1 TODO list for PostgreSQL
2 ========================
3 #A dash (-) marks changes that will appear in the upcoming 8.0 release.#
5 Bracketed items "[]" have more detail.
7 Current maintainer: Bruce Momjian (pgman@candle.pha.pa.us)
8 Last updated: Fri Aug 13 21:30:49 EDT 2004
10 The most recent version of this document can be viewed at the PostgreSQL web site, http://www.PostgreSQL.org.
12 Remove items before beta?
17 * Point-in-time data recovery using backup and write-ahead log,
18 * Create native Win32 port, http://momjian.postgresql.org/main/writings/pgsql/project/win32.html
24 * -Incremental backups
25 * Remove behavior of postmaster -o after making postmaster/postgres
27 * -Allow configuration files to be specified in a different directory
28 * Allow limits on per-db/user connections
29 * Add group object ownership, so groups can rename/drop/grant on objects,
30 so we can implement roles
31 * -Add the concept of dataspaces/tablespaces (Gavin)
32 * -Allow logging of only data definition(DDL), or DDL and modification statements
33 * -Allow log lines to include session-level information, like database and user
34 * Allow server log information to be output as INSERT statements
36 This would allow server log information to be easily loaded into
37 a database for analysis.
39 * Prevent default re-use of sysids for dropped users and groups
41 Currently, if a user is removed while he still owns objects, a new
42 user given might be given their user id and inherit the
43 previous users objects.
45 * Prevent dropping user that still owns objects, or auto-drop the objects
46 * Allow pooled connections to list all prepared queries
48 This would allow an application inheriting a pooled connection to know
49 the queries prepared in the current session.
51 * Allow major upgrades without dump/reload, perhaps using pg_upgrade
52 * Have SHOW ALL and pg_settings show descriptions for server-side variables
53 * -Allow external interfaces to extend the GUC variable set
54 * Allow GRANT/REVOKE permissions to be given to all schema objects with one
56 * Remove unreferenced table files created by transactions that were
57 in-progress when the server terminated abruptly
58 * Allow reporting of which objects are in which tablespaces
60 This item is difficult because a tablespace can contain objects from
61 multiple databases. There is a server-side function that returns the
62 databases which use a specific tablespace, so this requires a tool
63 that will call that function and connect to each database to find the
64 objects in each database for that tablespace.
66 * Allow database recovery where tablespaces can't be created
68 When a pg_dump is restored, all tablespaces will attempt to be created
69 in their original locations. If this fails, the user must be able to
70 adjust the restore process.
72 * Add "include file" functionality in postgresql.conf
73 * Add session start time and last statement time to pg_stat_activity
74 * Allow server logs to be remotely read using SQL commands
75 * Allow server configuration parameters to be remotely modified
76 * Allow administrators to safely terminate individual sessions
78 Right now, SIGTERM will terminate a session, but it is treated as
79 though the postmaster has paniced and shared memory might not be
80 cleaned up properly. A new signal is needed for safe termination.
82 * Un-comment all variables in postgresql.conf
84 By not showing commented-out variables, we discourage people from
85 thinking that re-commenting a variable returns it to its default.
87 * Allow point-in-time recovery to archive partially filled write-ahead
90 Currently only full WAL files are archived. This means that the most
91 recent transactions aren't available for recovery in case of a disk
94 * Create dump tool for write-ahead logs for use in determining
95 transaction id for point-in-time recovery
96 * Improve replication solutions
99 The proper solution to this will probably the use of a master/slave
100 replication solution like Sloney and a connection pooling tool like
105 You can use any of the master/slave replication servers to use a
106 standby server for data warehousing. To allow read/write queries to
107 multiple servers, you need multi-master replication like pgcluster.
109 o Allow replication over unreliable or non-persistent links
115 * Remove Money type, add money formatting for decimal type
116 * -Change factorial to return a numeric (Gavin)
117 * Change NUMERIC to enforce the maximum precision, and increase it
118 * Add function to return compressed length of TOAST data values
119 * Allow INET subnet tests with non-constants to be indexed
120 * Add transaction_timestamp(), statement_timestamp(), clock_timestamp() functionality
122 Current CURRENT_TIMESTAMP returns the start time of the current
123 transaction, and gettimeofday() returns the wallclock time. This will
124 make time reporting more consistent and will allow reporting of
125 the statement start time.
127 * Have sequence dependency track use of DEFAULT sequences,
129 * Disallow changing default expression of a SERIAL column (?)
130 * Allow infinite dates just like infinite timestamps
131 * -Allow pg_dump to dump sequences using NO_MAXVALUE and NO_MINVALUE
132 * -Prevent whole-row references from leaking memory, e.g. SELECT COUNT(tab.*)
133 * Have initdb set DateStyle based on locale?
134 * Add pg_get_acldef(), pg_get_typedefault(), and pg_get_attrdef()
135 * Allow to_char to print localized month names
136 * Allow functions to have a search path specified at creation time
137 * -Make LENGTH() of CHAR() not count trailing spaces
138 * Allow substring/replace() to get/set bit values
139 * Add GUC variable to allow output of interval values in ISO8601 format
140 * -Support composite types as table columns
141 * Fix data types where equality comparison isn't intuitive, e.g. box
145 o Allow nulls in arrays
146 o Allow MIN()/MAX() on arrays
147 o Delay resolution of array expression type so assignment coercion
148 can be performed on empty array expressions
149 o Modify array literal representation to handle array index lower bound
154 o Improve vacuum of large objects, like /contrib/vacuumlo (?)
155 o Add security checking for large objects
157 Currently large objects entries do not have owners. Permissions can
158 only be set at the pg_largeobject table level.
160 o Auto-delete large objects when referencing row is deleted
162 o Allow read/write into TOAST values like large objects
164 This requires the TOAST column to be stored EXTERNAL.
167 Multi-Language Support
168 ======================
170 * Add NCHAR (as distinguished from ordinary varchar),
171 * Allow locale to be set at database creation
173 Currently locale can only be set during initdb.
175 * Allow encoding on a per-column basis
177 Right now only one encoding is allowed per database.
179 * Optimize locale to have minimal performance impact when not used
180 * Support multiple simultaneous character sets, per SQL92
181 * Improve Unicode combined character handling (?)
182 * Add octet_length_server() and octet_length_client()
183 * Make octet_length_client() the same as octet_length()?
184 * -Prevent mismatch of frontend/backend encodings from converting bytea
185 data from being interpreted as encoded strings
186 * -Fix upper()/lower() to work for multibyte encodings
192 * Automatically create rules on views so they are updateable, per SQL92 [view]
193 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
194 * Allow NOTIFY in rules involving conditionals
195 * Have views on temporary tables exist in the temporary namespace
196 * Allow temporary views on non-temporary tables
197 * Allow RULE recompilation
203 * -Order duplicate index entries on creation by ctid for faster heap lookups
204 * Allow inherited tables to inherit index, UNIQUE constraint, and primary
205 key, foreign key [inheritance]
206 * UNIQUE INDEX on base column not honored on inserts/updates from
207 inherited table: INSERT INTO inherit_table (unique_index_col) VALUES
208 (dup) should fail [inheritance]
210 The main difficulty with this item is the problem of creating an index
211 that can spam more than one table.
213 * Add UNIQUE capability to non-btree indexes
214 * Add rtree index support for line, lseg, path, point
215 * Use indexes for MIN() and MAX()
217 MIN/MAX queries can already be rewritten as SELECT col FROM tab ORDER
218 BY col {DESC} LIMIT 1. Completing this item involves making this
219 transformation automatically.
221 * Use index to restrict rows returned by multi-key index when used with
222 non-consecutive keys to reduce heap accesses
224 For an index on col1,col2,col3, and a WHERE clause of col1 = 5 and
225 col3 = 9, spin though the index checking for col1 and col3 matches,
226 rather than just col1
228 * -Be smarter about insertion of already-ordered data into btree index
229 * Prevent index uniqueness checks when UPDATE does not modify the column
231 Uniqueness (index) checks are done when updating a column even if the
232 column is not modified by the UPDATE.
234 * Fetch heap pages matching index entries in sequential order [performance]
236 Rather than randomly accessing heap pages based on index entries, mark
237 heap pages needing access in a bitmap and do the lookups in sequential
238 order. Another method would be to sort heap ctids matching the index
239 before accessing the heap rows.
241 * Use bitmaps to combine existing indexes [performance]
243 Bitmap indexes allow single indexed columns to be combined to
244 dynamically create a composite index to match a specific query. Each
245 index is a bitmap, and the bitmaps are AND'ed or OR'ed to be combined.
247 * Allow use of indexes to search for NULLs
249 One solution is to create a partial index on an IS NULL expression.
251 * -Allow SELECT * FROM tab WHERE int2col = 4 to use int2col index, int8,
252 float4, numeric/decimal too
253 * Add concurrency to GIST
254 * Pack hash index buckets onto disk pages more efficiently
256 Currently no only one hash bucket can be stored on a page. Ideally
257 several hash buckets could be stored on a single page and greater
258 granularity used for the hash algorithm.
264 * Add BETWEEN ASYMMETRIC/SYMMETRIC
265 * Change LIMIT/OFFSET to use int8
266 * CREATE TABLE AS can not determine column lengths from expressions [atttypmod]
267 * Allow UPDATE to handle complex aggregates [update] (?)
268 * -Allow command blocks to ignore certain types of errors
269 * Allow backslash handling in quoted strings to be disabled for portability
271 The use of C-style backslashes (.e.g. \n, \r) in quoted strings is not
272 SQL-spec compliant, so allow such handling to be disabled.
274 * Allow an alias to be provided for the target table in UPDATE/DELETE
276 This is not SQL-spec but many DBMSs allow it.
278 * Allow additional tables to be specified in DELETE for joins
280 UPDATE already allows this (UPDATE...FROM) but we need similar
281 functionality in DELETE. It's been agreed that the keyword should
282 be USING, to avoid anything as confusing as DELETE FROM a FROM b.
284 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
285 * Allow REINDEX to rebuild all database indexes, remove /contrib/reindex
286 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
287 * Add a schema option to createlang
288 * -Allow savepoints / nested transactions [transactions] (Alvaro)
289 * -Use nested transactions to prevent syntax errors from aborting a transaction
290 * Allow UPDATE tab SET ROW (col, ...) = (...) for updating multiple columns
291 * Allow SET CONSTRAINTS to be qualified by schema/table name
292 * -Prevent COMMENT ON DATABASE from using a database name
294 * Allow TRUNCATE ... CASCADE/RESTRICT
295 * Allow PREPARE of cursors
296 * Allow PREPARE to automatically determine parameter types based on the SQL
298 * Allow finer control over the caching of prepared query plans
300 Currently, queries prepared via the libpq API are planned on first
301 execute using the supplied parameters --- allow SQL PREPARE to do the
302 same. Also, allow control over replanning prepared queries either
303 manually or automatically when statistics for execute parameters
304 differ dramatically from those used during planning.
306 * Allow LISTEN/NOTIFY to store info in memory rather than tables?
308 Currently LISTEN/NOTIFY information is stored in pg_listener. Storing
309 such information in memory would improve performance.
311 * -COMMENT ON [ CAST | CONVERSION | OPERATOR CLASS | LARGE OBJECT | LANGUAGE ]
313 * Dump large object comments in custom dump format
314 * Add optional textual message to NOTIFY
316 This would allow an informational message to be added to the notify
317 message, perhaps indicating the row modified or other custom
320 * -Allow more ISOLATION LEVELS to be accepted
321 * Allow CREATE TABLE foo (f1 INT CHECK (f1 > 0) CHECK (f1 < 10)) to work
322 by searching for non-conflicting constraint names, and prefix with
324 * Use more reliable method for CREATE DATABASE to get a consistent copy
327 Currently the system uses the operating system COPY command to create
330 * Add C code to copy directories for use in creating new databases
331 * -Have psql \dn show only visible temp schemas using current_schemas()
332 * -Have psql '\i ~/<tab><tab>' actually load files it displays from home dir
333 * Ignore temporary tables from other sessions when processing
335 * -Add GUC setting to make created tables default to WITHOUT OIDS
336 * Have pg_ctl look at PGHOST in case it is a socket directory?
337 * Allow column-level GRANT/REVOKE privileges
338 * Add a session mode to warn about non-standard SQL usage in queries
339 * Add MERGE command that does UPDATE/DELETE, or on failure, INSERT (rules, triggers?)
340 * Add ON COMMIT capability to CREATE TABLE AS SELECT
341 * Add NOVICE output level for helpful messages like automatic sequence/index creation
342 * Add COMMENT ON for all cluster global objects (users, groups,
343 databases and tablespaces)
346 o -ALTER TABLE ADD COLUMN does not honor DEFAULT and non-CHECK CONSTRAINT
347 o -ALTER TABLE ADD COLUMN column DEFAULT should fill existing
348 rows with DEFAULT value
349 o -ALTER TABLE ADD COLUMN column SERIAL doesn't create sequence because
351 o Have ALTER TABLE RENAME rename SERIAL sequence names
352 o -Allow ALTER TABLE to modify column lengths and change to binary
354 o -Add ALTER DATABASE ... OWNER TO newowner
355 o Add ALTER DOMAIN TYPE
356 o Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
357 o Allow ALTER TABLE to change constraint deferrability and actions
358 o Disallow dropping of an inherited constraint
359 o Allow objects to be moved to different schemas
360 o Allow ALTER TABLESPACE to move to different directories
361 o Allow databases, schemas, and indexes to be moved to different
363 o Allow moving system tables to other tablespaces, where possible
365 Currently non-global system tables must be in the default database
366 schema. Global system tables can never be moved.
368 o -Add ALTER DOMAIN, AGGREGATE, CONVERSION ... OWNER TO
369 o -Add ALTER SEQUENCE ... OWNER TO
370 o Add ALTER INDEX that works just like ALTER TABLE already does
374 o Automatically maintain clustering on a table
376 This would require some background daemon to restore clustering
377 during periods of low usage. It might also require tables to be only
378 paritally filled for easier reorganization.
380 o -Add ALTER TABLE table SET WITHOUT CLUSTER (Christopher)
381 o Add default clustering to system tables
383 To do this, determine the ideal cluster index for each system
384 table and set the cluster setting during initdb.
387 o -Allow dump/load of CSV format
388 o Allow COPY to report error lines and continue
390 This requires the use of a savepoint before each COPY line is
391 processed, with ROLLBACK on COPY failure.
393 o Allow COPY to understand \x as a hex byte
394 o Have COPY return the number of rows loaded/unloaded (?)
397 o Allow UPDATE/DELETE WHERE CURRENT OF cursor
399 This requires using the row ctid to map cursor rows back to the
400 original heap row. This become more complicated if WITH HOLD cursors
401 are to be supported because WITH HOLD cursors have a copy of the row
402 and no FOR UPDATE lock.
404 o Prevent DROP TABLE from dropping a row referenced by its own open
407 o Allow pooled connections to list all open WITH HOLD cursors
409 Because WITH HOLD cursors exist outside transactions, this allows
410 them to be listed so they can be closed.
413 o Allow INSERT/UPDATE of the system-generated oid value for a row
414 o Allow INSERT INTO tab (col1, ..) VALUES (val1, ..), (val2, ..)
415 o Allow INSERT/UPDATE ... RETURNING new.col or old.col
417 This is useful for returning the auto-generated key for an INSERT.
418 One complication is how to handle rules that run as part of
422 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
424 o Add SET PATH for schemas (?)
426 This is basically the same as SET search_path.
428 o Prevent conflicting SET options from being set
430 This requires a checking function to be called after the server
431 configuration file is read.
433 * SERVER-SIDE LANGUAGES
434 o Allow PL/PgSQL's RAISE function to take expressions (?)
436 Currently only constants are supported.
438 o Change PL/PgSQL to use palloc() instead of malloc()
439 o -Allow Java server-side programming
440 o Handle references to temporary tables that are created, destroyed,
441 then recreated during a session, and EXECUTE is not used
443 This requires the cached PL/PgSQL byte code to be invalidated when
444 an object referenced in the function is changed.
446 o Fix PL/pgSQL RENAME to work on variables other than OLD/NEW
447 o Improve PL/PgSQL exception handling using savepoints
448 o -Allow PL/pgSQL parameters to be specified by name and type during definition
449 o Allow function parameters to be passed by name,
450 get_employee_salary(emp_id => 12345, tax_year => 2001)
451 o Add Oracle-style packages
452 o Add table function support to pltcl, plperl, plpython (?)
453 o Allow PL/pgSQL to name columns by ordinal position, e.g. rec.(3)
454 o Allow PL/pgSQL EXECUTE query_var INTO record_var;
455 o Add capability to create and call PROCEDURES
456 o Allow PL/pgSQL to handle %TYPE arrays, e.g. tab.col%TYPE[]
462 * Add XML output to pg_dump and COPY
464 We already allow XML to be stored in the database, and XPath queries
465 can be used on that data using /contrib/xml2. It also supports XSLT
468 * -Allow psql \du to show users, and add \dg for groups
469 * Add a libpq function to support Parse/DescribeStatement capability
470 * Prevent libpq's PQfnumber() from lowercasing the column name (?)
471 * -Allow pg_dump to dump CREATE CONVERSION (Christopher)
472 * -Make pg_restore continue after errors, so it acts more like pg_dump scripts
473 * Have psql show current values for a sequence
474 * Allow pg_dumpall to use non-text output formats
475 * Have pg_dump use multi-statement transactions for INSERT dumps
476 * Move psql backslash database information into the backend, use mnemonic
479 This would allow non-psql clients to pull the same information out of
480 the database as psql.
482 * Allow pg_dump to use multiple -t and -n switches
484 This should be done by allowing a '-t schema.table' syntax.
486 * Fix oid2name and dbsize for tablespaces
487 * Consistently display privilege information for all objects in psql
492 Document differences between ecpg and the SQL standard and
493 information about the Informix-compatibility module.
495 o -Implement SET DESCRIPTOR
496 o Solve cardinality > 1 for input descriptors / variables (?)
497 o Improve error handling (?)
498 o Add a semantic check level, e.g. check if a table really exists
499 o fix handling of DB attributes that are arrays
500 o Use backend PREPARE/EXECUTE facility for ecpg where possible
502 o Fix nested C comments
503 o sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
504 o Make SET CONNECTION thread-aware, non-standard?
505 o Allow multidimensional arrays
508 Referential Integrity
509 =====================
511 * Add MATCH PARTIAL referential integrity
512 * Add deferred trigger queue file
514 Right now all deferred trigger information is stored in backend
515 memory. This could exhaust memory for very large trigger queues.
516 This item involves dumping large queues into files.
518 * Implement dirty reads or shared row locks and use them in RI triggers (?)
519 * Enforce referential integrity for system tables
520 * Change foreign key constraint for array -> element to mean element
522 * Allow DEFERRABLE UNIQUE constraints (?)
523 * Allow triggers to be disabled [trigger]
525 Currently the only way to disable triggers is to modify the system
528 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
530 If the dump is known to be valid, allow foreign keys to be added
531 without revalidating the data.
533 * Allow statement-level triggers to access modified rows
534 * Support triggers on columns
535 * Have AFTER triggers execute after the appropriate SQL statement in a
536 function, not at the end of the function
537 * -Print table names with constraint names in error messages, or make constraint
538 names unique within a schema
539 * -Issue NOTICE if foreign key data requires costly test to match primary key
540 * Remove CREATE CONSTRAINT TRIGGER
542 This was used in older releases to dump referential integrity
545 * Allow AFTER triggers on system tables
547 System tables are modified in many places in the backend without going
548 through the executor and therefore not causing triggers to fire. To
549 complete this item, the functions that modify system tables will have
556 * Flush cached query plans when the dependent objects change
557 * -Use dependency information to dump data in proper order
558 * -Have pg_dump -c clear the database using dependency information
559 * Track dependencies in function bodies and recompile/invalidate
565 * Add SQL99 WITH clause to SELECT
566 * Add SQL99 WITH RECURSIVE to SELECT
567 * Add pre-parsing phase that converts non-ANSI syntax to supported
570 This could allow SQL written for other databases to run without
573 * Allow plug-in modules to emulate features from other databases
574 * SQL*Net listener that makes PostgreSQL appear as an Oracle database
576 * Allow queries across databases or servers with transaction
579 Right now contrib/dblink can be used to issue such queries except it
580 does not have locking or transaction semantics. Two-phase commit is
581 needed to enable transaction semantics.
583 * Add two-phase commit
585 This will involve adding a way to respond to commit failure by either
586 taking the server into offline/readonly mode or notifying the
597 * Improve commit_delay handling to reduce fsync()
598 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
599 * Allow multiple blocks to be written to WAL with one write()
600 * Add an option to sync() before fsync()'ing checkpoint files
605 * Add free-behind capability for large sequential scans [fadvise]
606 * Consider use of open/fcntl(O_DIRECT) to minimize OS caching
607 * Cache last known per-tuple offsets to speed long tuple access
609 While column offsets are already cached, the cache can not be used if
610 the tuple has NULLs or TOAST columns because these values change the
611 typical column offsets. Caching of such offsets could be accomplished
612 by remembering the previous offsets and use them again if the row has
617 We could use a fixed row count and a +/- count to follow MVCC
618 visibility rules, or a single cached value could be used and
619 invalidated if anyone modifies the table. [count]
625 * Improve speed with indexes
627 For large table adjustements during vacuum, it is faster to reindex
628 rather than update the index.
630 * Reduce lock time by moving tuples with read lock, then write
631 lock and truncate table
633 Moved tuples are invisible to other backends so they don't require a
634 write lock. However, the read lock promotion to write lock could lead
635 to deadlock situations.
637 * -Provide automatic running of vacuum in the background in backend
638 rather than in /contrib (Matthew)
639 * Allow free space map to be auto-sized or warn when it is too small
641 The free space map is in shared memory so resizing is difficult.
643 * Maintain a map of recently-expired rows
645 This allows vacuum to reclaim free space without requiring
652 * Make locking of shared data structures more fine-grained
654 This requires that more locks be acquired but this would reduce lock
655 contention, improving concurrency.
657 * Add code to detect an SMP machine and handle spinlocks accordingly
658 from distributted.net, http://www1.distributed.net/source,
659 in client/common/cpucheck.cpp
661 On SMP machines, it is possible that locks might be released shortly,
662 while on non-SMP machines, the backend should sleep so the process
663 holding the lock can complete and release it.
665 * Improve SMP performance on i386 machines
667 i386-based SMP machines can generate excessive context switching
668 caused by lock failure in high concurrency situations. This may be
669 caused by CPU cache line invalidation inefficiencies.
671 * Research use of sched_yield() for spinlock acquisition failure
677 * Experiment with multi-threaded backend [thread]
679 This would prevent the overhead associated with process creation. Most
680 operating systems have trivial process creation time compared to
681 database startup overhead, but a few operating systems (WIn32,
682 Solaris) might benefit from threading.
684 * Add connection pooling [pool]
686 It is unclear if this should be done inside the backend code or done
687 by something external like pgpool. The passing of file descriptors to
688 existing backends is one of the difficulties with a backend approach.
694 * Eliminate need to write full pages to WAL before page modification [wal]
696 Currently, to protect against partial disk page writes, we write the
697 full page images to WAL before they are modified so we can correct any
698 partial page writes during recovery.
700 * Reduce WAL traffic so only modified values are written rather than
702 * Turn off after-change writes if fsync is disabled
704 If fsync is off, there is no purpose in writing full pages to WAL
706 * Add WAL index reliability improvement to non-btree indexes
707 * Allow the pg_xlog directory location to be specified during initdb
708 with a symlink back to the /data location
710 * Allow WAL information to recover corrupted pg_controldata
711 * Find a way to reduce rotational delay when repeatedly writing
714 Currently fsync of WAL requires the disk platter to perform a full
715 rotation to fsync again. One idea is to write the WAL to different
716 offsets that might reduce the rotational delay.
718 * Allow buffered WAL writes and fsync
720 Instead of guaranteeing recovery of all committed transactions, this
721 would provide improved performance by delaying WAL writes and fsync
722 so an abrupt operating system restart might lose a few seconds of
723 committed transactions but still be consistent. We could perhaps
724 remove the 'fsync' parameter (which results in an an inconsistent
725 database) in favor of this capability.
731 * Add missing optimizer selectivities for date, r-tree, etc
732 * Allow ORDER BY ... LIMIT 1 to select high/low value without sort or
733 index using a sequential scan for highest/lowest values
735 If only one value is needed, there is no need to sort the entire
736 table. Instead a sequential scan could get the matching value.
738 * Precompile SQL functions to avoid overhead
739 * Add utility to compute accurate random_page_cost value
740 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
741 * Allow sorting, temp files, temp tables to use multiple work directories
743 This allows the I/O load to be spread across multiple disk drives.
744 * Have EXPLAIN ANALYZE highlight poor optimizer estimates
745 * Use CHECK constraints to influence optimizer decisions
747 CHECK constraints contain information about the distribution of values
748 within the table. This is also useful for implementing subtables where
749 a tables content is distributed across several subtables.
755 * Do async I/O for faster random read-ahead of data
757 Async I/O allows multiple I/O requests to be sent to the disk with
758 results coming back asynchronously.
760 * Use mmap() rather than SYSV shared memory or to write WAL files (?) [mmap]
762 This would remove the requirement for SYSV SHM but would introduce
763 portability issues. Anonymous mmap is required to prevent I/O
766 * Add a script to ask system configuration questions and tune postgresql.conf
767 * -Use background process to write dirty shared buffers to disk
768 * Use a phantom command counter for nested subtransactions to reduce
775 * Add use of 'const' for variables in source tree
776 * Rename some /contrib modules from pg* to pg_*
777 * Move some things from /contrib into main tree
778 * Remove warnings created by -Wcast-align
779 * Move platform-specific ps status display info from ps_status.c to ports
780 * Improve access-permissions check on data directory in Cygwin (Tom)
781 * Add optional CRC checksum to heap and index pages
782 * -Change representation of whole-tuple parameters to functions
783 * Clarify use of 'application' and 'command' tags in SGML docs
784 * Better document ability to build only certain interfaces (Marc)
785 * Remove or relicense modules that are not under the BSD license, if possible
786 * Remove memory/file descriptor freeing before ereport(ERROR)
787 * Acquire lock on a relation before building a relcache entry for it
788 * Research interaction of setitimer() and sleep() used by statement_timeout
789 * -Add checks for fclose() failure (Tom)
790 * -Change CVS ID to PostgreSQL
791 * -Exit postmaster if postgresql.conf can not be opened
792 * Rename /scripts directory because they are all C programs now
793 * Promote debug_query_string into a server-side function current_query()
794 * Allow the identifier length to be increased via a configure option
795 * Allow binaries to be statically linked so they are more easily relocated
798 * Wire Protocol Changes
799 o Allow dynamic character set handling
800 o Add decoded type, length, precision
802 o Update clients to use data types, typmod, schema.table.column names of
803 result sets using new query protocol
806 ---------------------------------------------------------------------------
809 Developers who have claimed items are:
810 --------------------------------------
811 * Alvaro is Alvaro Herrera <alvherre@dcc.uchile.cl>
812 * Andrew is Andrew Dunstan <andrew@dunslane.net>
813 * Bruce is Bruce Momjian <pgman@candle.pha.pa.us> of Software Research Assoc.
814 * Christopher is Christopher Kings-Lynne <chriskl@familyhealth.com.au> of
815 Family Health Network
816 * Claudio is Claudio Natoli <claudio.natoli@memetrics.com>
817 * D'Arcy is D'Arcy J.M. Cain <darcy@druid.net> of The Cain Gang Ltd.
818 * Fabien is Fabien Coelho <coelho@cri.ensmp.fr>
819 * Gavin is Gavin Sherry <swm@linuxworld.com.au> of Alcove Systems Engineering
820 * Greg is Greg Sabino Mullane <greg@turnstep.com>
821 * Hiroshi is Hiroshi Inoue <Inoue@tpf.co.jp>
822 * Jan is Jan Wieck <JanWieck@Yahoo.com> of Afilias, Inc.
823 * Joe is Joe Conway <mail@joeconway.com>
824 * Karel is Karel Zak <zakkr@zf.jcu.cz>
826 * Magnus is Magnus Hagander <mha@sollentuna.net>
827 * Marc is Marc Fournier <scrappy@hub.org> of PostgreSQL, Inc.
828 * Matthew T. O'Connor <matthew@zeut.net>
829 * Michael is Michael Meskes <meskes@postgresql.org> of Credativ
830 * Neil is Neil Conway <neilc@samurai.com>
831 * Oleg is Oleg Bartunov <oleg@sai.msu.su>
832 * Peter is Peter Eisentraut <peter_e@gmx.net>
833 * Philip is Philip Warner <pjw@rhyme.com.au> of Albatross Consulting Pty. Ltd.
834 * Rod is Rod Taylor <pg@rbt.ca>
835 * Simon is Simon Riggs
836 * Stephan is Stephan Szabo <sszabo@megazone23.bigpanda.com>
837 * Tatsuo is Tatsuo Ishii <t-ishii@sra.co.jp> of Software Research Assoc.
839 * Tom is Tom Lane <tgl@sss.pgh.pa.us> of Red Hat