start))
-                            {
-                                filenum = 0;
-                                metaID = uncompress2(&s);
-                                metadata_length = uncompress2(&s);
-                                start = s;
-                            }
-                        }
-
-                        if(e->tfrequency)
-                            CompressCurrentLocEntry(sw_output, e);
-
-                        efree(worddata);
-                    }
-                    cur_index = cur_index->next;
@@ -330 +334 @@
-
-    /* 2002/09 MERGE fix jmruiz */
-
+e
-    /* walk the hash list to merge worddata */
-    for (word_count = 0, hash = 0; hash < VERYBIGHASHSIZE; hash++)
@@ -986 +990 @@
-static void dump_index_words(SWISH * sw, IndexFILE * indexf, SWISH *sw_output)
-{
-    int         j;
-    int         word_count = 0;
-2
-
+1
+,*previousword
-    DB_WORDID  *wordID;
-
@@ -998 +1001 @@
-    fflush(stdout);
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
-        {
-            /* Add resultword to output */
-            getentry(sw_output, resultword);
+            efree(previousword);
+            previousword = resultword;
+        }
+        else
-            efree(resultword);
-
-
-
-
-
+
+
+
+
-    }
-    printf("Getting words in index '%s': %6d words\n", indexf->line, word_count);
+
+    efree(previousword);
-
-    DB_EndReadWords(sw, indexf->DB);