1 <!-- doc/src/sgml/release-9.3.sgml -->
2 <!-- See header comment in release.sgml about typical markup -->
4 <sect1 id="release-9-3">
5 <title>Release 9.3</title>
8 <title>Release Date</title>
9 <simpara>2013-XX-XX, CURRENT AS OF 2013-04-20</simpara>
13 <title>Overview</title>
16 ADD HERE Major enhancements include:
19 <!-- This list duplicates items below, but without authors or details-->
22 The above items are explained in more detail in the sections below.
29 <title>Migration to Version 9.3</title>
32 A dump/restore using <link
33 linkend="APP-PG-DUMPALL"><application>pg_dumpall</></link>, or use
34 of <link linkend="pgupgrade"><application>pg_upgrade</></link>, is
35 required for those wishing to migrate data from any previous release.
39 Version 9.3 contains a number of changes that may affect compatibility
40 with previous releases. Observe the following incompatibilities:
44 <title>ADD HERE</title>
51 <title>Changes</title>
54 Below you will find a detailed account of the changes between
55 <productname>PostgreSQL</productname> 9.3 and the previous major
63 <title>Locking</title>
69 Prevent non-key-field row updates from locking foreign key rows
70 (Álvaro Herrera, Noah Misch, Andres Freund, Alexander
71 Shulgin, Marti Raudsepp)
75 This improves concurrency and reduces the probability of deadlocks.
76 UPDATEs on non-key columns use the new SELECT FOR NO KEY UPDATE lock
77 type, and foreign key checks use the new SELECT FOR KEY SHARE lock mode.
83 Add cache of local locks (Jeff Janes)
87 This speeds lock release at statement completion in transactions
88 that hold many locks; it is particularly useful for pg_dump.
97 <title>Indexes</title>
103 Add SP-GiST support for range data types (Alexander Korotkov)
109 Allow unlogged GiST indexes (Jeevan Chalke)
115 Improve concurrency of hash indexes (Robert Haas)
124 <title>Optimizer</title>
130 Collect and use histograms for range types (Alexander Korotkov)
136 Reduce optimizer overhead by discarding plans with unneeded cheaper
137 startup costs (Tom Lane)
143 Improve optimizer cost estimation for index access (Tom Lane)
152 <title>General Performance</title>
158 Add COPY FREEZE option to avoid the overhead of later marking
159 tuples as committed (Simon Riggs, Jeff Davis)
165 Improve performance of NUMERIC calculations (Kyotaro Horiguchi)
171 Improve grouping of sessions waiting for commit_delay (Peter
176 This improves the usefulness and behavior of commit_delay.
177 BACKWARD COMPATIBILITY
183 Allow in-memory sorts to use their full memory allocation (Jeff Janes)
187 Users who have set work_mem based on the previous behavior should
188 revisit that setting.
194 Improve performance for transactions creating, rebuilding, or
195 dropping many relations (Jeff Janes, Tomas Vondra)
201 Improve performance of the CREATE TABLE ... ON COMMIT DELETE ROWS
202 clause by only issuing delete if the temporary table was accessed
209 Have vacuum recheck visibility after it has removed expired tuples
214 This increases the chance of a page being marked as all-visible.
220 Split pgstat file in per-database and global files (Tomas Vondra)
224 This reduces the statistics management read and write overhead.
233 <title>Monitoring</title>
239 Add optional ability to checksum data pages and report corruption
240 (Simon Riggs, Jeff Davis, Greg Smith)
244 The checksum option can be set during initdb.
250 Allow pg_terminate_backend() to terminate other backends with
251 the same role (Dan Farina)
255 Previously, only superusers could terminate other sessions.
261 Allow the statistics collector to operate properly in cases where
262 the system clock goes backwards (Tom Lane)
266 Previously statistics collection would stop until the time again
267 reached the previously-stored latest time.
276 <title>Authentication</title>
282 Improve LDAP error reporting and documentation (Peter Eisentraut)
288 Add support for LDAP authentication to be specified in URL format
295 Change the ssl_ciphers parameter to start with DEFAULT, rather
296 than ALL, then remove insecure ciphers (Magnus Hagander)
300 It is assumed DEFAULT is more appropriate cipher set.
306 Parse/load pg_ident.conf once, not during each connection (Amit
311 This is similar to how pg_hba.conf is processed.
320 <title>Server Settings</title>
326 Dramatically reduce System V shared memory usage (Robert Haas)
330 Instead, on Unix-like systems, mmap() is used for shared memory.
331 This eliminates the requirement of increasing System V shared
332 memory kernel parameters.
338 Allow the postmaster to listen on multiple Unix-domain sockets
343 This renames configuration parameter unix_socket_directory to
344 unix_socket_directories, which accepts a list of directories.
350 Add configuration variable lock_timeout to limit lock wait duration
351 (Zoltán Böszörményi)
357 Allow a directory of configuration files to be processed (Magnus
358 Hagander, Greg Smith, Selena Deckelmann)
362 The directory is specified as include_dir in server configuration file.
368 Increase the maximum initdb-configured value for shared_buffers
369 to 128MB (Robert Haas)
373 This is the maximum value initdb attempts to set in
374 postgresql.conf; the previous value was 32MB.
380 Require superuser privileges to set commit_delay because it
381 can now potentially delay other sessions (Simon Riggs)
382 BACKWARD COMPATIBILITY
389 Remove the external PID file on postmaster exit (Peter Eisentraut)
395 Have "session id" in log_line_prefix (%c) always output four hex
396 digits after the period (Bruce Momjian) COMPATIBILITY
407 <title>Replication and Recovery</title>
413 Allow a streaming replication standbys to follow a timeline switch (Heikki Linnakangas)
417 This allows streaming standbys to feed from newly-promoted slaves.
418 Previously slaves required access to a WAL archive directory to
425 Add SQL functions pg_backup_in_progress() and pg_backup_start_time() (Gilles Darold)
429 These functions report the status of base backups.
435 Improve performance of streaming log shipping with
436 synchronous_commit disabled (Andres Freund)
442 Allow much faster promotion of a streaming standby to primary (Simon
443 Riggs, Kyotaro Horiguchi)
449 Add the last checkpoint's redo location to pg_controldata's output
454 This information is useful for determining the WAL files needed for
461 Allow tooling like pg_receivexlog to run on computers with different
462 architectures (Heikki Linnakangas)
468 Have pg_basebackup --write-recovery-conf output a minimal
469 recovery.conf (Zoltán Böszörményi, Magnus Hagander)
473 This simplifies setting up a standby server.
479 Allow pg_receivexlog and pg_basebackup -X to handle streaming
480 timeline switches (Heikki Linnakangas)
486 Add wal_receiver_timeout parameter to control the WAL receiver
487 timeout (Amit Kapila)
491 This allows more rapid detection of connection failure. No longer set
492 wal_receiver_status_interval?
498 Rename replication_timeout to wal_sender_timeout (Amit Kapila)
502 BACKWARD COMPATIBILITY
503 This setting controls the WAL sender timeout.
510 <title>Write-Ahead Log (WAL)</title>
516 Store WAL in a continuous stream, rather than skipping the last
517 16MB segment every 4GB (Heikki Linnakangas) BACKWARD COMPATIBLE BREAK
523 Restructure WAL files to better handle timeline changes during
524 recovery (Heikki Linnakangas)
530 Restructure WAL files to use a more compact storage format (Heikki Linnakangas)
541 <title>Queries</title>
547 Implement SQL-standard LATERAL for FROM-clause subqueries and
548 function calls (Tom Lane)
552 This is useful in allowing FROM subqueries and functions to reference
553 output from tables in the FROM clause. The LATERAL keyword is optional
560 Add support for piping COPY and psql \copy to/from an external program (Etsuro
567 Improve query string error location reporting (Tom Lane)
571 Only CREATE SCHEMA ... IF NOT EXISTS? If so, remove it.
580 <title>Object Manipulation</title>
586 Add support to event triggers (Dimitri Fontaine, Robert Haas,
587 Álvaro Herrera)
591 This allows C functions to be called when specific commands are run.
597 Allow a multi-row VALUES clause in rules to reference OLD/NEW (Tom Lane)
603 Add CREATE SCHEMA ... IF NOT EXIST clause (Fabrízio de
610 Throw an error if expiring tuple is again updated or deleted (Kevin Grittner)
611 DETAILS? BACKWARD INCOMPATIBILITY
617 Have REASSIGN OWNED also change ownership of shared objects (Álvaro
625 <title><link linkend="SQL-CREATETABLE"><command>CREATE TABLE</></link></title>
631 No longer output messages about implicit index and sequence creation (Robert Haas)
635 These messages now appear with DEBUG1-level output.
641 Allow CREATE TABLE IF NOT EXIST to succeed for a non-existent schema
646 Previously, it threw an error if the schema did not exist.
655 <title>Constraints</title>
661 Change ON UPDATE SET NULL/SET DEFAULT foreign key actions to affect
662 all referenced columns, not just those referenced in the UPDATE
667 Previously only columns referenced in the UPDATE were set to null or DEFAULT.
668 BACKWARD COMPATIBILITY CHANGE
674 Provide clients with constraint violation details as separate fields
679 This allows clients to retrieve table, column, data type, or constraint
680 name error details. Previously such information had to be extracted from
681 error strings. Client language support is required to access these
688 Internally store default foreign key matches (non-FULL, non-PARTIAL) as "simple" (Tom Lane)
692 These were previously stored as "<unspecified>".
693 This changes the value stored in system column
694 pg_constraint.confmatchtype. BACKWARD COMPATIBILITY CHANGE
703 <title><command>ALTER</></title>
709 Add the ALTER TYPE ... IF NOT EXISTS clause when adding enumerated type labels (Andrew
714 This is useful for conditional label creation in transaction blocks.
720 Add ALTER ROLE ALL SET to add settings to all users (Peter Eisentraut)
724 This allows settings to apply to all users in all databases. ALTER
725 DATABASE SET already allowed addition of settings for all users in a
726 single database. postgresql.conf has a similar effect.
732 Add support for ALTER RULE ... RENAME (Ali Dar)
741 <title><link linkend="rules-views"><command>VIEWs</></link></title>
747 Add a materialized view relations (Kevin Grittner)
751 Unlike ordinary views, where the base tables are read on every access,
752 materialized views create physical tables at creation or refresh time.
753 Access to the materialized view reads from these materialized physical
754 tables. There is no facility for incrementally refreshing materialized
755 views or auto-accessing them via base table access.
761 Make simple views auto-updatable (Dean Rasheed)
765 INSTEAD rules are still available for complex views.
771 Allow views and rules to handle cases where referenced tables are
772 renamed, or columns are renamed, added, or dropped (Tom Lane)
773 HOW DOES IT HANDLE A DROPPED REFERENCED COLUMN?
779 Add CREATE RECURSIVE VIEW syntax (Peter Eisentraut)
783 Internally this is translated into CREATE VIEW ... WITH RECURSIVE.
789 When converting a table to a view, remove its system columns (Tom Lane)
801 <title>Data Types</title>
807 Increase the maximum length of large objects from 2GB to 4TB (Nozomi
812 This change includes new libpq and server-side 64-bit-enabled functions.
818 Allow text timezone designations, e.g. "America/Chicago" when using the
819 ISO "T" timestamptz format (Bruce Momjian)
832 Add operators and functions to extract values from JSON data strings
839 Allow JSON data strings to be converted into records (Andrew Dunstan)
845 Add functions to convert values, records, and hstore data to JSON (Andrew Dunstan)
857 <title>Functions</title>
863 Add array_remove() and array_replace() functions (Marco Nenciarini, Gabriele Bartolini)
869 Allow concat() and format() to properly expand VARIADIC-labeled
870 arguments (Pavel Stehule)
876 Improve format() to handle field width and left/right alignment
883 Have to_char(), to_date(), and to_timestamp() properly handle
884 negative century designations (CC) (Bruce Momjian)
888 Previously the behavior was either wrong or inconsistent with
889 positive/AD handling, e.g. format mask 'IYYY-IW-DY'.
895 Have to_date() and to_timestamp() return proper results when mixing
896 ISO and Gregorian week/day designations (Bruce Momjian)
902 Cause pg_get_viewdef() to default to wrapping after every SELECT
903 target list and FROM entry (Marko Tiikkaja)
907 This causes view output, like from pg_dump, to be more narrow.
913 Have map_sql_value_to_xml_value() properly expand domains (Pavel
920 Force cached functions to be replanned if the search_path changes (Tom Lane)
924 Previously functions already run in the current session ignored
935 <title>Server-Side Languages</title>
941 Allow SPI functions to access the number of rows processed by COPY (Pavel
949 <title><link linkend="plpgsql">PL/pgSQL</link> Server-Side Language</title>
955 Allow PL/pgSQL to use RETURN with a composite-type expressions (Asif
960 Previously RETURN could only reference composite-type variables.
966 Allow PL/pgSQL to access the number of rows processed by COPY (Pavel
971 The command is GET DIAGNOSTICS x = ROW_COUNT.
977 Allow greater flexibility in where keywords can be used in PL/pgSQL (Tom Lane)
986 <title><link linkend="plpython">PL/Python</link> Server-Side Language</title>
992 Add PL/Python result object string handler (Peter Eisentraut)
996 This allows plpy.debug(rv) to output something reasonable.
1002 Have PL/Python convert oid values to a proper Python numeric type (Peter Eisentraut)
1008 Handle SPIErrors raised explicitly with PL/Python's RAISE the same as
1009 as internal SPI errors (Oskari Saarenmaa and Jan Urbanski)
1020 <title>Client Applications</title>
1026 Add command-line utility pg_isready to check if the server is ready
1027 to accept connections (Phil Sorber)
1033 Support multiple --table arguments for pg_restore, clusterdb,
1034 reindexdb, and vacuumdb (Josh Kupershmidt)
1038 This is similar to the pg_dump --table support.
1044 Add -d option to pg_dumpall, pg_basebackup, pg_receivexlog to
1045 specify the connection string (Amit Kapila)
1051 Allow pg_ctl --idempotent to a 'success' return code if the requested
1052 start/stop action fails, but the cluster is already in the requested
1053 state (Peter Eisentraut)
1059 Change pg_ctl to return an error code if start fails because the
1060 server is already running (Peter Eisentraut)
1066 Add libpq function PQconninfo() to return connection information
1067 (Zoltan Boszormenyi, Magnus Hagander)
1074 <title><link linkend="APP-PSQL"><application>psql</></link></title>
1080 Adjust function cost settings so psql tab completion and pattern
1081 searching is more efficient (Tom Lane)
1087 Improve psql tab completion (Jeff Janes, Peter Eisentraut)
1093 Allow the psql --single-transaction mode to work when reading from
1094 standard input (Fabien Coelho, Robert Haas)
1098 Previously this option only worked when reading from a file.
1104 Remove psql warning when connecting to an older server (Peter Eisentraut)
1108 The warning when connecting to a newer server was retained.
1115 <title>Backslash Commands</title>
1121 Add psql \watch command to repeatedly execute commands (Will
1128 Add psql command \gset to store query results in psql variables
1135 Add SSL information to psql's \conninfo command (Alastair Turner)
1141 Add "Security" label to psql \df+ output (Jon Erdman)
1147 Allow psql \l to accept a database name pattern (Peter Eisentraut)
1153 In psql, do not allow \connect to use defaults if there is no active
1154 connection (Bruce Momjian)
1158 This might be the case if the server had crashed.
1164 Properly reset state if psql's "\g file" command failed (Tom Lane)
1168 Previously failed commands discarded output from subsequent commands.
1177 <title>Output</title>
1183 Add a 'latex-longtable' output format to psql (Bruce Momjian)
1187 Longtable allows tables to span multiple pages.
1193 Add a border=3 output mode to the psql 'latex' format (Bruce Momjian)
1199 In psql tuples-only and expanded modes, no longer output "(No rows)"
1206 In psql, no longer print an empty line for unaligned, expanded
1207 output for zero rows (Peter Eisentraut)
1218 <title><link linkend="APP-PGDUMP"><application>pg_dump</></link></title>
1224 Add pg_dump --jobs to dump in parallel when using directory
1225 output format (Joachim Wieland)
1231 Have pg_dump output functions in a predictable order (Joel Jacobson)
1237 Fix tar files emitted by pg_dump and pg_basebackup to be POSIX
1238 conformant (Brian Weaver, Tom Lane)
1244 Add -d/--dbname option to pg_dump, for consistency with other client
1245 commands (Heikki Linnakangas)
1249 The database name could already be supplied last without a flag.
1258 <title><link linkend="APP-INITDB"><application>initdb</></link></title>
1264 Have initdb fsync the newly created data directory (Jeff Davis)
1268 This can be disabled by using --nosync.
1274 Add initdb --sync-only option to sync the data directory to durable
1275 storage (Bruce Momjian)
1279 This is used by pg_upgrade.
1285 Have initdb issue a warning about placing the data directory at the
1286 top of file system mount points (Bruce Momjian)
1297 <title>Source Code</title>
1303 Add an embedded list interface (Andres Freund)
1309 Add infrastructure to better support plug-in background worker
1310 processes (Álvaro Herrera)
1316 Create a centralized timeout API (Zoltán Böszörményi)
1322 Create libpgcommon and move pg_malloc() and other functions there
1323 (Álvaro Herrera, Andres Freund)
1327 This allows libpgport to be used solely for porting code.
1333 Standardize on naming of client-side memory allocation functions (Tom Lane)
1339 Add compiler designations to indicate some ereport() and elog()
1340 calls do not return (Peter Eisentraut, Andres Freund, Tom Lane,
1347 Allow options to be passed to the regression test output comparison
1348 utility via PG_REGRESS_DIFF_OPTS (Peter Eisentraut)
1354 Add isolation tests for CREATE INDEX CONCURRENTLY (Abhijit Menon-Sen)
1360 Remove typedefs for int2/int4 as they are better represented as int16/int32 (Peter Eisentraut)
1366 Fix install-strip on Mac OS X (Peter Eisentraut)
1372 Remove configure flag --disable-shared, as it is no longer used (Bruce Momjian)
1378 Rewrite pgindent in Perl (Andrew Dunstan)
1384 Add emacs macro to match Postgres perltidy formatting (Peter
1391 Run tool to check the keyword list when the backend grammar is changed (Tom Lane)
1397 Centralize flex and bison 'make' rules (Peter Eisentraut)
1401 This is useful for pgxs authors.
1407 Add Assert() functionality to client-side code for frontend code (Andrew Dunstan)
1413 Change many internal many backend functions to return OIDs rather
1414 than void (Dimitri Fontaine)
1418 This is useful for event triggers.
1424 Invent pre-commit/pre-prepare/pre-subcommit events for transaction
1425 callbacks (Tom Lane)
1429 Loadable modules that use transaction callbacks might need modification
1430 to handle these new event types.
1436 Add function pg_identify_object() to dump an object in
1437 machine-readable format (Álvaro Herrera)
1443 Add post-ALTER-object server hooks (KaiGai Kohei)
1449 Implement a generic binary heap and use it for Merge-Append
1450 operations (Abhijit Menon-Sen)
1456 Improve ability to detect official timezone abbreviation changes
1463 Add pkg-config support libpq and ecpg libraries (Peter Eisentraut)
1469 Remove src/tool/backend, now that the content is on the Postgres
1470 wiki (Bruce Momjian)
1476 Split out WAL reading as an independent facility (Heikki
1477 Linnakangas, Andres Freund)
1483 Allow PL/Python to support platform-specific include directories (Peter Eisentraut)
1489 Allow PL/Python on OS X to build against custom versions of Python
1499 <title>Additional Modules</title>
1505 Add a Postgres foreign data wrapper contrib module (Shigeru
1506 Hanada, KaiGai Kohei)
1510 This foreign data wrapper allows writes; potentially other
1511 foreign data wrappers can now support writes.
1517 Add pg_xlogdump contrib program (Andres Freund)
1523 Add support for indexing of regular-expression searches in
1524 contrib/pg_trgm (Alexander Korotkov)
1530 Improve pg_trgm handling of multibyte characters (Tom Lane)
1536 Add function to report the size of the GIN pending index insertion
1541 The function is pgstatginindex() in pgstattuple.
1547 Have oid2name, pgbench, and vacuumlo set fallback_application_name (Amit Kapila)
1553 Improve output of pg_test_timing (Bruce Momjian)
1559 Improve output of pg_test_fsync (Peter Geoghegan)
1565 Improve dblink option validator (Tom Lane)
1576 <title><link linkend="pgupgrade"><application>pg_upgrade</></link></title>
1582 Allow pg_upgrade --jobs to do parallelism (Bruce Momjian)
1586 This allows parallel schema dump/restore of databases, as well as
1587 parallel copy/link of data files per tablespace.
1593 Have pg_upgrade create unix-domain sockets in the current directory
1594 (Bruce Momjian, Tom Lane)
1598 This reduces the possibility that someone will accidentally connect
1605 Have pg_upgrade --check mode properly detect the location of
1606 non-default socket directories (Bruce Momjian, Tom Lane)
1612 Improve performance of pg_upgrade for databases with many tables (Bruce Momjian)
1618 Increase pg_upgrade logging content by showing executed command
1619 (Álvaro Herrera)
1625 Improve pg_upgrade's status display during copy/link (Bruce Momjian)
1634 <title><link linkend="pgbench"><application>pgbench</></link></title>
1640 Add --foreign-keys option to pgbench (Jeff Janes)
1644 This adds foreign key constraints to tables, for use in foreign
1645 key performance testing.
1651 Allow pgbench to aggregate performance statistics and produce output
1652 every --aggregate-interval specified seconds (Tomas Vondra)
1658 Add pg_bench --sampling-rate option to control the percentage of
1659 transactions logged (Tomas Vondra)
1665 Reduce and improve the status message output of pgbench's
1666 initialization mode (Robert Haas, Peter Eisentraut)
1672 Add pgbench -q mode to print one output line every five seconds (Tomas Vondra)
1676 Output pgbench elapsed and estimated remaining time during
1677 initialization logging (Tomas Vondra)
1683 Allow pgbench to use a larger scale factor (Greg Smith)
1694 <title>Documentation</title>
1700 Allow EPUB-format documentation to be created (Peter Eisentraut)
1706 Update FreeBSD kernel configuration documentation (Brad Davis)
1712 Improve WINDOW function documentation (Bruce Momjian, Tom Lane)
1718 Add instructions for setting up the documentation tool chain on
1719 Mac OS X (Peter Eisentraut)
1725 Improve commit_delay documentation (Peter Geoghegan)