Postgres Code Sprint - Recently Updated Pages

testing!

Anonymous — Mon, 27 Apr 2009 11:11:35 CDT

[Insert Name] Family Holiday Letter

n. More tips and image elements here.

[Insert Fun Intro]

Add a bunch of text. You could describe what the kids are up to, vacations you took, set backs, promotions, reunion info, fun family trivia, links to other sites or pages to check out.

[Insert Fun Headline]

The best holiday letters tend to be short and to the point summaries of the major things that happened during the year. In the online format, you can add links to pictures or pages of the people or subjects you are referring to, which is a fun way to dive deeper into the letter for those that are interested..

Be sure to replace the text here or you will end up with a really boring holiday letter. You could describe what the kids are up to, vacations you took, set backs, promotions, reunion info, fun family trivia, links to other sites or pages to check out.

Don't be afraid to use images. They really liven up your letter. You can even embed slideshows (first build a gallery in the photo section) or videos directly into your newsletter as well.

If you left or right align the image, the text will wrap around the image automatically.

[Insert Another Fun Headline]

Another great thing about using the web is that you can get comments on your letter if you like. You'll notice below that there are threads (like comments) that people can start, or you could pre-post a thread in your discussion forum and ask for people to comment via a link you add to the thread.

[Title for Picture Here]

Highlights of the year:

	Jimmy turns 2
	No more election
	Highlight 3
	Highlight 4
	Highlight 5

Don't forget to "sign" your names at the bottom of the letter.

Unsigned Integer Types

Anonymous — Thu, 28 Feb 2008 14:43:25 CST

This idea is to create unsigned versions of int2, int4 and int8. There have been numerous requests for this over time. Need to fully define the casting rules, and arithmetic rules for mixed math.

Interested Hackers:

Larry Rosenman

INSERT RETURNING

Anonymous — Thu, 13 Sep 2007 23:57:37 CDT

Several people have asked for the ability to return values from data-modifying commands (INSERT, UPDATE, DELETE) -- for example, getting the newly-generated value of a sequence in an INSERT command. Some variant of this functionality is supported by some other databases.

Omar Kilani has implemented a preliminary version of this functionality, but some additional work is needed to get the patch into shape for inclusion in the tree.

Ideas to consider:

How does this relate to JDBC requirements for getGeneratedKeys()?

A reasonable answer might just be that getGeneratedKeys() is braindamaged and not worth supporting.

What outstanding issues are there with the patch?

Related links:

Interested hackers: Gavin Sherry

http://www.gameroom.com/tok/index.htm

Anonymous — Tue, 16 Jan 2007 02:51:30 CST

Come Join Me for an Event

Celebrating Something Big

Who: You and a guest. Children are welcome!

When: July 4, 2006

Time: 7 p.m.

Where: Robinswood Park, 123 Main Street, Goodtown, WA

What: Light appetizers and drinks, music, and good friends.

Cost: Free!

Please RSVP!	Driving Directions
(Create subpages for your RSVPs) Wouldn't miss it! Sorry, can't make it Not sure	From Hwy 101: Take E Sunset Wy (exit 18) – go 3 mi. Bear left onto Highlands Dr – go 1.5 mi. Turn right at NE Federal Dr – go 2 mi. Turn right at NE Park Dr – go 1 mi.

[insert a map like this one!]

postgres code sprint

Anonymous — Sun, 10 Dec 2006 15:01:06 CST

There is no abstract available for this page revision.

SubDark

Anonymous — Fri, 18 Aug 2006 01:00:13 CDT

Come Join Me for an Event

Celebrating Something Big

Please RSVP!	Driving Directions
(Create subpages for your RSVPs) Wouldn't miss it! Sorry, can't make it Not sure	From Hwy 101: Take E Sunset Wy (exit 18) – go 3 mi. Bear left onto Highlands Dr – go 1.5 mi. Turn right at NE Federal Dr – go 2 mi. Turn right at NE Park Dr – go 1 mi.

[insert a map like this one!]

Project Ideas

Anonymous — Mon, 10 Jul 2006 16:15:00 CDT

This page is a free-form list of various ideas for projects to work on at the Postgres Code Sprint. If you have an idea for a project, feel free to add it to the list below, and create a sub-page for the project if needed. If you'd like to comment on a project, go right ahead; if think you might be interested in working on the project at the sprint, add your name to the project's description.

Per Tom, a good rule of thumb might be to limit code sprint projects to ideas where community input isn't warranted (because the feature is straightforward), or ideas for which a proposal has already been sent to the community, and a consensus has been reached about how the feature ought to behave.

Another source of good project ideas is the PostgreSQL TODO list, linked below.

Neilc is interested in working on:

Make the int8 type pass-by-value on 64-bit platforms

Implement CORRESPONDING BY, per SQL:2003
Consider using profile-guided optimization with recent versions of GCC
Generalize the planner knobs
Inline set-returning SQL functions
Fix ordering of function table entries in docs

Neilc's project ideas for anyone who's interested:

Implement UNIQUE predicate, per SQL:2003
Implement CREATE TABLE LIKE composite_type
Improve schema browsing in psql
Add a psql option to have it submit SELECT queries via a cursor, to avoid excessive memory consumption when manipulating large result sets (<- I'll look a bit into this (too shy to touch the backend :) - if you're doing this too, let me know, Chris Mair)
Finish INSERT ... RETURNING patch
Implement INSERT with multiple VALUES, per SQL:2003
Implement UNIQUE for hash indexes
Finish patch for column-level triggers
Enumerating each of the outstanding patches on Bruce's list, and applying or rejecting each of them

Gavin's suggestions:

Improving contended LWLocks
Preparing for 8.2
Make use of posix_fadvise for heap/index relations

Larry Rosenman's suggestion:

create unsigned integer type(s) (int2, int4 & int8).

Greg's suggestions and notes:

Will probably work on: column-level triggers, CREATE OR REPLACE enhancements
May work on: INSERT with multiple VALUES
Finalize my pg_dump patch
postgresql.conf analyzer
Asynchronous DBD::Pg
Improved logging of prepared statements
Overhaul of logging to allow fine granularity per database/user/subsection/etc.
More tid operators

alvherre's suggestions:

Will probably work on: finishing the relminxid patch
May work on: autovacuum maintenance-window feature
May work on: COPY (query) TO feature
Having BEGIN not start a transaction right away, but rather wait until the serial snapshot is set
Deferred UNIQUE constraints
Enhance pg_dump to support dumping multiple tables

Devrim will work on:

Enable multiple versions of PostgreSQL RPMs to be installed to the same server. Currently we do not allow that.

Hannu Krosing :

WAL progress reporting for up-to ms log-shipping
pg_dump instrumentation for structured splitting of dump file
add ctid to non-unique indexes when ordering + implement non-locking CREATE INDEX and REINDEX
discuss "ARCHIVE TABLES" + index-only queries.
haggle about getting CONCURRENT VACUUM into backend
add throttling knobs to more "maintenance" commands in addition to vacuum. Depending on context, "maintenance" commands can be DELETE, CREATE INDEX, UPDATE, COPY, INSERT ... SELECT FROM ..., CREATE TABLE ... AS SELECT ...,
tunability of new tuple placement (both insert and update)
clustering by set of bitmap indexes
TOAST tuples should be accessed by CTID
CICS

Relevant links:

PostgreSQL TODO List

Home

Anonymous — Mon, 10 Jul 2006 05:42:12 CDT

This wiki is devoted to planning for the PostgreSQL Anniversary Summit Code Sprint, scheduled for July 10th and 11th in Toronto, Canada. The day will start off at 9am at the Novotel and finish at 6pm.

There's not a whole lot of advance planning we can do, with one exception: I'd like everyone who's going to be attending the sprint to think about what projects they'd like to work on. I think the sprint will be a great opportunity to finish off outstanding patches or work together on ideas, but I think we'll all be more productive if we've done some background research and thinking before we start hacking. I've put together a list of potential project ideas -- if you're going to attend the sprint, feel free to consider those projects and add projects of your own.

The code sprint will be held at the Novotel Hotel in downtown Toronto.

You can find a google map for this here.

You can either meet Gavin Sherry at the Novotel at 9am or else meet at the entry to the Rogers building at Ryerson University at 8:30am -- where the PostgreSQL Summit was held. There will be people there to take you to the Novotel.

In case you get lost, the Novotel is 45 The Esplanade. It is near Fionn MacCools -- the location of the summit dinner. The Novotel web site can be found here.

If you have problems on the 10th or 11th -- such as locating the sprint room -- you can call Neil Conway on 6472342346 or Gavin Sherry on +61418487420.

Column Level Triggers

Anonymous — Mon, 26 Jun 2006 06:23:36 CDT

Support for column-level triggers, per the SQL spec. Greg Sabino Mullane has posted a patch that implements a preliminary version of this functionality, but more work is needed before it is ready for inclusion in the tree.

Related links:

Greg's preliminary patch

Interested hackers:

Greg. Note: the patch is quite old and incomplete. :)

posix_fadvise

neilc — Fri, 23 Jun 2006 01:32:14 CDT

POSIX allows us to pass hints to the kernel about proceeding activity on a file. Specifically, it allows us to recommend aggressive readahead or none at all. The Linux kernel developers (notably Andrew Morton) have suggested that Postgres ought to be using posix_fadvise if it expects to get good readahead behavior from the kernel, so it seems at least worth considering (and benchmarking).

Here's a brief patch which adds these hints to the system. It probably needs to be brought up to HEAD.

http://www.alcove.com.au/~swm/fadvise.diff

Preparing for 8.2

swm — Fri, 23 Jun 2006 01:28:31 CDT

We're coming up to 8.2. That means lots of things need to be taken care of: patches reviewed, testing, more docs and looking for performance regressions.

This might be a good opportunity to get through some of these significant tasks. Hopefully, it will reduce the beta period as well.

Contented locks

Anonymous — Fri, 23 Jun 2006 00:11:10 CDT

Recent instrumentation of the LW locking routines and benchmarking work being done by companies has revealed a few intersting contended locks -- particularly ProcArrayLock.

Interested hackers: Gavin Sherry

Int8 on 64-bit

neilc — Thu, 22 Jun 2006 23:38:13 CDT

Int8 is currently a pass-by-ref type: that is, a Datum containing an int8 value is really a pointer to a region of memory that holds the actual integer value. When sizeof(Datum) == 4, this is our only option: an int8 is too big to directly represented in a Datum.

But on 64-bit platforms, we can just store the int8 value in the Datum itself. This would be a significant performance win in some circumstances. The trickiness would be to avoid breaking existing code that uses int8 so that client code won't be aware of the exact representation of the int8 type.

A similar improvement could probably be made to float4 and float8 (we could even make float4 unconditionally pass-by-value, AFAICS).

Interested hackers:

Neil Conway

Unique Hash Indexes

neilc — Thu, 22 Jun 2006 18:53:06 CDT

This might be a fairly interesting project for someone who would like to learn about how access methods in PostgreSQL work. One downside is that the project wouldn't address some of the more fundamental problems with the current hash index implementation, so in a sense it is rearranging the deckchairs on the Titanic -- but it would be a fun project to hack on.

Basically, just add UNIQUE support for hash indexes. That just means walking through the chain of pages in a given hash bucket when doing an insertion, to verify that none of the existing members of the bucket are equal to the newly-inserted item. One consideration is how this would interact with the existing hash index locking protocol.

Neil C. actually implemented this functionality aeons ago, but never got the patch into shape for application to the tree. You could use that patch (linked below), or start from scratch.

Related links:

Version of the patch by Neil C.

Interested hackers:

Create Table Like Composite Type

neilc — Thu, 22 Jun 2006 18:51:50 CDT

We currently allow CREATE TABLE ... LIKE some_table; it would be useful to be able to specify a composite type in addition to just a table.

Interested hackers:

Profile Guided Optimization

neilc — Thu, 22 Jun 2006 18:16:38 CDT

Recent versions of GCC support profile-guided optimization: the idea is you compile the program with "profiling" hooks inserted by the compiler, and then run the program on a reasonable data set. This produces profiling data about the program's performance (e.g. how often a given branch is taken), which can be used by the compiler to generate better code when the program is recompiled.

The idea would be to have a configure flag that would enable support for PGO, if supported by the compiler. To use PGO, the user would perhaps build a special Makefile target, which would:

Compile/recompile the backend with profiling hooks
Run the regression tests
Recompile the backend to use the profiling data produced by #2

Neil C. implemented a primitive version of this idea earlier, but ran into some problems with the implementation of PGO in GCC 3.4. Presumably these have been fixed in GCC 4.1.

Actually getting this into the tree would depend in demonstrating that there is a real performance win to be gained from using PGO: getting benchmark data on PGO with typical PG workloads would be a useful first step.

Issues to consider:

Are the regression tests a representative set of input for PGO? If not, does it matter, and/or would another test workload be better? (e.g. pgbench)
What is the performance improvement to be gained from using PGO?
Should we support PGO only with GCC, or with other compilers as well? (Intel CC, MSVC++, etc.?)

Related links:

Primitive attempt at using PGO by Neil C
GCC docs:

Optimization Options: -fprofile-arcs, -fprofile-use, and related options

PGO with MSVC++

Interested hackers:

Neil Conway

UNIQUE predicate

neilc — Thu, 22 Jun 2006 18:01:38 CDT

This is a relatively small feature: SQL:2003 defines a UNIQUE predicate that takes a subquery and returns true if the subquery's result set is duplicate-free. The actual usefulness of this is debatable -- at the very least it is defined by the standard...

An implementation would want to make use of the existing machinery for performing grouping / duplicate elimination, so that this predicate can be evaluated efficiently for subqueries with arbitrarily large result sets.

Generalized Planner Knobs

neilc — Thu, 22 Jun 2006 17:28:42 CDT

The idea is to replace the existing boolean variables for controlling the operations considered by the planner ("enable_seqscan", etc.) with real-valued variables that just multiply the operation's cost by the value of the configuration parameter. This would allow a DBA to selectively favour or penalize the choice of a given operation by the planner.

Interested hackers:

Neil Conway

Related links:

Proposal on -hackers

CORRESPONDING BY

neilc — Thu, 22 Jun 2006 16:03:10 CDT

The CORRESPONDING BY clause is defined in section 7.13 of SQL:2003 (page 351 in my copy). It seems to be just syntax sugar for UNION.

Interested hackers:

Neil Conway

Postgres Code Sprint - Recently Updated Pages

testing!

n. More tips and image elements here.

Unsigned Integer Types

INSERT RETURNING

http://www.gameroom.com/tok/index.htm

Come Join Me for an Event

Celebrating Something Big

Please RSVP!

Driving Directions

postgres code sprint

SubDark

Come Join Me for an Event

Celebrating Something Big

Please RSVP!

Driving Directions

Project Ideas

Home

Column Level Triggers

posix_fadvise

Preparing for 8.2

Contented locks

Int8 on 64-bit

Unique Hash Indexes

Create Table Like Composite Type

Profile Guided Optimization

UNIQUE predicate

Generalized Planner Knobs

CORRESPONDING BY