4 Current maintainer: Bruce Momjian (pgman@candle.pha.pa.us)
5 Last updated: Tue Feb 28 14:47:56 EST 2006
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
24 * -%Allow pooled connections to list all prepared statements
26 This would allow an application inheriting a pooled connection to know
27 the statements prepared in the current session.
29 * Allow major upgrades without dump/reload, perhaps using pg_upgrade
31 * Check for unreferenced table files created by transactions that were
32 in-progress when the server terminated abruptly
33 * Allow administrators to safely terminate individual sessions either
34 via an SQL function or SIGTERM
36 Lock table corruption following SIGTERM of an individual backend
37 has been reported in 8.0. A possible cause was fixed in 8.1, but
38 it is unknown whether other problems exist. This item mostly
39 requires additional testing rather than of writing any new code.
41 * %Set proper permissions on non-system schemas during db creation
43 Currently all schemas are owned by the super-user because they are
44 copied from the template1 database.
46 * Support table partitioning that allows a single table to be stored
47 in subtables that are partitioned based on the primary key or a WHERE
49 * Add function to report the time of the most recent server reload
50 * Allow statistics collector information to be pulled from the collector
51 process directly, rather than requiring the collector to write a
52 filesystem file twice a second?
55 * Improve replication solutions
59 You can use any of the master/slave replication servers to use a
60 standby server for data warehousing. To allow read/write queries to
61 multiple servers, you need multi-master replication like pgcluster.
63 o Allow replication over unreliable or non-persistent links
68 o %Add "include file" functionality in postgresql.conf
69 o %Allow commenting of variables in postgresql.conf to restore them
72 Currently, if a variable is commented out, it keeps the
73 previous uncommented value until a server restarted.
75 o %Allow pg_hba.conf settings to be controlled via SQL
77 This would add a function to load the SQL table from
78 pg_hba.conf, and one to writes its contents to the flat file.
79 The table should have a line number that is a float so rows
80 can be inserted between existing rows, e.g. row 2.5 goes
81 between row 2 and row 3.
83 o Allow pg_hba.conf to specify host names along with IP addresses
85 Host name lookup could occur when the postmaster reads the
86 pg_hba.conf file, or when the backend starts. Another
87 solution would be to reverse lookup the connection IP and
88 check that hostname against the host names in pg_hba.conf.
89 We could also then check that the host name maps to the IP
92 o %Allow postgresql.conf file values to be changed via an SQL
93 API, perhaps using SET GLOBAL
94 o Allow the server to be stopped/restarted via an SQL API
95 o -Issue a warning if a change-on-restart-only postgresql.conf value
96 is modified and the server config files are reloaded
97 o Mark change-on-restart-only values in postgresql.conf
102 * Allow a database in tablespace t1 with tables created in
103 tablespace t2 to be used as a template for a new database created
104 with default tablespace t2
106 All objects in the default database tablespace must have default
107 tablespace specifications. This is because new databases are
108 created by copying directories. If you mix default tablespace
109 tables and tablespace-specified tables in the same directory,
110 creating a new database from such a mixed directory would create a
111 new database with tables that had incorrect explicit tablespaces.
112 To fix this would require modifying pg_class in the newly copied
113 database, which we don't currently do.
115 * Allow reporting of which objects are in which tablespaces
117 This item is difficult because a tablespace can contain objects
118 from multiple databases. There is a server-side function that
119 returns the databases which use a specific tablespace, so this
120 requires a tool that will call that function and connect to each
121 database to find the objects in each database for that tablespace.
123 o %Add a GUC variable to control the tablespace for temporary objects
126 It could start with a random tablespace from a supplied list and
127 cycle through the list.
129 o Allow WAL replay of CREATE TABLESPACE to work when the directory
130 structure on the recovery computer is different from the original
132 o Allow per-tablespace quotas
135 * Point-In-Time Recovery (PITR)
137 o Allow point-in-time recovery to archive partially filled
138 write-ahead logs [pitr]
140 Currently only full WAL files are archived. This means that the
141 most recent transactions aren't available for recovery in case
142 of a disk failure. This could be triggered by a user command or
145 o Automatically force archiving of partially-filled WAL files when
146 pg_stop_backup() is called or the server is stopped
148 Doing this will allow administrators to know more easily when
149 the archive contains all the files needed for point-in-time
152 o %Create dump tool for write-ahead logs for use in determining
153 transaction id for point-in-time recovery
154 o Allow a warm standby system to also allow read-only statements
157 This is useful for checking PITR recovery.
159 o Allow the PITR process to be debugged and data examined
165 * Allow server log information to be output as INSERT statements
167 This would allow server log information to be easily loaded into
168 a database for analysis.
170 * %Add ability to monitor the use of temporary sort files
171 * Allow server logs to be remotely read and removed using SQL commands
172 * Allow protocol-level BIND parameter values to be logged
178 * Improve the MONEY data type
180 Change the MONEY data type to use DECIMAL internally, with special
181 locale-aware output formatting.
183 * Change NUMERIC to enforce the maximum precision
184 * Add NUMERIC division operator that doesn't round?
186 Currently NUMERIC _rounds_ the result to the specified precision.
187 This means division can return a result that multiplied by the
188 divisor is greater than the dividend, e.g. this returns a value > 10:
190 SELECT (10::numeric(2,0) / 6::numeric(2,0))::numeric(2,0) * 6;
192 The positive modulus result returned by NUMERICs might be considered
193 inaccurate, in one sense.
195 * %Disallow changing default expression of a SERIAL column
196 * Fix data types where equality comparison isn't intuitive, e.g. box
197 * -Zero umasked bits in conversion from INET cast to CIDR
198 * -Prevent INET cast to CIDR from dropping netmask, SELECT '1.1.1.1'::inet::cidr
199 * -Allow INET + INT8 to increment the host part of the address or
200 throw an error on overflow
201 * -Add 'tid != tid ' operator for use in corruption recovery
202 * Allow user-defined types to specify a type modifier at table creation
208 o Allow infinite dates and intervals just like infinite timestamps
209 o Merge hardwired timezone names with the TZ database; allow either
210 kind everywhere a TZ name is currently taken
211 o Allow customization of the known set of TZ names (generalize the
212 present australian_timezones hack)
213 o Allow TIMESTAMP WITH TIME ZONE to store the original timezone
214 information, either zone name or offset from UTC [timezone]
216 If the TIMESTAMP value is stored with a time zone name, interval
217 computations should adjust based on the time zone rules.
219 o Fix SELECT '0.01 years'::interval, '0.01 months'::interval
220 o Fix SELECT INTERVAL '1' MONTH
221 o Add a GUC variable to allow output of interval values in ISO8601
223 o Improve timestamptz subtraction to be DST-aware
225 Currently, subtracting one date from another that crosses a
226 daylight savings time adjustment can return '1 day 1 hour', but
227 adding that back to the first date returns a time one hour in
228 the future. This is caused by the adjustment of '25 hours' to
229 '1 day 1 hour', and '1 day' is the same time the next day, even
230 if daylight savings adjustments are involved.
232 o Fix interval display to support values exceeding 2^31 hours
233 o Add overflow checking to timestamp and interval arithmetic
234 o Add ISO INTERVAL handling
235 o Add support for day-time syntax, INTERVAL '1 2:03:04' DAY TO
237 o Add support for year-month syntax, INTERVAL '50-6' YEAR TO MONTH
238 o For syntax that isn't uniquely ISO or PG syntax, like '1:30' or
239 '1', treat as ISO if there is a range specification clause,
240 and as PG if there no clause is present, e.g. interpret
241 '1:30' MINUTE TO SECOND as '1 minute 30 seconds', and
242 interpret '1:30' as '1 hour, 30 minutes'
243 o Interpret INTERVAL '1 year' MONTH as CAST (INTERVAL '1 year' AS
244 INTERVAL MONTH), and this should return '12 months'
245 o Round or truncate values to the requested precision, e.g.
246 INTERVAL '11 months' AS YEAR should return one or zero
247 o Support precision, CREATE TABLE foo (a INTERVAL MONTH(3))
252 o -Allow NULLs in arrays
253 o Delay resolution of array expression's data type so assignment
254 coercion can be performed on empty array expressions
259 o Improve vacuum of large objects, like /contrib/vacuumlo?
260 o Add security checking for large objects
261 o Auto-delete large objects when referencing row is deleted
263 /contrib/lo offers this functionality.
265 o Allow read/write into TOAST values like large objects
267 This requires the TOAST column to be stored EXTERNAL.
273 * Allow INET subnet tests using non-constants to be indexed
274 * Add transaction_timestamp(), statement_timestamp(), clock_timestamp()
277 Current CURRENT_TIMESTAMP returns the start time of the current
278 transaction, and gettimeofday() returns the wallclock time. This will
279 make time reporting more consistent and will allow reporting of
280 the statement start time.
282 * %Add pg_get_acldef(), pg_get_typedefault(), pg_get_attrdef(),
283 pg_get_tabledef(), pg_get_domaindef(), pg_get_functiondef()
284 * -Allow to_char() to print localized month names
285 * Allow functions to have a schema search path specified at creation time
286 * Allow substring/replace() to get/set bit values
287 * Allow to_char() on interval values to accumulate the highest unit
290 Some special format flag would be required to request such
291 accumulation. Such functionality could also be added to EXTRACT.
292 Prevent accumulation that crosses the month/day boundary because of
293 the uneven number of days in a month.
295 o to_char(INTERVAL '1 hour 5 minutes', 'MI') => 65
296 o to_char(INTERVAL '43 hours 20 minutes', 'MI' ) => 2600
297 o to_char(INTERVAL '43 hours 20 minutes', 'WK:DD:HR:MI') => 0:1:19:20
298 o to_char(INTERVAL '3 years 5 months','MM') => 41
300 * -Add sleep() function, remove from regress.c
301 * Allow user-defined functions retuning a domain value to enforce domain
303 * Add SPI_gettypmod() to return the typemod for a TupleDesc
306 Multi-Language Support
307 ======================
309 * Add NCHAR (as distinguished from ordinary varchar),
310 * Allow locale to be set at database creation
312 Currently locale can only be set during initdb. No global tables have
313 locale-aware columns. However, the database template used during
314 database creation might have locale-aware indexes. The indexes would
315 need to be reindexed to match the new locale.
317 * Allow encoding on a per-column basis
319 Right now only one encoding is allowed per database.
321 * Support multiple simultaneous character sets, per SQL92
322 * Improve UTF8 combined character handling?
323 * Add octet_length_server() and octet_length_client()
324 * Make octet_length_client() the same as octet_length()?
325 * Fix problems with wrong runtime encoding conversion for NLS message files
331 * Automatically create rules on views so they are updateable, per SQL99
333 We can only auto-create rules for simple views. For more complex
334 cases users will still have to write rules manually.
336 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
337 * Allow NOTIFY in rules involving conditionals
338 * Allow VIEW/RULE recompilation when the underlying tables change
340 Another issue is whether underlying table changes should be reflected
341 in the view, e.g. should SELECT * show additional columns if they
342 are added after the view is created.
348 * Change LIMIT/OFFSET and FETCH/MOVE to use int8
349 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
350 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
351 * %Allow SET CONSTRAINTS to be qualified by schema/table name
352 * %Allow TRUNCATE ... CASCADE/RESTRICT
354 This is like DELETE CASCADE, but truncates.
356 * %Add a separate TRUNCATE permission
358 Currently only the owner can TRUNCATE a table because triggers are not
359 called, and the table is locked in exclusive mode.
361 * Allow PREPARE of cursors
362 * -Allow PREPARE to automatically determine parameter types based on the SQL
364 * Allow finer control over the caching of prepared query plans
366 Currently, queries prepared via the libpq API are planned on first
367 execute using the supplied parameters --- allow SQL PREPARE to do the
368 same. Also, allow control over replanning prepared queries either
369 manually or automatically when statistics for execute parameters
370 differ dramatically from those used during planning.
372 * Allow LISTEN/NOTIFY to store info in memory rather than tables?
374 Currently LISTEN/NOTIFY information is stored in pg_listener. Storing
375 such information in memory would improve performance.
377 * Add optional textual message to NOTIFY
379 This would allow an informational message to be added to the notify
380 message, perhaps indicating the row modified or other custom
383 * Add a GUC variable to warn about non-standard SQL usage in queries
384 * Add SQL-standard MERGE command, typically used to merge two tables
387 This is similar to UPDATE, then for unmatched rows, INSERT.
388 Whether concurrent access allows modifications which could cause
389 row loss is implementation independent.
391 * Add REPLACE or UPSERT command that does UPDATE, or on failure, INSERT
394 To implement this cleanly requires that the table have a unique index
395 so duplicate checking can be easily performed. It is possible to
396 do it without a unique index if we require the user to LOCK the table
399 * Add NOVICE output level for helpful messages like automatic sequence/index
401 * -Add COMMENT ON for all cluster global objects (roles, databases
403 * -Make row-wise comparisons work per SQL spec
405 Right now, '(a, b) < (1, 2)' is processed as 'a < 1 and b < 2', but
406 the SQL standard requires it to be processed as a column-by-column
407 comparison, so the proper comparison is '(a < 1) OR (a = 1 AND b < 2)'.
409 * Add RESET CONNECTION command to reset all session state
411 This would include resetting of all variables (RESET ALL), dropping of
412 temporary tables, removing any NOTIFYs, cursors, open transactions,
413 prepared queries, currval()s, etc. This could be used for connection
414 pooling. We could also change RESET ALL to have this functionality.
415 The difficult of this features is allowing RESET ALL to not affect
416 changes made by the interface driver for its internal use. One idea
417 is for this to be a protocol-only feature. Another approach is to
418 notify the protocol when a RESET CONNECTION command is used.
420 * Add GUC to issue notice about statements that use unjoined tables
421 * Allow EXPLAIN to identify tables that were skipped because of
423 * Allow EXPLAIN output to be more easily processed by scripts
424 * Eventually enable escape_string_warning and standard_conforming_strings
425 * Simplify dropping roles that have objects in several databases
426 * Allow COMMENT ON to accept an expression rather than just a string
427 * Allow the count returned by SELECT, etc to be to represent as an int64
428 to allow a higher range of values
429 * Make CLUSTER preserve recently-dead tuples per MVCC requirements
430 * Add SQL99 WITH clause to SELECT
431 * Add SQL99 WITH RECURSIVE to SELECT
436 o Allow CREATE TABLE AS to determine column lengths for complex
437 expressions like SELECT col1 || col2
438 o Use more reliable method for CREATE DATABASE to get a consistent
440 o -Add ON COMMIT capability to CREATE TABLE AS ... SELECT
444 o Allow UPDATE to handle complex aggregates [update]?
445 o -Allow an alias to be provided for the target table in
447 o Allow UPDATE tab SET ROW (col, ...) = (...) for updating multiple
453 o %Have ALTER TABLE RENAME rename SERIAL sequence names
454 o Add ALTER DOMAIN to modify the underlying data type
455 o %Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
456 o %Allow ALTER TABLE to change constraint deferrability and actions
457 o Add missing object types for ALTER ... SET SCHEMA
458 o Allow ALTER TABLESPACE to move to different directories
459 o Allow databases to be moved to different tablespaces
460 o Allow moving system tables to other tablespaces, where possible
462 Currently non-global system tables must be in the default database
463 tablespace. Global system tables can never be moved.
465 o %Prevent child tables from altering or dropping constraints
466 like CHECK that were inherited from the parent table
467 o Have ALTER INDEX update the name of a constraint using that index
468 o Add ALTER TABLE RENAME CONSTRAINT, update index name also
473 o Automatically maintain clustering on a table
475 This might require some background daemon to maintain clustering
476 during periods of low usage. It might also require tables to be only
477 partially filled for easier reorganization. Another idea would
478 be to create a merged heap/index data file so an index lookup would
479 automatically access the heap data too. A third idea would be to
480 store heap rows in hashed groups, perhaps using a user-supplied
483 o %Add default clustering to system tables
485 To do this, determine the ideal cluster index for each system
486 table and set the cluster setting during initdb.
491 o Allow COPY to report error lines and continue
493 This requires the use of a savepoint before each COPY line is
494 processed, with ROLLBACK on COPY failure.
496 o %Have COPY return the number of rows loaded/unloaded?
497 o Allow COPY on a newly-created table to skip WAL logging
499 On crash recovery, the table involved in the COPY would
500 be removed or have its heap and index files truncated. One
501 issue is that no other backend should be able to add to
502 the table at the same time, which is something that is
505 o Allow COPY to output from views
507 Another idea would be to allow actual SELECT statements in a COPY.
512 o Allow column-level privileges
513 o %Allow GRANT/REVOKE permissions to be applied to all schema objects
516 The proposed syntax is:
517 GRANT SELECT ON ALL TABLES IN public TO phpuser;
518 GRANT SELECT ON NEW TABLES IN public TO phpuser;
520 * Allow GRANT/REVOKE permissions to be inherited by objects based on
523 * Allow SERIAL sequences to inherit permissions from the base table?
528 o Allow UPDATE/DELETE WHERE CURRENT OF cursor
530 This requires using the row ctid to map cursor rows back to the
531 original heap row. This become more complicated if WITH HOLD cursors
532 are to be supported because WITH HOLD cursors have a copy of the row
533 and no FOR UPDATE lock.
535 o Prevent DROP TABLE from dropping a row referenced by its own open
538 o -Allow pooled connections to list all open WITH HOLD cursors
540 Because WITH HOLD cursors exist outside transactions, this allows
541 them to be listed so they can be closed.
546 o Allow INSERT/UPDATE of the system-generated oid value for a row
547 o Allow INSERT INTO tab (col1, ..) VALUES (val1, ..), (val2, ..)
548 o Allow INSERT/UPDATE ... RETURNING new.col or old.col
550 This is useful for returning the auto-generated key for an INSERT.
551 One complication is how to handle rules that run as part of
557 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
559 o Add SET PATH for schemas?
561 This is basically the same as SET search_path.
564 * Server-Side Languages
566 o Fix PL/pgSQL RENAME to work on variables other than OLD/NEW
567 o Allow function parameters to be passed by name,
568 get_employee_salary(emp_id => 12345, tax_year => 2001)
569 o Add Oracle-style packages
570 o Add table function support to pltcl, plpython
571 o Add capability to create and call PROCEDURES
572 o Allow PL/pgSQL to handle %TYPE arrays, e.g. tab.col%TYPE[]
573 o Allow function argument names to be statements from PL/PgSQL
574 o Add MOVE to PL/pgSQL
575 o Add support for polymorphic arguments and return types to
576 languages other than PL/PgSQL
577 o Add support for OUT and INOUT parameters to languages other
579 o Add single-step debugging of PL/PgSQL functions
580 o Allow PL/PgSQL to support WITH HOLD cursors
586 * -Have initdb set the input DateStyle (MDY or DMY) based on locale
587 * Have pg_ctl look at PGHOST in case it is a socket directory?
588 * Allow pg_ctl to work properly with configuration files located outside
591 pg_ctl can not read the pid file because it isn't located in the
592 config directory but in the PGDATA directory. The solution is to
593 allow pg_ctl to read and understand postgresql.conf to find the
594 data_directory value.
599 o Have psql show current values for a sequence
600 o Move psql backslash database information into the backend, use
601 mnemonic commands? [psql]
603 This would allow non-psql clients to pull the same information out
604 of the database as psql.
606 o Fix psql's display of schema information (Neil)
607 o Allow psql \pset boolean variables to set to fixed values, rather
609 o Consistently display privilege information for all objects in psql
610 o -Improve psql's handling of multi-line statements
612 Currently, while \e saves a single statement as one entry, interactive
613 statements are saved one line at a time. Ideally all statements
614 would be saved like \e does.
616 o -Allow multi-line column values to align in the proper columns
618 If the second output column value is 'a\nb', the 'b' should appear
619 in the second display column, rather than the first column as it
622 o Display IN, INOUT, and OUT parameters in \df+
624 It probably requires psql to output newlines in the proper
625 column, which is already on the TODO list.
627 o Add auto-expanded mode so expanded output is used if the row
628 length is wider than the screen width.
630 Consider using auto-expanded mode for backslash commands like \df+.
632 o Prevent tab completion of SET TRANSACTION from querying the
633 database and therefore preventing the transaction isolation
634 level from being set.
636 Currently, SET <tab> causes a database lookup to check all
637 supported session variables. This query causes problems
638 because setting the transaction isolation level must be the
639 first statement of a transaction.
644 o %Have pg_dump use multi-statement transactions for INSERT dumps
645 o %Allow pg_dump to use multiple -t and -n switches [pg_dump]
646 o %Add dumping of comments on index columns and composite type columns
647 o %Add full object name to the tag field. eg. for operators we need
648 '=(integer, integer)', instead of just '='.
649 o Add pg_dumpall custom format dumps?
650 o %Add CSV output format
651 o Update pg_dump and psql to use the new COPY libpq API (Christopher)
652 o Remove unnecessary function pointer abstractions in pg_dump source
654 o Allow selection of individual object(s) of all types, not just
656 o In a selective dump, allow dumping of an object and all its
658 o Add options like pg_restore -l and -L to pg_dump
659 o Stop dumping CASCADE on DROP TYPE commands in clean mode
660 o Allow pg_dump --clean to drop roles that own objects or have
662 o Add -f to pg_dumpall
669 Document differences between ecpg and the SQL standard and
670 information about the Informix-compatibility module.
672 o Solve cardinality > 1 for input descriptors / variables?
673 o Add a semantic check level, e.g. check if a table really exists
674 o fix handling of DB attributes that are arrays
675 o Use backend PREPARE/EXECUTE facility for ecpg where possible
677 o Fix nested C comments
678 o %sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
679 o Make SET CONNECTION thread-aware, non-standard?
680 o Allow multidimensional arrays
681 o Add internationalized message strings
686 o Add a function to support Parse/DescribeStatement capability
687 o Add PQescapeIdentifier()
688 o Prevent PQfnumber() from lowercasing unquoted the column name
690 PQfnumber() should never have been doing lowercasing, but
691 historically it has so we need a way to prevent it
693 o Allow statement results to be automatically batched to the client
695 Currently, all statement results are transferred to the libpq
696 client before libpq makes the results available to the
697 application. This feature would allow the application to make
698 use of the first result rows while the rest are transferred, or
699 held on the server waiting for them to be requested by libpq.
700 One complexity is that a statement like SELECT 1/col could error
701 out mid-way through the result set.
704 Referential Integrity
705 =====================
707 * Add MATCH PARTIAL referential integrity
708 * Add deferred trigger queue file
710 Right now all deferred trigger information is stored in backend
711 memory. This could exhaust memory for very large trigger queues.
712 This item involves dumping large queues into files.
714 * Change foreign key constraint for array -> element to mean element
716 * Allow DEFERRABLE UNIQUE constraints?
717 * Allow triggers to be disabled in only the current session.
719 This is currently possible by starting a multi-statement transaction,
720 modifying the system tables, performing the desired SQL, restoring the
721 system tables, and committing the transaction. ALTER TABLE ...
722 TRIGGER requires a table lock so it is not ideal for this usage.
724 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
726 If the dump is known to be valid, allow foreign keys to be added
727 without revalidating the data.
729 * Allow statement-level triggers to access modified rows
730 * Support triggers on columns (Greg Sabino Mullane)
731 * Enforce referential integrity for system tables
732 * Allow AFTER triggers on system tables
734 System tables are modified in many places in the backend without going
735 through the executor and therefore not causing triggers to fire. To
736 complete this item, the functions that modify system tables will have
743 * Flush cached query plans when the dependent objects change,
744 when the cardinality of parameters changes dramatically, or
745 when new ANALYZE statistics are available
747 A more complex solution would be to save multiple plans for different
748 cardinality and use the appropriate plan based on the EXECUTE values.
750 * Track dependencies in function bodies and recompile/invalidate
752 This is particularly important for references to temporary tables
753 in PL/PgSQL because PL/PgSQL caches query plans. The only workaround
754 in PL/PgSQL is to use EXECUTE. One complexity is that a function
755 might itself drop and recreate dependent tables, causing it to
756 invalidate its own query plan.
762 * Add pre-parsing phase that converts non-ISO syntax to supported
765 This could allow SQL written for other databases to run without
768 * Allow plug-in modules to emulate features from other databases
769 * SQL*Net listener that makes PostgreSQL appear as an Oracle database
771 * Allow statements across databases or servers with transaction
774 This can be done using dblink and two-phase commit.
776 * Add the features of packages
778 o Make private objects accessible only to objects in the same schema
779 o Allow current_schema.objname to access current schema objects
780 o Add session variables
781 o Allow nested schemas
787 * Allow inherited tables to inherit index, UNIQUE constraint, and primary
789 * UNIQUE INDEX on base column not honored on INSERTs/UPDATEs from
790 inherited table: INSERT INTO inherit_table (unique_index_col) VALUES
793 The main difficulty with this item is the problem of creating an index
794 that can span more than one table.
796 * Allow SELECT ... FOR UPDATE on inherited tables
797 * Add UNIQUE capability to non-btree indexes
798 * Prevent index uniqueness checks when UPDATE does not modify the column
800 Uniqueness (index) checks are done when updating a column even if the
801 column is not modified by the UPDATE.
803 * Allow the creation of on-disk bitmap indexes which can be quickly
804 combined with other bitmap indexes
806 Such indexes could be more compact if there are only a few distinct values.
807 Such indexes can also be compressed. Keeping such indexes updated can be
810 * Allow use of indexes to search for NULLs
812 One solution is to create a partial index on an IS NULL expression.
814 * Allow accurate statistics to be collected on indexes with more than
815 one column or expression indexes, perhaps using per-index statistics
816 * Add fillfactor to control reserved free space during index creation
817 * Allow the creation of indexes with mixed ascending/descending specifiers
818 * Allow constraint_exclusion to work for UNIONs like it does for
819 inheritance, allow it to work for UPDATE and DELETE statements, and allow
820 it to be used for all statements with little performance impact
821 * Allow CREATE INDEX to take an additional parameter for use with
823 * Consider compressing indexes by storing key values duplicated in
824 several rows as a single index entry
826 This is difficult because it requires datatype-specific knowledge.
831 o Add more GIST index support for geometric data types
832 o Allow GIST indexes to create certain complex index types, like
833 digital trees (see Aoki)
837 o Pack hash index buckets onto disk pages more efficiently
839 Currently only one hash bucket can be stored on a page. Ideally
840 several hash buckets could be stored on a single page and greater
841 granularity used for the hash algorithm.
843 o Consider sorting hash buckets so entries can be found using a
844 binary search, rather than a linear scan
846 o In hash indexes, consider storing the hash value with or instead
849 o Add WAL logging for crash recovery
850 o Allow multi-column hash indexes
856 * Improve commit_delay handling to reduce fsync()
857 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
859 Ideally this requires a separate test program that can be run
860 at initdb time or optionally later. Consider O_SYNC when
863 * %Add an option to sync() before fsync()'ing checkpoint files
864 * Add program to test if fsync has a delay compared to non-fsync
870 * Allow free-behind capability for large sequential scans, perhaps using
873 Posix_fadvise() can control both sequential/random file caching and
874 free-behind behavior, but it is unclear how the setting affects other
875 backends that also have the file open, and the feature is not supported
876 on all operating systems.
880 We could use a fixed row count and a +/- count to follow MVCC
881 visibility rules, or a single cached value could be used and
882 invalidated if anyone modifies the table. Another idea is to
883 get a count directly from a unique index, but for this to be
884 faster than a sequential scan it must avoid access to the heap
885 to obtain tuple visibility information.
887 * Add estimated_count(*) to return an estimate of COUNT(*)
889 This would use the planner ANALYZE statistics to return an estimated
892 * Allow data to be pulled directly from indexes
894 Currently indexes do not have enough tuple visibility information
895 to allow data to be pulled from the index without also accessing
896 the heap. One way to allow this is to set a bit on index tuples
897 to indicate if a tuple is currently visible to all transactions
898 when the first valid heap lookup happens. This bit would have to
899 be cleared when a heap tuple is expired.
901 Another idea is to maintain a bitmap of heap pages where all rows
902 are visible to all backends, and allow index lookups to reference
903 that bitmap to avoid heap lookups, perhaps the same bitmap we might
904 add someday to determine which heap pages need vacuuming. Frequently
905 accessed bitmaps would have to be stored in shared memory. One 8k
906 page of bitmaps could track 512MB of heap pages.
908 * Consider automatic caching of statements at various levels:
914 * Allow sequential scans to take advantage of other concurrent
915 sequential scans, also called "Synchronised Scanning"
917 One possible implementation is to start sequential scans from the lowest
918 numbered buffer in the shared cache, and when reaching the end wrap
919 around to the beginning, rather than always starting sequential scans
920 at the start of the table.
926 * Improve speed with indexes
928 For large table adjustments during VACUUM FULL, it is faster to
929 reindex rather than update the index.
931 * Reduce lock time during VACUUM FULL by moving tuples with read lock,
932 then write lock and truncate table
934 Moved tuples are invisible to other backends so they don't require a
935 write lock. However, the read lock promotion to write lock could lead
936 to deadlock situations.
938 * Auto-fill the free space map by scanning the buffer cache or by
939 checking pages written by the background writer
940 * Create a bitmap of pages that need vacuuming
942 Instead of sequentially scanning the entire table, have the background
943 writer or some other process record pages that have expired rows, then
944 VACUUM can look at just those pages rather than the entire table. In
945 the event of a system crash, the bitmap would probably be invalidated.
946 One complexity is that index entries still have to be vacuumed, and
947 doing this without an index scan (by using the heap values to find the
948 index entry) might be slow and unreliable, especially for user-defined
951 * -Add system view to show free space map contents
956 o Use free-space map information to guide refilling
957 o %Issue log message to suggest VACUUM FULL if a table is nearly
959 o Improve xid wraparound detection by recording per-table rather
961 o Consider logging activity either to the logs or a system view
967 * Fix priority ordering of read and write light-weight locks (Neil)
970 Startup Time Improvements
971 =========================
973 * Experiment with multi-threaded backend [thread]
975 This would prevent the overhead associated with process creation. Most
976 operating systems have trivial process creation time compared to
977 database startup overhead, but a few operating systems (Win32,
978 Solaris) might benefit from threading. Also explore the idea of
979 a single session using multiple threads to execute a statement faster.
981 * Add connection pooling
983 It is unclear if this should be done inside the backend code or done
984 by something external like pgpool. The passing of file descriptors to
985 existing backends is one of the difficulties with a backend approach.
991 * Eliminate need to write full pages to WAL before page modification [wal]
993 Currently, to protect against partial disk page writes, we write
994 full page images to WAL before they are modified so we can correct any
995 partial page writes during recovery. These pages can also be
996 eliminated from point-in-time archive files.
998 o When off, write CRC to WAL and check file system blocks
1001 If CRC check fails during recovery, remember the page in case
1002 a later CRC for that page properly matches.
1004 o Write full pages during file system write and not when
1005 the page is modified in the buffer cache
1007 This allows most full page writes to happen in the background
1008 writer. It might cause problems for applying WAL on recovery
1009 into a partially-written page, but later the full page will be
1012 * Allow WAL traffic to be streamed to another server for stand-by
1014 * Reduce WAL traffic so only modified values are written rather than
1016 * Allow the pg_xlog directory location to be specified during initdb
1017 with a symlink back to the /data location
1018 * Allow WAL information to recover corrupted pg_controldata
1019 * Find a way to reduce rotational delay when repeatedly writing
1022 Currently fsync of WAL requires the disk platter to perform a full
1023 rotation to fsync again. One idea is to write the WAL to different
1024 offsets that might reduce the rotational delay.
1026 * Allow buffered WAL writes and fsync
1028 Instead of guaranteeing recovery of all committed transactions, this
1029 would provide improved performance by delaying WAL writes and fsync
1030 so an abrupt operating system restart might lose a few seconds of
1031 committed transactions but still be consistent. We could perhaps
1032 remove the 'fsync' parameter (which results in an an inconsistent
1033 database) in favor of this capability.
1035 * Allow WAL logging to be turned off for a table, but the table
1036 might be dropped or truncated during crash recovery [walcontrol]
1038 Allow tables to bypass WAL writes and just fsync() dirty pages on
1039 commit. This should be implemented using ALTER TABLE, e.g. ALTER
1040 TABLE PERSISTENCE [ DROP | TRUNCATE | DEFAULT ]. Tables using
1041 non-default logging should not use referential integrity with
1042 default-logging tables. A table without dirty buffers during a
1043 crash could perhaps avoid the drop/truncate.
1045 * Allow WAL logging to be turned off for a table, but the table would
1046 avoid being truncated/dropped [walcontrol]
1048 To do this, only a single writer can modify the table, and writes
1049 must happen only on new pages so the new pages can be removed during
1050 crash recovery. Readers can continue accessing the table. Such
1051 tables probably cannot have indexes. One complexity is the handling
1052 of indexes on TOAST tables.
1055 Optimizer / Executor
1056 ====================
1058 * Improve selectivity functions for geometric operators
1059 * Allow ORDER BY ... LIMIT # to select high/low value without sort or
1060 index using a sequential scan for highest/lowest values
1062 Right now, if no index exists, ORDER BY ... LIMIT # requires we sort
1063 all values to return the high/low value. Instead The idea is to do a
1064 sequential scan to find the high/low value, thus avoiding the sort.
1065 MIN/MAX already does this, but not for LIMIT > 1.
1067 * Precompile SQL functions to avoid overhead
1068 * Create utility to compute accurate random_page_cost value
1069 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
1070 * Have EXPLAIN ANALYZE highlight poor optimizer estimates
1071 * Consider using hash buckets to do DISTINCT, rather than sorting
1073 This would be beneficial when there are few distinct values. This is
1074 already used by GROUP BY.
1076 * Log statements where the optimizer row estimates were dramatically
1077 different from the number of rows actually found?
1080 Miscellaneous Performance
1081 =========================
1083 * Do async I/O for faster random read-ahead of data
1085 Async I/O allows multiple I/O requests to be sent to the disk with
1086 results coming back asynchronously.
1088 * Use mmap() rather than SYSV shared memory or to write WAL files?
1090 This would remove the requirement for SYSV SHM but would introduce
1091 portability issues. Anonymous mmap (or mmap to /dev/zero) is required
1092 to prevent I/O overhead.
1094 * Consider mmap()'ing files into a backend?
1096 Doing I/O to large tables would consume a lot of address space or
1097 require frequent mapping/unmapping. Extending the file also causes
1098 mapping problems that might require mapping only individual pages,
1099 leading to thousands of mappings. Another problem is that there is no
1100 way to _prevent_ I/O to disk from the dirty shared buffers so changes
1101 could hit disk before WAL is written.
1103 * Add a script to ask system configuration questions and tune postgresql.conf
1104 * Merge xmin/xmax/cmin/cmax back into three header fields
1106 Before subtransactions, there used to be only three fields needed to
1107 store these four values. This was possible because only the current
1108 transaction looks at the cmin/cmax values. If the current transaction
1109 created and expired the row the fields stored where xmin (same as
1110 xmax), cmin, cmax, and if the transaction was expiring a row from a
1111 another transaction, the fields stored were xmin (cmin was not
1112 needed), xmax, and cmax. Such a system worked because a transaction
1113 could only see rows from another completed transaction. However,
1114 subtransactions can see rows from outer transactions, and once the
1115 subtransaction completes, the outer transaction continues, requiring
1116 the storage of all four fields. With subtransactions, an outer
1117 transaction can create a row, a subtransaction expire it, and when the
1118 subtransaction completes, the outer transaction still has to have
1119 proper visibility of the row's cmin, for example, for cursors.
1121 One possible solution is to create a phantom cid which represents a
1122 cmin/cmax pair and is stored in local memory. Another idea is to
1123 store both cmin and cmax only in local memory.
1125 * Research storing disk pages with no alignment/padding
1131 * Add use of 'const' for variables in source tree
1132 * Rename some /contrib modules from pg* to pg_*
1133 * Move some things from /contrib into main tree
1134 * Move some /contrib modules out to their own project sites
1135 * %Remove warnings created by -Wcast-align
1136 * Move platform-specific ps status display info from ps_status.c to ports
1137 * Add optional CRC checksum to heap and index pages
1138 * Improve documentation to build only interfaces (Marc)
1139 * Remove or relicense modules that are not under the BSD license, if possible
1140 * %Remove memory/file descriptor freeing before ereport(ERROR)
1141 * Acquire lock on a relation before building a relcache entry for it
1142 * %Promote debug_query_string into a server-side function current_query()
1143 * %Allow the identifier length to be increased via a configure option
1144 * Allow cross-compiling by generating the zic database on the target system
1145 * Improve NLS maintenance of libpgport messages linked onto applications
1146 * Allow ecpg to work with MSVC and BCC
1147 * Add xpath_array() to /contrib/xml2 to return results as an array
1148 * Allow building in directories containing spaces
1150 This is probably not possible because 'gmake' and other compiler tools
1151 do not fully support quoting of paths with spaces.
1153 * -Allow installing to directories containing spaces
1155 This is possible if proper quoting is added to the makefiles for the
1156 install targets. Because PostgreSQL supports relocatable installs, it
1157 is already possible to install into a directory that doesn't contain
1158 spaces and then copy the install to a directory with spaces.
1160 * Fix sgmltools so PDFs can be generated with bookmarks
1161 * -%Clean up compiler warnings (especially with gcc version 4)
1162 * Use UTF8 encoding for NLS messages so all server encodings can
1164 * Update Bonjour to work with newer cross-platform SDK
1165 * -Remove BeOS and QNX-specific code
1166 * Split out libpq pgpass and environment documentation sections to make
1167 it easier for non-developers to find
1172 o Remove configure.in check for link failure when cause is found
1173 o Remove readdir() errno patch when runtime/mingwex/dirent.c rev
1175 o Remove psql newline patch when we find out why mingw outputs an
1177 o Allow psql to use readline once non-US code pages work with
1179 o Re-enable timezone output on log_line_prefix '%t' when a
1180 shorter timezone string is available
1181 o Fix problem with shared memory on the Win32 Terminal Server
1182 o Improve signal handling,
1183 http://archives.postgresql.org/pgsql-patches/2005-06/msg00027.php
1184 o Add long file support for binary pg_dump output
1186 While Win32 supports 64-bit files, the MinGW API does not,
1187 meaning we have to build an fseeko replacement on top of the
1188 Win32 API, and we have to make sure MinGW handles it. Another
1189 option is to wait for the MinGW project to fix it, or use the
1190 code from the LibGW32C project as a guide.
1193 * Wire Protocol Changes
1195 o Allow dynamic character set handling
1196 o Add decoded type, length, precision
1198 o Update clients to use data types, typmod, schema.table.column names
1199 of result sets using new statement protocol
1202 ---------------------------------------------------------------------------
1205 Developers who have claimed items are:
1206 --------------------------------------
1207 * Alvaro is Alvaro Herrera <alvherre@dcc.uchile.cl>
1208 * Andrew is Andrew Dunstan <andrew@dunslane.net>
1209 * Bruce is Bruce Momjian <pgman@candle.pha.pa.us> of Software Research Assoc.
1210 * Christopher is Christopher Kings-Lynne <chriskl@familyhealth.com.au> of
1211 Family Health Network
1212 * D'Arcy is D'Arcy J.M. Cain <darcy@druid.net> of The Cain Gang Ltd.
1213 * Fabien is Fabien Coelho <coelho@cri.ensmp.fr>
1214 * Gavin is Gavin Sherry <swm@linuxworld.com.au> of Alcove Systems Engineering
1215 * Greg is Greg Sabino Mullane <greg@turnstep.com>
1216 * Jan is Jan Wieck <JanWieck@Yahoo.com> of Afilias, Inc.
1217 * Joe is Joe Conway <mail@joeconway.com>
1218 * Karel is Karel Zak <zakkr@zf.jcu.cz>
1219 * Magnus is Magnus Hagander <mha@sollentuna.net>
1220 * Marc is Marc Fournier <scrappy@hub.org> of PostgreSQL, Inc.
1221 * Matthew T. O'Connor <matthew@zeut.net>
1222 * Michael is Michael Meskes <meskes@postgresql.org> of Credativ
1223 * Neil is Neil Conway <neilc@samurai.com>
1224 * Oleg is Oleg Bartunov <oleg@sai.msu.su>
1225 * Peter is Peter Eisentraut <peter_e@gmx.net>
1226 * Philip is Philip Warner <pjw@rhyme.com.au> of Albatross Consulting Pty. Ltd.
1227 * Rod is Rod Taylor <pg@rbt.ca>
1228 * Simon is Simon Riggs <simon@2ndquadrant.com>
1229 * Stephan is Stephan Szabo <sszabo@megazone23.bigpanda.com>
1230 * Tatsuo is Tatsuo Ishii <t-ishii@sra.co.jp> of Software Research Assoc.
1231 * Teodor is Teodor Sigaev <teodor@sigaev.ru>
1232 * Tom is Tom Lane <tgl@sss.pgh.pa.us> of Red Hat