Skip to content

Wöchentlicher PostgreSQL Newsletter - 19. Juli 2009

Der Originalartikel befindet sich unter:

== Wöchentlicher PostgreSQL Newsletter - 19. Juli 2009 ==

Das erste Commitfest für die Entwicklung von 8.5 hat begonnen!

Die Italienische PostgreSQL Usergroup ist dem PostgreSQL
Übersetzungsprojekt beigetreten. Eine Arbeitsgruppe von 9 Personen hat
die Reise begonnen, die Italienische Sektion von
grüner zu gestalten. Mehr Informationen in Italienisch unter:

== PostgreSQL Produkt Neuigkeiten ==

Archiveopteryx 3.1.2, ein PostgreSQL-basierender E-Mail Handler, ist

Benetl 2.9, ein ETL für Dateien, die PostgreSQL nutzen, ist erschienen.

Npgsql2 2.0.6, ein .NET data provider für PostgreSQL, ist erschienen.

PostgreDAC ver. 2.5.2 ist erschienen.

== PostgreSQL Jobs im Juli ==

== PostgreSQL Lokal ==

Das OpenSQL Camp findet am 22. und 23. August in St. Augustin,
Deutschland, in der Nähe von Bonn und Köln, statt. Einsendeschluß für
den CfP ist der 19. Juli 2009. Vorträge einreichen!

Die Deutsche PostgreSQL Usergruppe wird einen Dev-Room auf der
FrOSCon am Sonntag, dem 23. August 2009 haben. Der Call for Papers
ist eröffnet:

Der CfP für PyCon Argentina ist bis zum 29. Juni 2009 23:59 ART
geöffnet. Die Konferenz selbst findet in Buenos Aires am 4. und 5.
September 2009 statt.

Andreas (ads) Scherbaum wird einen Kurs "PostgreSQL im
Unternehmenseinsatz" an der Volkshochschule Magdeburg, Deutschland,
vom 7. bis zum 11. Septemper 2009 geben. Details unter:

Es findet eine Konferenz in Athens, Georgia, USA am 19. September
2009 statt. Der CfP ist eröffnet.

Es findet eine Konferenz in Seattle, Washington, USA am 16.-18.
Oktober 2009 statt. Der CfP ist eröffnet.

PGCon Brasilien wird am 23.-24. Oktober 2009 auf dem Unicamp in
Campinas, Sao Paulo, stattfinden. Der CfP ist eröffnet!

PGDay.EU 2009 wird an der Telecom ParisTech Universität in Paris,
Frankreich, am 6. und 7. Nivember 2009 stattfinden. Der CfP ist
eröffnet. Vorträge einreichen!

Die 10. jährliche JPUG Konferenz hat den Aufruf für Vorträge gestartet.
Die Konferenz findet am 20-21. November 2009 in Tokio, Japan, statt.

== PostgreSQL in den News ==

Planet PostgreSQL:

Dieser wöchentliche PostgreSQL Newsletter wurde erstellt von David
Fetter und Josh Berkus.

Sende Neuigkeiten und Ankündigungen bis Sonntag, 15 Uhr Pazifischer
Zeit. Bitte sende englische Beiträge an, deutsche an, italienische an

== Angewandte Patches ==

Peter Eisentraut committed:

- Update information schema to SQL:2008:  Add yes_or_no domain for
  "boolean" data.  Add new columns for VIEWS view.  Slight section

- Rearrangement of the HTML docs build rules.  Set up proper makefile
  dependencies in the documentation build rules, especially around the
  HTML/index build.  The problem we've had with all previous solutions
  is that we have used the same file name, such as HTML.index or
  bookindex.sgml, to mean different things at different stages of the
  build, and make can't distinguish that.  The solution here is that
  the first jade run produces HTML.index, but does not require
  bookindex.sgml at all, and produces no other html output (the latter
  an idea from Alvaro).  The second jade run includes bookindex.sgml,
  but does not recreate HTML.index.  That way, when you change an sgml
  file, jade is run twice and at the end all dependencies are
  satisfied.  Omitting the html output in the first stage also makes
  the full build a lot faster.  When you run one of the print format
  targets, only the first jade run is run, then the print
  target-specific commands.  If an HTML build has completed
  previously, the first jade run is skipped because the dependencies
  have already been satisfied.  The draft and check targets for quick
  builds and syntax verification are still there.

- Make backend header files C++ safe.  This alters various incidental
  uses of C++ key words to use other similar identifiers, so that a
  C++ compiler won't choke outright.  You still (probably) need extern
  "C" { }; around the inclusion of backend headers.  Based on a patch
  by Kurt Harriman.  Also add a script cpluspluscheck to check for C++
  compatibility in the future.  As of right now, this passes without
  error for me.

Bruce Momjian committed:

- In pgsql/contrib/vacuumlo/vacuumlo.c, vacuumlo query cleanups.  No
  need for VACUUM ANAYZE of newly created/populated temp table, just
  use analyze.  No need to apologize for using subquery in DELETE

Tom Lane committed:

- In pgsql/src/backend/utils/cache/plancache.c, do a conditional
  SPI_push/SPI_pop when replanning a query in RevalidateCachedPlan.
  This is to avoid a "SPI_ERROR_CONNECT" failure when the planner
  calls a SPI-using function and we are already inside one.  The
  alternative fix is to expect callers of RevalidateCachedPlan to do
  this, which seems likely to result in additional hard-to-detect bugs
  of omission.  Per reports from Frank van Vugt and Marek Lewczuk.
  Back-patch to 8.3. It's much harder to trigger the bug in 8.3, due
  to a smaller set of cases in which plans can be invalidated, but it
  could happen.  (I think perhaps only a SI reset event could make 8.3
  fail here, but that's certainly within the realm of possibility.)

- Tweak the core scanner so that it can be used by plpgsql too.
  Changes: Pass in the keyword lookup array instead of having it be
  hardwired.  (This incidentally allows elimination of some duplicate
  coding in ecpg.) Re-order the token declarations in gram.y so that
  non-keyword tokens have numbers that won't change when keywords are
  added or removed.  Add ".." and ":=" to the set of tokens recognized
  by scan.l.  (Since these combinations are nowhere legal in core SQL,
  this does not change anything except the precise wording of the
  error you get when you write this.)

- In pgsql/src/interfaces/ecpg/preproc/ecpg.tokens, remove duplicate
  definition of TYPECAST token.  (Apparently, some but not all
  versions of Bison will warn about this.)

- Add erand48() to the set of functions supported by our src/port/
  library, and extend configure to test for it properly instead of
  hard-wiring an assumption that everybody but Windows has the rand48
  functions.  (We do cheat to the extent of assuming that probing for
  erand48 will do for the entire rand48 family.) erand48() is unused
  as of this commit, but a followon patch will cause GEQO to depend on
  it.  Andres Freund, additional hacking by Tom Lane.

- Make GEQO's planning deterministic by having it start from a
  predictable random number seed each time.  This is how it used to
  work years ago, but we got rid of the seed reset because it was
  resetting the main random() sequence and thus having undesirable
  effects on the rest of the system.  To fix, establish a private
  random number state for each execution of geqo(), and initialize the
  state using the new GUC variable geqo_seed.  People who want to
  experiment with different random searches can do so by changing
  geqo_seed, but you'll always get the same plan for the same value of
  geqo_seed (if holding all other planner inputs constant, of course).
  The new state is kept in PlannerInfo by adding a "void *" field
  reserved for use by join_search hooks.  Most of the rather bulky
  code changes in this commit are just arranging to pass PlannerInfo
  around to all the GEQO functions (many of which formerly didn't
  receive it).  Andres Freund, with some editorialization by Tom Lane.

- Repair bug #4926 "too few pathkeys for mergeclauses".  This example
  shows that the sanity checking I added to create_mergejoin_plan() in
  8.3 was a few bricks shy of a load: the mergeclauses could reference
  pathkeys in a noncanonical order such as x,y,x, not only cases like
  x,x,y which is all that the code had allowed for.  The odd cases
  only turn up when using redundant clauses in an outer join
  condition, which is why no one had noticed before.

- Fix error cleanup failure caused by 8.4 changes in plpgsql to try to
  avoid memory leakage in error recovery.  We were calling
  FreeExprContext, and therefore invoking ExprContextCallback
  callbacks, in both normal and error exits from subtransactions.
  However this isn't very safe, as shown in recent trouble report from
  Frank van Vugt, in which releasing a tupledesc refcount failed.
  It's also unnecessary, since the resources that callbacks might wish
  to release should be cleaned up by other error recovery mechanisms
  (ie the resource owners).  We only really want FreeExprContext to
  release memory attached to the exprcontext in the error-exit case.
  So, add a bool parameter to FreeExprContext to tell it not to call
  the callbacks.  A more general solution would be to pass the
  isCommit bool parameter on to the callbacks, so they could do only
  safe things during error exit.  But that would make the patch
  significantly more invasive and possibly break third-party code that
  registers ExprContextCallback callbacks.  We might want to do that
  later in HEAD, but for now I'll just do what seems reasonable to

- Fix a thinko in join_is_legal: when we decide we can implement a
  semijoin by unique-ifying the RHS and then inner-joining to some
  other relation, that is not grounds for violating the RHS of some
  other outer join.  Noticed while regression-testing new GEQO code,
  which will blindly follow any path that join_is_legal says is legal,
  and then complain later if that leads to a dead end.  I'm not
  certain that this can result in any visible failure in 8.4: the
  mistake may always be masked by the fact that subsequent attempts to
  join the rest of the RHS of the other join will fail.  But I'm not
  certain it can't, either, and it's definitely not operating as
  intended.  So back-patch.  The added regression test depends on the
  new no-failures-allowed logic that I'm about to commit in GEQO, so
  no point back-patching that.

- Rewrite GEQO's gimme_tree function so that it always finds a legal
  join sequence, even when the input "tour" doesn't lead directly to
  such a sequence.  The stack logic that was added in 2004 only
  supported cases where relations that had to be joined to each other
  (due to join order restrictions) were adjacent in the tour.
  However, relying on a random search to figure that out is
  tremendously inefficient in large join problems, and could even fail
  completely (leading to "failed to make a valid plan" errors) if
  random_init_pool ran out of patience.  It seems better to make the
  tour-to-plan transformation a little bit fuzzier so that every tour
  can form a legal plan, even though this means that apparently
  different tours will sometimes yield the same plan.  In the same
  vein, get rid of the logic that knew that tours (a,b,c,d,...) are
  the same as tours (b,a,c,d,...), and therefore insisted the latter
  are invalid.  The chance of generating two tours that differ only in
  this way isn't that high, and throwing out 50% of possible tours to
  avoid such duplication seems more likely to waste valuable genetic-
  refinement generations than to do anything useful.  This leaves us
  with no cases in which geqo_eval will deem a tour invalid, so get
  rid of assorted kluges that tried to deal with such cases, in
  particular the undocumented assumption that DBL_MAX is an impossible
  plan cost.  This is all per testing of Robert Haas'
  lets-remove-the-collapse-limits patch.  That idea has crashed and
  burned, at least for now, but we still got something useful out of
  it.  It's possible we should back-patch this change, since the
  "failed to make a valid plan" error can happen in existing releases;
  but I'd rather not until it has gotten more testing.

- Teach simplify_boolean_equality to simplify the forms foo <> true
  and foo <> false, along with its previous duties of simplifying foo
  = true and foo = false.  (All of these are equivalent to just foo or
  NOT foo as the case may be.)  It's not clear how often this is
  really useful; but it costs almost nothing to do, and it seems some
  people think we should be smart about such cases.  Per recent bug

- In pgsql/src/bin/psql/describe.c, remove unnecessary and
  version-sensitive dependence on the exact set of column names to be
  found in a sequence.  Per gripe from Bruce Momjian.

Heikki Linnakangas committed:

- In pgsql/contrib/pg_standby/Makefile, pg_standby doesn't need to be
  linked with libpq. Fujii Masao.

Andrew Dunstan committed:

- DROP IF EXISTS for columns and constraints. Andres Freund.

== Abgelehnte Patches (bis jetzt) ==

No one was disappointed this week :-)

== Eingesandte Patches ==

Heikki Linnakangas sent in two revisions of a patch which helps set up
index-only scans.

Robert Haas sent in an updated patch for Hot Standby.

Petr (PJMODOS) Jelinek sent in three revisions of a patch to add

ITAGAKI Takahiro sent in two revisions of a patch to add FORCE QUOTE
to COPY...CSV.

KaiGai Kohei sent in three more revisions of his SE-PostgreSQL patch.

Jeff Davis sent in another revision of his patch to make more
generalized constraints.

Andres Freund sent in a patch to refactor *_collapse_limit and

Jaime Casanova sent in a revision of Taro (higepon) Minowa's patch to
allow pg_dump* to dump comments on commposite-type and index columns.

ITAGAKI Takahiro sent in an update for the sampling profiler patch.

Tsutomu Yamada sent in two revisions of a patch to fix the "could not
reattach to shared memory" bug on Windows.

Fujii Masao sent in two more revisions of the Synchronous Replication

Oleg Bartunov sent in a patch which adds prefix support for synonym
dictionaries in full-text search.

Oleg Bartunov sent in a patch to add filtering dictionaries and a new
contrib module, unaccent, which removes accents, using same.

Sergey V. Karpov sent in a patch which extends the functionality of
dict_xsyn extended synonym dictionary.

Fernando Ike de Oliveira sent in two revisions of a patch to add \dL
for PL languages to psql.

Zoltan Boszormenyi sent in another revision of his patch set to
improve ECPG's Informix compatibility layer.

Tom Lane sent in a WIP patch to make PL/pgsql use the core lexer.

Caleb Welton sent in another revision of his patch to add bytea to

Andrew (RhodiumToad) Gierth sent in another revision of the hstore
patch with fixes and new tests for some edge case bugs.

Kevin Grittner sent in a patch to revise parallel pg_restore's
scheduling heuristic.

Steve Prentice sent in another revision of the patch to make PL/pgsql
treat any word following AS as a label.

Petr (PJMODOS) Jelinek sent in two revisions of a patch to implement

Richard Huxton sent in a patch to create a navigation menu for the

Fujii Masao sent in another revision of the patch to multiplex
SIGUSR1, which is infrastructure for the Synchronous Replication

Zoltan Boszormenyi sent in a patch to add struct support for SELECT

Peter Eisentraut sent in a patch to make PL/Python use errcontext()
instead of passing the function names around to use in error messages.

Alan Li sent in a patch to optimize MIN and MAX queries on partitioned

Andreas Wenk sent in a patch to include the \du+ option in \? output
in psql.

Bernd Helmle sent in a revision of Pavel Stehule's patch to allow
mixed and named notation support for PL/pgsql functions.

Dean Rasheed sent in another revision of the deferrable UNIQUE
constraints patch.

Jeremy Kerr sent in another revision of the patch to avoid manual
shift-and-test logic in AllocSetFreeIndex.

Andreas Freund sent in another revision of his patch to allow
machine-readable EXPLAIN output.

Andreas Freund sent in two more revisions of his generic EXPLAIN
options patch.


No Trackbacks


Display comments as Linear | Threaded

No comments

Add Comment

Enclosing asterisks marks text as bold (*word*), underscore are made via _word_.
E-Mail addresses will not be displayed and will only be used for E-Mail notifications.
To leave a comment you must approve it via e-mail, which will be sent to your address after submission.
Form options