1. BTREE_VERSION_1: using bti_itup->t_tid as unique identifier for a given

index tuple (logical position within A LEVEL). bti_oid & bti_dummy
taken off from BTItemData.
2. Fix for multi-column indices (nbtsearch.c):
   _bt_binsrch() - for searches on internal pages having keysize <
	number of attrs we point at the last item < the scankey, not at the
	first item = the scankey;
   _bt_moveright() - if keysize < number of attrs we compare scankey with
	_last_ item on current page to decide should we move right or
	not.

src/backend/access/nbtree/nbtinsert.c

@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *    $Header: /cvsroot/pgsql/src/backend/access/nbtree/nbtinsert.c,v 1.11 1997/03/24 08:48:09 vadim Exp $
+ *    $Header: /cvsroot/pgsql/src/backend/access/nbtree/nbtinsert.c,v 1.12 1997/04/16 01:48:11 vadim Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -33,7 +33,7 @@ static OffsetNumber _bt_findsplitloc(Relation rel, Page page, OffsetNumber start
 static void _bt_newroot(Relation rel, Buffer lbuf, Buffer rbuf);
 static OffsetNumber _bt_pgaddtup(Relation rel, Buffer buf, int keysz, ScanKey itup_scankey, Size itemsize, BTItem btitem, BTItem afteritem);
 static bool _bt_goesonpg(Relation rel, Buffer buf, Size keysz, ScanKey scankey, BTItem afteritem);
-static void _bt_updateitem(Relation rel, Size keysz, Buffer buf, Oid bti_oid, BTItem newItem);
+static void _bt_updateitem(Relation rel, Size keysz, Buffer buf, BTItem oldItem, BTItem newItem);
 static bool _bt_isequal (TupleDesc itupdesc, Page page, OffsetNumber offnum, int keysz, ScanKey scankey);
 
 /*
@@ -357,7 +357,7 @@ _bt_insertonpg(Relation rel,
        			DOUBLEALIGN (IndexTupleDSize (stack->bts_btitem->bti_itup)) ) 
        		{
 		    _bt_updateitem(rel, keysz, pbuf, 
-		    		stack->bts_btitem->bti_oid, lowLeftItem);
+		    		stack->bts_btitem, lowLeftItem);
 		    _bt_relbuf(rel, buf, BT_WRITE);
 		    _bt_relbuf(rel, rbuf, BT_WRITE);
 		}
@@ -644,23 +644,14 @@ _bt_findsplitloc(Relation rel,
     OffsetNumber saferight;
     ItemId nxtitemid, safeitemid;
     BTItem safeitem, nxtitem;
-    IndexTuple safetup, nxttup;
     Size nbytes;
-    TupleDesc itupdesc;
     int natts;
-    int attno;
-    Datum attsafe;
-    Datum attnext;
-    bool null;
     
-    itupdesc = RelationGetTupleDescriptor(rel);
     natts = rel->rd_rel->relnatts;
-    
     saferight = start;
     safeitemid = PageGetItemId(page, saferight);
     nbytes = ItemIdGetLength(safeitemid) + sizeof(ItemIdData);
     safeitem = (BTItem) PageGetItem(page, safeitemid);
-    safetup = &(safeitem->bti_itup);
     
     i = OffsetNumberNext(start);
     
@@ -670,26 +661,17 @@ _bt_findsplitloc(Relation rel,
 	nxtitemid = PageGetItemId(page, i);
 	nbytes += (ItemIdGetLength(nxtitemid) + sizeof(ItemIdData));
 	nxtitem = (BTItem) PageGetItem(page, nxtitemid);
-	nxttup = &(nxtitem->bti_itup);
-	
-	/* test against last known safe item */
-	for (attno = 1; attno <= natts; attno++) {
-	    attsafe = index_getattr(safetup, attno, itupdesc, &null);
-	    attnext = index_getattr(nxttup, attno, itupdesc, &null);
 
 	/* 
-	     *  If the tuple we're looking at isn't equal to the last safe one
+	 * Test against last known safe item:
-	     *  we saw, then it's our new safe tuple.
+	 * if the tuple we're looking at isn't equal to the last safe 
+	 * one we saw, then it's our new safe tuple.
 	 */
-	    
+	if ( !_bt_itemcmp (rel, natts, 
-	    if (!_bt_invokestrat(rel, attno, BTEqualStrategyNumber,
+			safeitem, nxtitem, BTEqualStrategyNumber) )
-				 attsafe, attnext)) {
+	{
-		safetup = nxttup;
+	    safeitem = nxtitem;
 	    saferight = i;
-		
-		/* break is for the attno for loop */
-		break;
-	    }
 	}
 	i = OffsetNumberNext(i);
     }
@@ -793,7 +775,7 @@ _bt_newroot(Relation rel, Buffer lbuf, Buffer rbuf)
     _bt_wrtbuf(rel, rootbuf);
     
     /* update metadata page with new root block number */
-    _bt_metaproot(rel, rootbknum);
+    _bt_metaproot(rel, rootbknum, 0);
 }
 
 /*
@@ -820,7 +802,6 @@ _bt_pgaddtup(Relation rel,
     Page page;
     BTPageOpaque opaque;
     BTItem chkitem;
-    Oid afteroid;
     
     page = BufferGetPage(buf);
     opaque = (BTPageOpaque) PageGetSpecialPointer(page);
@@ -829,14 +810,13 @@ _bt_pgaddtup(Relation rel,
     if (afteritem == (BTItem) NULL) {
 	itup_off = _bt_binsrch(rel, buf, keysz, itup_scankey, BT_INSERTION);
     } else {
-	afteroid = afteritem->bti_oid;
 	itup_off = first;
 	
 	do {
 	    chkitem =
 		(BTItem) PageGetItem(page, PageGetItemId(page, itup_off));
 	    itup_off = OffsetNumberNext(itup_off);
-	} while (chkitem->bti_oid != afteroid);
+	} while ( ! BTItemSame (chkitem, afteritem) );
     }
 
     (void) PageAddItem(page, (Item) btitem, itemsize, itup_off, LP_USED);
@@ -870,7 +850,6 @@ _bt_goesonpg(Relation rel,
     BTPageOpaque opaque;
     BTItem chkitem;
     OffsetNumber offnum, maxoff;
-    Oid afteroid;
     bool found;
     
     page = BufferGetPage(buf);
@@ -908,7 +887,6 @@ _bt_goesonpg(Relation rel,
 	return (false);
     
     /* damn, have to work for it.  i hate that. */
-    afteroid = afteritem->bti_oid;
     maxoff = PageGetMaxOffsetNumber(page);
     
     /*
@@ -924,7 +902,8 @@ _bt_goesonpg(Relation rel,
 	 offnum <= maxoff;
 	 offnum = OffsetNumberNext(offnum)) {
 	chkitem = (BTItem) PageGetItem(page, PageGetItemId(page, offnum));
-	if (chkitem->bti_oid == afteroid) {
+
+	if ( BTItemSame (chkitem, afteritem) ) {
 	    found = true;
 	    break;
 	}
@@ -1029,7 +1008,7 @@ static void
 _bt_updateitem(Relation rel,
 	       Size keysz,
 	       Buffer buf,
-	       Oid bti_oid,
+	       BTItem oldItem,
 	       BTItem newItem)
 {
     Page page;
@@ -1050,10 +1029,10 @@ _bt_updateitem(Relation rel,
     do {
 	item = (BTItem) PageGetItem(page, PageGetItemId(page, i));
 	i = OffsetNumberNext(i);
-    } while (i <= maxoff && item->bti_oid != bti_oid);
+    } while (i <= maxoff && ! BTItemSame (item, oldItem));
     
     /* this should never happen (in theory) */
-    if (item->bti_oid != bti_oid) {
+    if ( ! BTItemSame (item, oldItem) ) {
 	elog(FATAL, "_bt_getstackbuf was lying!!");
     }
     

src/backend/access/nbtree/nbtpage.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *    $Header: /cvsroot/pgsql/src/backend/access/nbtree/nbtpage.c,v 1.6 1996/11/05 10:35:30 scrappy Exp $
+ *    $Header: /cvsroot/pgsql/src/backend/access/nbtree/nbtpage.c,v 1.7 1997/04/16 01:48:15 vadim Exp $
  *
  *  NOTES
  *     Postgres btree pages look like ordinary relation pages.  The opaque
@@ -38,12 +38,20 @@
 
 #define BTREE_METAPAGE	0
 #define BTREE_MAGIC	0x053162
+
+#ifdef BTREE_VERSION_1
+#define BTREE_VERSION	1
+#else
 #define BTREE_VERSION	0
+#endif
 
 typedef struct BTMetaPageData {
     uint32	btm_magic;
     uint32	btm_version;
     BlockNumber	btm_root;
+#ifdef BTREE_VERSION_1
+    int32	btm_level;
+#endif
 } BTMetaPageData;
 
 #define	BTPageGetMeta(p) \
@@ -95,6 +103,9 @@ _bt_metapinit(Relation rel)
     metad.btm_magic = BTREE_MAGIC;
     metad.btm_version = BTREE_VERSION;
     metad.btm_root = P_NONE;
+#ifdef BTREE_VERSION_1
+    metad.btm_level = 0;
+#endif
     memmove((char *) BTPageGetMeta(pg), (char *) &metad, sizeof(metad));
     
     op = (BTPageOpaque) PageGetSpecialPointer(pg);
@@ -180,6 +191,17 @@ _bt_getroot(Relation rel, int access)
     Assert(metaopaque->btpo_flags & BTP_META);
     metad = BTPageGetMeta(metapg);
 
+    if (metad->btm_magic != BTREE_MAGIC) {
+	elog(WARN, "Index %s is not a btree",
+	     RelationGetRelationName(rel));
+    }
+    
+    if (metad->btm_version != BTREE_VERSION) {
+	elog(WARN, "Version mismatch on %s:  version %d file, version %d code",
+	     RelationGetRelationName(rel),
+	     metad->btm_version, BTREE_VERSION);
+    }
+    
     /* if no root page initialized yet, do it */
     if (metad->btm_root == P_NONE) {
 	
@@ -209,6 +231,9 @@ _bt_getroot(Relation rel, int access)
 	    rootblkno = BufferGetBlockNumber(rootbuf);
 	    rootpg = BufferGetPage(rootbuf);
 	    metad->btm_root = rootblkno;
+#ifdef BTREE_VERSION_1
+    	    metad->btm_level = 1;
+#endif
 	    _bt_pageinit(rootpg, BufferGetPageSize(rootbuf));
 	    rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpg);
 	    rootopaque->btpo_flags |= (BTP_LEAF | BTP_ROOT);
@@ -387,7 +412,7 @@ _bt_pageinit(Page page, Size size)
  *	a reference to or lock on the metapage.
  */
 void
-_bt_metaproot(Relation rel, BlockNumber rootbknum)
+_bt_metaproot(Relation rel, BlockNumber rootbknum, int level)
 {
     Buffer metabuf;
     Page metap;
@@ -400,6 +425,12 @@ _bt_metaproot(Relation rel, BlockNumber rootbknum)
     Assert(metaopaque->btpo_flags & BTP_META);
     metad = BTPageGetMeta(metap);
     metad->btm_root = rootbknum;
+#ifdef BTREE_VERSION_1
+    if ( level == 0 )			/* called from _do_insert */
+    	metad->btm_level += 1;
+    else
+    	metad->btm_level = level;	/* called from btsort */
+#endif
     _bt_wrtbuf(rel, metabuf);
 }
 
@@ -434,7 +465,7 @@ _bt_getstackbuf(Relation rel, BTStack stack, int access)
 	item = (BTItem) PageGetItem(page, itemid);
 	
 	/* if the item is where we left it, we're done */
-	if (item->bti_oid == stack->bts_btitem->bti_oid)
+	if ( BTItemSame (item, stack->bts_btitem) )
 	    return (buf);
 	
 	/* if the item has just moved right on this page, we're done */
@@ -445,7 +476,7 @@ _bt_getstackbuf(Relation rel, BTStack stack, int access)
 	    item = (BTItem) PageGetItem(page, itemid);
 	    
 	    /* if the item is where we left it, we're done */
-	    if (item->bti_oid == stack->bts_btitem->bti_oid)
+	    if ( BTItemSame (item, stack->bts_btitem) )
 		return (buf);
 	}
     }
@@ -471,7 +502,7 @@ _bt_getstackbuf(Relation rel, BTStack stack, int access)
 	     offnum = OffsetNumberNext(offnum)) {
 	    itemid = PageGetItemId(page, offnum);
 	    item = (BTItem) PageGetItem(page, itemid);
-	    if (item->bti_oid == stack->bts_btitem->bti_oid)
+	    if ( BTItemSame (item, stack->bts_btitem) )
 		return (buf);
 	}
     }

src/backend/access/nbtree/nbtsearch.c

@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *    $Header: /cvsroot/pgsql/src/backend/access/nbtree/nbtsearch.c,v 1.16 1997/03/24 08:48:12 vadim Exp $
+ *    $Header: /cvsroot/pgsql/src/backend/access/nbtree/nbtsearch.c,v 1.17 1997/04/16 01:48:17 vadim Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -160,6 +160,7 @@ _bt_moveright(Relation rel,
     ItemId hikey;
     ItemId itemid;
     BlockNumber rblkno;
+    int natts = rel->rd_rel->relnatts;
         
     page = BufferGetPage(buf);
     opaque = (BTPageOpaque) PageGetSpecialPointer(page);
@@ -178,22 +179,43 @@ _bt_moveright(Relation rel,
      */
     
     if (_bt_skeycmp(rel, keysz, scankey, page, hikey,
-		    BTGreaterEqualStrategyNumber)) {
+		    BTGreaterEqualStrategyNumber))
-	
+    {
 	/* move right as long as we need to */
-	do {
+	do
+	{
+	    OffsetNumber offmax;
 	    /*
 	     *  If this page consists of all duplicate keys (hikey and first
 	     *  key on the page have the same value), then we don't need to
 	     *  step right.
+	     *
+	     *  NOTE for multi-column indices: we may do scan using
+	     *  keys not for all attrs. But we handle duplicates
+	     *  using all attrs in _bt_insert/_bt_spool code.
+	     *  And so we've to compare scankey with _last_ item
+	     *  on this page to do not lose "good" tuples if number
+	     *  of attrs > keysize. Example: (2,0) - last items on 
+	     *  this page, (2,1) - first item on next page (hikey), 
+	     *  our scankey is x = 2. Scankey >= (2,1) because of 
+	     *  we compare first attrs only, but we shouldn't to move
+	     *  right of here.		- vadim 04/15/97
 	     */
-	    if (PageGetMaxOffsetNumber(page) > P_HIKEY) {
+	    if ( (offmax = PageGetMaxOffsetNumber(page)) > P_HIKEY)
+	    {
 		itemid = PageGetItemId(page, P_FIRSTKEY);
 		if (_bt_skeycmp(rel, keysz, scankey, page, itemid,
 				BTEqualStrategyNumber)) {
 		    /* break is for the "move right" while loop */
 		    break;
 		}
+		else if ( natts > keysz )
+		{
+		    itemid = PageGetItemId(page, offmax);
+		    if (_bt_skeycmp(rel, keysz, scankey, page, itemid,
+				BTLessEqualStrategyNumber))
+		    	break;
+		}
 	    }
 	    
 	    /* step right one page */
@@ -346,6 +368,7 @@ _bt_binsrch(Relation rel,
     Page page;
     BTPageOpaque opaque;
     OffsetNumber low, mid, high;
+    int natts = rel->rd_rel->relnatts;
     int result;
     
     page = BufferGetPage(buf);
@@ -379,7 +402,26 @@ _bt_binsrch(Relation rel,
 	else if (result < 0)
 	    high = mid - 1;
 	else
-	    return (_bt_firsteq(rel, itupdesc, page, keysz, scankey, mid));
+	{
+	    mid = _bt_firsteq(rel, itupdesc, page, keysz, scankey, mid);
+	    /*
+	     *  NOTE for multi-column indices: we may do scan using
+	     *  keys not for all attrs. But we handle duplicates using
+	     *  all attrs in _bt_insert/_bt_spool code. And so while
+	     *  searching on internal pages having number of attrs > keysize
+	     *  we want to point at the last item < the scankey, not at the 
+	     *  first item = the scankey (!!!), and let _bt_moveright
+	     *  decide later whether to move right or not (see comments and 
+	     *  example there). Note also that INSERTions are not affected 
+	     *  by this code (natts == keysz).		- vadim 04/15/97
+	     */
+	    if ( natts == keysz || opaque->btpo_flags & BTP_LEAF )
+	    	return (mid);
+    	    low = P_RIGHTMOST(opaque) ? P_HIKEY : P_FIRSTKEY;
+    	    if ( mid == low )
+    	    	return (mid);
+	    return (OffsetNumberPrev(mid));
+	}
     }
 	
     /*
@@ -389,35 +431,45 @@ _bt_binsrch(Relation rel,
      *  For non-insertion searches on internal pages, we want to point at
      *  the last key <, or first key =, the scankey on the page.  This
      *  guarantees that we'll descend the tree correctly.
+     *  (NOTE comments above for multi-column indices).
      *
      *  For all other cases, we want to point at the first key >=
      *  the scankey on the page.  This guarantees that scans and
      *  insertions will happen correctly.
      */
 	
-	opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+    if (!(opaque->btpo_flags & BTP_LEAF) && srchtype == BT_DESCENT)
-	if (!(opaque->btpo_flags & BTP_LEAF) && srchtype == BT_DESCENT) {
+    {	/*
-	    
-	    /*
 	 *  We want the last key <, or first key ==, the scan key.
 	 */
-	    
 	result = _bt_compare(rel, itupdesc, page, keysz, scankey, high);
 	
-	    if (result == 0) {
+	if (result == 0)
-		return (_bt_firsteq(rel, itupdesc, page, keysz, scankey, high));
+	{
-	    } else if (result > 0) {
+	    mid = _bt_firsteq(rel, itupdesc, page, keysz, scankey, high);
+	    /*
+	     *  If natts > keysz we want last item < the scan key.
+	     *  See comments above for multi-column indices.
+	     */
+	    if ( natts == keysz )
+	    	return (mid);
+    	    low = P_RIGHTMOST(opaque) ? P_HIKEY : P_FIRSTKEY;
+    	    if ( mid == low )
+    	    	return (mid);
+	    return (OffsetNumberPrev(mid));
+	}
+	else if (result > 0)
 	    return (high);
-	    } else {
+	else
 	    return (low);
     } 
-	} else {
+    else	/* we want the first key >= the scan key */
-	    
+    {
-	    /* we want the first key >= the scan key */
 	result = _bt_compare(rel, itupdesc, page, keysz, scankey, low);
-	    if (result <= 0) {
+	if (result <= 0)
 	    return (low);
-	    } else {
+	else
+	{
 	    if (low == high)
 		return (OffsetNumberNext(low));
 		
@@ -1107,7 +1159,7 @@ _bt_twostep(IndexScanDesc scan, Buffer *bufP, ScanDirection dir)
     while (offnum <= maxoff) {
 	itemid = PageGetItemId(page, offnum);
 	btitem = (BTItem) PageGetItem(page, itemid);
-	if (btitem->bti_oid == svitem->bti_oid) {
+	if ( BTItemSame (btitem, svitem) ) {
 	    pfree(svitem);
 	    ItemPointerSet(current, blkno, offnum);
 	    return (false);

src/backend/access/nbtree/nbtsort.c

@@ -5,7 +5,7 @@
  *
  *
  * IDENTIFICATION
- *    $Id: nbtsort.c,v 1.13 1997/03/24 08:48:15 vadim Exp $
+ *    $Id: nbtsort.c,v 1.14 1997/04/16 01:48:27 vadim Exp $
  *
  * NOTES
  *
@@ -1021,9 +1021,13 @@ _bt_buildadd(Relation index, void *pstate, BTItem bti, int flags)
     }
 #endif /* FASTBUILD_DEBUG && FASTBUILD_MERGE */
 #endif
-    if (last_bti == (BTItem) NULL) {
+    if (last_bti == (BTItem) NULL)
+    {
 	first_off = P_FIRSTKEY;
-    } else if (!_bt_itemcmp(index, _bt_nattr, bti, last_bti, BTEqualStrategyNumber)) {
+    }
+    else if ( !_bt_itemcmp(index, _bt_nattr, 
+    			bti, last_bti, BTEqualStrategyNumber) )
+    {
 	first_off = off;
     }
     last_off = off;
@@ -1061,7 +1065,7 @@ _bt_uppershutdown(Relation index, BTPageState *state)
 	if (s->btps_doupper) {
 	    if (s->btps_next == (BTPageState *) NULL) {
 		opaque->btpo_flags |= BTP_ROOT;
-		_bt_metaproot(index, blkno);
+		_bt_metaproot(index, blkno, s->btps_level + 1);
 	    } else {
 		bti = _bt_minitem(s->btps_page, blkno, 0);
 		(void) _bt_buildadd(index, s->btps_next, bti, 0);

src/backend/access/nbtree/nbtutils.c

@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *    $Header: /cvsroot/pgsql/src/backend/access/nbtree/nbtutils.c,v 1.9 1997/03/24 08:48:16 vadim Exp $
+ *    $Header: /cvsroot/pgsql/src/backend/access/nbtree/nbtutils.c,v 1.10 1997/04/16 01:48:29 vadim Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -297,7 +297,9 @@ _bt_formitem(IndexTuple itup)
     btitem = (BTItem) palloc(nbytes_btitem);
     memmove((char *) &(btitem->bti_itup), (char *) itup, tuplen);
     
+#ifndef BTREE_VERSION_1
     btitem->bti_oid = newoid();
+#endif
     return (btitem);
 }