Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commit6312c08

Browse files
nbtree: Use raw PageAddItem() for retail inserts.
Only internal page splits need to call _bt_pgaddtup() instead ofPageAddItem(), and only for data items, one of which will end up at thefirst offset (or first offset after the high key offset) on the newright page. This data item alone will need to be truncated in_bt_pgaddtup().Since there is no reason why retail inserts ever need to truncate theincoming item, use a raw PageAddItem() call there instead. Even_bt_split() uses raw PageAddItem() calls for left page and right pagehigh keys. Clearly the _bt_pgaddtup() shim function wasn't reallyencapsulating anything. _bt_pgaddtup() should now be thought of as a_bt_split() helper function.Note that the assertions from commitd1e241c verify that retail insertsnever insert an item at an internal page's negative infinity offset.This invariant could only ever be violated as a result of a basic logicerror in nbtinsert.c.
1 parentd41202f commit6312c08

File tree

1 file changed

+19
-7
lines changed

1 file changed

+19
-7
lines changed

‎src/backend/access/nbtree/nbtinsert.c‎

Lines changed: 19 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -1249,7 +1249,8 @@ _bt_insertonpg(Relation rel,
12491249
if (postingoff!=0)
12501250
memcpy(oposting,nposting,MAXALIGN(IndexTupleSize(nposting)));
12511251

1252-
if (!_bt_pgaddtup(page,itemsz,itup,newitemoff))
1252+
if (PageAddItem(page, (Item)itup,itemsz,newitemoff, false,
1253+
false)==InvalidOffsetNumber)
12531254
elog(PANIC,"failed to add new item to block %u in index \"%s\"",
12541255
BufferGetBlockNumber(buf),RelationGetRelationName(rel));
12551256

@@ -2528,19 +2529,30 @@ _bt_newroot(Relation rel, Buffer lbuf, Buffer rbuf)
25282529
}
25292530

25302531
/*
2531-
*_bt_pgaddtup() -- add atupleto a particular pagein the index.
2532+
*_bt_pgaddtup() -- add adata itemto a particular pageduring split.
25322533
*
25332534
*This routine adds the tuple to the page as requested. It does
25342535
*not affect pin/lock status, but you'd better have a write lock
25352536
*and pin on the target buffer! Don't forget to write and release
25362537
*the buffer afterwards, either.
25372538
*
2538-
*Themaindifference between this routine and a bare PageAddItem call
2539-
*isthat this code knows that the leftmostindex tuple ona non-leaf
2539+
*The difference between this routine and a bare PageAddItem call is
2540+
*that this code knows that the leftmostdata item onan internal
25402541
*btree page has a key that must be treated as minus infinity.
2541-
*Therefore, it truncates away all attributes. CAUTION: this works
2542-
*ONLY if we insert the tuples in order, so that the given itup_off
2543-
*does represent the final position of the tuple!
2542+
*Therefore, it truncates away all attributes. This extra step is
2543+
*only needed during internal page splits.
2544+
*
2545+
*Truncation of an internal page data item can be thought of as one
2546+
*of the steps used to "move" a boundary separator key during an
2547+
*internal page split. Conceptually, _bt_split() caller splits
2548+
*internal pages "inside" the firstright data item: firstright's
2549+
*separator key is used as the high key for the left page, while its
2550+
*downlink is used within the first data item (also the negative
2551+
*infinity item) for the right page. Each distinct separator key
2552+
*should appear no more than once per level of the tree.
2553+
*
2554+
*CAUTION: this works ONLY if we insert the tuples in order, so that
2555+
*the given itup_off does represent the final position of the tuple!
25442556
*/
25452557
staticbool
25462558
_bt_pgaddtup(Pagepage,

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp