2 TODO list for PostgreSQL
3 ========================
4 #A hyphen (-) marks changes that will appear in the upcoming 8.1 release.#
6 Bracketed items "[]" have more detail.
8 Current maintainer: Bruce Momjian (pgman@candle.pha.pa.us)
9 Last updated: Wed Nov 10 00:47:03 EST 2004
11 The most recent version of this document can be viewed at the PostgreSQL web
12 site, http://www.PostgreSQL.org.
19 * Remove behavior of postmaster -o after making postmaster/postgres
21 * Allow limits on per-db/user connections
22 * Add group object ownership, so groups can rename/drop/grant on objects,
23 so we can implement roles
24 * Allow server log information to be output as INSERT statements
26 This would allow server log information to be easily loaded into
27 a database for analysis.
29 * Prevent default re-use of sysids for dropped users and groups
31 Currently, if a user is removed while he still owns objects, a new
32 user given might be given their user id and inherit the
33 previous users objects.
35 * Prevent dropping user that still owns objects, or auto-drop the objects
36 * Allow pooled connections to list all prepared queries
38 This would allow an application inheriting a pooled connection to know
39 the queries prepared in the current session.
41 * Allow major upgrades without dump/reload, perhaps using pg_upgrade
42 * Have SHOW ALL and pg_settings show descriptions for server-side variables
43 * Allow GRANT/REVOKE permissions to be given to all schema objects with one
45 * Remove unreferenced table files created by transactions that were
46 in-progress when the server terminated abruptly
47 * Allow reporting of which objects are in which tablespaces
49 This item is difficult because a tablespace can contain objects from
50 multiple databases. There is a server-side function that returns the
51 databases which use a specific tablespace, so this requires a tool
52 that will call that function and connect to each database to find the
53 objects in each database for that tablespace.
55 * Allow a database in tablespace t1 with tables created in tablespace t2
56 to be used as a template for a new database created with default
59 All objects in the default database tablespace must have default tablespace
60 specifications. This is because new databases are created by copying
61 directories. If you mix default tablespace tables and tablespace-specified
62 tables in the same directory, creating a new database from such a mixed
63 directory would create a new database with tables that had incorrect
64 explicit tablespaces. To fix this would require modifying pg_class in the
65 newly copied database, which we don't currently do.
67 * Add a GUC variable to control the tablespace for temporary objects and
70 It could start with a random tablespace from a supplied list and cycle
73 * Add "include file" functionality in postgresql.conf
74 * Add session start time and last statement time to pg_stat_activity
75 * Allow server logs to be remotely read using SQL commands
76 * Allow server configuration parameters to be remotely modified
77 * Allow administrators to safely terminate individual sessions
79 Right now, SIGTERM will terminate a session, but it is treated as
80 though the postmaster has paniced and shared memory might not be
81 cleaned up properly. A new signal is needed for safe termination.
83 * Un-comment all variables in postgresql.conf
85 By not showing commented-out variables, we discourage people from
86 thinking that re-commenting a variable returns it to its default.
87 This has to address environment variables that are then overridden
88 by config file values. Another option is to allow commented values
89 to return to their default values.
91 * Allow point-in-time recovery to archive partially filled write-ahead
94 Currently only full WAL files are archived. This means that the most
95 recent transactions aren't available for recovery in case of a disk
98 * Create dump tool for write-ahead logs for use in determining
99 transaction id for point-in-time recovery
100 * Set proper permissions on non-system schemas during db creation
102 Currently all schemas are owned by the super-user because they are
103 copied from the template1 database.
105 * Add a function that returns the 'uptime' of the postmaster
106 * Improve replication solutions
109 The proper solution to this will probably the use of a master/slave
110 replication solution like Sloney and a connection pooling tool like
115 You can use any of the master/slave replication servers to use a
116 standby server for data warehousing. To allow read/write queries to
117 multiple servers, you need multi-master replication like pgcluster.
119 o Allow replication over unreliable or non-persistent links
125 * Remove Money type, add money formatting for decimal type
126 * Change NUMERIC to enforce the maximum precision, and increase it
127 * Add function to return compressed length of TOAST data values
128 * Allow INET subnet tests with non-constants to be indexed
129 * Add transaction_timestamp(), statement_timestamp(), clock_timestamp()
132 Current CURRENT_TIMESTAMP returns the start time of the current
133 transaction, and gettimeofday() returns the wallclock time. This will
134 make time reporting more consistent and will allow reporting of
135 the statement start time.
137 * Have sequence dependency track use of DEFAULT sequences,
139 * Disallow changing default expression of a SERIAL column (?)
140 * Allow infinite dates just like infinite timestamps
141 * Have initdb set DateStyle based on locale?
142 * Add pg_get_acldef(), pg_get_typedefault(), and pg_get_attrdef()
143 * Allow to_char to print localized month names
144 * Allow functions to have a search path specified at creation time
145 * Allow substring/replace() to get/set bit values
146 * Add GUC variable to allow output of interval values in ISO8601 format
147 * Fix data types where equality comparison isn't intuitive, e.g. box
148 * Merge hardwired timezone names with the TZ database; allow either kind
149 everywhere a TZ name is currently taken
150 * Allow customization of the known set of TZ names (generalize the
151 present australian_timezones hack)
152 * Allow TIMESTAMP WITH TIME ZONE to store the original timezone
153 information, either by name or offset from UTC
154 * Prevent inet cast to cidr if the unmasked bits are not zero, or
158 o Allow nulls in arrays
159 o Allow MIN()/MAX() on arrays
160 o Delay resolution of array expression type so assignment coercion
161 can be performed on empty array expressions
162 o Modify array literal representation to handle array index lower bound
167 o Improve vacuum of large objects, like /contrib/vacuumlo (?)
168 o Add security checking for large objects
170 Currently large objects entries do not have owners. Permissions can
171 only be set at the pg_largeobject table level.
173 o Auto-delete large objects when referencing row is deleted
175 o Allow read/write into TOAST values like large objects
177 This requires the TOAST column to be stored EXTERNAL.
180 Multi-Language Support
181 ======================
183 * Add NCHAR (as distinguished from ordinary varchar),
184 * Allow locale to be set at database creation
186 Currently locale can only be set during initdb.
188 * Allow encoding on a per-column basis
190 Right now only one encoding is allowed per database.
192 * Optimize locale to have minimal performance impact when not used
193 * Support multiple simultaneous character sets, per SQL92
194 * Improve Unicode combined character handling (?)
195 * Add octet_length_server() and octet_length_client()
196 * Make octet_length_client() the same as octet_length()?
202 * Automatically create rules on views so they are updateable, per SQL99
204 We can only auto-create rules for simple views. For more complex
205 cases users will still have to write rules.
207 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
208 * Allow NOTIFY in rules involving conditionals
209 * Have views on temporary tables exist in the temporary namespace
210 * Allow temporary views on non-temporary tables
211 * Allow RULE recompilation
217 * Allow inherited tables to inherit index, UNIQUE constraint, and primary
218 key, foreign key [inheritance]
219 * UNIQUE INDEX on base column not honored on inserts/updates from
220 inherited table: INSERT INTO inherit_table (unique_index_col) VALUES
221 (dup) should fail [inheritance]
223 The main difficulty with this item is the problem of creating an index
224 that can span more than one table.
226 * Add UNIQUE capability to non-btree indexes
227 * Add rtree index support for line, lseg, path, point
228 * Use indexes for MIN() and MAX()
230 MIN/MAX queries can already be rewritten as SELECT col FROM tab ORDER
231 BY col {DESC} LIMIT 1. Completing this item involves making this
232 transformation automatically.
234 * Use index to restrict rows returned by multi-key index when used with
235 non-consecutive keys to reduce heap accesses
237 For an index on col1,col2,col3, and a WHERE clause of col1 = 5 and
238 col3 = 9, spin though the index checking for col1 and col3 matches,
239 rather than just col1; also called skip-scanning.
241 * Prevent index uniqueness checks when UPDATE does not modify the column
243 Uniqueness (index) checks are done when updating a column even if the
244 column is not modified by the UPDATE.
246 * Fetch heap pages matching index entries in sequential order [performance]
248 Rather than randomly accessing heap pages based on index entries, mark
249 heap pages needing access in a bitmap and do the lookups in sequential
250 order. Another method would be to sort heap ctids matching the index
251 before accessing the heap rows.
253 * Allow non-bitmap indexes to be combined by creating bitmaps in memory
255 Bitmap indexes index single columns that can be combined with other bitmap
256 indexes to dynamically create a composite index to match a specific query.
257 Each index is a bitmap, and the bitmaps are bitwise AND'ed or OR'ed to be
258 combined. They can index by tid or can be lossy requiring a scan of the
259 heap page to find matching rows, or perhaps use a mixed solution where
260 tids are recorded for pages with only a few matches and per-page bitmaps
261 are used for more dense pages. Another idea is to use a 32-bit bitmap
262 for every page and set a bit based on the item number mod(32).
264 * Allow the creation of on-disk bitmap indexes which can be quickly
265 combined with other bitmap indexes
267 Such indexes could be more compact if there are only a few distinct values.
268 Such indexes can also be compressed. Keeping such indexes updated can be
271 * Allow use of indexes to search for NULLs
273 One solution is to create a partial index on an IS NULL expression.
275 * Add concurrency to GIST
276 * Pack hash index buckets onto disk pages more efficiently
278 Currently no only one hash bucket can be stored on a page. Ideally
279 several hash buckets could be stored on a single page and greater
280 granularity used for the hash algorithm.
282 * Allow accurate statistics to be collected on indexes with more than
283 one column or expression indexes, perhaps using per-index statistics
285 * Add fillfactor to control reserved free space during index creation
290 * Add BETWEEN ASYMMETRIC/SYMMETRIC
291 * Change LIMIT/OFFSET to use int8
292 * Allow CREATE TABLE AS to determine column lengths for complex
293 expressions like SELECT col1 || col2
294 * Allow UPDATE to handle complex aggregates [update] (?)
295 * Allow backslash handling in quoted strings to be disabled for portability
297 The use of C-style backslashes (.e.g. \n, \r) in quoted strings is not
298 SQL-spec compliant, so allow such handling to be disabled.
300 * Allow an alias to be provided for the target table in UPDATE/DELETE
302 This is not SQL-spec but many DBMSs allow it.
304 * Allow additional tables to be specified in DELETE for joins
306 UPDATE already allows this (UPDATE...FROM) but we need similar
307 functionality in DELETE. It's been agreed that the keyword should
308 be USING, to avoid anything as confusing as DELETE FROM a FROM b.
310 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
311 * Allow REINDEX to rebuild all database indexes, remove /contrib/reindex
312 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
313 * Add a schema option to createlang
314 * Allow UPDATE tab SET ROW (col, ...) = (...) for updating multiple columns
315 * Allow SET CONSTRAINTS to be qualified by schema/table name
316 * Allow TRUNCATE ... CASCADE/RESTRICT
317 * Allow PREPARE of cursors
318 * Allow PREPARE to automatically determine parameter types based on the SQL
320 * Allow finer control over the caching of prepared query plans
322 Currently, queries prepared via the libpq API are planned on first
323 execute using the supplied parameters --- allow SQL PREPARE to do the
324 same. Also, allow control over replanning prepared queries either
325 manually or automatically when statistics for execute parameters
326 differ dramatically from those used during planning.
328 * Allow LISTEN/NOTIFY to store info in memory rather than tables?
330 Currently LISTEN/NOTIFY information is stored in pg_listener. Storing
331 such information in memory would improve performance.
333 * Dump large object comments in custom dump format
334 * Add optional textual message to NOTIFY
336 This would allow an informational message to be added to the notify
337 message, perhaps indicating the row modified or other custom
340 * Allow CREATE TABLE foo (f1 INT CHECK (f1 > 0) CHECK (f1 < 10)) to work
341 by searching for non-conflicting constraint names, and prefix with
343 * Use more reliable method for CREATE DATABASE to get a consistent copy
346 Currently the system uses the operating system COPY command to create
349 * Add C code to copy directories for use in creating new databases
350 * Ignore temporary tables from other sessions when processing
352 * Have pg_ctl look at PGHOST in case it is a socket directory?
353 * Allow column-level GRANT/REVOKE privileges
354 * Add a session mode to warn about non-standard SQL usage in queries
355 * Add MERGE command that does UPDATE/DELETE, or on failure, INSERT (rules, triggers?)
356 * Add ON COMMIT capability to CREATE TABLE AS SELECT
357 * Add NOVICE output level for helpful messages like automatic sequence/index creation
358 * Add COMMENT ON for all cluster global objects (users, groups,
359 databases and tablespaces)
360 * Add an option to automatically use savepoints for each statement in a
361 multi-statement transaction.
363 When enabled, this would allow errors in multi-statement transactions
364 to be automatically ignored.
366 * Make row-wise comparisons work per SQL spec
367 * Add RESET CONNECTION command to reset all session state
369 This would include resetting of all variables (RESET ALL), dropping of
370 all temporary tables, removal of any NOTIFYs, etc. This could be used
371 for connection pooling. We could also change RESET ALL to have this
375 o Have ALTER TABLE RENAME rename SERIAL sequence names
376 o Add ALTER DOMAIN TYPE
377 o Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
378 o Allow ALTER TABLE to change constraint deferrability and actions
379 o Disallow dropping of an inherited constraint
380 o Allow objects to be moved to different schemas
381 o Allow ALTER TABLESPACE to move to different directories
382 o Allow databases and schemas to be moved to different tablespaces
384 One complexity is whether moving a schema should move all existing
385 schema objects or just define the location for future object creation.
387 o Allow moving system tables to other tablespaces, where possible
389 Currently non-global system tables must be in the default database
390 schema. Global system tables can never be moved.
394 o Automatically maintain clustering on a table
396 This might require some background daemon to maintain clustering
397 during periods of low usage. It might also require tables to be only
398 paritally filled for easier reorganization. Another idea would
399 be to create a merged heap/index data file so an index lookup would
400 automatically access the heap data too. A third idea would be to
401 store heap rows in hashed groups, perhaps using a user-supplied
404 o Add default clustering to system tables
406 To do this, determine the ideal cluster index for each system
407 table and set the cluster setting during initdb.
410 o Allow COPY to report error lines and continue
412 This requires the use of a savepoint before each COPY line is
413 processed, with ROLLBACK on COPY failure.
415 o Allow COPY to understand \x as a hex byte
416 o Have COPY return the number of rows loaded/unloaded (?)
417 o Allow COPY to optionally include column headings as the first line
420 o Allow UPDATE/DELETE WHERE CURRENT OF cursor
422 This requires using the row ctid to map cursor rows back to the
423 original heap row. This become more complicated if WITH HOLD cursors
424 are to be supported because WITH HOLD cursors have a copy of the row
425 and no FOR UPDATE lock.
427 o Prevent DROP TABLE from dropping a row referenced by its own open
430 o Allow pooled connections to list all open WITH HOLD cursors
432 Because WITH HOLD cursors exist outside transactions, this allows
433 them to be listed so they can be closed.
436 o Allow INSERT/UPDATE of the system-generated oid value for a row
437 o Allow INSERT INTO tab (col1, ..) VALUES (val1, ..), (val2, ..)
438 o Allow INSERT/UPDATE ... RETURNING new.col or old.col
440 This is useful for returning the auto-generated key for an INSERT.
441 One complication is how to handle rules that run as part of
445 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
447 o Add SET PATH for schemas (?)
449 This is basically the same as SET search_path.
451 o Prevent conflicting SET options from being set
453 This requires a checking function to be called after the server
454 configuration file is read.
456 * SERVER-SIDE LANGUAGES
457 o Allow PL/PgSQL's RAISE function to take expressions (?)
459 Currently only constants are supported.
461 o Change PL/PgSQL to use palloc() instead of malloc()
462 o Handle references to temporary tables that are created, destroyed,
463 then recreated during a session, and EXECUTE is not used
465 This requires the cached PL/PgSQL byte code to be invalidated when
466 an object referenced in the function is changed.
468 o Fix PL/pgSQL RENAME to work on variables other than OLD/NEW
469 o Improve PL/PgSQL exception handling using savepoints
470 o Allow function parameters to be passed by name,
471 get_employee_salary(emp_id => 12345, tax_year => 2001)
472 o Add Oracle-style packages
473 o Add table function support to pltcl, plperl, plpython (?)
474 o Allow PL/pgSQL to name columns by ordinal position, e.g. rec.(3)
475 o Allow PL/pgSQL EXECUTE query_var INTO record_var;
476 o Add capability to create and call PROCEDURES
477 o Allow PL/pgSQL to handle %TYPE arrays, e.g. tab.col%TYPE[]
483 * Add XML output to pg_dump and COPY
485 We already allow XML to be stored in the database, and XPath queries
486 can be used on that data using /contrib/xml2. It also supports XSLT
489 * Add a libpq function to support Parse/DescribeStatement capability
490 * Prevent libpq's PQfnumber() from lowercasing the column name (?)
491 * Allow libpq to access SQLSTATE so pg_ctl can test for connection failure
493 This would be used for checking if the server is up.
495 * Have psql show current values for a sequence
496 * Move psql backslash database information into the backend, use mnemonic
499 This would allow non-psql clients to pull the same information out of
500 the database as psql.
502 * Consistently display privilege information for all objects in psql
505 o Have pg_dump use multi-statement transactions for INSERT dumps
506 o Allow pg_dump to use multiple -t and -n switches
508 This should be done by allowing a '-t schema.table' syntax.
510 o Add dumping of comments on composite type columns
511 o Add dumping of comments on index columns
512 o Replace crude DELETE FROM method of pg_dumpall for cleaning of
513 users and groups with separate DROP commands
514 o Add dumping and restoring of LOB comments
515 o Stop dumping CASCADE on DROP TYPE commands in clean mode
516 o Add full object name to the tag field. eg. for operators we need
517 '=(integer, integer)', instead of just '='.
518 o Add pg_dumpall custom format dumps.
520 This is probably best done by combining pg_dump and pg_dumpall
521 into a single binary.
523 o Add CSV output format
528 Document differences between ecpg and the SQL standard and
529 information about the Informix-compatibility module.
531 o Solve cardinality > 1 for input descriptors / variables (?)
532 o Improve error handling (?)
533 o Add a semantic check level, e.g. check if a table really exists
534 o fix handling of DB attributes that are arrays
535 o Use backend PREPARE/EXECUTE facility for ecpg where possible
537 o Fix nested C comments
538 o sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
539 o Make SET CONNECTION thread-aware, non-standard?
540 o Allow multidimensional arrays
543 Referential Integrity
544 =====================
546 * Add MATCH PARTIAL referential integrity
547 * Add deferred trigger queue file
549 Right now all deferred trigger information is stored in backend
550 memory. This could exhaust memory for very large trigger queues.
551 This item involves dumping large queues into files.
553 * Implement dirty reads or shared row locks and use them in RI triggers
555 Adding shared locks requires recording the table/rows numbers in a
556 shared area, and this could potentially be a large amount of data.
557 One idea is to store the table/row numbers in a separate table and set
558 a bit on the row indicating looking in this new table is required to
559 find any shared row locks.
561 * Enforce referential integrity for system tables
562 * Change foreign key constraint for array -> element to mean element
564 * Allow DEFERRABLE UNIQUE constraints (?)
565 * Allow triggers to be disabled [trigger]
567 Currently the only way to disable triggers is to modify the system
570 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
572 If the dump is known to be valid, allow foreign keys to be added
573 without revalidating the data.
575 * Allow statement-level triggers to access modified rows
576 * Support triggers on columns
577 * Remove CREATE CONSTRAINT TRIGGER
579 This was used in older releases to dump referential integrity
582 * Allow AFTER triggers on system tables
584 System tables are modified in many places in the backend without going
585 through the executor and therefore not causing triggers to fire. To
586 complete this item, the functions that modify system tables will have
593 * Flush cached query plans when the dependent objects change
594 * Track dependencies in function bodies and recompile/invalidate
600 * Add SQL99 WITH clause to SELECT
601 * Add SQL99 WITH RECURSIVE to SELECT
602 * Add pre-parsing phase that converts non-ANSI syntax to supported
605 This could allow SQL written for other databases to run without
608 * Allow plug-in modules to emulate features from other databases
609 * SQL*Net listener that makes PostgreSQL appear as an Oracle database
611 * Allow queries across databases or servers with transaction
614 Right now contrib/dblink can be used to issue such queries except it
615 does not have locking or transaction semantics. Two-phase commit is
616 needed to enable transaction semantics.
618 * Add two-phase commit
620 This will involve adding a way to respond to commit failure by either
621 taking the server into offline/readonly mode or notifying the
632 * Improve commit_delay handling to reduce fsync()
633 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
634 * Allow multiple blocks to be written to WAL with one write()
635 * Add an option to sync() before fsync()'ing checkpoint files
640 * Allow free-behind capability for large sequential scans, perhaps using
643 Posix_fadvise() can control both sequential/random file caching and
644 free-behind behavior, but it is unclear how the setting affects other
645 backends that also have the file open, and the feature is not supported
646 on all operating systems.
648 * Consider use of open/fcntl(O_DIRECT) to minimize OS caching
649 * Cache last known per-tuple offsets to speed long tuple access
651 While column offsets are already cached, the cache can not be used if
652 the tuple has NULLs or TOAST columns because these values change the
653 typical column offsets. Caching of such offsets could be accomplished
654 by remembering the previous offsets and use them again if the row has
659 We could use a fixed row count and a +/- count to follow MVCC
660 visibility rules, or a single cached value could be used and
661 invalidated if anyone modifies the table.
663 * Consider automatic caching of queries at various levels:
672 * Improve speed with indexes
674 For large table adjustements during vacuum, it is faster to reindex
675 rather than update the index.
677 * Reduce lock time by moving tuples with read lock, then write
678 lock and truncate table
680 Moved tuples are invisible to other backends so they don't require a
681 write lock. However, the read lock promotion to write lock could lead
682 to deadlock situations.
684 * Allow free space map to be auto-sized or warn when it is too small
686 The free space map is in shared memory so resizing is difficult.
688 * Maintain a map of recently-expired rows
690 This allows vacuum to reclaim free space without requiring
697 * Make locking of shared data structures more fine-grained
699 This requires that more locks be acquired but this would reduce lock
700 contention, improving concurrency.
702 * Add code to detect an SMP machine and handle spinlocks accordingly
703 from distributted.net, http://www1.distributed.net/source,
704 in client/common/cpucheck.cpp
706 On SMP machines, it is possible that locks might be released shortly,
707 while on non-SMP machines, the backend should sleep so the process
708 holding the lock can complete and release it.
710 * Improve SMP performance on i386 machines
712 i386-based SMP machines can generate excessive context switching
713 caused by lock failure in high concurrency situations. This may be
714 caused by CPU cache line invalidation inefficiencies.
716 * Research use of sched_yield() for spinlock acquisition failure
722 * Experiment with multi-threaded backend [thread]
724 This would prevent the overhead associated with process creation. Most
725 operating systems have trivial process creation time compared to
726 database startup overhead, but a few operating systems (WIn32,
727 Solaris) might benefit from threading.
729 * Add connection pooling [pool]
731 It is unclear if this should be done inside the backend code or done
732 by something external like pgpool. The passing of file descriptors to
733 existing backends is one of the difficulties with a backend approach.
739 * Eliminate need to write full pages to WAL before page modification [wal]
741 Currently, to protect against partial disk page writes, we write the
742 full page images to WAL before they are modified so we can correct any
743 partial page writes during recovery. These pages can also be
744 eliminated from point-in-time archive files.
746 * Reduce WAL traffic so only modified values are written rather than
748 * Turn off after-change writes if fsync is disabled
750 If fsync is off, there is no purpose in writing full pages to WAL
752 * Add WAL index reliability improvement to non-btree indexes
753 * Allow the pg_xlog directory location to be specified during initdb
754 with a symlink back to the /data location
756 * Allow WAL information to recover corrupted pg_controldata
757 * Find a way to reduce rotational delay when repeatedly writing
760 Currently fsync of WAL requires the disk platter to perform a full
761 rotation to fsync again. One idea is to write the WAL to different
762 offsets that might reduce the rotational delay.
764 * Allow buffered WAL writes and fsync
766 Instead of guaranteeing recovery of all committed transactions, this
767 would provide improved performance by delaying WAL writes and fsync
768 so an abrupt operating system restart might lose a few seconds of
769 committed transactions but still be consistent. We could perhaps
770 remove the 'fsync' parameter (which results in an an inconsistent
771 database) in favor of this capability.
777 * Add missing optimizer selectivities for date, r-tree, etc
778 * Allow ORDER BY ... LIMIT 1 to select high/low value without sort or
779 index using a sequential scan for highest/lowest values
781 If only one value is needed, there is no need to sort the entire
782 table. Instead a sequential scan could get the matching value.
784 * Precompile SQL functions to avoid overhead
785 * Add utility to compute accurate random_page_cost value
786 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
787 * Allow sorting, temp files, temp tables to use multiple work directories
789 This allows the I/O load to be spread across multiple disk drives.
790 * Have EXPLAIN ANALYZE highlight poor optimizer estimates
791 * Use CHECK constraints to influence optimizer decisions
793 CHECK constraints contain information about the distribution of values
794 within the table. This is also useful for implementing subtables where
795 a tables content is distributed across several subtables.
796 * Consider using hash buckets to do DISTINCT, rather than sorting
798 This would be beneficial when there are few distinct values.
804 * Do async I/O for faster random read-ahead of data
806 Async I/O allows multiple I/O requests to be sent to the disk with
807 results coming back asynchronously.
809 * Use mmap() rather than SYSV shared memory or to write WAL files (?) [mmap]
811 This would remove the requirement for SYSV SHM but would introduce
812 portability issues. Anonymous mmap (or mmap to /dev/zero) is required
813 to prevent I/O overhead.
815 * Consider mmap()'ing files into a backend?
817 Doing I/O to large tables would consume a lot of address space or
818 require frequent mapping/unmapping. Extending the file also causes
819 mapping problems that might require mapping only individual pages,
820 leading to thousands of mappings. Another problem is that there is no
821 way to _prevent_ I/O to disk from the dirty shared buffers so changes
822 could hit disk before WAL is written.
824 * Add a script to ask system configuration questions and tune postgresql.conf
825 * Use a phantom command counter for nested subtransactions to reduce
827 * Consider parallel processing a single query
829 This would involve using multiple threads or processes to do optimization,
830 sorting, or execution of single query. The major advantage of such a
831 feature would be to allow multiple CPUs to work together to process a
834 * Research the use of larger pages sizes
840 * Add use of 'const' for variables in source tree
841 * Rename some /contrib modules from pg* to pg_*
842 * Move some things from /contrib into main tree
843 * Remove warnings created by -Wcast-align
844 * Move platform-specific ps status display info from ps_status.c to ports
845 * Improve access-permissions check on data directory in Cygwin (Tom)
846 * Add optional CRC checksum to heap and index pages
847 * Clarify use of 'application' and 'command' tags in SGML docs
848 * Better document ability to build only certain interfaces (Marc)
849 * Remove or relicense modules that are not under the BSD license, if possible
850 * Remove memory/file descriptor freeing before ereport(ERROR)
851 * Acquire lock on a relation before building a relcache entry for it
852 * Research interaction of setitimer() and sleep() used by statement_timeout
853 * Rename /scripts directory because they are all C programs now
854 * Promote debug_query_string into a server-side function current_query()
855 * Allow the identifier length to be increased via a configure option
856 * Allow binaries to be statically linked so they are more easily relocated
857 * Move some /contrib modules out to their own project sites
858 * Remove Win32 rename/unlink looping if unnecessary
859 * Remove kerberos4 from source tree?
862 o Remove per-backend parameter file and move into shared memory?
863 o Remove configure.in check for link failure when cause is found
864 o Remove readdir() errno patch when runtime/mingwex/dirent.c rev
866 o Remove psql newline patch when we find out why mingw outputs an
868 o Allow psql to use readline once non-US code pages work with
870 o Re-enable timezone output on log_line_prefix '%t' when a
871 shorter timezone string is available
873 * Wire Protocol Changes
874 o Allow dynamic character set handling
875 o Add decoded type, length, precision
877 o Update clients to use data types, typmod, schema.table.column names of
878 result sets using new query protocol
881 ---------------------------------------------------------------------------
884 Developers who have claimed items are:
885 --------------------------------------
886 * Alvaro is Alvaro Herrera <alvherre@dcc.uchile.cl>
887 * Andrew is Andrew Dunstan <andrew@dunslane.net>
888 * Bruce is Bruce Momjian <pgman@candle.pha.pa.us> of Software Research Assoc.
889 * Christopher is Christopher Kings-Lynne <chriskl@familyhealth.com.au> of
890 Family Health Network
891 * Claudio is Claudio Natoli <claudio.natoli@memetrics.com>
892 * D'Arcy is D'Arcy J.M. Cain <darcy@druid.net> of The Cain Gang Ltd.
893 * Fabien is Fabien Coelho <coelho@cri.ensmp.fr>
894 * Gavin is Gavin Sherry <swm@linuxworld.com.au> of Alcove Systems Engineering
895 * Greg is Greg Sabino Mullane <greg@turnstep.com>
896 * Hiroshi is Hiroshi Inoue <Inoue@tpf.co.jp>
897 * Jan is Jan Wieck <JanWieck@Yahoo.com> of Afilias, Inc.
898 * Joe is Joe Conway <mail@joeconway.com>
899 * Karel is Karel Zak <zakkr@zf.jcu.cz>
901 * Magnus is Magnus Hagander <mha@sollentuna.net>
902 * Marc is Marc Fournier <scrappy@hub.org> of PostgreSQL, Inc.
903 * Matthew T. O'Connor <matthew@zeut.net>
904 * Michael is Michael Meskes <meskes@postgresql.org> of Credativ
905 * Neil is Neil Conway <neilc@samurai.com>
906 * Oleg is Oleg Bartunov <oleg@sai.msu.su>
907 * Peter is Peter Eisentraut <peter_e@gmx.net>
908 * Philip is Philip Warner <pjw@rhyme.com.au> of Albatross Consulting Pty. Ltd.
909 * Rod is Rod Taylor <pg@rbt.ca>
910 * Simon is Simon Riggs
911 * Stephan is Stephan Szabo <sszabo@megazone23.bigpanda.com>
912 * Tatsuo is Tatsuo Ishii <t-ishii@sra.co.jp> of Software Research Assoc.
914 * Tom is Tom Lane <tgl@sss.pgh.pa.us> of Red Hat