4 Current maintainer: Bruce Momjian (pgman@candle.pha.pa.us)
5 Last updated: Sat Dec 17 12:03:45 EST 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 8.2 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 statements
27 This would allow an application inheriting a pooled connection to know
28 the statements 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
50 * Add function to report the time of the most recent server reload
53 * Improve replication solutions
57 You can use any of the master/slave replication servers to use a
58 standby server for data warehousing. To allow read/write queries to
59 multiple servers, you need multi-master replication like pgcluster.
61 o Allow replication over unreliable or non-persistent links
66 o %Add "include file" functionality in postgresql.conf
67 o %Allow commenting of variables in postgresql.conf to restore them
70 Currently, if a variable is commented out, it keeps the
71 previous uncommented value until a server restarted.
73 o %Allow pg_hba.conf settings to be controlled via SQL
75 This would add a function to load the SQL table from
76 pg_hba.conf, and one to writes its contents to the flat file.
77 The table should have a line number that is a float so rows
78 can be inserted between existing rows, e.g. row 2.5 goes
79 between row 2 and row 3.
81 o %Allow postgresql.conf file values to be changed via an SQL
82 API, perhaps using SET GLOBAL
83 o Allow the server to be stopped/restarted via an SQL API
84 o Issue a warning if a change-on-restart-only postgresql.conf value
85 is modified and the server config files are reloaded
86 o Mark change-on-restart-only values in postgresql.conf
91 * Allow a database in tablespace t1 with tables created in
92 tablespace t2 to be used as a template for a new database created
93 with default tablespace t2
95 All objects in the default database tablespace must have default
96 tablespace specifications. This is because new databases are
97 created by copying directories. If you mix default tablespace
98 tables and tablespace-specified tables in the same directory,
99 creating a new database from such a mixed directory would create a
100 new database with tables that had incorrect explicit tablespaces.
101 To fix this would require modifying pg_class in the newly copied
102 database, which we don't currently do.
104 * Allow reporting of which objects are in which tablespaces
106 This item is difficult because a tablespace can contain objects
107 from multiple databases. There is a server-side function that
108 returns the databases which use a specific tablespace, so this
109 requires a tool that will call that function and connect to each
110 database to find the objects in each database for that tablespace.
112 o %Add a GUC variable to control the tablespace for temporary objects
115 It could start with a random tablespace from a supplied list and
116 cycle through the list.
118 o Allow WAL replay of CREATE TABLESPACE to work when the directory
119 structure on the recovery computer is different from the original
121 o Allow per-tablespace quotas
124 * Point-In-Time Recovery (PITR)
126 o Allow point-in-time recovery to archive partially filled
127 write-ahead logs [pitr]
129 Currently only full WAL files are archived. This means that the
130 most recent transactions aren't available for recovery in case
131 of a disk failure. This could be triggered by a user command or
134 o Automatically force archiving of partially-filled WAL files when
135 pg_stop_backup() is called or the server is stopped
137 Doing this will allow administrators to know more easily when
138 the archive contains all the files needed for point-in-time
141 o %Create dump tool for write-ahead logs for use in determining
142 transaction id for point-in-time recovery
143 o Allow a warm standby system to also allow read-only statements
146 This is useful for checking PITR recovery.
148 o Allow the PITR process to be debugged and data examined
154 * Allow server log information to be output as INSERT statements
156 This would allow server log information to be easily loaded into
157 a database for analysis.
159 * %Add ability to monitor the use of temporary sort files
160 * Allow server logs to be remotely read and removed using SQL commands
161 * Allow protocol-level BIND parameter values to be logged
167 * Improve the MONEY data type
169 Change the MONEY data type to use DECIMAL internally, with special
170 locale-aware output formatting.
172 * Change NUMERIC to enforce the maximum precision
173 * Add NUMERIC division operator that doesn't round?
175 Currently NUMERIC _rounds_ the result to the specified precision.
176 This means division can return a result that multiplied by the
177 divisor is greater than the dividend, e.g. this returns a value > 10:
179 SELECT (10::numeric(2,0) / 6::numeric(2,0))::numeric(2,0) * 6;
181 The positive modulus result returned by NUMERICs might be considered
182 inaccurate, in one sense.
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 dropping 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 Merge hardwired timezone names with the TZ database; allow either
200 kind everywhere a TZ name is currently taken
201 o Allow customization of the known set of TZ names (generalize the
202 present australian_timezones hack)
203 o Allow TIMESTAMP WITH TIME ZONE to store the original timezone
204 information, either zone name or offset from UTC [timezone]
206 If the TIMESTAMP value is stored with a time zone name, interval
207 computations should adjust based on the time zone rules.
209 o Fix SELECT '0.01 years'::interval, '0.01 months'::interval
210 o Fix SELECT INTERVAL '1' MONTH
211 o Add a GUC variable to allow output of interval values in ISO8601
213 o Improve timestamptz subtraction to be DST-aware
215 Currently, subtracting one date from another that crosses a
216 daylight savings time adjustment can return '1 day 1 hour', but
217 adding that back to the first date returns a time one hour in
218 the future. This is caused by the adjustment of '25 hours' to
219 '1 day 1 hour', and '1 day' is the same time the next day, even
220 if daylight savings adjustments are involved.
222 o Fix interval display to support values exceeding 2^31 hours
223 o Add overflow checking to timestamp and interval arithmetic
224 o Add ISO INTERVAL handling
225 o Add support for day-time syntax, INTERVAL '1 2:03:04' DAY TO
227 o Add support for year-month syntax, INTERVAL '50-6' YEAR TO MONTH
228 o For syntax that isn't uniquely ISO or PG syntax, like '1:30' or
229 '1', treat as ISO if there is a range specification clause,
230 and as PG if there no clause is present, e.g. interpret
231 '1:30' MINUTE TO SECOND as '1 minute 30 seconds', and
232 interpret '1:30' as '1 hour, 30 minutes'
233 o Interpret INTERVAL '1 year' MONTH as CAST (INTERVAL '1 year' AS
234 INTERVAL MONTH), and this should return '12 months'
235 o Round or truncate values to the requested precision, e.g.
236 INTERVAL '11 months' AS YEAR should return one or zero
237 o Support precision, CREATE TABLE foo (a INTERVAL MONTH(3))
242 o -Allow NULLs in arrays
243 o Delay resolution of array expression's data type so assignment
244 coercion can be performed on empty array expressions
249 o Improve vacuum of large objects, like /contrib/vacuumlo?
250 o Add security checking for large objects
251 o Auto-delete large objects when referencing row is deleted
253 /contrib/lo offers this functionality.
255 o Allow read/write into TOAST values like large objects
257 This requires the TOAST column to be stored EXTERNAL.
263 * Allow INET subnet tests using non-constants to be indexed
264 * Add transaction_timestamp(), statement_timestamp(), clock_timestamp()
267 Current CURRENT_TIMESTAMP returns the start time of the current
268 transaction, and gettimeofday() returns the wallclock time. This will
269 make time reporting more consistent and will allow reporting of
270 the statement start time.
272 * %Add pg_get_acldef(), pg_get_typedefault(), pg_get_attrdef(),
273 pg_get_tabledef(), pg_get_domaindef(), pg_get_functiondef()
274 * Allow to_char() to print localized month names
275 * Allow functions to have a schema search path specified at creation time
276 * Allow substring/replace() to get/set bit values
277 * Allow to_char() on interval values to accumulate the highest unit
280 Some special format flag would be required to request such
281 accumulation. Such functionality could also be added to EXTRACT.
282 Prevent accumulation that crosses the month/day boundary because of
283 the uneven number of days in a month.
285 o to_char(INTERVAL '1 hour 5 minutes', 'MI') => 65
286 o to_char(INTERVAL '43 hours 20 minutes', 'MI' ) => 2600
287 o to_char(INTERVAL '43 hours 20 minutes', 'WK:DD:HR:MI') => 0:1:19:20
288 o to_char(INTERVAL '3 years 5 months','MM') => 41
290 * Add sleep() function, remove from regress.c
291 * Allow user-defined functions retuning a domain value to enforce domain
293 * Add SPI_gettypmod() to return the typemod for a TupleDesc
296 Multi-Language Support
297 ======================
299 * Add NCHAR (as distinguished from ordinary varchar),
300 * Allow locale to be set at database creation
302 Currently locale can only be set during initdb. No global tables have
303 locale-aware columns. However, the database template used during
304 database creation might have locale-aware indexes. The indexes would
305 need to be reindexed to match the new locale.
307 * Allow encoding on a per-column basis
309 Right now only one encoding is allowed per database.
311 * Support multiple simultaneous character sets, per SQL92
312 * Improve UTF8 combined character handling?
313 * Add octet_length_server() and octet_length_client()
314 * Make octet_length_client() the same as octet_length()?
315 * Fix problems with wrong runtime encoding conversion for NLS message files
321 * %Automatically create rules on views so they are updateable, per SQL99
323 We can only auto-create rules for simple views. For more complex
324 cases users will still have to write rules.
326 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
327 * Allow NOTIFY in rules involving conditionals
328 * Allow VIEW/RULE recompilation when the underlying tables change
330 Another issue is whether underlying table changes should be reflected
331 in the view, e.g. should SELECT * show additional columns if they
332 are added after the view is created.
338 * Change LIMIT/OFFSET and FETCH/MOVE to use int8
339 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
340 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
341 * %Allow SET CONSTRAINTS to be qualified by schema/table name
342 * %Allow TRUNCATE ... CASCADE/RESTRICT
344 This is like DELETE CASCADE, but truncates.
346 * %Add a separate TRUNCATE permission
348 Currently only the owner can TRUNCATE a table because triggers are not
349 called, and the table is locked in exclusive mode.
351 * Allow PREPARE of cursors
352 * Allow PREPARE to automatically determine parameter types based on the SQL
354 * Allow finer control over the caching of prepared query plans
356 Currently, queries prepared via the libpq API are planned on first
357 execute using the supplied parameters --- allow SQL PREPARE to do the
358 same. Also, allow control over replanning prepared queries either
359 manually or automatically when statistics for execute parameters
360 differ dramatically from those used during planning.
362 * Allow LISTEN/NOTIFY to store info in memory rather than tables?
364 Currently LISTEN/NOTIFY information is stored in pg_listener. Storing
365 such information in memory would improve performance.
367 * Add optional textual message to NOTIFY
369 This would allow an informational message to be added to the notify
370 message, perhaps indicating the row modified or other custom
373 * Add a GUC variable to warn about non-standard SQL usage in queries
374 * Add SQL-standard MERGE command, typically used to merge two tables
377 This is similar to UPDATE, then for unmatched rows, INSERT.
378 Whether concurrent access allows modifications which could cause
379 row loss is implementation independent.
381 * Add REPLACE or UPSERT command that does UPDATE, or on failure, INSERT
384 To implement this cleanly requires that the table have a unique index
385 so duplicate checking can be easily performed. It is possible to
386 do it without a unique index if we require the user to LOCK the table
389 * Add NOVICE output level for helpful messages like automatic sequence/index
391 * %Add COMMENT ON for all cluster global objects (roles, databases
393 * %Make row-wise comparisons work per SQL spec
394 * Add RESET CONNECTION command to reset all session state
396 This would include resetting of all variables (RESET ALL), dropping of
397 temporary tables, removing any NOTIFYs, cursors, open transactions,
398 prepared queries, currval()s, etc. This could be used for connection
399 pooling. We could also change RESET ALL to have this functionality.
400 The difficult of this features is allowing RESET ALL to not affect
401 changes made by the interface driver for its internal use. One idea
402 is for this to be a protocol-only feature. Another approach is to
403 notify the protocol when a RESET CONNECTION command is used.
405 * Add GUC to issue notice about statements that use unjoined tables
406 * Allow EXPLAIN to identify tables that were skipped because of
408 * Allow EXPLAIN output to be more easily processed by scripts
409 * Eventually enable escape_string_warning and standard_conforming_strings
410 * Simplify dropping roles that have objects in several databases
411 * Allow COMMENT ON to accept an expression rather than just a string
412 * Allow the count returned by SELECT, etc to be to represent as an int64
413 to allow a higher range of values
418 o Allow CREATE TABLE AS to determine column lengths for complex
419 expressions like SELECT col1 || col2
420 o Use more reliable method for CREATE DATABASE to get a consistent
422 o Add ON COMMIT capability to CREATE TABLE AS ... SELECT
426 o Allow UPDATE to handle complex aggregates [update]?
427 o Allow an alias to be provided for the target table in
430 This is not SQL-spec but many DBMSs allow it.
432 o Allow UPDATE tab SET ROW (col, ...) = (...) for updating multiple
438 o %Have ALTER TABLE RENAME rename SERIAL sequence names
439 o Add ALTER DOMAIN to modify the underlying data type
440 o %Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
441 o %Allow ALTER TABLE to change constraint deferrability and actions
442 o Add missing object types for ALTER ... SET SCHEMA
443 o Allow ALTER TABLESPACE to move to different directories
444 o Allow databases to be moved to different tablespaces
445 o Allow moving system tables to other tablespaces, where possible
447 Currently non-global system tables must be in the default database
448 tablespace. Global system tables can never be moved.
450 o %Disallow dropping of an inherited constraint
451 o %Prevent child tables from altering or dropping constraints
452 like CHECK that were inherited from the parent table
453 o Have ALTER INDEX update the name of a constraint using that index
454 o Add ALTER TABLE RENAME CONSTRAINT, update index name also
459 o Automatically maintain clustering on a table
461 This might require some background daemon to maintain clustering
462 during periods of low usage. It might also require tables to be only
463 partially filled for easier reorganization. Another idea would
464 be to create a merged heap/index data file so an index lookup would
465 automatically access the heap data too. A third idea would be to
466 store heap rows in hashed groups, perhaps using a user-supplied
469 o %Add default clustering to system tables
471 To do this, determine the ideal cluster index for each system
472 table and set the cluster setting during initdb.
477 o Allow COPY to report error lines and continue
479 This requires the use of a savepoint before each COPY line is
480 processed, with ROLLBACK on COPY failure.
482 o %Have COPY return the number of rows loaded/unloaded?
483 o Allow COPY on a newly-created table to skip WAL logging
485 On crash recovery, the table involved in the COPY would
486 be removed or have its heap and index files truncated. One
487 issue is that no other backend should be able to add to
488 the table at the same time, which is something that is
491 o Allow COPY to output from views
493 Another idea would be to allow actual SELECT statements in a COPY.
498 o Allow column-level privileges
499 o %Allow GRANT/REVOKE permissions to be applied to all schema objects
502 The proposed syntax is:
503 GRANT SELECT ON ALL TABLES IN public TO phpuser;
504 GRANT SELECT ON NEW TABLES IN public TO phpuser;
506 * Allow GRANT/REVOKE permissions to be inherited by objects based on
512 o Allow UPDATE/DELETE WHERE CURRENT OF cursor
514 This requires using the row ctid to map cursor rows back to the
515 original heap row. This become more complicated if WITH HOLD cursors
516 are to be supported because WITH HOLD cursors have a copy of the row
517 and no FOR UPDATE lock.
519 o Prevent DROP TABLE from dropping a row referenced by its own open
522 o %Allow pooled connections to list all open WITH HOLD cursors
524 Because WITH HOLD cursors exist outside transactions, this allows
525 them to be listed so they can be closed.
530 o Allow INSERT/UPDATE of the system-generated oid value for a row
531 o Allow INSERT INTO tab (col1, ..) VALUES (val1, ..), (val2, ..)
532 o Allow INSERT/UPDATE ... RETURNING new.col or old.col
534 This is useful for returning the auto-generated key for an INSERT.
535 One complication is how to handle rules that run as part of
541 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
543 o Add SET PATH for schemas?
545 This is basically the same as SET search_path.
548 * Server-Side Languages
550 o Fix PL/pgSQL RENAME to work on variables other than OLD/NEW
551 o Allow function parameters to be passed by name,
552 get_employee_salary(emp_id => 12345, tax_year => 2001)
553 o Add Oracle-style packages
554 o Add table function support to pltcl, plpython
555 o Add capability to create and call PROCEDURES
556 o Allow PL/pgSQL to handle %TYPE arrays, e.g. tab.col%TYPE[]
557 o Allow function argument names to be statements from PL/PgSQL
558 o Add MOVE to PL/pgSQL
559 o Add support for polymorphic arguments and return types to
560 languages other than PL/PgSQL
561 o Add support for OUT and INOUT parameters to languages other
563 o Add single-step debugging of PL/PgSQL functions
564 o Allow PL/PgSQL to support WITH HOLD cursors
570 * -Have initdb set the input DateStyle (MDY or DMY) based on locale
571 * Have pg_ctl look at PGHOST in case it is a socket directory?
572 * Allow pg_ctl to work properly with configuration files located outside
575 pg_ctl can not read the pid file because it isn't located in the
576 config directory but in the PGDATA directory. The solution is to
577 allow pg_ctl to read and understand postgresql.conf to find the
578 data_directory value.
583 o Have psql show current values for a sequence
584 o Move psql backslash database information into the backend, use
585 mnemonic commands? [psql]
587 This would allow non-psql clients to pull the same information out
588 of the database as psql.
590 o Fix psql's display of schema information (Neil)
591 o Allow psql \pset boolean variables to set to fixed values, rather
593 o Consistently display privilege information for all objects in psql
594 o Improve psql's handling of multi-line statements
596 Currently, while \e saves a single statement as one entry, interactive
597 statements are saved one line at a time. Ideally all statements
598 would be saved like \e does.
600 o Allow multi-line column values to align in the proper columns
602 If the second output column value is 'a\nb', the 'b' should appear
603 in the second display column, rather than the first column as it
606 o Display IN, INOUT, and OUT parameters in \df+
608 It probably requires psql to output newlines in the proper
609 column, which is already on the TODO list.
611 o Add auto-expanded mode so expanded output is used if the row
612 length is wider than the screen width.
614 Consider using auto-expanded mode for backslash commands like \df+.
619 o %Have pg_dump use multi-statement transactions for INSERT dumps
620 o %Allow pg_dump to use multiple -t and -n switches [pg_dump]
621 o %Add dumping of comments on index columns and composite type columns
622 o %Add full object name to the tag field. eg. for operators we need
623 '=(integer, integer)', instead of just '='.
624 o Add pg_dumpall custom format dumps?
625 o %Add CSV output format
626 o Update pg_dump and psql to use the new COPY libpq API (Christopher)
627 o Remove unnecessary function pointer abstractions in pg_dump source
629 o Allow selection of individual object(s) of all types, not just
631 o In a selective dump, allow dumping of an object and all its
633 o Add options like pg_restore -l and -L to pg_dump
634 o Stop dumping CASCADE on DROP TYPE commands in clean mode
635 o Allow pg_dump --clean to drop roles that own objects or have
637 o Add -f to pg_dumpall
644 Document differences between ecpg and the SQL standard and
645 information about the Informix-compatibility module.
647 o Solve cardinality > 1 for input descriptors / variables?
648 o Add a semantic check level, e.g. check if a table really exists
649 o fix handling of DB attributes that are arrays
650 o Use backend PREPARE/EXECUTE facility for ecpg where possible
652 o Fix nested C comments
653 o %sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
654 o Make SET CONNECTION thread-aware, non-standard?
655 o Allow multidimensional arrays
656 o Add internationalized message strings
661 o Add a function to support Parse/DescribeStatement capability
662 o Add PQescapeIdentifier()
663 o Prevent PQfnumber() from lowercasing unquoted the column name
665 PQfnumber() should never have been doing lowercasing, but
666 historically it has so we need a way to prevent it
668 o Allow statement results to be automatically batched to the client
670 Currently, all statement results are transferred to the libpq
671 client before libpq makes the results available to the
672 application. This feature would allow the application to make
673 use of the first result rows while the rest are transferred, or
674 held on the server waiting for them to be requested by libpq.
675 One complexity is that a statement like SELECT 1/col could error
676 out mid-way through the result set.
679 Referential Integrity
680 =====================
682 * Add MATCH PARTIAL referential integrity
683 * Add deferred trigger queue file
685 Right now all deferred trigger information is stored in backend
686 memory. This could exhaust memory for very large trigger queues.
687 This item involves dumping large queues into files.
689 * Change foreign key constraint for array -> element to mean element
691 * Allow DEFERRABLE UNIQUE constraints?
692 * Allow triggers to be disabled in only the current session.
694 This is currently possible by starting a multi-statement transaction,
695 modifying the system tables, performing the desired SQL, restoring the
696 system tables, and committing the transaction. ALTER TABLE ...
697 TRIGGER requires a table lock so it is not ideal for this usage.
699 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
701 If the dump is known to be valid, allow foreign keys to be added
702 without revalidating the data.
704 * Allow statement-level triggers to access modified rows
705 * Support triggers on columns (Greg Sabino Mullane)
706 * Enforce referential integrity for system tables
707 * Allow AFTER triggers on system tables
709 System tables are modified in many places in the backend without going
710 through the executor and therefore not causing triggers to fire. To
711 complete this item, the functions that modify system tables will have
718 * Flush cached query plans when the dependent objects change or
719 when the cardinality of parameters changes dramatically
720 * Track dependencies in function bodies and recompile/invalidate
722 This is particularly important for references to temporary tables
723 in PL/PgSQL because PL/PgSQL caches query plans. The only workaround
724 in PL/PgSQL is to use EXECUTE. One complexity is that a function
725 might itself drop and recreate dependent tables, causing it to
726 invalidate its own query plan.
732 * Add SQL99 WITH clause to SELECT
733 * Add SQL99 WITH RECURSIVE to SELECT
734 * Add pre-parsing phase that converts non-ISO syntax to supported
737 This could allow SQL written for other databases to run without
740 * Allow plug-in modules to emulate features from other databases
741 * SQL*Net listener that makes PostgreSQL appear as an Oracle database
743 * Allow statements across databases or servers with transaction
746 This can be done using dblink and two-phase commit.
748 * Add the features of packages
750 o Make private objects accessible only to objects in the same schema
751 o Allow current_schema.objname to access current schema objects
752 o Add session variables
753 o Allow nested schemas
759 * Allow inherited tables to inherit index, UNIQUE constraint, and primary
761 * UNIQUE INDEX on base column not honored on INSERTs/UPDATEs from
762 inherited table: INSERT INTO inherit_table (unique_index_col) VALUES
765 The main difficulty with this item is the problem of creating an index
766 that can span more than one table.
768 * Allow SELECT ... FOR UPDATE on inherited tables
769 * Add UNIQUE capability to non-btree indexes
770 * Prevent index uniqueness checks when UPDATE does not modify the column
772 Uniqueness (index) checks are done when updating a column even if the
773 column is not modified by the UPDATE.
775 * Allow the creation of on-disk bitmap indexes which can be quickly
776 combined with other bitmap indexes
778 Such indexes could be more compact if there are only a few distinct values.
779 Such indexes can also be compressed. Keeping such indexes updated can be
782 * Allow use of indexes to search for NULLs
784 One solution is to create a partial index on an IS NULL expression.
786 * Allow accurate statistics to be collected on indexes with more than
787 one column or expression indexes, perhaps using per-index statistics
788 * Add fillfactor to control reserved free space during index creation
789 * Allow the creation of indexes with mixed ascending/descending specifiers
790 * Allow constraint_exclusion to work for UNIONs like it does for
791 inheritance, allow it to work for UPDATE and DELETE statements, and allow
792 it to be used for all statements with little performance impact
793 * Allow CREATE INDEX to take an additional parameter for use with
795 * Consider compressing indexes by storing key values duplicated in
796 several rows as a single index entry
798 This is difficult because it requires datatype-specific knowledge.
803 o Add more GIST index support for geometric data types
804 o Allow GIST indexes to create certain complex index types, like
805 digital trees (see Aoki)
809 o Pack hash index buckets onto disk pages more efficiently
811 Currently only one hash bucket can be stored on a page. Ideally
812 several hash buckets could be stored on a single page and greater
813 granularity used for the hash algorithm.
815 o Consider sorting hash buckets so entries can be found using a
816 binary search, rather than a linear scan
818 o In hash indexes, consider storing the hash value with or instead
821 o Add WAL logging for crash recovery
822 o Allow multi-column hash indexes
828 * Improve commit_delay handling to reduce fsync()
829 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
831 Ideally this requires a separate test program that can be run
832 at initdb time or optionally later. Consider O_SYNC when
835 * %Add an option to sync() before fsync()'ing checkpoint files
836 * Add program to test if fsync has a delay compared to non-fsync
842 * Allow free-behind capability for large sequential scans, perhaps using
845 Posix_fadvise() can control both sequential/random file caching and
846 free-behind behavior, but it is unclear how the setting affects other
847 backends that also have the file open, and the feature is not supported
848 on all operating systems.
852 We could use a fixed row count and a +/- count to follow MVCC
853 visibility rules, or a single cached value could be used and
854 invalidated if anyone modifies the table. Another idea is to
855 get a count directly from a unique index, but for this to be
856 faster than a sequential scan it must avoid access to the heap
857 to obtain tuple visibility information.
859 * Add estimated_count(*) to return an estimate of COUNT(*)
861 This would use the planner ANALYZE statistics to return an estimated
864 * Allow data to be pulled directly from indexes
866 Currently indexes do not have enough tuple visibility information
867 to allow data to be pulled from the index without also accessing
868 the heap. One way to allow this is to set a bit on index tuples
869 to indicate if a tuple is currently visible to all transactions
870 when the first valid heap lookup happens. This bit would have to
871 be cleared when a heap tuple is expired.
873 Another idea is to maintain a bitmap of heap pages where all rows
874 are visible to all backends, and allow index lookups to reference
875 that bitmap to avoid heap lookups, perhaps the same bitmap we might
876 add someday to determine which heap pages need vacuuming. Frequently
877 accessed bitmaps would have to be stored in shared memory. One 8k
878 page of bitmaps could track 512MB of heap pages.
880 * Consider automatic caching of statements at various levels:
886 * Allow sequential scans to take advantage of other concurrent
887 sequential scans, also called "Synchronised Scanning"
889 One possible implementation is to start sequential scans from the lowest
890 numbered buffer in the shared cache, and when reaching the end wrap
891 around to the beginning, rather than always starting sequential scans
892 at the start of the table.
898 * Improve speed with indexes
900 For large table adjustments during VACUUM FULL, it is faster to
901 reindex rather than update the index.
903 * Reduce lock time during VACUUM FULL by moving tuples with read lock,
904 then write lock and truncate table
906 Moved tuples are invisible to other backends so they don't require a
907 write lock. However, the read lock promotion to write lock could lead
908 to deadlock situations.
910 * Auto-fill the free space map by scanning the buffer cache or by
911 checking pages written by the background writer
912 * Create a bitmap of pages that need vacuuming
914 Instead of sequentially scanning the entire table, have the background
915 writer or some other process record pages that have expired rows, then
916 VACUUM can look at just those pages rather than the entire table. In
917 the event of a system crash, the bitmap would probably be invalidated.
918 One complexity is that index entries still have to be vacuumed, and
919 doing this without an index scan (by using the heap values to find the
920 index entry) might be slow and unreliable, especially for user-defined
923 * %Add system view to show free space map contents
928 o Use free-space map information to guide refilling
929 o %Issue log message to suggest VACUUM FULL if a table is nearly
931 o Improve xid wraparound detection by recording per-table rather
933 o Consider logging activity either to the logs or a system view
939 * Fix priority ordering of read and write light-weight locks (Neil)
942 Startup Time Improvements
943 =========================
945 * Experiment with multi-threaded backend [thread]
947 This would prevent the overhead associated with process creation. Most
948 operating systems have trivial process creation time compared to
949 database startup overhead, but a few operating systems (Win32,
950 Solaris) might benefit from threading. Also explore the idea of
951 a single session using multiple threads to execute a statement faster.
953 * Add connection pooling
955 It is unclear if this should be done inside the backend code or done
956 by something external like pgpool. The passing of file descriptors to
957 existing backends is one of the difficulties with a backend approach.
963 * Eliminate need to write full pages to WAL before page modification [wal]
965 Currently, to protect against partial disk page writes, we write
966 full page images to WAL before they are modified so we can correct any
967 partial page writes during recovery. These pages can also be
968 eliminated from point-in-time archive files.
970 o When off, write CRC to WAL and check file system blocks
973 If CRC check fails during recovery, remember the page in case
974 a later CRC for that page properly matches.
976 o Write full pages during file system write and not when
977 the page is modified in the buffer cache
979 This allows most full page writes to happen in the background
980 writer. It might cause problems for applying WAL on recovery
981 into a partially-written page, but later the full page will be
984 * Allow WAL traffic to be streamed to another server for stand-by
986 * Reduce WAL traffic so only modified values are written rather than
988 * Allow the pg_xlog directory location to be specified during initdb
989 with a symlink back to the /data location
990 * Allow WAL information to recover corrupted pg_controldata
991 * Find a way to reduce rotational delay when repeatedly writing
994 Currently fsync of WAL requires the disk platter to perform a full
995 rotation to fsync again. One idea is to write the WAL to different
996 offsets that might reduce the rotational delay.
998 * Allow buffered WAL writes and fsync
1000 Instead of guaranteeing recovery of all committed transactions, this
1001 would provide improved performance by delaying WAL writes and fsync
1002 so an abrupt operating system restart might lose a few seconds of
1003 committed transactions but still be consistent. We could perhaps
1004 remove the 'fsync' parameter (which results in an an inconsistent
1005 database) in favor of this capability.
1008 Optimizer / Executor
1009 ====================
1011 * Add missing optimizer selectivities for date, r-tree, etc
1012 * Allow ORDER BY ... LIMIT # to select high/low value without sort or
1013 index using a sequential scan for highest/lowest values
1015 Right now, if no index exists, ORDER BY ... LIMIT # requires we sort
1016 all values to return the high/low value. Instead The idea is to do a
1017 sequential scan to find the high/low value, thus avoiding the sort.
1018 MIN/MAX already does this, but not for LIMIT > 1.
1020 * Precompile SQL functions to avoid overhead
1021 * Create utility to compute accurate random_page_cost value
1022 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
1023 * Have EXPLAIN ANALYZE highlight poor optimizer estimates
1024 * Consider using hash buckets to do DISTINCT, rather than sorting
1026 This would be beneficial when there are few distinct values. This is
1027 already used by GROUP BY.
1029 * Log statements where the optimizer row estimates were dramatically
1030 different from the number of rows actually found?
1031 * Allow star join optimizations
1033 While our bitmap scan allows multiple indexes to be joined to get
1034 to heap rows, a star joins allows multiple dimension _tables_ to
1035 be joined to index into a larger main fact table. The join is
1036 usually performed by either creating a cartesian product of all
1037 the dimmension tables and doing a single join on that product or
1038 using subselects to create bitmaps of each dimmension table match
1039 and merge the bitmaps to perform the join on the fact table. Some
1040 of these algorithms might be patented.
1043 Miscellaneous Performance
1044 =========================
1046 * Do async I/O for faster random read-ahead of data
1048 Async I/O allows multiple I/O requests to be sent to the disk with
1049 results coming back asynchronously.
1051 * Use mmap() rather than SYSV shared memory or to write WAL files?
1053 This would remove the requirement for SYSV SHM but would introduce
1054 portability issues. Anonymous mmap (or mmap to /dev/zero) is required
1055 to prevent I/O overhead.
1057 * Consider mmap()'ing files into a backend?
1059 Doing I/O to large tables would consume a lot of address space or
1060 require frequent mapping/unmapping. Extending the file also causes
1061 mapping problems that might require mapping only individual pages,
1062 leading to thousands of mappings. Another problem is that there is no
1063 way to _prevent_ I/O to disk from the dirty shared buffers so changes
1064 could hit disk before WAL is written.
1066 * Add a script to ask system configuration questions and tune postgresql.conf
1067 * Merge xmin/xmax/cmin/cmax back into three header fields
1069 Before subtransactions, there used to be only three fields needed to
1070 store these four values. This was possible because only the current
1071 transaction looks at the cmin/cmax values. If the current transaction
1072 created and expired the row the fields stored where xmin (same as
1073 xmax), cmin, cmax, and if the transaction was expiring a row from a
1074 another transaction, the fields stored were xmin (cmin was not
1075 needed), xmax, and cmax. Such a system worked because a transaction
1076 could only see rows from another completed transaction. However,
1077 subtransactions can see rows from outer transactions, and once the
1078 subtransaction completes, the outer transaction continues, requiring
1079 the storage of all four fields. With subtransactions, an outer
1080 transaction can create a row, a subtransaction expire it, and when the
1081 subtransaction completes, the outer transaction still has to have
1082 proper visibility of the row's cmin, for example, for cursors.
1084 One possible solution is to create a phantom cid which represents a
1085 cmin/cmax pair and is stored in local memory. Another idea is to
1086 store both cmin and cmax only in local memory.
1088 * Research storing disk pages with no alignment/padding
1094 * Add use of 'const' for variables in source tree
1095 * Rename some /contrib modules from pg* to pg_*
1096 * Move some things from /contrib into main tree
1097 * Move some /contrib modules out to their own project sites
1098 * %Remove warnings created by -Wcast-align
1099 * Move platform-specific ps status display info from ps_status.c to ports
1100 * Add optional CRC checksum to heap and index pages
1101 * Improve documentation to build only interfaces (Marc)
1102 * Remove or relicense modules that are not under the BSD license, if possible
1103 * %Remove memory/file descriptor freeing before ereport(ERROR)
1104 * Acquire lock on a relation before building a relcache entry for it
1105 * %Promote debug_query_string into a server-side function current_query()
1106 * %Allow the identifier length to be increased via a configure option
1107 * Allow cross-compiling by generating the zic database on the target system
1108 * Improve NLS maintenace of libpgport messages linked onto applications
1109 * Allow ecpg to work with MSVC and BCC
1110 * Add xpath_array() to /contrib/xml2 to return results as an array
1111 * Allow building in directories containing spaces
1113 This is probably not possible because 'gmake' and other compiler tools
1114 do not fully support quoting of paths with spaces.
1116 * -Allow installing to directories containing spaces
1118 This is possible if proper quoting is added to the makefiles for the
1119 install targets. Because PostgreSQL supports relocatable installs, it
1120 is already possible to install into a directory that doesn't contain
1121 spaces and then copy the install to a directory with spaces.
1123 * Fix sgmltools so PDFs can be generated with bookmarks
1124 * %Clean up compiler warnings (especially with gcc version 4)
1125 * Use UTF8 encoding for NLS messages so all server encodings can
1127 * Update Bonjour to work with newer cross-platform SDK
1128 * Remove BeOS and QNX-specific code
1133 o Remove configure.in check for link failure when cause is found
1134 o Remove readdir() errno patch when runtime/mingwex/dirent.c rev
1136 o Remove psql newline patch when we find out why mingw outputs an
1138 o Allow psql to use readline once non-US code pages work with
1140 o Re-enable timezone output on log_line_prefix '%t' when a
1141 shorter timezone string is available
1142 o Fix problem with shared memory on the Win32 Terminal Server
1143 o Improve signal handling,
1144 http://archives.postgresql.org/pgsql-patches/2005-06/msg00027.php
1145 o Add long file support for binary pg_dump output
1147 While Win32 supports 64-bit files, the MinGW API does not,
1148 meaning we have to build an fseeko replacement on top of the
1149 Win32 API, and we have to make sure MinGW handles it. Another
1150 option is to wait for the MinGW project to fix it, or use the
1151 code from the LibGW32C project as a guide.
1154 * Wire Protocol Changes
1156 o Allow dynamic character set handling
1157 o Add decoded type, length, precision
1159 o Update clients to use data types, typmod, schema.table.column names
1160 of result sets using new statement protocol
1163 ---------------------------------------------------------------------------
1166 Developers who have claimed items are:
1167 --------------------------------------
1168 * Alvaro is Alvaro Herrera <alvherre@dcc.uchile.cl>
1169 * Andrew is Andrew Dunstan <andrew@dunslane.net>
1170 * Bruce is Bruce Momjian <pgman@candle.pha.pa.us> of Software Research Assoc.
1171 * Christopher is Christopher Kings-Lynne <chriskl@familyhealth.com.au> of
1172 Family Health Network
1173 * D'Arcy is D'Arcy J.M. Cain <darcy@druid.net> of The Cain Gang Ltd.
1174 * Fabien is Fabien Coelho <coelho@cri.ensmp.fr>
1175 * Gavin is Gavin Sherry <swm@linuxworld.com.au> of Alcove Systems Engineering
1176 * Greg is Greg Sabino Mullane <greg@turnstep.com>
1177 * Jan is Jan Wieck <JanWieck@Yahoo.com> of Afilias, Inc.
1178 * Joe is Joe Conway <mail@joeconway.com>
1179 * Karel is Karel Zak <zakkr@zf.jcu.cz>
1180 * Magnus is Magnus Hagander <mha@sollentuna.net>
1181 * Marc is Marc Fournier <scrappy@hub.org> of PostgreSQL, Inc.
1182 * Matthew T. O'Connor <matthew@zeut.net>
1183 * Michael is Michael Meskes <meskes@postgresql.org> of Credativ
1184 * Neil is Neil Conway <neilc@samurai.com>
1185 * Oleg is Oleg Bartunov <oleg@sai.msu.su>
1186 * Peter is Peter Eisentraut <peter_e@gmx.net>
1187 * Philip is Philip Warner <pjw@rhyme.com.au> of Albatross Consulting Pty. Ltd.
1188 * Rod is Rod Taylor <pg@rbt.ca>
1189 * Simon is Simon Riggs <simon@2ndquadrant.com>
1190 * Stephan is Stephan Szabo <sszabo@megazone23.bigpanda.com>
1191 * Tatsuo is Tatsuo Ishii <t-ishii@sra.co.jp> of Software Research Assoc.
1192 * Teodor is Teodor Sigaev <teodor@sigaev.ru>
1193 * Tom is Tom Lane <tgl@sss.pgh.pa.us> of Red Hat