Skip to content

Wöchentlicher PostgreSQL Newsletter - 24. Oktober 2010

Der Originalartikel befindet sich unter:

== Wöchentlicher PostgreSQL Newsletter - 24. Oktober 2010 ==

Neue Umfrage: was denkst du über Hot Standby / Streaming Replication?

Andreas Scherbaum wurde in einem Podcast über PostgreSQL interviewt:

== PostgreSQL Produkt Neuigkeiten ==

Another PostgreSQL Diff Tool 2.3, ein Werkzeug um Datenbankdiffs
zu erstellen, ist erschienen.

plproxyrc, ein Werkzeug um PL/Proxy zu konfigurieren,
ist erschienen.

partmanteau, ein Partition Management Werkzeug für PostgreSQL,
ist erschienen.

pgpool-II 3.0.1, ein Connection Pooler und mehr, und
pgpoolAdmin 3.0.1, ein GUI Administration Werkzeug dafür,
sind erschienen.

== PostgreSQL Jobs im Oktober ==

== PostgreSQL Lokal ==

Die Deadline für den CfP für die MySQL Conf ist der 25. Oktober 2010.
Maile Selena Deckelmann selenamarie <AT> gmail <DOT> com für Feedback,
Hilfe bei EInreichungen oder um Ideen zu finden.

Frühbucherregistrierung für JDCon West 2010 ist jetzt möglich. Scott
McNealy, früher bei Sun Microsystems, wird die Keynote halten.

Die Registrierung für PGDay.EU 2010, welcher vom 6. bis 8. Dezember in
Stuttgart stattfindet, ist offen.

PgDay.IT 2010 wird am 10. Dezember in Rom stattfinden. Der Call for
Papers ist jetzt offen.

== PostgreSQL in den News ==

Planet PostgreSQL:

Dieser wöchentliche PostgreSQL Newsletter wurde erstellt von David Fetter.

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

== Reviews ==

Jeff Davis reviewed the patch to implement a truly SERIALIAZABLE
isolation level.

ITAGAKI Takahiro reviewed several versions of the patch to add

== Applied Patches ==

Robert Haas pushed:

- Change example pg_hba.conf in docs to match altered
  pg_hba.conf.sample Peter Eisentraut's recent patch to allow host
  names in pg_hba.conf changed the contents of pg_hba.conf.sample
  Fujii Masao.;a=commitdiff;h=9c73e20f38e21f68d92f45238cd6314b241dd96e

- Unbreak comments on composite type attributes.  Report and diagnosis
  by Peter Eisentraut.;a=commitdiff;h=262c1a42dcd36831a964099659026fb9cec0eecc

- Add levenshtein_less_equal, optimized version for small distances.
  Alexander Korotkov, heavily revised by me.;a=commitdiff;h=604ab08145d0dde91279c430e08d570717c12b30

- Add some caveats to the contrib/isn docs.;a=commitdiff;h=4b6623a7e8f0f0116ff68ebfd4c98830fe59419d

- Updates to contrib/isn ISBN tables.  Jan Otto, reviewed by Peter

- Correct a mistake in levenshtein_less_equal() multibyte character
  handling.  Spotted by Alexander Korotkov.  Along the way, remove a
  misleading comment line.;a=commitdiff;h=dd1587089c3dadd4bc1c85e8e75edf3a24686934

Bruce Momjian pushed:

- Document the tablespace directory "should" be empty, rather than
  "must" be empty.  Because of binary migration usage, it might not be

- In pg_upgrade, rename macro EXEC_EXT to SHELL_EXT for clarity.
  Backpatch to 9.0.X.;a=commitdiff;h=9df96f9bb1cb8bc0350f02bbd0a113480e84224a

- Add mention of using tools/fsync to test fsync methods. Restructure
  recent wal_sync_method doc paragraph to be clearer.;a=commitdiff;h=f75d6a1b19ec334a9b1455c4b8030a09d582581d

- Add removal of PG_VERSION to optional old cluster deletion script.
  Backpatch to 9.0.X.;a=commitdiff;h=51eeccd6cc2e0fcb95112e0e4a7a20d0f95c181a

- In pg_upgrade, rename SHELL_EXT to SCRIPT_EXT, for clarity.;a=commitdiff;h=722d5beeb266ae83f548fc3953df700a71f30134

- Restructure the pg_upgrade code to use several global structures
  rather than packing everything into 'ctx' and passing that to every

- Pgindent run on pg_upgrade source after restructuring.;a=commitdiff;h=07456b45e6359e8a15807e3f71c0f2e6a76fa4e9

- In pg_upgrade, use cached copy of directory listing, rather than
  calling scandir() with a pattern for every table.  Optimization
  after report of pg_upgrade slowness with 150k tables.;a=commitdiff;h=ad06db2610f479d559ee4a3b9e3241d40be07440

- Cache most recent relfilenode lookups, for speed, after report of
  pg_upgrade slowness for 150k tables.;a=commitdiff;h=333b4a45f1d984bd7f32aecb1ba551feed088ec4

- Fix for new pg_upgrade cache code.;a=commitdiff;h=1c72614be710127b198411f49301a56453e3abc4

- Further speed up pg_upgrade lookups.;a=commitdiff;h=3325c9bddb332487914c5c96d838286350044814

- Rename pg_upgrade 'log' to 'log_opts', to avoid platform naming

- Clean up pg_upgrade cache lookup code;  remove useless NULL pointer

- Mention limited usefulness of .pgpass database field.;a=commitdiff;h=71be8db5dfada851c45d6f4d1d7364456fd40be5

Peter Eisentraut pushed:

- Support key word 'all' in host column of pg_hba.conf;a=commitdiff;h=bc8624b15d8055cdead310352e5943da18500d16

- Support suffix matching of host names in pg_hba.conf.  A name
  starting with a dot can be used to match a suffix of the actual host
  name (e.g., matches;a=commitdiff;h=24b29ca8f9dc4a5e5f873f0fcb56438c526700f6

Tom Lane pushed:

- Fix incorrect generation of whole-row variables in planner.  A
  couple of places in the planner need to generate whole-row Vars, and
  were cutting corners by setting vartype = RECORDOID in the Vars,
  even in cases where there's an identifiable named composite type for
  the RTE being referenced.  While we mostly got away with this, it
  failed when there was also a parser-generated whole-row reference to
  the same RTE, because the two Vars weren't equal() due to the
  difference in vartype.  Fix by providing a subroutine the planner
  can call to generate whole-row Vars the same way the parser does.
  Per bug #5716 from Andrew Tipton.  Back-patch to 9.0 where one of
  the bogus calls was introduced (the other one is new in HEAD).;a=commitdiff;h=6e74a91b2bf0e0032ccd60dd99d6cf47c190c428

- Update storage.sgml to describe the 9.0 tablespace directory layout.;a=commitdiff;h=c33bfb8b9b0d955b813d99e8f9f5c62c31f7387c

- Fix ecpg test building process to not generate *.dSYM junk on Macs.
  The trick is to not try to build executables directly from .c files,
  but to always build the intermediate .o files.  For obscure reasons,
  Darwin's version of gcc will leave debug cruft behind in the first
  case but not the second.  Per complaint from Robert Haas.;a=commitdiff;h=c75e1436467f32a06b5ab9d594d2a390e7f4594d

- Don't try to fetch database name when SetTransactionIdLimit() is
  executed outside a transaction.  This repairs brain fade in my patch
  of 2009-08-30: the reason we had been storing oldest-database name,
  not OID, in ShmemVariableCache was of course to avoid having to do a
  catalog lookup at times when it might be unsafe.  This error
  explains why Aleksandr Dushein is having trouble getting out of an
  XID wraparound state in bug #5718, though not how he got into that
  state in the first place.  I suspect pg_upgrade is at fault there.;a=commitdiff;h=def30e84c41389225ee9e56cb7c722980bab9746

- Remove obsolete comment, per Josh Kupershmidt.;a=commitdiff;h=572ab1a542c170ddd2e4c30ef472e13f531b64a4

- Improve handling of domains over arrays.  This patch eliminates
  various bizarre behaviors caused by sloppy thinking about the
  difference between a domain type and its underlying array type.  In
  particular, the operation of updating one element of such an array
  has to be considered as yielding a value of the underlying array
  type, not a value of the domain, because there's no assurance that
  the domain's CHECK constraints are still satisfied.  If we're
  intending to store the result back into a domain column, we have to
  re-cast to the domain type so that constraints are re-checked.  For
  similar reasons, such a domain can't be blindly matched to an
  ANYARRAY polymorphic parameter, because the polymorphic function is
  likely to apply array-ish operations that could invalidate the
  domain constraints.  For the moment, we just forbid such matching.
  We might later wish to insert an automatic downcast to the
  underlying array type, but such a change should also change matching
  of domains to ANYELEMENT for consistency.  To ensure that all such
  logic is rechecked, this patch removes the original hack of setting
  a domain's pg_type.typelem field to match its base type; the typelem
  will always be zero instead.  In those places where it's really okay
  to look through the domain type with no other logic changes, use the
  newly added get_base_element_type function in place of
  get_element_type.  catversion bumped due to change in pg_type
  contents.  Per bug #5717 from Richard Huxton and subsequent

- Allow new values to be added to an existing enum type.  After much
  expenditure of effort, we've got this to the point where the
  performance penalty is pretty minimal in typical cases.  Andrew
  Dunstan, reviewed by Brendan Jurd, Dean Rasheed, and Tom Lane.;a=commitdiff;h=84c123be1de8a9955741e20c9f945571e40c545e

Alvaro Herrera pushed:

- Remove AtStart_Cache() call in CommandCounterIncrement().  This call
  was present in the aboriginal code from Berkeley, and has never been
  touched; it may very well be that it was there to mask effects of
  bugs in other places and it may no longer be necessary.  The removal
  has been foreseen in a code comment since 2007; this seems to be a
  good time to test this hypothesis.;a=commitdiff;h=17a16663d0c1c70c5c70ba97de6fe2c9c6a3bc07

Heikki Linnakangas pushed:

- If pk is NULL, the backend would segfault when accessing ->algo and
  the following NULL check was never reached.  This problem was found
  by Coccinelle (null_ref.cocci from coccicheck).  Marti Raudsepp.;a=commitdiff;h=18e752f226300292c6182b68b2568dc2d66db8bd

- Make OFF keyword unreserved.  It's not hard to imagine wanting to
  use 'off' as a variable or column name, and it's not reserved in
  recent versions of the SQL spec either.  This became particularly
  annoying in 9.0, before that PL/pgSQL replaced variable names in
  queries with parameter markers, so it was possible to use OFF and
  many other backend parser keywords as variable names.  Because of
  that, backpatch to 9.0.;a=commitdiff;h=5c84fe4607907b22e45433195fec800d23ac1f49

- Add semicolon, missed in previous patch.  And update the keyword
  list in the docs to reflect that OFF is now unreserved.  Spotted by
  Tom Lane.;a=commitdiff;h=57b80b4c466f224b6623e38081cea6e6d2fe0744

Andrew Dunstan pushed:

- Remove unnecessary use of trigger flag to hash plperl functions.;a=commitdiff;h=2d01ec0708d571eef926f3f5795aa73759df5d9a

== Abgelehnte Patches (bis jetzt) ==

No one was disappointed this week :-)

== Eingesandte Patches ==

Dimitri Fontaine sent in eleven more revisions of the patch to add

Zoltan Boszormenyi sent in a patch to make the process of planning
execution of queries over large numbers of faster.

Marko (johto) Tiikkaja and Alvaro Herrera traded patches to fix an
issue with taking snapshots.

Josh Kupershmidt sent in a patch to add tab completion for \z and \dg
to psql.

Marti Raudsepp sent in a patch to keep psql's ON_ERROR_STOP from
leaking file descriptors based on a coccicheck run.

Alvaro Herrera sent in a patch to fix an issue with the interaction of
a crashing plancache with subtransactions.

Robert Haas sent in another revision of the patch to make truly

Pavel Stehule sent in another revision of the patch to add a format()

Jesper Krogh sent in a patch to clarify the docs on the behavior of
windowing functions with LIMIT.

Robert Haas sent in another WIP patch for K-nearest-neighbor GiST.

Peter Eisentraut sent in a WIP patch to refactor the type and typmod
lookups.  This will support, among other things, collations.


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