From: Adam Dickmeiss <adam@indexdata.dk>
Date: Mon, 13 May 1996 14:23:03 +0000 (+0000)
Subject: Work on compaction of set/use bytes in dictionary.
X-Git-Tag: ZEBRA.1.0~464
X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=commitdiff_plain;h=bad279c5dd1ae783ef9393775f0483cc3228cdbb

Work on compaction of set/use bytes in dictionary.
---

diff --git a/index/Makefile b/index/Makefile
index ab81dec..9246d15 100644
--- a/index/Makefile
+++ b/index/Makefile
@@ -1,7 +1,7 @@
 # Copyright (C) 1995, Index Data I/S 
 # All rights reserved.
 # Sebastian Hammer, Adam Dickmeiss
-# $Id: Makefile,v 1.33 1996-05-01 13:46:35 adam Exp $
+# $Id: Makefile,v 1.34 1996-05-13 14:23:03 adam Exp $
 
 SHELL=/bin/sh
 RANLIB=ranlib
@@ -20,7 +20,7 @@ TPROG3=zebrasrv
 DEFS=$(INCLUDE)
 O1 = main.o dir.o dirs.o trav.o extract.o kinput.o kcompare.o \
  symtab.o text.o recctrl.o structrec.o recindex.o regxread.o recstat.o \
- lockutil.o lockidx.o
+ lockutil.o lockidx.o zinfo.o
 O2 = kdump.o
 O3 = zserver.o kcompare.o zrpn.o zsets.o text.o recctrl.o structrec.o  \
 	attribute.o recindex.o zlogs.o regxread.o lockutil.o locksrv.o
diff --git a/index/extract.c b/index/extract.c
index 42e86de..62a7613 100644
--- a/index/extract.c
+++ b/index/extract.c
@@ -4,7 +4,10 @@
  * Sebastian Hammer, Adam Dickmeiss
  *
  * $Log: extract.c,v $
- * Revision 1.56  1996-05-09 09:54:42  adam
+ * Revision 1.57  1996-05-13 14:23:04  adam
+ * Work on compaction of set/use bytes in dictionary.
+ *
+ * Revision 1.56  1996/05/09  09:54:42  adam
  * Server supports maps from one logical attributes to a list of physical
  * attributes.
  * The extraction process doesn't make space consuming 'any' keys.
@@ -213,7 +216,7 @@
 #include <recctrl.h>
 #include "index.h"
 
-#include "recindex.h"
+#include "zinfo.h"
 
 static Dict matchDict;
 
@@ -230,6 +233,8 @@ static int records_updated = 0;
 static int records_deleted = 0;
 static int records_processed = 0;
 
+static ZebTargetInfo *zti = NULL;
+
 static void logRecord (int showFlag)
 {
     if (!showFlag)
@@ -260,6 +265,9 @@ void key_open (int mem)
     }
     assert (!records);
     records = rec_open (1);
+#if 1
+    zti = zebTargetInfo_open (records, 1);
+#endif
 }
 
 struct encode_info {
@@ -373,6 +381,9 @@ int key_close (void)
 {
     key_flush ();
     xfree (key_buf);
+#if 1
+    zebTargetInfo_close (zti, 1);
+#endif
     rec_close (&records);
     dict_close (matchDict);
 
@@ -461,6 +472,14 @@ static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys,
     char attrSet = -1;
     short attrUse = -1;
     int off = 0;
+
+#if 1
+    if (zebTargetInfo_curDatabase (zti, databaseName))
+    {
+        if (zebTargetInfo_newDatabase (zti, databaseName))
+            abort ();
+    }
+#endif
     while (off < reckeys->buf_used)
     {
         const char *src = reckeys->buf + off;
@@ -483,8 +502,16 @@ static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys,
             key_flush ();
         ++ptr_i;
         key_buf[ptr_top-ptr_i] = (char*)key_buf + key_buf_used;
+#if 1
+        lead = zebTargetInfo_lookupSU (zti, attrSet, attrUse);
+        if (lead < 0)
+            lead = zebTargetInfo_addSU (zti, attrSet, attrUse);
+        assert (lead > 0);
+        ((char*) key_buf) [key_buf_used++] = lead;
+#else
         key_buf_used += index_word_prefix ((char*)key_buf + key_buf_used,
                                            attrSet, attrUse, databaseName);
+#endif
         while (*src)
             ((char*)key_buf) [key_buf_used++] = index_char_cvt (*src++);
         src++;
diff --git a/index/kcompare.c b/index/kcompare.c
index 878f491..3ca6daa 100644
--- a/index/kcompare.c
+++ b/index/kcompare.c
@@ -4,7 +4,10 @@
  * Sebastian Hammer, Adam Dickmeiss
  *
  * $Log: kcompare.c,v $
- * Revision 1.15  1995-11-20 16:59:46  adam
+ * Revision 1.16  1996-05-13 14:23:05  adam
+ * Work on compaction of set/use bytes in dictionary.
+ *
+ * Revision 1.15  1995/11/20  16:59:46  adam
  * New update method: the 'old' keys are saved for each records.
  *
  * Revision 1.14  1995/10/30  15:08:08  adam
@@ -126,6 +129,7 @@ int index_char_cvt (int c)
     return tolower (c);
 }
 
+#if 1
 int index_word_prefix (char *string, int attset_ordinal,
                        int local_attribute,
                        const char *databaseName)
@@ -137,4 +141,4 @@ int index_word_prefix (char *string, int attset_ordinal,
         string[i] = index_char_cvt (string[i]);
     return i;
 }
-
+#endif
diff --git a/index/recindex.c b/index/recindex.c
index ab5e687..936f6e0 100644
--- a/index/recindex.c
+++ b/index/recindex.c
@@ -4,7 +4,10 @@
  * Sebastian Hammer, Adam Dickmeiss
  *
  * $Log: recindex.c,v $
- * Revision 1.14  1996-02-01 20:48:15  adam
+ * 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.
  *
@@ -552,6 +555,9 @@ 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);
diff --git a/index/zinfo.c b/index/zinfo.c
new file mode 100644
index 0000000..4f34867
--- /dev/null
+++ b/index/zinfo.c
@@ -0,0 +1,263 @@
+/*
+ * Copyright (C) 1994-1996, Index Data I/S 
+ * All rights reserved.
+ * Sebastian Hammer, Adam Dickmeiss
+ *
+ * $Log: zinfo.c,v $
+ * Revision 1.1  1996-05-13 14:23:07  adam
+ * Work on compaction of set/use bytes in dictionary.
+ *
+ */
+
+#include <stdlib.h>
+#include <assert.h>
+
+#include "zinfo.h"
+
+struct zebSUInfo {
+    int set;
+    int use;
+    int ordinal;
+};
+
+struct zebSUInfoB {
+    struct zebSUInfo info;
+    struct zebSUInfoB *next;
+};
+
+struct zebDatabaseInfo {
+    struct zebSUInfoB *SUInfo;
+    char *databaseName;
+    int sysno;
+    int readFlag;
+    struct zebDatabaseInfo *next;
+};
+
+struct zebTargetInfo {
+    int  dictNum;
+    int  dirty;
+    Records records;
+    struct zebDatabaseInfo *databaseInfo;
+    struct zebDatabaseInfo *curDatabaseInfo;
+};
+
+void zebTargetInfo_close (ZebTargetInfo *zti, int writeFlag)
+{
+    struct zebDatabaseInfo *zdi, *zdi1;
+    
+    if (writeFlag)
+    {
+        Record grec;
+        char p0[4096], *p = p0;
+
+        memcpy (p, &zti->dictNum, sizeof(zti->dictNum));
+        p += sizeof(zti->dictNum);
+        for (zdi = zti->databaseInfo; zdi; zdi=zdi->next)
+        {
+            if (zdi->readFlag || !zdi->sysno)
+            {
+                char q0[4096], *q = q0;
+                struct zebSUInfoB *zsui;
+                Record drec;
+                int no = 0;
+                
+                if (zdi->sysno)
+                    drec = rec_get (zti->records, zdi->sysno);
+                else
+                {
+                    drec = rec_new (zti->records);
+                    zdi->sysno = drec->sysno;
+                }
+                assert (drec);
+                for (zsui = zdi->SUInfo; zsui; zsui=zsui->next)
+                    no++;
+                memcpy (q, &no, sizeof(no));
+                q += sizeof(no);
+                for (zsui = zdi->SUInfo; zsui; zsui=zsui->next)
+                {
+                    memcpy (q, &zsui->info, sizeof(zsui->info));
+                    q += sizeof(zsui->info);
+                }
+                xfree (drec->info[0]);
+                drec->size[0] = q-q0;
+                drec->info[0] = xmalloc (drec->size[0]);
+                memcpy (drec->info[0], q0, drec->size[0]);
+                rec_put (zti->records, &drec);
+            }
+            strcpy (p, zdi->databaseName);
+            p += strlen(p)+1;
+            memcpy (p, &zdi->sysno, sizeof(zdi->sysno));
+            p += sizeof(zdi->sysno);
+        }
+        *p = '\0';
+        grec = rec_get (zti->records, 1);
+        xfree (grec->info[0]);
+        grec->size[0] = p-p0;
+        grec->info[0] = xmalloc (grec->size[0]);
+        memcpy (grec->info[0], p0, grec->size[0]);
+        rec_put (zti->records, &grec);
+    }
+    for (zdi = zti->databaseInfo; zdi; zdi = zdi1)
+    {
+        struct zebSUInfoB *zsui, *zsui1;
+
+        zdi1 = zdi->next;
+        for (zsui = zdi->SUInfo; zsui; zsui = zsui1)
+        {
+            zsui1 = zsui->next;
+            xfree (zsui);
+        }
+        xfree (zdi->databaseName);
+        xfree (zdi);
+    }
+    xfree (zti);
+}
+
+ZebTargetInfo *zebTargetInfo_open (Records records, int writeFlag)
+{
+    Record rec;
+    ZebTargetInfo *zti;
+    struct zebDatabaseInfo **zdi;
+
+    zti = xmalloc (sizeof(*zti));
+    zti->dirty = 0;
+    zti->curDatabaseInfo = NULL;
+    zti->records = records;
+
+    zdi = &zti->databaseInfo;
+    
+    rec = rec_get (records, 1);
+    if (rec)
+    {
+        const char *p;
+
+        p = rec->info[0];
+
+        memcpy (&zti->dictNum, p, sizeof(zti->dictNum));
+        p += sizeof(zti->dictNum);
+        while (*p)
+        {
+            *zdi = xmalloc (sizeof(**zdi));
+            (*zdi)->SUInfo = NULL;
+            (*zdi)->databaseName = xstrdup (p);
+            p += strlen(p)+1;
+            memcpy (&(*zdi)->sysno, p, sizeof((*zdi)->sysno));
+            p += sizeof((*zdi)->sysno);
+            (*zdi)->readFlag = 1;
+            zdi = &(*zdi)->next;
+        }
+        assert (p - rec->info[0] == rec->size[0]-1);
+    }
+    else
+    {
+        zti->dictNum = 1;
+        if (writeFlag)
+        {
+            rec = rec_new (records);
+            rec->info[0] = xmalloc (1+sizeof(zti->dictNum));
+            memcpy (rec->info[0], &zti->dictNum, sizeof(zti->dictNum));
+            rec->info[0][sizeof(zti->dictNum)] = '\0';
+            rec_put (records, &rec);
+        }
+    }
+    *zdi = NULL;
+    rec_rm (&rec);
+    return zti;
+}
+
+static void zebTargetInfo_readDatabase (ZebTargetInfo *zti,
+                                        struct zebDatabaseInfo *zdi)
+{
+    const char *p;
+    struct zebSUInfoB **zsuip = &zdi->SUInfo;
+    int i, no;
+    Record rec;
+
+    rec = rec_get (zti->records, zdi->sysno);
+    assert (rec);
+    p = rec->info[0];
+    memcpy (&no, p, sizeof(no));
+    p += sizeof(no);
+    for (i = 0; i<no; i++)
+    {
+        *zsuip = xmalloc (sizeof(**zsuip));
+        memcpy (&(*zsuip)->info, p, sizeof((*zsuip)->info));
+        p += sizeof((*zsuip)->info);
+        zsuip = &(*zsuip)->next;
+    }
+    *zsuip = NULL;
+    zdi->readFlag = 0;
+    rec_rm (&rec);
+}
+
+int zebTargetInfo_curDatabase (ZebTargetInfo *zti, const char *database)
+{
+    struct zebDatabaseInfo *zdi;
+    
+    assert (zti);
+    if (zti->curDatabaseInfo &&
+        !strcmp (zti->curDatabaseInfo->databaseName, database))
+        return 0;
+    for (zdi = zti->databaseInfo; zdi; zdi=zdi->next)
+    {
+        if (!strcmp (zdi->databaseName, database))
+            break;
+    }
+    if (!zdi)
+        return -1;
+    if (zdi->readFlag)
+        zebTargetInfo_readDatabase (zti, zdi);
+    zti->curDatabaseInfo = zdi;
+    return 0;
+}
+
+int zebTargetInfo_newDatabase (ZebTargetInfo *zti, const char *database)
+{
+    struct zebDatabaseInfo *zdi;
+
+    assert (zti);
+    for (zdi = zti->databaseInfo; zdi; zdi=zdi->next)
+    {
+        if (!strcmp (zdi->databaseName, database))
+            break;
+    }
+    if (zdi)
+        return -1;
+    zdi = xmalloc (sizeof(*zdi));
+    zdi->next = zti->databaseInfo;
+    zti->databaseInfo = zdi;
+    zdi->sysno = 0;
+    zdi->readFlag = 0;
+    zdi->databaseName = xstrdup (database);
+    zdi->SUInfo = NULL;
+    zti->curDatabaseInfo = zdi;
+    return 0;
+}
+
+int zebTargetInfo_lookupSU (ZebTargetInfo *zti, int set, int use)
+{
+    struct zebSUInfoB *zsui;
+
+    assert (zti->curDatabaseInfo);
+    for (zsui = zti->curDatabaseInfo->SUInfo; zsui; zsui=zsui->next)
+        if (zsui->info.use == use && zsui->info.set == set)
+            return zsui->info.ordinal;
+    return -1;
+}
+
+int zebTargetInfo_addSU (ZebTargetInfo *zti, int set, int use)
+{
+    struct zebSUInfoB *zsui;
+
+    assert (zti->curDatabaseInfo);
+    for (zsui = zti->curDatabaseInfo->SUInfo; zsui; zsui=zsui->next)
+        if (zsui->info.use == use && zsui->info.set == set)
+            return -1;
+    zsui = xmalloc (sizeof(*zsui));
+    zsui->next = zti->curDatabaseInfo->SUInfo;
+    zti->curDatabaseInfo->SUInfo = zsui;
+    zsui->info.set = set;
+    zsui->info.use = use;
+    zsui->info.ordinal = (zti->dictNum)++;
+    return zsui->info.ordinal;
+}
diff --git a/index/zinfo.h b/index/zinfo.h
new file mode 100644
index 0000000..09368d4
--- /dev/null
+++ b/index/zinfo.h
@@ -0,0 +1,21 @@
+/*
+ * Copyright (C) 1994-1996, Index Data I/S 
+ * All rights reserved.
+ * Sebastian Hammer, Adam Dickmeiss
+ *
+ * $Log: zinfo.h,v $
+ * Revision 1.1  1996-05-13 14:23:07  adam
+ * Work on compaction of set/use bytes in dictionary.
+ *
+ */
+
+#include "recindex.h"
+
+typedef struct zebTargetInfo ZebTargetInfo;
+
+ZebTargetInfo *zebTargetInfo_open (Records records, int writeFlag);
+void zebTargetInfo_close (ZebTargetInfo *zti, int writeFlag);
+int zebTargetInfo_curDatabase (ZebTargetInfo *zti, const char *database);
+int zebTargetInfo_newDatabase (ZebTargetInfo *zti, const char *database);
+int zebTargetInfo_lookupSU (ZebTargetInfo *zti, int set, int use);
+int zebTargetInfo_addSU (ZebTargetInfo *zti, int set, int use);