4 Current maintainer: Bruce Momjian (pgman@candle.pha.pa.us)
5 Last updated: Fri Sep 16 13:37:02 EDT 2005
7 The most recent version of this document can be viewed at
8 http://www.postgresql.org/docs/faqs.TODO.html.
10 #A hyphen, "-", marks changes that will appear in the upcoming 8.1 release.#
11 #A percent sign, "%", marks items that are easier to implement.#
13 Bracketed items, "[]", have more detail.
15 This list contains all known PostgreSQL bugs and feature requests. If
16 you would like to work on an item, please read the Developer's FAQ
23 * %Remove behavior of postmaster -o after making postmaster/postgres
25 * %Allow pooled connections to list all prepared queries
27 This would allow an application inheriting a pooled connection to know
28 the queries prepared in the current session.
30 * Allow major upgrades without dump/reload, perhaps using pg_upgrade
32 * Check for unreferenced table files created by transactions that were
33 in-progress when the server terminated abruptly
34 * Allow administrators to safely terminate individual sessions either
35 via an SQL function or SIGTERM
37 Lock table corruption following SIGTERM of an individual backend
38 has been reported in 8.0. A possible cause was fixed in 8.1, but
39 it is unknown whether other problems exist. This item mostly
40 requires additional testing rather than of writing any new code.
42 * %Set proper permissions on non-system schemas during db creation
44 Currently all schemas are owned by the super-user because they are
45 copied from the template1 database.
47 * Support table partitioning that allows a single table to be stored
48 in subtables that are partitioned based on the primary key or a WHERE
52 * Improve replication solutions
56 You can use any of the master/slave replication servers to use a
57 standby server for data warehousing. To allow read/write queries to
58 multiple servers, you need multi-master replication like pgcluster.
60 o Allow replication over unreliable or non-persistent links
65 o %Add "include file" functionality in postgresql.conf
66 o %Allow commenting of variables in postgresql.conf to restore them
69 Currently, if a variable is commented out, it keeps the
70 previous uncommented value until a server restarted.
72 o %Allow pg_hba.conf settings to be controlled via SQL
74 This would add a function to load the SQL table from
75 pg_hba.conf, and one to writes its contents to the flat file.
76 The table should have a line number that is a float so rows
77 can be inserted between existing rows, e.g. row 2.5 goes
78 between row 2 and row 3.
80 o %Allow postgresql.conf file values to be changed via an SQL
81 API, perhaps using SET GLOBAL
82 o Allow the server to be stopped/restarted via an SQL API
83 o Issue a warning if a change-on-restart-only postgresql.conf value
84 is modified and the server config files are reloaded
85 o Mark change-on-restart-only values in postgresql.conf
90 * Allow a database in tablespace t1 with tables created in
91 tablespace t2 to be used as a template for a new database created
92 with default tablespace t2
94 All objects in the default database tablespace must have default
95 tablespace specifications. This is because new databases are
96 created by copying directories. If you mix default tablespace
97 tables and tablespace-specified tables in the same directory,
98 creating a new database from such a mixed directory would create a
99 new database with tables that had incorrect explicit tablespaces.
100 To fix this would require modifying pg_class in the newly copied
101 database, which we don't currently do.
103 * Allow reporting of which objects are in which tablespaces
105 This item is difficult because a tablespace can contain objects
106 from multiple databases. There is a server-side function that
107 returns the databases which use a specific tablespace, so this
108 requires a tool that will call that function and connect to each
109 database to find the objects in each database for that tablespace.
111 o %Add a GUC variable to control the tablespace for temporary objects
114 It could start with a random tablespace from a supplied list and
115 cycle through the list.
117 o Allow WAL replay of CREATE TABLESPACE to work when the directory
118 structure on the recovery computer is different from the original
120 o Allow per-tablespace quotas
123 * Point-In-Time Recovery (PITR)
125 o Allow point-in-time recovery to archive partially filled
126 write-ahead logs [pitr]
128 Currently only full WAL files are archived. This means that the
129 most recent transactions aren't available for recovery in case
130 of a disk failure. This could be triggered by a user command or
133 o Automatically force archiving of partially-filled WAL files when
134 pg_stop_backup() is called or the server is stopped
136 Doing this will allow administrators to know more easily when
137 the archive contins all the files needed for point-in-time
140 o %Create dump tool for write-ahead logs for use in determining
141 transaction id for point-in-time recovery
142 o Allow a warm standby system to also allow read-only queries
145 This is useful for checking PITR recovery.
147 o Allow the PITR process to be debugged and data examined
153 * Allow server log information to be output as INSERT statements
155 This would allow server log information to be easily loaded into
156 a database for analysis.
158 * %Add ability to monitor the use of temporary sort files
159 * Allow server logs to be remotely read and removed using SQL commands
165 * Improve the MONEY data type
167 Change the MONEY data type to use DECIMAL internally, with special
168 locale-aware output formatting.
170 * Change NUMERIC to enforce the maximum precision, and increase it
171 * Add NUMERIC division operator that doesn't round?
173 Currently NUMERIC _rounds_ the result to the specified precision.
174 This means division can return a result that multiplied by the
175 divisor is greater than the dividend, e.g. this returns a value > 10:
177 SELECT (10::numeric(2,0) / 6::numeric(2,0))::numeric(2,0) * 6;
179 The positive modulus result returned by NUMERICs might be considered
180 inaccurate, in one sense.
182 * Have sequence dependency track use of DEFAULT sequences,
184 * %Disallow changing default expression of a SERIAL column?
185 * Fix data types where equality comparison isn't intuitive, e.g. box
186 * %Prevent INET cast to CIDR if the unmasked bits are not zero, or
188 * %Prevent INET cast to CIDR from droping netmask, SELECT '1.1.1.1'::inet::cidr
189 * Allow INET + INT4 to increment the host part of the address, or
190 throw an error on overflow
191 * %Add 'tid != tid ' operator for use in corruption recovery
192 * Allow user-defined types to specify a type modifier at table creation
198 o Allow infinite dates just like infinite timestamps
199 o Add a GUC variable to allow output of interval values in ISO8601
201 o Merge hardwired timezone names with the TZ database; allow either
202 kind everywhere a TZ name is currently taken
203 o Allow customization of the known set of TZ names (generalize the
204 present australian_timezones hack)
205 o Allow TIMESTAMP WITH TIME ZONE to store the original timezone
206 information, either zone name or offset from UTC [timezone]
208 If the TIMESTAMP value is stored with a time zone name, interval
209 computations should adjust based on the time zone rules.
211 o Fix SELECT '0.01 years'::interval, '0.01 months'::interval
212 o Add ISO INTERVAL handling
213 o Add support for day-time syntax, INTERVAL '1 2:03:04' DAY TO
215 o Add support for year-month syntax, INTERVAL '50-6' YEAR TO MONTH
216 o For syntax that isn't uniquely ISO or PG syntax, like '1:30' or
217 '1', treat as ISO if there is a range specification clause,
218 and as PG if there no clause is present, e.g. interpret
219 '1:30' MINUTE TO SECOND as '1 minute 30 seconds', and
220 interpret '1:30' as '1 hour, 30 minutes'
221 o Interpret INTERVAL '1 year' MONTH as CAST (INTERVAL '1 year' AS
222 INTERVAL MONTH), and this should return '12 months'
223 o Round or truncate values to the requested precision, e.g.
224 INTERVAL '11 months' AS YEAR should return one or zero
225 o Support precision, CREATE TABLE foo (a INTERVAL MONTH(3))
230 o Allow NULLs in arrays
231 o Delay resolution of array expression's data type so assignment
232 coercion can be performed on empty array expressions
237 o Improve vacuum of large objects, like /contrib/vacuumlo?
238 o Add security checking for large objects
239 o Auto-delete large objects when referencing row is deleted
241 /contrib/lo offers this functionality.
243 o Allow read/write into TOAST values like large objects
245 This requires the TOAST column to be stored EXTERNAL.
251 * Allow INET subnet tests using non-constants to be indexed
252 * Add transaction_timestamp(), statement_timestamp(), clock_timestamp()
255 Current CURRENT_TIMESTAMP returns the start time of the current
256 transaction, and gettimeofday() returns the wallclock time. This will
257 make time reporting more consistent and will allow reporting of
258 the statement start time.
260 * %Add pg_get_acldef(), pg_get_typedefault(), and pg_get_attrdef()
261 * Allow to_char() to print localized month names
262 * Allow functions to have a schema search path specified at creation time
263 * Allow substring/replace() to get/set bit values
264 * Allow to_char() on interval values to accumulate the highest unit
267 Some special format flag would be required to request such
268 accumulation. Such functionality could also be added to EXTRACT.
269 Prevent accumulation that crosses the month/day boundary because of
270 the uneven number of days in a month.
272 o to_char(INTERVAL '1 hour 5 minutes', 'MI') => 65
273 o to_char(INTERVAL '43 hours 20 minutes', 'MI' ) => 2600
274 o to_char(INTERVAL '43 hours 20 minutes', 'WK:DD:HR:MI') => 0:1:19:20
275 o to_char(INTERVAL '3 years 5 months','MM') => 41
277 * Add sleep() function, remove from regress.c
280 Multi-Language Support
281 ======================
283 * Add NCHAR (as distinguished from ordinary varchar),
284 * Allow locale to be set at database creation
286 Currently locale can only be set during initdb. No global tables have
287 locale-aware columns. However, the database template used during
288 database creation might have locale-aware indexes. The indexes would
289 need to be reindexed to match the new locale.
291 * Allow encoding on a per-column basis
293 Right now only one encoding is allowed per database.
295 * Support multiple simultaneous character sets, per SQL92
296 * Improve UTF8 combined character handling?
297 * Add octet_length_server() and octet_length_client()
298 * Make octet_length_client() the same as octet_length()?
299 * Fix problems with wrong runtime encoding conversion for NLS message files
305 * %Automatically create rules on views so they are updateable, per SQL99
307 We can only auto-create rules for simple views. For more complex
308 cases users will still have to write rules.
310 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
311 * Allow NOTIFY in rules involving conditionals
312 * Allow VIEW/RULE recompilation when the underlying tables change
314 Another issue is whether underlying table changes should be reflected
315 in the view, e.g. should SELECT * show additional columns if they
316 are added after the view is created.
322 * Change LIMIT/OFFSET and FETCH/MOVE to use int8
323 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
324 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
325 * %Allow SET CONSTRAINTS to be qualified by schema/table name
326 * %Allow TRUNCATE ... CASCADE/RESTRICT
328 This is like DELETE CASCADE, but truncates.
330 * %Add a separate TRUNCATE permission
332 Currently only the owner can TRUNCATE a table because triggers are not
333 called, and the table is locked in exclusive mode.
335 * Allow PREPARE of cursors
336 * Allow PREPARE to automatically determine parameter types based on the SQL
338 * Allow finer control over the caching of prepared query plans
340 Currently, queries prepared via the libpq API are planned on first
341 execute using the supplied parameters --- allow SQL PREPARE to do the
342 same. Also, allow control over replanning prepared queries either
343 manually or automatically when statistics for execute parameters
344 differ dramatically from those used during planning.
346 * Allow LISTEN/NOTIFY to store info in memory rather than tables?
348 Currently LISTEN/NOTIFY information is stored in pg_listener. Storing
349 such information in memory would improve performance.
351 * Add optional textual message to NOTIFY
353 This would allow an informational message to be added to the notify
354 message, perhaps indicating the row modified or other custom
357 * Add a GUC variable to warn about non-standard SQL usage in queries
358 * Add MERGE command that does UPDATE/DELETE, or on failure, INSERT (rules,
360 * Add NOVICE output level for helpful messages like automatic sequence/index
362 * %Add COMMENT ON for all cluster global objects (roles, databases
364 * %Make row-wise comparisons work per SQL spec
365 * Add RESET CONNECTION command to reset all session state
367 This would include resetting of all variables (RESET ALL), dropping of
368 temporary tables, removing any NOTIFYs, cursors, open transactions,
369 prepared queries, currval()s, etc. This could be used for connection
370 pooling. We could also change RESET ALL to have this functionality.
371 The difficult of this features is allowing RESET ALL to not affect
372 changes made by the interface driver for its internal use. One idea
373 is for this to be a protocol-only feature. Another approach is to
374 notify the protocol when a RESET CONNECTION command is used.
376 * Add GUC to issue notice about queries that use unjoined tables
377 * Allow EXPLAIN to identify tables that were skipped because of
379 * Allow EXPLAIN output to be more easily processed by scripts
380 * Eventually enable escape_string_warning and standard_conforming_strings
385 o Allow CREATE TABLE AS to determine column lengths for complex
386 expressions like SELECT col1 || col2
387 o Use more reliable method for CREATE DATABASE to get a consistent
389 o Add ON COMMIT capability to CREATE TABLE AS ... SELECT
393 o Allow UPDATE to handle complex aggregates [update]?
394 o Allow an alias to be provided for the target table in
397 This is not SQL-spec but many DBMSs allow it.
399 o Allow UPDATE tab SET ROW (col, ...) = (...) for updating multiple
405 o %Have ALTER TABLE RENAME rename SERIAL sequence names
406 o Add ALTER DOMAIN to modify the underlying data type
407 o %Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
408 o %Allow ALTER TABLE to change constraint deferrability and actions
409 o Add missing object types for ALTER ... SET SCHEMA
410 o Allow ALTER TABLESPACE to move to different directories
411 o Allow databases to be moved to different tablespaces
412 o Allow moving system tables to other tablespaces, where possible
414 Currently non-global system tables must be in the default database
415 tablespace. Global system tables can never be moved.
417 o %Disallow dropping of an inherited constraint
418 o %Prevent child tables from altering or dropping constraints
419 like CHECK that were inherited from the parent table
424 o Automatically maintain clustering on a table
426 This might require some background daemon to maintain clustering
427 during periods of low usage. It might also require tables to be only
428 paritally filled for easier reorganization. Another idea would
429 be to create a merged heap/index data file so an index lookup would
430 automatically access the heap data too. A third idea would be to
431 store heap rows in hashed groups, perhaps using a user-supplied
434 o %Add default clustering to system tables
436 To do this, determine the ideal cluster index for each system
437 table and set the cluster setting during initdb.
442 o Allow COPY to report error lines and continue
444 This requires the use of a savepoint before each COPY line is
445 processed, with ROLLBACK on COPY failure.
447 o %Have COPY return the number of rows loaded/unloaded?
448 o Allow COPY on a newly-created table to skip WAL logging
450 On crash recovery, the table involved in the COPY would
451 be removed or have its heap and index files truncated. One
452 issue is that no other backend should be able to add to
453 the table at the same time, which is something that is
458 o Allow column-level privileges
459 o %Allow GRANT/REVOKE permissions to be applied to all schema objects
462 The proposed syntax is:
463 GRANT SELECT ON ALL TABLES IN public TO phpuser;
464 GRANT SELECT ON NEW TABLES IN public TO phpuser;
466 * Allow GRANT/REVOKE permissions to be inherited by objects based on
472 o Allow UPDATE/DELETE WHERE CURRENT OF cursor
474 This requires using the row ctid to map cursor rows back to the
475 original heap row. This become more complicated if WITH HOLD cursors
476 are to be supported because WITH HOLD cursors have a copy of the row
477 and no FOR UPDATE lock.
479 o Prevent DROP TABLE from dropping a row referenced by its own open
482 o %Allow pooled connections to list all open WITH HOLD cursors
484 Because WITH HOLD cursors exist outside transactions, this allows
485 them to be listed so they can be closed.
490 o Allow INSERT/UPDATE of the system-generated oid value for a row
491 o Allow INSERT INTO tab (col1, ..) VALUES (val1, ..), (val2, ..)
492 o Allow INSERT/UPDATE ... RETURNING new.col or old.col
494 This is useful for returning the auto-generated key for an INSERT.
495 One complication is how to handle rules that run as part of
501 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
503 o Add SET PATH for schemas?
505 This is basically the same as SET search_path.
508 * Server-Side Languages
510 o Fix PL/pgSQL RENAME to work on variables other than OLD/NEW
511 o Allow function parameters to be passed by name,
512 get_employee_salary(emp_id => 12345, tax_year => 2001)
513 o Add Oracle-style packages
514 o Add table function support to pltcl, plpython
515 o Add capability to create and call PROCEDURES
516 o Allow PL/pgSQL to handle %TYPE arrays, e.g. tab.col%TYPE[]
517 o Allow function argument names to be queries from PL/PgSQL
518 o Add MOVE to PL/pgSQL
519 o Add support for polymorphic arguments and return types to
520 languages other than PL/PgSQL
521 o Add support for OUT and INOUT parameters to languages other
523 o Add single-step debugging of PL/PgSQL functions
529 * Add a libpq function to support Parse/DescribeStatement capability
530 * Prevent libpq's PQfnumber() from lowercasing the column name?
531 * Add PQescapeIdentifier() to libpq
532 * Have initdb set the input DateStyle (MDY or DMY) based on locale?
533 * Have pg_ctl look at PGHOST in case it is a socket directory?
534 * Allow pg_ctl to work properly with configuration files located outside
537 pg_ctl can not read the pid file because it isn't located in the
538 config directory but in the PGDATA directory. The solution is to
539 allow pg_ctl to read and understand postgresql.conf to find the
540 data_directory value.
545 o Have psql show current values for a sequence
546 o Move psql backslash database information into the backend, use
547 mnemonic commands? [psql]
549 This would allow non-psql clients to pull the same information out
550 of the database as psql.
552 o Fix psql's display of schema information (Neil)
553 o Allow psql \pset boolean variables to set to fixed values, rather
555 o Consistently display privilege information for all objects in psql
556 o Improve psql's handling of multi-line queries
558 Currently, while \e saves a single query as one entry, interactive
559 queries are saved one line at a time. Ideally all queries
560 whould be saved like \e does.
562 o Allow multi-line column values to align in the proper columns
564 If the second output column value is 'a\nb', the 'b' should appear
565 in the second display column, rather than the first column as it
571 o %Have pg_dump use multi-statement transactions for INSERT dumps
572 o %Allow pg_dump to use multiple -t and -n switches [pg_dump]
573 o %Add dumping of comments on composite type columns
574 o %Add dumping of comments on index columns
575 o %Replace crude DELETE FROM method of pg_dumpall --clean for
576 cleaning of roles with separate DROP commands
577 o Stop dumping CASCADE on DROP TYPE commands in clean mode
578 o %Add full object name to the tag field. eg. for operators we need
579 '=(integer, integer)', instead of just '='.
580 o Add pg_dumpall custom format dumps?
581 o %Add CSV output format
582 o Update pg_dump and psql to use the new COPY libpq API (Christopher)
583 o Remove unnecessary function pointer abstractions in pg_dump source
591 Document differences between ecpg and the SQL standard and
592 information about the Informix-compatibility module.
594 o Solve cardinality > 1 for input descriptors / variables?
595 o Add a semantic check level, e.g. check if a table really exists
596 o fix handling of DB attributes that are arrays
597 o Use backend PREPARE/EXECUTE facility for ecpg where possible
599 o Fix nested C comments
600 o %sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
601 o Make SET CONNECTION thread-aware, non-standard?
602 o Allow multidimensional arrays
603 o Add internationalized message strings
606 Referential Integrity
607 =====================
609 * Add MATCH PARTIAL referential integrity
610 * Add deferred trigger queue file
612 Right now all deferred trigger information is stored in backend
613 memory. This could exhaust memory for very large trigger queues.
614 This item involves dumping large queues into files.
616 * Change foreign key constraint for array -> element to mean element
618 * Allow DEFERRABLE UNIQUE constraints?
619 * Allow triggers to be disabled in only the current session.
621 This is currently possible by starting a multi-statement transaction,
622 modifying the system tables, performing the desired SQL, restoring the
623 system tables, and committing the transaction. ALTER TABLE ...
624 TRIGGER requires a table lock so it is not ideal for this usage.
626 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
628 If the dump is known to be valid, allow foreign keys to be added
629 without revalidating the data.
631 * Allow statement-level triggers to access modified rows
632 * Support triggers on columns (Greg Sabino Mullane)
633 * Enforce referential integrity for system tables
634 * Allow AFTER triggers on system tables
636 System tables are modified in many places in the backend without going
637 through the executor and therefore not causing triggers to fire. To
638 complete this item, the functions that modify system tables will have
645 * Flush cached query plans when the dependent objects change
646 * Track dependencies in function bodies and recompile/invalidate
648 This is particularly important for references to temporary tables
649 in PL/PgSQL because PL/PgSQL caches query plans. The only workaround
650 in PL/PgSQL is to use EXECUTE. One complexity is that a function
651 might itself drop and recreate dependent tables, causing it to
652 invalidate its own query plan.
658 * Add SQL99 WITH clause to SELECT
659 * Add SQL99 WITH RECURSIVE to SELECT
660 * Add pre-parsing phase that converts non-ISO syntax to supported
663 This could allow SQL written for other databases to run without
666 * Allow plug-in modules to emulate features from other databases
667 * SQL*Net listener that makes PostgreSQL appear as an Oracle database
669 * Allow queries across databases or servers with transaction
672 This can be done using dblink and two-phase commit.
674 * Add the features of packages
676 o Make private objects accessable only to objects in the same schema
677 o Allow current_schema.objname to access current schema objects
678 o Add session variables
679 o Allow nested schemas
685 * Allow inherited tables to inherit index, UNIQUE constraint, and primary
687 * UNIQUE INDEX on base column not honored on INSERTs/UPDATEs from
688 inherited table: INSERT INTO inherit_table (unique_index_col) VALUES
691 The main difficulty with this item is the problem of creating an index
692 that can span more than one table.
694 * Allow SELECT ... FOR UPDATE on inherited tables
695 * Add UNIQUE capability to non-btree indexes
696 * Prevent index uniqueness checks when UPDATE does not modify the column
698 Uniqueness (index) checks are done when updating a column even if the
699 column is not modified by the UPDATE.
701 * Allow the creation of on-disk bitmap indexes which can be quickly
702 combined with other bitmap indexes
704 Such indexes could be more compact if there are only a few distinct values.
705 Such indexes can also be compressed. Keeping such indexes updated can be
708 * Allow use of indexes to search for NULLs
710 One solution is to create a partial index on an IS NULL expression.
712 * Allow accurate statistics to be collected on indexes with more than
713 one column or expression indexes, perhaps using per-index statistics
714 * Add fillfactor to control reserved free space during index creation
715 * Allow the creation of indexes with mixed ascending/descending specifiers
716 * Allow constraint_exclusion to work for UNIONs like it does for
717 inheritance, allow it to work for UPDATE and DELETE queries, and allow
718 it to be used for all queries with little performance impact
723 o Add more GIST index support for geometric data types
724 o Allow GIST indexes to create certain complex index types, like
725 digital trees (see Aoki)
729 o Pack hash index buckets onto disk pages more efficiently
731 Currently only one hash bucket can be stored on a page. Ideally
732 several hash buckets could be stored on a single page and greater
733 granularity used for the hash algorithm.
735 o Consider sorting hash buckets so entries can be found using a
736 binary search, rather than a linear scan
738 o In hash indexes, consider storing the hash value with or instead
741 o Add WAL logging for crash recovery
742 o Allow multi-column hash indexes
748 * Improve commit_delay handling to reduce fsync()
749 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
751 Ideally this requires a separate test program that can be run
752 at initdb time or optionally later.
754 * %Add an option to sync() before fsync()'ing checkpoint files
755 * Add program to test if fsync has a delay compared to non-fsync
761 * Allow free-behind capability for large sequential scans, perhaps using
764 Posix_fadvise() can control both sequential/random file caching and
765 free-behind behavior, but it is unclear how the setting affects other
766 backends that also have the file open, and the feature is not supported
767 on all operating systems.
771 We could use a fixed row count and a +/- count to follow MVCC
772 visibility rules, or a single cached value could be used and
773 invalidated if anyone modifies the table. Another idea is to
774 get a count directly from a unique index, but for this to be
775 faster than a sequential scan it must avoid access to the heap
776 to obtain tuple visibility information.
778 * Allow data to be pulled directly from indexes
780 Currently indexes do not have enough tuple visibility information
781 to allow data to be pulled from the index without also accessing
782 the heap. One way to allow this is to set a bit to index tuples
783 to indicate if a tuple is currently visible to all transactions
784 when the first valid heap lookup happens. This bit would have to
785 be cleared when a heap tuple is expired.
788 * Consider automatic caching of queries at various levels:
794 * Allow sequential scans to take advantage of other concurrent
795 sequentiqal scans, also called "Synchronised Scanning"
797 One possible implementation is to start sequential scans from the lowest
798 numbered buffer in the shared cache, and when reaching the end wrap
799 around to the beginning, rather than always starting sequential scans
800 at the start of the table.
806 * Improve speed with indexes
808 For large table adjustements during VACUUM FULL, it is faster to
809 reindex rather than update the index.
811 * Reduce lock time during VACUUM FULL by moving tuples with read lock,
812 then write lock and truncate table
814 Moved tuples are invisible to other backends so they don't require a
815 write lock. However, the read lock promotion to write lock could lead
816 to deadlock situations.
818 * Auto-fill the free space map by scanning the buffer cache or by
819 checking pages written by the background writer
820 * Create a bitmap of pages that need vacuuming
822 Instead of sequentially scanning the entire table, have the background
823 writer or some other process record pages that have expired rows, then
824 VACUUM can look at just those pages rather than the entire table. In
825 the event of a system crash, the bitmap would probably be invalidated.
826 One complexity is that index entries still have to be vacuumed, and
827 doing this without an index scan (by using the heap values to find the
828 index entry) might be slow and unreliable, especially for user-defined
831 * %Add system view to show free space map contents
836 o Use free-space map information to guide refilling
837 o %Issue log message to suggest VACUUM FULL if a table is nearly
839 o Improve xid wraparound detection by recording per-table rather
846 * Add code to detect an SMP machine and handle spinlocks accordingly
847 from distributted.net, http://www1.distributed.net/source,
848 in client/common/cpucheck.cpp
850 On SMP machines, it is possible that locks might be released shortly,
851 while on non-SMP machines, the backend should sleep so the process
852 holding the lock can complete and release it.
854 * Research use of sched_yield() for spinlock acquisition failure
855 * Fix priority ordering of read and write light-weight locks (Neil)
858 Startup Time Improvements
859 =========================
861 * Experiment with multi-threaded backend [thread]
863 This would prevent the overhead associated with process creation. Most
864 operating systems have trivial process creation time compared to
865 database startup overhead, but a few operating systems (WIn32,
866 Solaris) might benefit from threading. Also explore the idea of
867 a single session using multiple threads to execute a query faster.
869 * Add connection pooling
871 It is unclear if this should be done inside the backend code or done
872 by something external like pgpool. The passing of file descriptors to
873 existing backends is one of the difficulties with a backend approach.
879 * Eliminate need to write full pages to WAL before page modification [wal]
881 Currently, to protect against partial disk page writes, we write
882 full page images to WAL before they are modified so we can correct any
883 partial page writes during recovery. These pages can also be
884 eliminated from point-in-time archive files.
886 o When off, write CRC to WAL and check file system blocks
889 If CRC check fails during recovery, remember the page in case
890 a later CRC for that page properly matches.
892 o Write full pages during file system write and not when
893 the page is modified in the buffer cache
895 This allows most full page writes to happen in the background
896 writer. It might cause problems for applying WAL on recovery
897 into a partially-written page, but later the full page will be
900 * Reduce WAL traffic so only modified values are written rather than
902 * Allow the pg_xlog directory location to be specified during initdb
903 with a symlink back to the /data location
904 * Allow WAL information to recover corrupted pg_controldata
905 * Find a way to reduce rotational delay when repeatedly writing
908 Currently fsync of WAL requires the disk platter to perform a full
909 rotation to fsync again. One idea is to write the WAL to different
910 offsets that might reduce the rotational delay.
912 * Allow buffered WAL writes and fsync
914 Instead of guaranteeing recovery of all committed transactions, this
915 would provide improved performance by delaying WAL writes and fsync
916 so an abrupt operating system restart might lose a few seconds of
917 committed transactions but still be consistent. We could perhaps
918 remove the 'fsync' parameter (which results in an an inconsistent
919 database) in favor of this capability.
925 * Add missing optimizer selectivities for date, r-tree, etc
926 * Allow ORDER BY ... LIMIT # to select high/low value without sort or
927 index using a sequential scan for highest/lowest values
929 Right now, if no index exists, ORDER BY ... LIMIT # requires we sort
930 all values to return the high/low value. Instead The idea is to do a
931 sequential scan to find the high/low value, thus avoiding the sort.
932 MIN/MAX already does this, but not for LIMIT > 1.
934 * Precompile SQL functions to avoid overhead
935 * Create utility to compute accurate random_page_cost value
936 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
937 * Have EXPLAIN ANALYZE highlight poor optimizer estimates
938 * Consider using hash buckets to do DISTINCT, rather than sorting
940 This would be beneficial when there are few distinct values.
942 * Log queries where the optimizer row estimates were dramatically
943 different from the number of rows actually found?
946 Miscellaneous Performance
947 =========================
949 * Do async I/O for faster random read-ahead of data
951 Async I/O allows multiple I/O requests to be sent to the disk with
952 results coming back asynchronously.
954 * Use mmap() rather than SYSV shared memory or to write WAL files?
956 This would remove the requirement for SYSV SHM but would introduce
957 portability issues. Anonymous mmap (or mmap to /dev/zero) is required
958 to prevent I/O overhead.
960 * Consider mmap()'ing files into a backend?
962 Doing I/O to large tables would consume a lot of address space or
963 require frequent mapping/unmapping. Extending the file also causes
964 mapping problems that might require mapping only individual pages,
965 leading to thousands of mappings. Another problem is that there is no
966 way to _prevent_ I/O to disk from the dirty shared buffers so changes
967 could hit disk before WAL is written.
969 * Add a script to ask system configuration questions and tune postgresql.conf
970 * Merge xmin/xmax/cmin/cmax back into three header fields
972 Before subtransactions, there used to be only three fields needed to
973 store these four values. This was possible because only the current
974 transaction looks at the cmin/cmax values. If the current transaction
975 created and expired the row the fields stored where xmin (same as
976 xmax), cmin, cmax, and if the transaction was expiring a row from a
977 another transaction, the fields stored were xmin (cmin was not
978 needed), xmax, and cmax. Such a system worked because a transaction
979 could only see rows from another completed transaction. However,
980 subtransactions can see rows from outer transactions, and once the
981 subtransaction completes, the outer transaction continues, requiring
982 the storage of all four fields. With subtransactions, an outer
983 transaction can create a row, a subtransaction expire it, and when the
984 subtransaction completes, the outer transaction still has to have
985 proper visibility of the row's cmin, for example, for cursors.
987 One possible solution is to create a phantom cid which represents a
988 cmin/cmax pair and is stored in local memory. Another idea is to
989 store both cmin and cmax only in local memory.
991 * Research storing disk pages with no alignment/padding
996 * Add use of 'const' for variables in source tree
997 * Rename some /contrib modules from pg* to pg_*
998 * Move some things from /contrib into main tree
999 * Move some /contrib modules out to their own project sites
1000 * %Remove warnings created by -Wcast-align
1001 * Move platform-specific ps status display info from ps_status.c to ports
1002 * Add optional CRC checksum to heap and index pages
1003 * Improve documentation to build only interfaces (Marc)
1004 * Remove or relicense modules that are not under the BSD license, if possible
1005 * %Remove memory/file descriptor freeing before ereport(ERROR)
1006 * Acquire lock on a relation before building a relcache entry for it
1007 * %Promote debug_query_string into a server-side function current_query()
1008 * %Allow the identifier length to be increased via a configure option
1009 * Remove Win32 rename/unlink looping if unnecessary
1010 * Allow cross-compiling by generating the zic database on the target system
1011 * Improve NLS maintenace of libpgport messages linked onto applications
1012 * Allow ecpg to work with MSVC and BCC
1013 * Add xpath_array() to /contrib/xml2 to return results as an array
1014 * Allow building in directories containing spaces
1016 This is probably not possible because 'gmake' and other compiler tools
1017 do not fully support quoting of paths with spaces.
1019 * Allow installing to directories containing spaces
1021 This is possible if proper quoting is added to the makefiles for the
1022 install targets. Because PostgreSQL supports relocatable installs, it
1023 is already possible to install into a directory that doesn't contain
1024 spaces and then copy the install to a directory with spaces.
1026 * Fix sgmltools so PDFs can be generated with bookmarks
1027 * %Clean up compiler warnings (especially with gcc version 4)
1028 * Add function to return the thread safety status of libpq and ecpg
1029 * Use UTF8 encoding for NLS messages so all server encodings can
1031 * Update Bonjour to work with newer cross-platform SDK
1035 o Remove configure.in check for link failure when cause is found
1036 o Remove readdir() errno patch when runtime/mingwex/dirent.c rev
1038 o Remove psql newline patch when we find out why mingw outputs an
1040 o Allow psql to use readline once non-US code pages work with
1042 o Re-enable timezone output on log_line_prefix '%t' when a
1043 shorter timezone string is available
1044 o Fix problem with shared memory on the Win32 Terminal Server
1047 * Wire Protocol Changes
1049 o Allow dynamic character set handling
1050 o Add decoded type, length, precision
1052 o Update clients to use data types, typmod, schema.table.column names
1053 of result sets using new query protocol
1056 ---------------------------------------------------------------------------
1059 Developers who have claimed items are:
1060 --------------------------------------
1061 * Alvaro is Alvaro Herrera <alvherre@dcc.uchile.cl>
1062 * Andrew is Andrew Dunstan <andrew@dunslane.net>
1063 * Bruce is Bruce Momjian <pgman@candle.pha.pa.us> of Software Research Assoc.
1064 * Christopher is Christopher Kings-Lynne <chriskl@familyhealth.com.au> of
1065 Family Health Network
1066 * Claudio is Claudio Natoli <claudio.natoli@memetrics.com>
1067 * D'Arcy is D'Arcy J.M. Cain <darcy@druid.net> of The Cain Gang Ltd.
1068 * Fabien is Fabien Coelho <coelho@cri.ensmp.fr>
1069 * Gavin is Gavin Sherry <swm@linuxworld.com.au> of Alcove Systems Engineering
1070 * Greg is Greg Sabino Mullane <greg@turnstep.com>
1071 * Hiroshi is Hiroshi Inoue <Inoue@tpf.co.jp>
1072 * Jan is Jan Wieck <JanWieck@Yahoo.com> of Afilias, Inc.
1073 * Joe is Joe Conway <mail@joeconway.com>
1074 * Karel is Karel Zak <zakkr@zf.jcu.cz>
1075 * Magnus is Magnus Hagander <mha@sollentuna.net>
1076 * Marc is Marc Fournier <scrappy@hub.org> of PostgreSQL, Inc.
1077 * Matthew T. O'Connor <matthew@zeut.net>
1078 * Michael is Michael Meskes <meskes@postgresql.org> of Credativ
1079 * Neil is Neil Conway <neilc@samurai.com>
1080 * Oleg is Oleg Bartunov <oleg@sai.msu.su>
1081 * Peter is Peter Eisentraut <peter_e@gmx.net>
1082 * Philip is Philip Warner <pjw@rhyme.com.au> of Albatross Consulting Pty. Ltd.
1083 * Rod is Rod Taylor <pg@rbt.ca>
1084 * Simon is Simon Riggs <simon@2ndquadrant.com>
1085 * Stephan is Stephan Szabo <sszabo@megazone23.bigpanda.com>
1086 * Tatsuo is Tatsuo Ishii <t-ishii@sra.co.jp> of Software Research Assoc.
1087 * Tom is Tom Lane <tgl@sss.pgh.pa.us> of Red Hat