X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=index%2Fextract.c;h=18cac3d597c58577d9346d69cc32ed0805d2f728;hp=ce51945bbdbbaac8e922173d4b63f34a536f9b94;hb=7c3a0352f0492609a3b6b26b63a72b0b2d207aab;hpb=8ec67e89869d6ab86fcdf35563f3f7ebfe02947a

diff --git a/index/extract.c b/index/extract.c
index ce51945..18cac3d 100644
--- a/index/extract.c
+++ b/index/extract.c
@@ -1,12 +1,28 @@
-/*
- * Copyright (C) 1994-2002, Index Data 
- * All rights reserved.
- * Sebastian Hammer, Adam Dickmeiss
- *
- * $Id: extract.c,v 1.117 2002-04-15 14:05:43 adam Exp $
- */
+/* $Id: extract.c,v 1.161 2004-09-14 14:38:07 quinn Exp $
+   Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002,2003,2004
+   Index Data Aps
+
+This file is part of the Zebra server.
+
+Zebra is free software; you can redistribute it and/or modify it under
+the terms of the GNU General Public License as published by the Free
+Software Foundation; either version 2, or (at your option) any later
+version.
+
+Zebra is distributed in the hope that it will be useful, but WITHOUT ANY
+WARRANTY; without even the implied warranty of MERCHANTABILITY or
+FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+for more details.
+
+You should have received a copy of the GNU General Public License
+along with Zebra; see the file LICENSE.zebra.  If not, write to the
+Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
+02111-1307, USA.
+*/
+
 #include <stdio.h>
 #include <assert.h>
+#include <ctype.h>
 #ifdef WIN32
 #include <io.h>
 #else
@@ -24,12 +40,40 @@
 #define PRINTF_OFF_T "%ld"
 #endif
 
+#define USE_SHELLSORT 0
+
+#if USE_SHELLSORT
+static void shellsort(void *ar, int r, size_t s,
+                      int (*cmp)(const void *a, const void *b))
+{
+    char *a = ar;
+    char v[100];
+    int h, i, j, k;
+    static const int incs[16] = { 1391376, 463792, 198768, 86961, 33936,
+                                  13776, 4592, 1968, 861, 336, 
+                                  112, 48, 21, 7, 3, 1 };
+    for ( k = 0; k < 16; k++)
+        for (h = incs[k], i = h; i < r; i++)
+        { 
+            memcpy (v, a+s*i, s);
+            j = i;
+            while (j > h && (*cmp)(a + s*(j-h), v) > 0)
+            {
+                memcpy (a + s*j, a + s*(j-h), s);
+                j -= h;
+            }
+            memcpy (a+s*j, v, s);
+        } 
+}
+#endif
+
 static void logRecord (ZebraHandle zh)
 {
     ++zh->records_processed;
     if (!(zh->records_processed % 1000))
     {
-        logf (LOG_LOG, "Records: %7d i/u/d %d/%d/%d", 
+        logf (LOG_LOG, "Records: "ZINT_FORMAT" i/u/d "
+			ZINT_FORMAT"/"ZINT_FORMAT"/"ZINT_FORMAT, 
               zh->records_processed, zh->records_inserted, zh->records_updated,
               zh->records_deleted);
     }
@@ -49,6 +93,51 @@ static const char **searchRecordKey (ZebraHandle zh,
                                      struct recKeys *reckeys,
 				     int attrSetS, int attrUseS)
 {
+#if IT_KEY_NEW
+/* #error searchRecordKey does not work yet in this mode.. */
+    static const char *ws[32];
+    void *decode_handle = iscz1_start();
+    int off = 0;
+    int startSeq = -1;
+    int seqno = 0;
+    int i;
+
+    for (i = 0; i<32; i++)
+        ws[i] = NULL;
+
+    while (off < reckeys->buf_used)
+    {
+        const char *src = reckeys->buf + off;
+        struct it_key key;
+	char *dst = (char*) &key;
+	int attrSet, attrUse;
+
+	iscz1_decode(decode_handle, &dst, &src);
+	assert(key.len < 4 && key.len > 2);
+
+	attrSet = (int) key.mem[0];
+	attrUse = (int) key.mem[1];
+	seqno = (int) key.mem[2];
+
+	if (attrUseS == attrUse && attrSetS == attrSet)
+        {
+            int woff;
+
+            if (startSeq == -1)
+                startSeq = seqno;
+            woff = seqno - startSeq;
+            if (woff >= 0 && woff < 31)
+                ws[woff] = src;
+        }
+
+        while (*src++)
+	    ;
+        off = src - reckeys->buf;
+    }
+    iscz1_stop(decode_handle);
+    assert (off == reckeys->buf_used);
+    return ws;
+#else
     static const char *ws[32];
     int off = 0;
     int startSeq = -1;
@@ -127,6 +216,7 @@ static const char **searchRecordKey (ZebraHandle zh,
     }
     assert (off == reckeys->buf_used);
     return ws;
+#endif
 }
 
 struct file_read_info {
@@ -216,7 +306,7 @@ static void file_end (void *handle, off_t offset)
 }
 
 static char *fileMatchStr (ZebraHandle zh,
-                           struct recKeys *reckeys, struct recordGroup *rGroup,
+                           struct recKeys *reckeys,
                            const char *fname, const char *spec)
 {
     static char dstBuf[2048];      /* static here ??? */
@@ -275,7 +365,7 @@ static char *fileMatchStr (ZebraHandle zh,
             else
             {
                 logf (LOG_WARN, "Missing ) in match criteria %s in group %s",
-                      spec, rGroup->groupName ? rGroup->groupName : "none");
+                      spec, zh->m_group ? zh->m_group : "none");
                 return NULL;
             }
             s++;
@@ -315,13 +405,14 @@ static char *fileMatchStr (ZebraHandle zh,
             s = s1;
 
             if (!strcmp (special, "group"))
-                spec_src = rGroup->groupName;
+                spec_src = zh->m_group;
             else if (!strcmp (special, "database"))
-                spec_src = rGroup->databaseName;
-            else if (!strcmp (special, "filename"))
+                spec_src = zh->basenames[0];
+            else if (!strcmp (special, "filename")) {
                 spec_src = fname;
+	    }
             else if (!strcmp (special, "type"))
-                spec_src = rGroup->recordType;
+                spec_src = zh->m_record_type;
             else 
                 spec_src = NULL;
             if (spec_src)
@@ -350,7 +441,7 @@ static char *fileMatchStr (ZebraHandle zh,
         else
         {
             logf (LOG_WARN, "Syntax error in match criteria %s in group %s",
-                  spec, rGroup->groupName ? rGroup->groupName : "none");
+                  spec, zh->m_group ? zh->m_group : "none");
             return NULL;
         }
         *dst++ = 1;
@@ -358,7 +449,7 @@ static char *fileMatchStr (ZebraHandle zh,
     if (dst == dstBuf)
     {
         logf (LOG_WARN, "No match criteria for record %s in group %s",
-              fname, rGroup->groupName ? rGroup->groupName : "none");
+              fname, zh->m_group ? zh->m_group : "none");
         return NULL;
     }
     *dst = '\0';
@@ -370,33 +461,50 @@ struct recordLogInfo {
     int recordOffset;
     struct recordGroup *rGroup;
 };
-     
-static void recordLogPreamble (int level, const char *msg, void *info)
+
+void create_rec_keys_codec(struct recKeys *keys)
 {
-    struct recordLogInfo *p = (struct recordLogInfo *) info;
-    FILE *outf = yaz_log_file ();
-
-    if (level & LOG_LOG)
-        return ;
-    fprintf (outf, "File %s, offset %d, type %s\n",
-             p->rGroup->recordType, p->recordOffset, p->fname);
-    log_event_start (NULL, NULL);
+    keys->buf_used = 0;
+#if IT_KEY_NEW
+    iscz1_reset(keys->codec_handle);
+#else
+    keys->prevAttrUse = -1;
+    keys->prevAttrSet = -1;
+    keys->prevSeqNo = 0;
+#endif
 }
-
-
-static int recordExtract (ZebraHandle zh,
-                          SYSNO *sysno, const char *fname,
-                          struct recordGroup *rGroup, int deleteFlag,
-                          struct file_read_info *fi,
-			  RecType recType, char *subType, void *clientData)
+     
+static int file_extract_record(ZebraHandle zh,
+			       SYSNO *sysno, const char *fname,
+			       int deleteFlag,
+			       struct file_read_info *fi,
+			       int force_update)
 {
     RecordAttr *recordAttr;
     int r;
-    char *matchStr;
+    const char *matchStr;
     SYSNO sysnotmp;
     Record rec;
-    struct recordLogInfo logInfo;
     off_t recordOffset = 0;
+    RecType recType;
+    void *clientData;
+    char subType[128];
+    
+    if (!(recType =
+	  recType_byName (zh->reg->recTypes, zh->m_record_type, subType,
+			  &clientData)))
+    {
+        logf (LOG_WARN, "No such record type: %s", zh->m_record_type);
+        return 0;
+    }
+
+    /* announce database */
+    if (zebraExplain_curDatabase (zh->reg->zei, zh->basenames[0]))
+    {
+        if (zebraExplain_newDatabase (zh->reg->zei, zh->basenames[0],
+				      zh->m_explain_database))
+	    return 0;
+    }
 
     if (fi->fd != -1)
     {
@@ -405,11 +513,9 @@ static int recordExtract (ZebraHandle zh,
         /* we are going to read from a file, so prepare the extraction */
 	int i;
 
-	zh->reg->keys.buf_used = 0;
-	zh->reg->keys.prevAttrUse = -1;
-	zh->reg->keys.prevAttrSet = -1;
-	zh->reg->keys.prevSeqNo = 0;
-	zh->reg->sortKeys = 0;
+	create_rec_keys_codec(&zh->reg->keys);
+
+	zh->reg->sortKeys.buf_used = 0;
 	
 	recordOffset = fi->file_moffset;
 	extractCtrl.offset = fi->file_moffset;
@@ -432,29 +538,41 @@ static int recordExtract (ZebraHandle zh,
 		extractCtrl.seqno[i] = 0;
 	}
 	extractCtrl.zebra_maps = zh->reg->zebra_maps;
-	extractCtrl.flagShowRecords = !rGroup->flagRw;
+	extractCtrl.flagShowRecords = !zh->m_flag_rw;
 
-        if (!rGroup->flagRw)
+        if (!zh->m_flag_rw)
             printf ("File: %s " PRINTF_OFF_T "\n", fname, recordOffset);
-
-        logInfo.fname = fname;
-        logInfo.recordOffset = recordOffset;
-        logInfo.rGroup = rGroup;
-        log_event_start (recordLogPreamble, &logInfo);
+        if (zh->m_flag_rw)
+        {
+            char msg[512];
+            sprintf (msg, "%s:" PRINTF_OFF_T , fname, recordOffset);
+            yaz_log_init_prefix2 (msg);
+        }
 
         r = (*recType->extract)(clientData, &extractCtrl);
 
-        log_event_start (NULL, NULL);
-
+        yaz_log_init_prefix2 (0);
 	if (r == RECCTRL_EXTRACT_EOF)
 	    return 0;
-	else if (r == RECCTRL_EXTRACT_ERROR)
+	else if (r == RECCTRL_EXTRACT_ERROR_GENERIC)
+	{
+            /* error occured during extraction ... */
+            if (zh->m_flag_rw &&
+		zh->records_processed < zh->m_file_verbose_limit)
+            {
+                logf (LOG_WARN, "fail %s %s " PRINTF_OFF_T, zh->m_record_type,
+                      fname, recordOffset);
+            }
+            return 0;
+        }
+	else if (r == RECCTRL_EXTRACT_ERROR_NO_SUCH_FILTER)
 	{
             /* error occured during extraction ... */
-            if (rGroup->flagRw &&
-		zh->records_processed < rGroup->fileVerboseLimit)
+            if (zh->m_flag_rw &&
+		zh->records_processed < zh->m_file_verbose_limit)
             {
-                logf (LOG_WARN, "fail %s %s " PRINTF_OFF_T, rGroup->recordType,
+                logf (LOG_WARN, "no filter for %s %s " 
+                      PRINTF_OFF_T, zh->m_record_type,
                       fname, recordOffset);
             }
             return 0;
@@ -463,10 +581,10 @@ static int recordExtract (ZebraHandle zh,
         {
             /* the extraction process returned no information - the record
                is probably empty - unless flagShowRecords is in use */
-            if (!rGroup->flagRw)
+            if (!zh->m_flag_rw)
                 return 1;
 	    
-	    logf (LOG_WARN, "empty %s %s " PRINTF_OFF_T, rGroup->recordType,
+	    logf (LOG_WARN, "empty %s %s " PRINTF_OFF_T, zh->m_record_type,
 		  fname, recordOffset);
             return 1;
         }
@@ -479,17 +597,20 @@ static int recordExtract (ZebraHandle zh,
     {
         sysnotmp = 0;
         sysno = &sysnotmp;
-        if (rGroup->recordId && *rGroup->recordId)
+        if (zh->m_record_id && *zh->m_record_id)
         {
             char *rinfo;
         
-            matchStr = fileMatchStr (zh, &zh->reg->keys, rGroup, fname, 
-                                     rGroup->recordId);
+            matchStr = fileMatchStr (zh, &zh->reg->keys, fname, 
+				     zh->m_record_id);
             if (matchStr)
             {
                 rinfo = dict_lookup (zh->reg->matchDict, matchStr);
                 if (rinfo)
+		{
+		    assert(*rinfo == sizeof(*sysno));
                     memcpy (sysno, rinfo+1, sizeof(*sysno));
+		}
             }
             else
             {
@@ -504,13 +625,13 @@ static int recordExtract (ZebraHandle zh,
         /* new record */
         if (deleteFlag)
         {
-	    logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, rGroup->recordType,
+	    logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, zh->m_record_type,
 		  fname, recordOffset);
             logf (LOG_WARN, "cannot delete record above (seems new)");
             return 1;
         }
-        if (zh->records_processed < rGroup->fileVerboseLimit)
-            logf (LOG_LOG, "add %s %s " PRINTF_OFF_T, rGroup->recordType,
+        if (zh->records_processed < zh->m_file_verbose_limit)
+            logf (LOG_LOG, "add %s %s " PRINTF_OFF_T, zh->m_record_type,
                   fname, recordOffset);
         rec = rec_new (zh->reg->records);
 
@@ -522,8 +643,8 @@ static int recordExtract (ZebraHandle zh,
         {
             dict_insert (zh->reg->matchDict, matchStr, sizeof(*sysno), sysno);
         }
-        extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys);
 	extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys);
+        extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys);
 
         zh->records_inserted++;
     }
@@ -531,17 +652,19 @@ static int recordExtract (ZebraHandle zh,
     {
         /* record already exists */
         struct recKeys delkeys;
+        struct sortKeys sortKeys;
 
         rec = rec_get (zh->reg->records, *sysno);
         assert (rec);
 	
 	recordAttr = rec_init_attr (zh->reg->zei, rec);
 
-	if (recordAttr->runNumber ==
+	if (!force_update && recordAttr->runNumber ==
             zebraExplain_runNumberIncrement (zh->reg->zei, 0))
 	{
-	    logf (LOG_LOG, "skipped %s %s " PRINTF_OFF_T, rGroup->recordType,
-		  fname, recordOffset);
+            yaz_log (LOG_LOG, "run number = %d", recordAttr->runNumber);
+	    yaz_log (LOG_LOG, "skipped %s %s " PRINTF_OFF_T,
+                     zh->m_record_type, fname, recordOffset);
 	    extract_flushSortKeys (zh, *sysno, -1, &zh->reg->sortKeys);
 	    rec_rm (&rec);
 	    logRecord (zh);
@@ -549,7 +672,11 @@ static int recordExtract (ZebraHandle zh,
 	}
         delkeys.buf_used = rec->size[recInfo_delKeys];
 	delkeys.buf = rec->info[recInfo_delKeys];
-	extract_flushSortKeys (zh, *sysno, 0, &zh->reg->sortKeys);
+
+        sortKeys.buf_used = rec->size[recInfo_sortKeys];
+        sortKeys.buf = rec->info[recInfo_sortKeys];
+
+	extract_flushSortKeys (zh, *sysno, 0, &sortKeys);
         extract_flushRecordKeys (zh, *sysno, 0, &delkeys);
         if (deleteFlag)
         {
@@ -557,14 +684,14 @@ static int recordExtract (ZebraHandle zh,
             if (!delkeys.buf_used)
             {
                 logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T,
-                      rGroup->recordType, fname, recordOffset);
+                      zh->m_record_type, fname, recordOffset);
                 logf (LOG_WARN, "cannot delete file above, storeKeys false");
             }
             else
             {
-                if (zh->records_processed < rGroup->fileVerboseLimit)
+                if (zh->records_processed < zh->m_file_verbose_limit)
                     logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T,
-                         rGroup->recordType, fname, recordOffset);
+                         zh->m_record_type, fname, recordOffset);
                 zh->records_deleted++;
                 if (matchStr)
                     dict_delete (zh->reg->matchDict, matchStr);
@@ -580,14 +707,15 @@ static int recordExtract (ZebraHandle zh,
             if (!delkeys.buf_used)
             {
                 logf (LOG_LOG, "update %s %s " PRINTF_OFF_T,
-                      rGroup->recordType, fname, recordOffset);
+                      zh->m_record_type, fname, recordOffset);
                 logf (LOG_WARN, "cannot update file above, storeKeys false");
             }
             else
             {
-                if (zh->records_processed < rGroup->fileVerboseLimit)
+                if (zh->records_processed < zh->m_file_verbose_limit)
                     logf (LOG_LOG, "update %s %s " PRINTF_OFF_T,
-                        rGroup->recordType, fname, recordOffset);
+                        zh->m_record_type, fname, recordOffset);
+                extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys);
                 extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys);
                 zh->records_updated++;
             }
@@ -596,7 +724,7 @@ static int recordExtract (ZebraHandle zh,
     /* update file type */
     xfree (rec->info[recInfo_fileType]);
     rec->info[recInfo_fileType] =
-        rec_strdup (rGroup->recordType, &rec->size[recInfo_fileType]);
+        rec_strdup (zh->m_record_type, &rec->size[recInfo_fileType]);
 
     /* update filename */
     xfree (rec->info[recInfo_filename]);
@@ -605,19 +733,12 @@ static int recordExtract (ZebraHandle zh,
 
     /* update delete keys */
     xfree (rec->info[recInfo_delKeys]);
-    if (zh->reg->keys.buf_used > 0 && rGroup->flagStoreKeys == 1)
+    if (zh->reg->keys.buf_used > 0 && zh->m_store_keys == 1)
     {
-#if 1
         rec->size[recInfo_delKeys] = zh->reg->keys.buf_used;
         rec->info[recInfo_delKeys] = zh->reg->keys.buf;
         zh->reg->keys.buf = NULL;
         zh->reg->keys.buf_max = 0;
-#else
-        rec->info[recInfo_delKeys] = xmalloc (reckeys.buf_used);
-        rec->size[recInfo_delKeys] = reckeys.buf_used;
-        memcpy (rec->info[recInfo_delKeys], reckeys.buf,
-                rec->size[recInfo_delKeys]);
-#endif
     }
     else
     {
@@ -625,6 +746,14 @@ static int recordExtract (ZebraHandle zh,
         rec->size[recInfo_delKeys] = 0;
     }
 
+    /* update sort keys */
+    xfree (rec->info[recInfo_sortKeys]);
+
+    rec->size[recInfo_sortKeys] = zh->reg->sortKeys.buf_used;
+    rec->info[recInfo_sortKeys] = zh->reg->sortKeys.buf;
+    zh->reg->sortKeys.buf = NULL;
+    zh->reg->sortKeys.buf_max = 0;
+
     /* save file size of original record */
     zebraExplain_recordBytesIncrement (zh->reg->zei,
                                        - recordAttr->recordSize);
@@ -640,7 +769,7 @@ static int recordExtract (ZebraHandle zh,
 
     /* update store data */
     xfree (rec->info[recInfo_storeData]);
-    if (rGroup->flagStoreData == 1)
+    if (zh->m_store_data)
     {
         rec->size[recInfo_storeData] = recordAttr->recordSize;
         rec->info[recInfo_storeData] = (char *)
@@ -667,7 +796,7 @@ static int recordExtract (ZebraHandle zh,
     /* update database name */
     xfree (rec->info[recInfo_databaseName]);
     rec->info[recInfo_databaseName] =
-        rec_strdup (rGroup->databaseName, &rec->size[recInfo_databaseName]); 
+        rec_strdup (zh->basenames[0], &rec->size[recInfo_databaseName]); 
 
     /* update offset */
     recordAttr->recordOffset = recordOffset;
@@ -679,26 +808,20 @@ static int recordExtract (ZebraHandle zh,
 }
 
 int fileExtract (ZebraHandle zh, SYSNO *sysno, const char *fname, 
-                 const struct recordGroup *rGroupP, int deleteFlag)
+		 int deleteFlag)
 {
     int r, i, fd;
     char gprefix[128];
     char ext[128];
     char ext_res[128];
-    char subType[128];
-    RecType recType;
-    struct recordGroup rGroupM;
-    struct recordGroup *rGroup = &rGroupM;
     struct file_read_info *fi;
-    void *clientData;
+    const char *original_record_type = 0;
 
-    memcpy (rGroup, rGroupP, sizeof(*rGroupP));
-   
-    if (!rGroup->groupName || !*rGroup->groupName)
+    if (!zh->m_group || !*zh->m_group)
         *gprefix = '\0';
     else
-        sprintf (gprefix, "%s.", rGroup->groupName);
-
+        sprintf (gprefix, "%s.", zh->m_group);
+    
     logf (LOG_DEBUG, "fileExtract %s", fname);
 
     /* determine file extension */
@@ -712,98 +835,24 @@ int fileExtract (ZebraHandle zh, SYSNO *sysno, const char *fname,
             break;
         }
     /* determine file type - depending on extension */
-    if (!rGroup->recordType)
+    original_record_type = zh->m_record_type;
+    if (!zh->m_record_type)
     {
         sprintf (ext_res, "%srecordType.%s", gprefix, ext);
-        if (!(rGroup->recordType = res_get (zh->res, ext_res)))
-        {
-            sprintf (ext_res, "%srecordType", gprefix);
-            rGroup->recordType = res_get (zh->res, ext_res);
-        }
+        zh->m_record_type = res_get (zh->res, ext_res);
     }
-    if (!rGroup->recordType)
+    if (!zh->m_record_type)
     {
-        if (zh->records_processed < rGroup->fileVerboseLimit)
+	if (zh->records_processed < zh->m_file_verbose_limit)
             logf (LOG_LOG, "? %s", fname);
         return 0;
     }
-    if (!*rGroup->recordType)
-	return 0;
-    if (!(recType =
-	  recType_byName (zh->reg->recTypes, rGroup->recordType, subType,
-			  &clientData)))
-    {
-        logf (LOG_WARN, "No such record type: %s", rGroup->recordType);
-        return 0;
-    }
-
     /* determine match criteria */
-    if (!rGroup->recordId)
+    if (!zh->m_record_id)
     {
         sprintf (ext_res, "%srecordId.%s", gprefix, ext);
-        rGroup->recordId = res_get (zh->res, ext_res);
-    }
-
-    /* determine database name */
-    if (!rGroup->databaseName)
-    {
-        sprintf (ext_res, "%sdatabase.%s", gprefix, ext);
-        if (!(rGroup->databaseName = res_get (zh->res, ext_res)))
-        {
-            sprintf (ext_res, "%sdatabase", gprefix);
-            rGroup->databaseName = res_get (zh->res, ext_res);
-        }
-    }
-    if (!rGroup->databaseName)
-        rGroup->databaseName = "Default";
-
-    /* determine if explain database */
-    
-    sprintf (ext_res, "%sexplainDatabase", gprefix);
-    rGroup->explainDatabase =
-	atoi (res_get_def (zh->res, ext_res, "0"));
-
-    /* announce database */
-    if (zebraExplain_curDatabase (zh->reg->zei, rGroup->databaseName))
-    {
-        if (zebraExplain_newDatabase (zh->reg->zei, rGroup->databaseName,
-				      rGroup->explainDatabase))
-	    return 0;
-    }
-
-    if (rGroup->flagStoreData == -1)
-    {
-        const char *sval;
-        sprintf (ext_res, "%sstoreData.%s", gprefix, ext);
-        if (!(sval = res_get (zh->res, ext_res)))
-        {
-            sprintf (ext_res, "%sstoreData", gprefix);
-            sval = res_get (zh->res, ext_res);
-        }
-        if (sval)
-            rGroup->flagStoreData = atoi (sval);
-    }
-    if (rGroup->flagStoreData == -1)
-        rGroup->flagStoreData = 0;
-
-    if (rGroup->flagStoreKeys == -1)
-    {
-        const char *sval;
-
-        sprintf (ext_res, "%sstoreKeys.%s", gprefix, ext);
-        sval = res_get (zh->res, ext_res);
-	if (!sval)
-        {
-            sprintf (ext_res, "%sstoreKeys", gprefix);
-            sval = res_get (zh->res, ext_res);
-        }
-	if (!sval)
-	    sval = res_get (zh->res, "storeKeys");
-        if (sval)
-            rGroup->flagStoreKeys = atoi (sval);
+        zh->m_record_id = res_get (zh->res, ext_res);
     }
-    if (rGroup->flagStoreKeys == -1)
-        rGroup->flagStoreKeys = 0;
 
     if (sysno && deleteFlag)
         fd = -1;
@@ -824,6 +873,7 @@ int fileExtract (ZebraHandle zh, SYSNO *sysno, const char *fname,
         if ((fd = open (full_rep, O_BINARY|O_RDONLY)) == -1)
         {
             logf (LOG_WARN|LOG_ERRNO, "open %s", full_rep);
+	    zh->m_record_type = original_record_type;
             return 0;
         }
     }
@@ -831,34 +881,46 @@ int fileExtract (ZebraHandle zh, SYSNO *sysno, const char *fname,
     do
     {
         file_begin (fi);
-        r = recordExtract (zh, sysno, fname, rGroup, deleteFlag, fi,
-                           recType, subType, clientData);
+        r = file_extract_record (zh, sysno, fname, deleteFlag, fi, 1);
     } while (r && !sysno && fi->file_more);
     file_read_stop (fi);
     if (fd != -1)
         close (fd);
+    zh->m_record_type = original_record_type;
     return r;
 }
 
-
-int extract_rec_in_mem (ZebraHandle zh, const char *recordType,
-                        const char *buf, size_t buf_size,
-                        const char *databaseName, int delete_flag,
-                        int test_mode, int *sysno,
-                        int store_keys, int store_data,
-                        const char *match_criteria)
+/*
+  If sysno is provided, then it's used to identify the reocord.
+  If not, and match_criteria is provided, then sysno is guessed
+  If not, and a record is provided, then sysno is got from there
+  
+ */
+int buffer_extract_record (ZebraHandle zh, 
+			   const char *buf, size_t buf_size,
+			   int delete_flag,
+			   int test_mode, 
+			   const char *recordType,
+			   SYSNO *sysno,
+			   const char *match_criteria,
+			   const char *fname,
+			   int force_update,
+			   int allow_update)
 {
     RecordAttr *recordAttr;
     struct recExtractCtrl extractCtrl;
     int i, r;
-    char *matchStr = 0;
-    RecType recType;
+    const char *matchStr = 0;
+    RecType recType = NULL;
     char subType[1024];
     void *clientData;
-    const char *fname = "<no file>";
     Record rec;
     long recordOffset = 0;
     struct zebra_fetch_control fc;
+    const char *pr_fname = fname;  /* filename to print .. */
+
+    if (!pr_fname)
+	pr_fname = "<no file>";  /* make it printable if file is omitted */
 
     fc.fd = -1;
     fc.record_int_buf = buf;
@@ -874,26 +936,37 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType,
     extractCtrl.endf = zebra_record_int_end;
     extractCtrl.fh = &fc;
 
-    /* announce database */
-    if (zebraExplain_curDatabase (zh->reg->zei, databaseName))
+    create_rec_keys_codec(&zh->reg->keys);
+
+    zh->reg->sortKeys.buf_used = 0;
+
+    if (zebraExplain_curDatabase (zh->reg->zei, zh->basenames[0]))
     {
-        if (zebraExplain_newDatabase (zh->reg->zei, databaseName, 0))
-	    return 0;
+        if (zebraExplain_newDatabase (zh->reg->zei, zh->basenames[0], 
+				      zh->m_explain_database))
+            return 0;
     }
-    if (!(recType =
-	  recType_byName (zh->reg->recTypes, recordType, subType,
-			  &clientData)))
-    {
-        logf (LOG_WARN, "No such record type: %s", recordType);
+    
+    if (recordType && *recordType) {
+        logf (LOG_DEBUG, "Record type explicitly specified: %s", recordType);
+        recType = recType_byName (zh->reg->recTypes, recordType, subType,
+                                  &clientData);
+    } else {
+        if (!(zh->m_record_type)) {
+            logf (LOG_WARN, "No such record type defined");
+            return 0;
+        }
+        logf (LOG_DEBUG, "Get record type from rgroup: %s",zh->m_record_type);
+        recType = recType_byName (zh->reg->recTypes, zh->m_record_type, subType,
+                                  &clientData);
+        recordType = zh->m_record_type;
+    }
+    
+    if (!recType) {
+        logf (LOG_WARN, "No such record type: %s", zh->m_record_type);
         return 0;
     }
-
-    zh->reg->keys.buf_used = 0;
-    zh->reg->keys.prevAttrUse = -1;
-    zh->reg->keys.prevAttrSet = -1;
-    zh->reg->keys.prevSeqNo = 0;
-    zh->reg->sortKeys = 0;
-
+    
     extractCtrl.subType = subType;
     extractCtrl.init = extract_init;
     extractCtrl.tokenAdd = extract_token_add;
@@ -914,19 +987,16 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType,
 
     if (r == RECCTRL_EXTRACT_EOF)
 	return 0;
-    else if (r == RECCTRL_EXTRACT_ERROR)
+    else if (r == RECCTRL_EXTRACT_ERROR_GENERIC)
     {
 	/* error occured during extraction ... */
-#if 1
-	yaz_log (LOG_WARN, "extract error");
-#else
-	if (rGroup->flagRw &&
-	    zh->records_processed < rGroup->fileVerboseLimit)
-	{
-	    logf (LOG_WARN, "fail %s %s %ld", rGroup->recordType,
-		  fname, (long) recordOffset);
-	}
-#endif
+	yaz_log (LOG_WARN, "extract error: generic");
+	return 0;
+    }
+    else if (r == RECCTRL_EXTRACT_ERROR_NO_SUCH_FILTER)
+    {
+	/* error occured during extraction ... */
+	yaz_log (LOG_WARN, "extract error: no such filter");
 	return 0;
     }
     if (zh->reg->keys.buf_used == 0)
@@ -940,6 +1010,32 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType,
 	return 1;
     }
     /* match criteria */
+    matchStr = NULL;
+
+    if (! *sysno) {
+        char *rinfo;
+        if (match_criteria && *match_criteria) {
+            matchStr = match_criteria;
+        } else {
+            if (zh->m_record_id && *zh->m_record_id) {
+                matchStr = fileMatchStr (zh, &zh->reg->keys, pr_fname, 
+                                         zh->m_record_id);
+		if (!matchStr)
+                {
+                    logf (LOG_WARN, "Bad match criteria (recordID)");
+		    return 1;
+                }
+            }
+        }
+        if (matchStr) {
+            rinfo = dict_lookup (zh->reg->matchDict, matchStr);
+            if (rinfo)
+	    {
+		assert(*rinfo == sizeof(*sysno));
+                memcpy (sysno, rinfo+1, sizeof(*sysno));
+	    }
+        }
+    }
 
     if (! *sysno)
     {
@@ -947,11 +1043,11 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType,
         if (delete_flag)
         {
 	    logf (LOG_LOG, "delete %s %s %ld", recordType,
-		  fname, (long) recordOffset);
+		  pr_fname, (long) recordOffset);
             logf (LOG_WARN, "cannot delete record above (seems new)");
             return 1;
         }
-	logf (LOG_LOG, "add %s %s %ld", recordType, fname,
+	logf (LOG_LOG, "add %s %s %ld", recordType, pr_fname,
 	      (long) recordOffset);
         rec = rec_new (zh->reg->records);
 
@@ -964,30 +1060,49 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType,
             dict_insert (zh->reg->matchDict, matchStr,
                          sizeof(*sysno), sysno);
         }
-        extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys);
 	extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys);
-    }
+        extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys);
+
+        zh->records_inserted++;
+    } 
     else
     {
         /* record already exists */
         struct recKeys delkeys;
+        struct sortKeys sortKeys;
+
+	if (!allow_update) {
+	      logf (LOG_LOG, "skipped %s %s %ld", 
+		    recordType, pr_fname, (long) recordOffset);
+	      logRecord(zh);
+	      return -1;
+	}
 
         rec = rec_get (zh->reg->records, *sysno);
         assert (rec);
 	
 	recordAttr = rec_init_attr (zh->reg->zei, rec);
-
-	if (recordAttr->runNumber ==
-	    zebraExplain_runNumberIncrement (zh->reg->zei, 0))
-	{
-	    logf (LOG_LOG, "skipped %s %s %ld", recordType,
-		  fname, (long) recordOffset);
-	    rec_rm (&rec);
-	    return 1;
+	
+	if (!force_update) {
+	    if (recordAttr->runNumber ==
+		zebraExplain_runNumberIncrement (zh->reg->zei, 0))
+	    {
+		logf (LOG_LOG, "skipped %s %s %ld", recordType,
+		      pr_fname, (long) recordOffset);
+		extract_flushSortKeys (zh, *sysno, -1, &zh->reg->sortKeys);
+		rec_rm (&rec);
+		logRecord(zh);
+		return -1;
+	    }
 	}
+
         delkeys.buf_used = rec->size[recInfo_delKeys];
 	delkeys.buf = rec->info[recInfo_delKeys];
-	extract_flushSortKeys (zh, *sysno, 0, &zh->reg->sortKeys);
+
+        sortKeys.buf_used = rec->size[recInfo_sortKeys];
+        sortKeys.buf = rec->info[recInfo_sortKeys];
+
+	extract_flushSortKeys (zh, *sysno, 0, &sortKeys);
         extract_flushRecordKeys (zh, *sysno, 0, &delkeys);
         if (delete_flag)
         {
@@ -995,21 +1110,21 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType,
             if (!delkeys.buf_used)
             {
                 logf (LOG_LOG, "delete %s %s %ld", recordType,
-                      fname, (long) recordOffset);
+                      pr_fname, (long) recordOffset);
                 logf (LOG_WARN, "cannot delete file above, storeKeys false");
             }
             else
             {
 		logf (LOG_LOG, "delete %s %s %ld", recordType,
-		      fname, (long) recordOffset);
-#if 0
+		      pr_fname, (long) recordOffset);
+                zh->records_deleted++;
                 if (matchStr)
-                    dict_delete (matchDict, matchStr);
-#endif
+                    dict_delete (zh->reg->matchDict, matchStr);
                 rec_del (zh->reg->records, &rec);
             }
 	    rec_rm (&rec);
-            return 1;
+            logRecord(zh);
+            return 0;
         }
         else
         {
@@ -1017,14 +1132,16 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType,
             if (!delkeys.buf_used)
             {
                 logf (LOG_LOG, "update %s %s %ld", recordType,
-                      fname, (long) recordOffset);
+                      pr_fname, (long) recordOffset);
                 logf (LOG_WARN, "cannot update file above, storeKeys false");
             }
             else
             {
 		logf (LOG_LOG, "update %s %s %ld", recordType,
-		      fname, (long) recordOffset);
+		      pr_fname, (long) recordOffset);
+                extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys);
                 extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys);
+                zh->records_updated++;
             }
         }
     }
@@ -1040,7 +1157,7 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType,
 
     /* update delete keys */
     xfree (rec->info[recInfo_delKeys]);
-    if (zh->reg->keys.buf_used > 0 && store_keys == 1)
+    if (zh->reg->keys.buf_used > 0 && zh->m_store_keys == 1)
     {
         rec->size[recInfo_delKeys] = zh->reg->keys.buf_used;
         rec->info[recInfo_delKeys] = zh->reg->keys.buf;
@@ -1053,6 +1170,14 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType,
         rec->size[recInfo_delKeys] = 0;
     }
 
+    /* update sort keys */
+    xfree (rec->info[recInfo_sortKeys]);
+
+    rec->size[recInfo_sortKeys] = zh->reg->sortKeys.buf_used;
+    rec->info[recInfo_sortKeys] = zh->reg->sortKeys.buf;
+    zh->reg->sortKeys.buf = NULL;
+    zh->reg->sortKeys.buf_max = 0;
+
     /* save file size of original record */
     zebraExplain_recordBytesIncrement (zh->reg->zei,
 				       - recordAttr->recordSize);
@@ -1072,28 +1197,12 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType,
 
     /* update store data */
     xfree (rec->info[recInfo_storeData]);
-    if (store_data == 1)
+    if (zh->m_store_data)
     {
         rec->size[recInfo_storeData] = recordAttr->recordSize;
         rec->info[recInfo_storeData] = (char *)
 	    xmalloc (recordAttr->recordSize);
-#if 1
         memcpy (rec->info[recInfo_storeData], buf, recordAttr->recordSize);
-#else
-        if (lseek (fi->fd, recordOffset, SEEK_SET) < 0)
-        {
-            logf (LOG_ERRNO|LOG_FATAL, "seek to %ld in %s",
-                  (long) recordOffset, fname);
-            exit (1);
-        }
-        if (read (fi->fd, rec->info[recInfo_storeData], recordAttr->recordSize)
-	    < recordAttr->recordSize)
-        {
-            logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s",
-                  recordAttr->recordSize, fname);
-            exit (1);
-        }
-#endif
     }
     else
     {
@@ -1103,14 +1212,14 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType,
     /* update database name */
     xfree (rec->info[recInfo_databaseName]);
     rec->info[recInfo_databaseName] =
-        rec_strdup (databaseName, &rec->size[recInfo_databaseName]); 
+        rec_strdup (zh->basenames[0], &rec->size[recInfo_databaseName]); 
 
     /* update offset */
     recordAttr->recordOffset = recordOffset;
     
     /* commit this record */
     rec_put (zh->reg->records, &rec);
-
+    logRecord(zh);
     return 0;
 }
 
@@ -1129,11 +1238,9 @@ int explain_extract (void *handle, Record rec, data1_node *n)
             abort ();
     }
 
-    zh->reg->keys.buf_used = 0;
-    zh->reg->keys.prevAttrUse = -1;
-    zh->reg->keys.prevAttrSet = -1;
-    zh->reg->keys.prevSeqNo = 0;
-    zh->reg->sortKeys = 0;
+    create_rec_keys_codec(&zh->reg->keys);
+
+    zh->reg->sortKeys.buf_used = 0;
     
     extractCtrl.init = extract_init;
     extractCtrl.tokenAdd = extract_token_add;
@@ -1144,17 +1251,22 @@ int explain_extract (void *handle, Record rec, data1_node *n)
     extractCtrl.zebra_maps = zh->reg->zebra_maps;
     extractCtrl.flagShowRecords = 0;
     extractCtrl.handle = handle;
-    
-    grs_extract_tree(&extractCtrl, n);
+
+    if (n)
+	grs_extract_tree(&extractCtrl, n);
 
     if (rec->size[recInfo_delKeys])
     {
 	struct recKeys delkeys;
-	struct sortKey *sortKeys = 0;
+	struct sortKeys sortkeys;
 
 	delkeys.buf_used = rec->size[recInfo_delKeys];
 	delkeys.buf = rec->info[recInfo_delKeys];
-	extract_flushSortKeys (zh, rec->sysno, 0, &sortKeys);
+
+	sortkeys.buf_used = rec->size[recInfo_sortKeys];
+	sortkeys.buf = rec->info[recInfo_sortKeys];
+
+	extract_flushSortKeys (zh, rec->sysno, 0, &sortkeys);
 	extract_flushRecordKeys (zh, rec->sysno, 0, &delkeys);
     }
     extract_flushRecordKeys (zh, rec->sysno, 1, &zh->reg->keys);
@@ -1165,25 +1277,43 @@ int explain_extract (void *handle, Record rec, data1_node *n)
     rec->info[recInfo_delKeys] = zh->reg->keys.buf;
     zh->reg->keys.buf = NULL;
     zh->reg->keys.buf_max = 0;
+
+    xfree (rec->info[recInfo_sortKeys]);
+    rec->size[recInfo_sortKeys] = zh->reg->sortKeys.buf_used;
+    rec->info[recInfo_sortKeys] = zh->reg->sortKeys.buf;
+    zh->reg->sortKeys.buf = NULL;
+    zh->reg->sortKeys.buf_max = 0;
+
     return 0;
 }
 
 void extract_flushRecordKeys (ZebraHandle zh, SYSNO sysno,
                               int cmd, struct recKeys *reckeys)
 {
+#if IT_KEY_NEW
+    void *decode_handle = iscz1_start();
+#else
+    int seqno = 0;
 #if SU_SCHEME
 #else
     unsigned char attrSet = (unsigned char) -1;
     unsigned short attrUse = (unsigned short) -1;
 #endif
-    int seqno = 0;
+#endif
     int off = 0;
     int ch = 0;
     ZebraExplainInfo zei = zh->reg->zei;
 
     if (!zh->reg->key_buf)
     {
-	int mem = 8*1024*1024;
+	int mem= 1024*1024* atoi( res_get_def( zh->res, "memmax", "8"));
+	if (mem <= 0)
+	{
+	    logf(LOG_WARN, "Invalid memory setting, using default 8 MB");
+	    mem= 1024*1024*8;
+	}
+	/* FIXME: That "8" should be in a default settings include */
+	/* not hard-coded here! -H */
 	zh->reg->key_buf = (char**) xmalloc (mem);
 	zh->reg->ptr_top = mem/sizeof(char*);
 	zh->reg->ptr_i = 0;
@@ -1191,6 +1321,52 @@ void extract_flushRecordKeys (ZebraHandle zh, SYSNO sysno,
 	zh->reg->key_file_no = 0;
     }
     zebraExplain_recordCountIncrement (zei, cmd ? 1 : -1);
+#if IT_KEY_NEW
+    while (off < reckeys->buf_used)
+    {
+        const char *src = reckeys->buf + off;
+        struct it_key key;
+	char *dst = (char*) &key;
+	int attrSet, attrUse;
+
+	iscz1_decode(decode_handle, &dst, &src);
+	assert(key.len < 4 && key.len > 2);
+
+	attrSet = (int) key.mem[0];
+	attrUse = (int) key.mem[1]; /* sequence in mem[2] */
+
+        if (zh->reg->key_buf_used + 1024 > 
+            (zh->reg->ptr_top -zh->reg->ptr_i)*sizeof(char*))
+            extract_flushWriteKeys (zh,0);
+        ++(zh->reg->ptr_i);
+        assert(zh->reg->ptr_i > 0);
+        (zh->reg->key_buf)[zh->reg->ptr_top - zh->reg->ptr_i] =
+	    (char*)zh->reg->key_buf + zh->reg->key_buf_used;
+
+        ch = zebraExplain_lookupSU (zei, attrSet, attrUse);
+        if (ch < 0)
+            ch = zebraExplain_addSU (zei, attrSet, attrUse);
+
+        assert (ch > 0);
+	zh->reg->key_buf_used +=
+	    key_SU_encode (ch,((char*)zh->reg->key_buf) +
+                           zh->reg->key_buf_used);
+        while (*src)
+            ((char*)zh->reg->key_buf) [(zh->reg->key_buf_used)++] = *src++;
+        src++;
+        ((char*)(zh->reg->key_buf))[(zh->reg->key_buf_used)++] = '\0';
+        ((char*)(zh->reg->key_buf))[(zh->reg->key_buf_used)++] = cmd;
+
+        key.len = 2;
+	key.mem[0] = sysno;
+	key.mem[1] = key.mem[2];  /* sequence .. */
+	
+        memcpy ((char*)zh->reg->key_buf + zh->reg->key_buf_used,
+		&key, sizeof(key));
+        (zh->reg->key_buf_used) += sizeof(key);
+        off = src - reckeys->buf;
+    }
+#else
     while (off < reckeys->buf_used)
     {
         const char *src = reckeys->buf + off;
@@ -1219,8 +1395,10 @@ void extract_flushRecordKeys (ZebraHandle zh, SYSNO sysno,
 #endif
         if (zh->reg->key_buf_used + 1024 > 
             (zh->reg->ptr_top -zh->reg->ptr_i)*sizeof(char*))
-            extract_flushWriteKeys (zh);
+            extract_flushWriteKeys (zh,0);
+        assert(zh->reg->ptr_i >= 0);
         ++(zh->reg->ptr_i);
+        assert(zh->reg->ptr_i > 0);
         (zh->reg->key_buf)[zh->reg->ptr_top - zh->reg->ptr_i] =
 	    (char*)zh->reg->key_buf + zh->reg->key_buf_used;
 #if SU_SCHEME
@@ -1253,27 +1431,73 @@ void extract_flushRecordKeys (ZebraHandle zh, SYSNO sysno,
         (zh->reg->key_buf_used) += sizeof(key);
         off = src - reckeys->buf;
     }
+#endif
     assert (off == reckeys->buf_used);
+#if IT_KEY_NEW
+    iscz1_stop(decode_handle);
+#endif
 }
 
-void extract_flushWriteKeys (ZebraHandle zh)
+void extract_flushWriteKeys (ZebraHandle zh, int final)
+        /* optimizing: if final=1, and no files written yet */
+        /* push the keys directly to merge, sidestepping the */
+        /* temp file altogether. Speeds small updates */
 {
     FILE *outf;
     char out_fname[200];
     char *prevcp, *cp;
     struct encode_info encode_info;
     int ptr_i = zh->reg->ptr_i;
+    int temp_policy;
 #if SORT_EXTRA
     int i;
 #endif
     if (!zh->reg->key_buf || ptr_i <= 0)
+    {
+        logf (LOG_DEBUG, "  nothing to flush section=%d buf=%p i=%d",
+               zh->reg->key_file_no, zh->reg->key_buf, ptr_i);
+        logf (LOG_DEBUG, "  buf=%p ",
+               zh->reg->key_buf);
+        logf (LOG_DEBUG, "  ptr=%d ",zh->reg->ptr_i);
+        logf (LOG_DEBUG, "  reg=%p ",zh->reg);
+               
         return;
+    }
 
     (zh->reg->key_file_no)++;
     logf (LOG_LOG, "sorting section %d", (zh->reg->key_file_no));
+    logf (LOG_DEBUG, "  sort_buff at %p n=%d",
+                    zh->reg->key_buf + zh->reg->ptr_top - ptr_i,ptr_i);
 #if !SORT_EXTRA
-    qsort (zh->reg->key_buf + zh->reg->ptr_top - ptr_i, ptr_i, sizeof(char*),
-	    key_qsort_compare);
+    qsort (zh->reg->key_buf + zh->reg->ptr_top - ptr_i, ptr_i,
+               sizeof(char*), key_qsort_compare);
+
+    /* zebra.cfg: tempfiles:  
+       Y: always use temp files (old way) 
+       A: use temp files, if more than one (auto) 
+          = if this is both the last and the first 
+       N: never bother with temp files (new) */
+
+    temp_policy=toupper(res_get_def(zh->res,"tempfiles","auto")[0]);
+    if (temp_policy != 'Y' && temp_policy != 'N' && temp_policy != 'A') {
+        logf (LOG_WARN, "Illegal tempfiles setting '%c'. using 'Auto' ", 
+                        temp_policy);
+        temp_policy='A';
+    }
+
+    if (   ( temp_policy =='N' )   ||     /* always from memory */
+         ( ( temp_policy =='A' ) &&       /* automatic */
+             (zh->reg->key_file_no == 1) &&  /* this is first time */
+             (final) ) )                     /* and last (=only) time */
+    { /* go directly from memory */
+        zh->reg->key_file_no =0; /* signal not to read files */
+        zebra_index_merge(zh); 
+        zh->reg->ptr_i = 0;
+        zh->reg->key_buf_used = 0; 
+        return; 
+    }
+
+    /* Not doing directly from memory, write into a temp file */
     extract_get_fname_tmp (zh, out_fname, zh->reg->key_file_no);
 
     if (!(outf = fopen (out_fname, "wb")))
@@ -1292,6 +1516,7 @@ void extract_flushWriteKeys (ZebraHandle zh)
         cp = (zh->reg->key_buf)[zh->reg->ptr_top - ptr_i];
         if (strcmp (cp, prevcp))
         {
+            encode_key_flush ( &encode_info, outf);
             encode_key_init (&encode_info);
             encode_key_write (cp, &encode_info, outf);
             prevcp = cp;
@@ -1299,6 +1524,7 @@ void extract_flushWriteKeys (ZebraHandle zh)
         else
             encode_key_write (cp + strlen(cp), &encode_info, outf);
     }
+    encode_key_flush ( &encode_info, outf);
 #else
     qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_x_compare);
     extract_get_fname_tmp (out_fname, key_file_no);
@@ -1330,6 +1556,7 @@ void extract_flushWriteKeys (ZebraHandle zh)
                 cp = key_buf[ptr_top-ptr_i];
                 encode_key_write (cp+key_y_len, &encode_info, outf);
             }
+            encode_key_flush ( &encode_info, outf);
             if (!i)
                 break;
             prevcp = key_buf[ptr_top-ptr_i];
@@ -1345,18 +1572,22 @@ void extract_flushWriteKeys (ZebraHandle zh)
     zh->reg->key_buf_used = 0;
 }
 
-void extract_add_index_string (RecWord *p, const char *string,
-                               int length)
+void extract_add_index_string (RecWord *p, const char *str, int length)
 {
     char *dst;
+    ZebraHandle zh = p->extractCtrl->handle;
+    struct recKeys *keys = &zh->reg->keys;
+#if IT_KEY_NEW
+    struct it_key key;
+    const char *src = (char*) &key;
+#else
     unsigned char attrSet;
     unsigned short attrUse;
     int lead = 0;
     int diff = 0;
     int *pseqno = &p->seqno;
-    ZebraHandle zh = p->extractCtrl->handle;
     ZebraExplainInfo zei = zh->reg->zei;
-    struct recKeys *keys = &zh->reg->keys;
+#endif
     
     if (keys->buf_used+1024 > keys->buf_max)
     {
@@ -1370,6 +1601,29 @@ void extract_add_index_string (RecWord *p, const char *string,
     }
     dst = keys->buf + keys->buf_used;
 
+#if IT_KEY_NEW
+    key.len = 3;
+    key.mem[0] = p->attrSet;
+    key.mem[1] = p->attrUse;
+    key.mem[2] = p->seqno;
+
+#if 0
+    /* just for debugging .. */
+    yaz_log(LOG_LOG, "set=%d use=%d seqno=%d", p->attrSet, p->attrUse,
+	    p->seqno);
+#endif
+
+    iscz1_encode(keys->codec_handle, &dst, &src);
+
+    *dst++ = p->reg_type;
+    memcpy (dst, str, length);
+    dst += length;
+    *dst++ = '\0';
+#else
+    /* leader byte is encoded as follows:
+       bit 0 : 1 if attrset is unchanged; 0 if attrset is changed
+       bit 1 : 1 if attruse is unchanged; 0 if attruse is changed
+    */
     attrSet = p->attrSet;
     if (keys->buf_used > 0 && keys->prevAttrSet == attrSet)
         lead |= 1;
@@ -1380,13 +1634,13 @@ void extract_add_index_string (RecWord *p, const char *string,
         lead |= 2;
     else
         keys->prevAttrUse = attrUse;
-#if 1
+
     diff = 1 + *pseqno - keys->prevSeqNo;
     if (diff >= 1 && diff <= 15)
         lead |= (diff << 2);
     else
         diff = 0;
-#endif
+
     keys->prevSeqNo = *pseqno;
     
     *dst++ = lead;
@@ -1418,7 +1672,7 @@ void extract_add_index_string (RecWord *p, const char *string,
     }
 #endif
     *dst++ = p->reg_type;
-    memcpy (dst, string, length);
+    memcpy (dst, str, length);
     dst += length;
     *dst++ = '\0';
 
@@ -1427,29 +1681,45 @@ void extract_add_index_string (RecWord *p, const char *string,
         memcpy (dst, pseqno, sizeof(*pseqno));
         dst += sizeof(*pseqno);
     }
+#endif
     keys->buf_used = dst - keys->buf;
 }
 
-static void extract_add_sort_string (RecWord *p, const char *string,
+static void extract_add_sort_string (RecWord *p, const char *str,
 				     int length)
 {
-    struct sortKey *sk;
     ZebraHandle zh = p->extractCtrl->handle;
+    struct sortKeys *sk = &zh->reg->sortKeys;
+    int off = 0;
 
-    for (sk = zh->reg->sortKeys; sk; sk = sk->next)
-	if (sk->attrSet == p->attrSet && sk->attrUse == p->attrUse)
-	    return;
-
-    sk = (struct sortKey *) xmalloc (sizeof(*sk));
-    sk->next = zh->reg->sortKeys;
-    zh->reg->sortKeys = sk;
-
-    sk->string = (char *) xmalloc (length);
-    sk->length = length;
-    memcpy (sk->string, string, length);
-
-    sk->attrSet = p->attrSet;
-    sk->attrUse = p->attrUse;
+    while (off < sk->buf_used)
+    {
+        int set, use, slen;
+
+        off += key_SU_decode(&set, sk->buf + off);
+        off += key_SU_decode(&use, sk->buf + off);
+        off += key_SU_decode(&slen, sk->buf + off);
+        off += slen;
+        if (p->attrSet == set && p->attrUse == use)
+            return;
+    }
+    assert (off == sk->buf_used);
+    
+    if (sk->buf_used + IT_MAX_WORD > sk->buf_max)
+    {
+        char *b;
+        
+        b = (char *) xmalloc (sk->buf_max += 128000);
+        if (sk->buf_used > 0)
+            memcpy (b, sk->buf, sk->buf_used);
+        xfree (sk->buf);
+        sk->buf = b;
+    }
+    off += key_SU_encode(p->attrSet, sk->buf + off);
+    off += key_SU_encode(p->attrUse, sk->buf + off);
+    off += key_SU_encode(length, sk->buf + off);
+    memcpy (sk->buf + off, str, length);
+    sk->buf_used = off + length;
 }
 
 void extract_add_string (RecWord *p, const char *string, int length)
@@ -1468,7 +1738,7 @@ static void extract_add_incomplete_field (RecWord *p)
     const char **map = 0;
 
     if (remain > 0)
-	map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain);
+	map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain, 0);
 
     while (map)
     {
@@ -1480,7 +1750,7 @@ static void extract_add_incomplete_field (RecWord *p)
 	{
 	    remain = p->length - (b - p->string);
 	    if (remain > 0)
-		map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain);
+		map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain, 0);
 	    else
 		map = 0;
 	}
@@ -1495,7 +1765,7 @@ static void extract_add_incomplete_field (RecWord *p)
 		buf[i++] = *(cp++);
 	    remain = p->length - (b - p->string);
 	    if (remain > 0)
-		map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain);
+		map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain, 0);
 	    else
 		map = 0;
 	}
@@ -1512,17 +1782,24 @@ static void extract_add_complete_field (RecWord *p)
     char buf[IT_MAX_WORD+1];
     const char **map = 0;
     int i = 0, remain = p->length;
+    int first; /* first position */
+
+yaz_log(LOG_DEBUG, "Complete field, w='%s'", p->string);
 
     if (remain > 0)
-	map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, remain);
+	map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, remain, 1);
 
     while (remain > 0 && i < IT_MAX_WORD)
     {
 	while (map && *map && **map == *CHR_SPACE)
 	{
 	    remain = p->length - (b - p->string);
+
 	    if (remain > 0)
-		map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain);
+	    {
+		first = i ? 0 : 1;
+		map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain, first);
+	    }
 	    else
 		map = 0;
 	}
@@ -1535,14 +1812,24 @@ static void extract_add_complete_field (RecWord *p)
 	{
 	    const char *cp = *map;
 
-	    if (i >= IT_MAX_WORD)
-		break;
-	    while (i < IT_MAX_WORD && *cp)
-		buf[i++] = *(cp++);
+	    if (**map == *CHR_CUT)
+	    {
+		i = 0;
+	    }
+	    else
+	    {
+		if (i >= IT_MAX_WORD)
+		    break;
+yaz_log(LOG_DEBUG, "Adding string to index '%d'", *map);
+		while (i < IT_MAX_WORD && *cp)
+		    buf[i++] = *(cp++);
+	    }
 	    remain = p->length  - (b - p->string);
 	    if (remain > 0)
+	    {
 		map = zebra_maps_input (p->zebra_maps, p->reg_type, &b,
-					remain);
+					remain, 0);
+	    }
 	    else
 		map = 0;
 	}
@@ -1555,6 +1842,12 @@ static void extract_add_complete_field (RecWord *p)
 void extract_token_add (RecWord *p)
 {
     WRBUF wrbuf;
+#if 0
+    yaz_log (LOG_LOG, "token_add "
+	     "reg_type=%c attrSet=%d attrUse=%d seqno=%d s=%.*s",
+             p->reg_type, p->attrSet, p->attrUse, p->seqno, p->length,
+             p->string);
+#endif
     if ((wrbuf = zebra_replace(p->zebra_maps, p->reg_type, 0,
 			       p->string, p->length)))
     {
@@ -1574,22 +1867,28 @@ void extract_schema_add (struct recExtractCtrl *p, Odr_oid *oid)
 }
 
 void extract_flushSortKeys (ZebraHandle zh, SYSNO sysno,
-                            int cmd, struct sortKey **skp)
+                            int cmd, struct sortKeys *sk)
 {
-    struct sortKey *sk = *skp;
     SortIdx sortIdx = zh->reg->sortIdx;
+    int off = 0;
 
     sortIdx_sysno (sortIdx, sysno);
-    while (sk)
+
+    while (off < sk->buf_used)
     {
-	struct sortKey *sk_next = sk->next;
-	sortIdx_type (sortIdx, sk->attrUse);
-	sortIdx_add (sortIdx, sk->string, sk->length);
-	xfree (sk->string);
-	xfree (sk);
-	sk = sk_next;
+        int set, use, slen;
+        
+        off += key_SU_decode(&set, sk->buf + off);
+        off += key_SU_decode(&use, sk->buf + off);
+        off += key_SU_decode(&slen, sk->buf + off);
+        
+        sortIdx_type(sortIdx, use);
+        if (cmd == 1)
+            sortIdx_add(sortIdx, sk->buf + off, slen);
+        else
+            sortIdx_add(sortIdx, "", 1);
+        off += slen;
     }
-    *skp = 0;
 }
 
 void encode_key_init (struct encode_info *i)
@@ -1597,8 +1896,17 @@ void encode_key_init (struct encode_info *i)
     i->sysno = 0;
     i->seqno = 0;
     i->cmd = -1;
+    i->prevsys=0;
+    i->prevseq=0;
+    i->prevcmd=-1;
+    i->keylen=0;
+#if IT_KEY_NEW
+    i->encode_handle = iscz1_start();
+#endif
 }
 
+#if IT_KEY_NEW
+#else
 char *encode_key_int (int d, char *bp)
 {
     if (d <= 63)
@@ -1623,15 +1931,31 @@ char *encode_key_int (int d, char *bp)
     }
     return bp;
 }
+#endif
 
+#define OLDENCODE 1
+
+#ifdef OLDENCODE
+/* this is the old encode_key_write 
+ * may be deleted once we are confident that the new works
+ * HL 15-oct-2002
+ */
 void encode_key_write (char *k, struct encode_info *i, FILE *outf)
 {
     struct it_key key;
-    char *bp = i->buf;
+    char *bp = i->buf, *bp0;
+    const char *src = (char *) &key;
 
+    /* copy term to output buf */
     while ((*bp++ = *k++))
         ;
-    memcpy (&key, k+1, sizeof(struct it_key));
+    /* and copy & align key so we can mangle */
+    memcpy (&key, k+1, sizeof(struct it_key));  /* *k is insert/delete */
+#if IT_KEY_NEW
+    bp0 = bp++;
+    iscz1_encode(i->encode_handle, &bp, &src);
+    *bp0 = (*k * 128) + bp - bp0 - 1; /* length and insert/delete combined */
+#else
     bp = encode_key_int ( (key.sysno - i->sysno) * 2 + *k, bp);
     if (i->sysno != key.sysno)
     {
@@ -1643,6 +1967,7 @@ void encode_key_write (char *k, struct encode_info *i, FILE *outf)
     bp = encode_key_int (key.seqno - i->seqno, bp);
     i->seqno = key.seqno;
     i->cmd = *k;
+#endif
     if (fwrite (i->buf, bp - i->buf, 1, outf) != 1)
     {
         logf (LOG_FATAL|LOG_ERRNO, "fwrite");
@@ -1650,3 +1975,111 @@ void encode_key_write (char *k, struct encode_info *i, FILE *outf)
     }
 }
 
+void encode_key_flush (struct encode_info *i, FILE *outf)
+{ /* dummy routine */
+#if IT_KEY_NEW
+    iscz1_stop(i->encode_handle);
+#endif
+}
+
+#else
+
+/* new encode_key_write
+ * The idea is to buffer one more key, and compare them
+ * If we are going to delete and insert the same key, 
+ * we may as well not bother. Should make a difference in 
+ * updates with small modifications (appending to a mbox)
+ */
+void encode_key_write (char *k, struct encode_info *i, FILE *outf)
+{
+    struct it_key key;
+    char *bp; 
+
+    if (*k)  /* first time for new key */
+    {
+        bp = i->buf;
+        while ((*bp++ = *k++))
+            ;
+	i->keylen= bp - i->buf -1;    
+	assert(i->keylen+1+sizeof(struct it_key) < ENCODE_BUFLEN);
+    }
+    else
+    {
+	bp=i->buf + i->keylen;
+	*bp++=0;
+	k++;
+    }
+
+    memcpy (&key, k+1, sizeof(struct it_key));
+    if (0==i->prevsys) /* no previous filter, fill up */
+    {
+        i->prevsys=key.sysno;
+	i->prevseq=key.seqno;
+	i->prevcmd=*k;
+    }
+    else if ( (i->prevsys==key.sysno) &&
+              (i->prevseq==key.seqno) &&
+	      (i->prevcmd!=*k) )
+    { /* same numbers, diff cmd, they cancel out */
+        i->prevsys=0;
+    }
+    else 
+    { /* different stuff, write previous, move buf */
+        bp = encode_key_int ( (i->prevsys - i->sysno) * 2 + i->prevcmd, bp);
+	if (i->sysno != i->prevsys)
+	{
+	    i->sysno = i->prevsys;
+	    i->seqno = 0;
+        }
+        else if (!i->seqno && !i->prevseq && i->cmd == i->prevcmd)
+	{
+	    return; /* ??? Filters some sort of duplicates away */
+	            /* ??? Can this ever happen   -H 15oct02 */
+	}
+        bp = encode_key_int (i->prevseq - i->seqno, bp);
+        i->seqno = i->prevseq;
+        i->cmd = i->prevcmd;
+        if (fwrite (i->buf, bp - i->buf, 1, outf) != 1)
+        {
+            logf (LOG_FATAL|LOG_ERRNO, "fwrite");
+            exit (1);
+        }
+        i->keylen=0; /* ok, it's written, forget it */
+	i->prevsys=key.sysno;
+	i->prevseq=key.seqno;
+	i->prevcmd=*k;
+    }
+}
+
+void encode_key_flush (struct encode_info *i, FILE *outf)
+{ /* flush the last key from i */
+    char *bp =i->buf + i->keylen;
+    if (0==i->prevsys)
+    {
+        return; /* nothing to flush */
+    }
+    *bp++=0;
+    bp = encode_key_int ( (i->prevsys - i->sysno) * 2 + i->prevcmd, bp);
+    if (i->sysno != i->prevsys)
+    {
+        i->sysno = i->prevsys;
+        i->seqno = 0;
+    }
+    else if (!i->seqno && !i->prevseq && i->cmd == i->prevcmd)
+    {
+        return; /* ??? Filters some sort of duplicates away */
+                /* ??? Can this ever happen   -H 15oct02 */
+    }
+    bp = encode_key_int (i->prevseq - i->seqno, bp);
+    i->seqno = i->prevseq;
+    i->cmd = i->prevcmd;
+    if (fwrite (i->buf, bp - i->buf, 1, outf) != 1)
+    {
+        logf (LOG_FATAL|LOG_ERRNO, "fwrite");
+        exit (1);
+    }
+    i->keylen=0; /* ok, it's written, forget it */
+    i->prevsys=0; /* forget the values too */
+    i->prevseq=0;
+}
+#endif