8000 docs: PG 12 relnotes, update btree items · postgres/postgres@4bfb79f · GitHub
[go: up one dir, main page]
Skip to content

Commit 4bfb79f

Browse files
bmomjianbmomjian
committed
docs: PG 12 relnotes, update btree items
Reported-by: Peter Geoghegan Discussion: https://postgr.es/m/CAH2-Wzn-aH4ToZEWR05ELSSp7bO_JMn=sMPfUhNruwVmCXKh-w@mail.gmail.com
1 parent 3d99a81 commit 4bfb79f

File tree

1 file changed

+50
-27
lines changed

1 file changed

+50
-27
lines changed

‎doc/src/sgml/release-12.sgml

Lines changed: 50 additions & 27 deletions
Original file line numberDiff line numberDiff line change
@@ -606,22 +606,64 @@ Author: Tom Lane <tgl@sss.pgh.pa.us>
606606

607607
<listitem>
608608
<!--
609-
Author: Alexander Korotkov <akorotkov@postgresql.org>
610-
2018-07-28 [d2086b08b] Reduce path length for locking leaf B-tree pages during
609+
Author: Peter Geoghegan <pg@bowt.ie>
610+
2019-03-20 [dd299df81] Make heap TID a tiebreaker nbtree index column.
611+
Author: Peter Geoghegan <pg@bowt.ie>
612+
2019-03-20 [fab250243] Consider secondary factors during nbtree splits.
611613
Author: Peter Geoghegan <pg@bowt.ie>
612614
2019-03-25 [f21668f32] Add "split after new tuple" nbtree optimization.
613615
-->
614616

615617
<para>
616-
Improve speed of btree index insertions (Peter Geoghegan,
617-
Alexander Korotkov)
618+
Improve performance and space utilization of btree indexes with
619+
many duplicates (Peter Geoghegan, Heikki Linnakangas)
620+
</para>
621+
622+
<para>
623+
Previously, duplicate index entries were stored unordered within
624+
their duplicate groups. This caused overhead during index
625+
inserts, wasted space due to excessive page splits, and reduced
626+
<command>VACUUM</command>'s ability to recycle entire pages.
627+
Duplicate index entries are now sorted in heap-storage order.
628+
</para>
629+
630+
<para>
631+
Indexes <application>pg_upgraded</application> from previous
632+
releases will not have these benefits.
633+
</para>
634+
</listitem>
635+
636+
<listitem>
637+
<!--
638+
see commits above
639+
-->
640+
641+
<para>
642+
Allow multi-column btree indexes to be smaller (Peter Geoghegan,
643+
Heikki Linnakangas)
644+
</para>
645+
646+
<para>
647+
Internal pages and min/max leaf page indicators now only store
648+
index keys until the change key, rather than all indexed keys.
649+
This also improves the locality of index access.
618650
</para>
619651

620652
<para>
621-
The new code improves the space-efficiency of page splits,
622-
reduces locking overhead, and gives better performance for
623-
<command>UPDATE</command>s and <command>DELETE</command>s on
624-
indexes with many duplicates.
653+
Indexes <application>pg_upgraded</application> from previous
654+
releases will not have these benefits.
655+
</para>
656+
</listitem>
657+
658+
<listitem>
659+
<!--
660+
Author: Alexander Korotkov <akorotkov@postgresql.org>
661+
2018-07-28 [d2086b08b] Reduce path length for locking leaf B-tree pages during
662+
-->
663+
664+
<para>
665+
Improve speed of btree index insertions by reducing locking
666+
overhead (Alexander Korotkov)
625667
</para>
626668
</listitem>
627669

@@ -680,25 +722,6 @@ Author: Tom Lane <tgl@sss.pgh.pa.us>
680722

681723
<listitem>
682724
<!--
683-
Author: Peter Geoghegan <pg@bowt.ie>
684-
2019-03-20 [dd299df81] Make heap TID a tiebreaker nbtree index column.
685-
Author: Peter Geoghegan <pg@bowt.ie>
686-
2019-03-20 [fab250243] Consider secondary factors during nbtree splits.
687-
-->
688-
689-
<para>
690-
Have new btree indexes sort duplicate index entries in heap-storage
691-
order (Peter Geoghegan, Heikki Linnakangas)
692-
</para>
693-
694-
<para>
695-
Indexes <application>pg_upgraded</application> from previous
696-
releases will not have this ordering.
697-
</para>
698-
</listitem>
699-
700-
<listitem>
701-
<!--
702725
Author: Heikki Linnakangas <heikki.linnakangas@iki.fi>
703726
2019-03-05 [fe280694d] Scan GiST indexes in physical order during VACUUM.
704727
-->

0 commit comments

Comments
 (0)
0