4 Current maintainer: Bruce Momjian (pgman@candle.pha.pa.us)
5 Last updated: Sat Mar 4 00:02:00 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 to_date() and to_timestamp() accept localized month names
286 * Add missing parameter handling in to_char()
288 http://archives.postgresql.org/pgsql-hackers/2005-12/msg00948.php
290 * Allow functions to have a schema search path specified at creation time
291 * Allow substring/replace() to get/set bit values
292 * Allow to_char() on interval values to accumulate the highest unit
295 Some special format flag would be required to request such
296 accumulation. Such functionality could also be added to EXTRACT.
297 Prevent accumulation that crosses the month/day boundary because of
298 the uneven number of days in a month.
300 o to_char(INTERVAL '1 hour 5 minutes', 'MI') => 65
301 o to_char(INTERVAL '43 hours 20 minutes', 'MI' ) => 2600
302 o to_char(INTERVAL '43 hours 20 minutes', 'WK:DD:HR:MI') => 0:1:19:20
303 o to_char(INTERVAL '3 years 5 months','MM') => 41
305 * -Add sleep() function, remove from regress.c
306 * Allow user-defined functions retuning a domain value to enforce domain
308 * Add SPI_gettypmod() to return the typemod for a TupleDesc
311 Multi-Language Support
312 ======================
314 * Add NCHAR (as distinguished from ordinary varchar),
315 * Allow locale to be set at database creation
317 Currently locale can only be set during initdb. No global tables have
318 locale-aware columns. However, the database template used during
319 database creation might have locale-aware indexes. The indexes would
320 need to be reindexed to match the new locale.
322 * Allow encoding on a per-column basis
324 Right now only one encoding is allowed per database.
326 * Support multiple simultaneous character sets, per SQL92
327 * Improve UTF8 combined character handling?
328 * Add octet_length_server() and octet_length_client()
329 * Make octet_length_client() the same as octet_length()?
330 * Fix problems with wrong runtime encoding conversion for NLS message files
336 * Automatically create rules on views so they are updateable, per SQL99
338 We can only auto-create rules for simple views. For more complex
339 cases users will still have to write rules manually.
341 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
342 * Allow NOTIFY in rules involving conditionals
343 * Allow VIEW/RULE recompilation when the underlying tables change
345 Another issue is whether underlying table changes should be reflected
346 in the view, e.g. should SELECT * show additional columns if they
347 are added after the view is created.
353 * Change LIMIT/OFFSET and FETCH/MOVE to use int8
354 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
355 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
356 * %Allow SET CONSTRAINTS to be qualified by schema/table name
357 * -Allow TRUNCATE ... CASCADE/RESTRICT
359 This is like DELETE CASCADE, but truncates.
361 * %Add a separate TRUNCATE permission
363 Currently only the owner can TRUNCATE a table because triggers are not
364 called, and the table is locked in exclusive mode.
366 * Allow PREPARE of cursors
367 * -Allow PREPARE to automatically determine parameter types based on the SQL
369 * Allow finer control over the caching of prepared query plans
371 Currently, queries prepared via the libpq API are planned on first
372 execute using the supplied parameters --- allow SQL PREPARE to do the
373 same. Also, allow control over replanning prepared queries either
374 manually or automatically when statistics for execute parameters
375 differ dramatically from those used during planning.
377 * Allow LISTEN/NOTIFY to store info in memory rather than tables?
379 Currently LISTEN/NOTIFY information is stored in pg_listener. Storing
380 such information in memory would improve performance.
382 * Add optional textual message to NOTIFY
384 This would allow an informational message to be added to the notify
385 message, perhaps indicating the row modified or other custom
388 * Add a GUC variable to warn about non-standard SQL usage in queries
389 * Add SQL-standard MERGE command, typically used to merge two tables
392 This is similar to UPDATE, then for unmatched rows, INSERT.
393 Whether concurrent access allows modifications which could cause
394 row loss is implementation independent.
396 * Add REPLACE or UPSERT command that does UPDATE, or on failure, INSERT
399 To implement this cleanly requires that the table have a unique index
400 so duplicate checking can be easily performed. It is possible to
401 do it without a unique index if we require the user to LOCK the table
404 * Add NOVICE output level for helpful messages like automatic sequence/index
406 * -Add COMMENT ON for all cluster global objects (roles, databases
408 * -Make row-wise comparisons work per SQL spec
410 Right now, '(a, b) < (1, 2)' is processed as 'a < 1 and b < 2', but
411 the SQL standard requires it to be processed as a column-by-column
412 comparison, so the proper comparison is '(a < 1) OR (a = 1 AND b < 2)'.
414 * Add RESET CONNECTION command to reset all session state
416 This would include resetting of all variables (RESET ALL), dropping of
417 temporary tables, removing any NOTIFYs, cursors, open transactions,
418 prepared queries, currval()s, etc. This could be used for connection
419 pooling. We could also change RESET ALL to have this functionality.
420 The difficult of this features is allowing RESET ALL to not affect
421 changes made by the interface driver for its internal use. One idea
422 is for this to be a protocol-only feature. Another approach is to
423 notify the protocol when a RESET CONNECTION command is used.
425 * Add GUC to issue notice about statements that use unjoined tables
426 * Allow EXPLAIN to identify tables that were skipped because of
428 * Allow EXPLAIN output to be more easily processed by scripts
429 * Eventually enable escape_string_warning and standard_conforming_strings
430 * Simplify dropping roles that have objects in several databases
431 * Allow COMMENT ON to accept an expression rather than just a string
432 * Allow the count returned by SELECT, etc to be to represent as an int64
433 to allow a higher range of values
434 * Make CLUSTER preserve recently-dead tuples per MVCC requirements
435 * Add SQL99 WITH clause to SELECT
436 * Add SQL99 WITH RECURSIVE to SELECT
441 o Allow CREATE TABLE AS to determine column lengths for complex
442 expressions like SELECT col1 || col2
443 o Use more reliable method for CREATE DATABASE to get a consistent
445 o -Add ON COMMIT capability to CREATE TABLE AS ... SELECT
449 o Allow UPDATE to handle complex aggregates [update]?
450 o -Allow an alias to be provided for the target table in
452 o Allow UPDATE tab SET ROW (col, ...) = (...) for updating multiple
458 o %Have ALTER TABLE RENAME rename SERIAL sequence names
459 o Add ALTER DOMAIN to modify the underlying data type
460 o %Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
461 o %Allow ALTER TABLE to change constraint deferrability and actions
462 o Add missing object types for ALTER ... SET SCHEMA
463 o Allow ALTER TABLESPACE to move to different directories
464 o Allow databases to be moved to different tablespaces
465 o Allow moving system tables to other tablespaces, where possible
467 Currently non-global system tables must be in the default database
468 tablespace. Global system tables can never be moved.
470 o %Prevent child tables from altering or dropping constraints
471 like CHECK that were inherited from the parent table
472 o Have ALTER INDEX update the name of a constraint using that index
473 o Add ALTER TABLE RENAME CONSTRAINT, update index name also
478 o Automatically maintain clustering on a table
480 This might require some background daemon to maintain clustering
481 during periods of low usage. It might also require tables to be only
482 partially filled for easier reorganization. Another idea would
483 be to create a merged heap/index data file so an index lookup would
484 automatically access the heap data too. A third idea would be to
485 store heap rows in hashed groups, perhaps using a user-supplied
488 o %Add default clustering to system tables
490 To do this, determine the ideal cluster index for each system
491 table and set the cluster setting during initdb.
496 o Allow COPY to report error lines and continue
498 This requires the use of a savepoint before each COPY line is
499 processed, with ROLLBACK on COPY failure.
501 o -Have COPY return the number of rows loaded/unloaded?
502 o Allow COPY on a newly-created table to skip WAL logging
504 On crash recovery, the table involved in the COPY would
505 be removed or have its heap and index files truncated. One
506 issue is that no other backend should be able to add to
507 the table at the same time, which is something that is
510 o Allow COPY to output from views
512 Another idea would be to allow actual SELECT statements in a COPY.
517 o Allow column-level privileges
518 o %Allow GRANT/REVOKE permissions to be applied to all schema objects
521 The proposed syntax is:
522 GRANT SELECT ON ALL TABLES IN public TO phpuser;
523 GRANT SELECT ON NEW TABLES IN public TO phpuser;
525 * Allow GRANT/REVOKE permissions to be inherited by objects based on
528 * Allow SERIAL sequences to inherit permissions from the base table?
533 o Allow UPDATE/DELETE WHERE CURRENT OF cursor
535 This requires using the row ctid to map cursor rows back to the
536 original heap row. This become more complicated if WITH HOLD cursors
537 are to be supported because WITH HOLD cursors have a copy of the row
538 and no FOR UPDATE lock.
540 o Prevent DROP TABLE from dropping a row referenced by its own open
543 o -Allow pooled connections to list all open WITH HOLD cursors
545 Because WITH HOLD cursors exist outside transactions, this allows
546 them to be listed so they can be closed.
551 o Allow INSERT/UPDATE of the system-generated oid value for a row
552 o Allow INSERT INTO tab (col1, ..) VALUES (val1, ..), (val2, ..)
553 o Allow INSERT/UPDATE ... RETURNING new.col or old.col
555 This is useful for returning the auto-generated key for an INSERT.
556 One complication is how to handle rules that run as part of
562 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
564 o Add SET PATH for schemas?
566 This is basically the same as SET search_path.
569 * Server-Side Languages
571 o Fix PL/pgSQL RENAME to work on variables other than OLD/NEW
572 o Allow function parameters to be passed by name,
573 get_employee_salary(emp_id => 12345, tax_year => 2001)
574 o Add Oracle-style packages
575 o Add table function support to pltcl, plpython
576 o Add capability to create and call PROCEDURES
577 o Allow PL/pgSQL to handle %TYPE arrays, e.g. tab.col%TYPE[]
578 o Allow function argument names to be statements from PL/PgSQL
579 o Add MOVE to PL/pgSQL
580 o Add support for polymorphic arguments and return types to
581 languages other than PL/PgSQL
582 o Add support for OUT and INOUT parameters to languages other
584 o Add single-step debugging of PL/PgSQL functions
585 o Allow PL/PgSQL to support WITH HOLD cursors
591 * -Have initdb set the input DateStyle (MDY or DMY) based on locale
592 * Have pg_ctl look at PGHOST in case it is a socket directory?
593 * Allow pg_ctl to work properly with configuration files located outside
596 pg_ctl can not read the pid file because it isn't located in the
597 config directory but in the PGDATA directory. The solution is to
598 allow pg_ctl to read and understand postgresql.conf to find the
599 data_directory value.
604 o Have psql show current values for a sequence
605 o Move psql backslash database information into the backend, use
606 mnemonic commands? [psql]
608 This would allow non-psql clients to pull the same information out
609 of the database as psql.
611 o Fix psql's display of schema information (Neil)
612 o Allow psql \pset boolean variables to set to fixed values, rather
614 o Consistently display privilege information for all objects in psql
615 o -Improve psql's handling of multi-line statements
617 Currently, while \e saves a single statement as one entry, interactive
618 statements are saved one line at a time. Ideally all statements
619 would be saved like \e does.
621 o -Allow multi-line column values to align in the proper columns
623 If the second output column value is 'a\nb', the 'b' should appear
624 in the second display column, rather than the first column as it
627 o Display IN, INOUT, and OUT parameters in \df+
629 It probably requires psql to output newlines in the proper
630 column, which is already on the TODO list.
632 o Add auto-expanded mode so expanded output is used if the row
633 length is wider than the screen width.
635 Consider using auto-expanded mode for backslash commands like \df+.
637 o Prevent tab completion of SET TRANSACTION from querying the
638 database and therefore preventing the transaction isolation
639 level from being set.
641 Currently, SET <tab> causes a database lookup to check all
642 supported session variables. This query causes problems
643 because setting the transaction isolation level must be the
644 first statement of a transaction.
649 o %Have pg_dump use multi-statement transactions for INSERT dumps
650 o %Allow pg_dump to use multiple -t and -n switches [pg_dump]
651 o %Add dumping of comments on index columns and composite type columns
652 o %Add full object name to the tag field. eg. for operators we need
653 '=(integer, integer)', instead of just '='.
654 o Add pg_dumpall custom format dumps?
655 o %Add CSV output format
656 o Update pg_dump and psql to use the new COPY libpq API (Christopher)
657 o Remove unnecessary function pointer abstractions in pg_dump source
659 o Allow selection of individual object(s) of all types, not just
661 o In a selective dump, allow dumping of an object and all its
663 o Add options like pg_restore -l and -L to pg_dump
664 o Stop dumping CASCADE on DROP TYPE commands in clean mode
665 o Allow pg_dump --clean to drop roles that own objects or have
667 o Add -f to pg_dumpall
674 Document differences between ecpg and the SQL standard and
675 information about the Informix-compatibility module.
677 o Solve cardinality > 1 for input descriptors / variables?
678 o Add a semantic check level, e.g. check if a table really exists
679 o fix handling of DB attributes that are arrays
680 o Use backend PREPARE/EXECUTE facility for ecpg where possible
682 o Fix nested C comments
683 o %sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
684 o Make SET CONNECTION thread-aware, non-standard?
685 o Allow multidimensional arrays
686 o Add internationalized message strings
691 o Add a function to support Parse/DescribeStatement capability
692 o Add PQescapeIdentifier()
693 o Prevent PQfnumber() from lowercasing unquoted the column name
695 PQfnumber() should never have been doing lowercasing, but
696 historically it has so we need a way to prevent it
698 o Allow statement results to be automatically batched to the client
700 Currently, all statement results are transferred to the libpq
701 client before libpq makes the results available to the
702 application. This feature would allow the application to make
703 use of the first result rows while the rest are transferred, or
704 held on the server waiting for them to be requested by libpq.
705 One complexity is that a statement like SELECT 1/col could error
706 out mid-way through the result set.
709 Referential Integrity
710 =====================
712 * Add MATCH PARTIAL referential integrity
713 * Add deferred trigger queue file
715 Right now all deferred trigger information is stored in backend
716 memory. This could exhaust memory for very large trigger queues.
717 This item involves dumping large queues into files.
719 * Change foreign key constraint for array -> element to mean element
721 * Allow DEFERRABLE UNIQUE constraints?
722 * Allow triggers to be disabled in only the current session.
724 This is currently possible by starting a multi-statement transaction,
725 modifying the system tables, performing the desired SQL, restoring the
726 system tables, and committing the transaction. ALTER TABLE ...
727 TRIGGER requires a table lock so it is not ideal for this usage.
729 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
731 If the dump is known to be valid, allow foreign keys to be added
732 without revalidating the data.
734 * Allow statement-level triggers to access modified rows
735 * Support triggers on columns (Greg Sabino Mullane)
736 * Enforce referential integrity for system tables
737 * Allow AFTER triggers on system tables
739 System tables are modified in many places in the backend without going
740 through the executor and therefore not causing triggers to fire. To
741 complete this item, the functions that modify system tables will have
748 * Flush cached query plans when the dependent objects change,
749 when the cardinality of parameters changes dramatically, or
750 when new ANALYZE statistics are available
752 A more complex solution would be to save multiple plans for different
753 cardinality and use the appropriate plan based on the EXECUTE values.
755 * Track dependencies in function bodies and recompile/invalidate
757 This is particularly important for references to temporary tables
758 in PL/PgSQL because PL/PgSQL caches query plans. The only workaround
759 in PL/PgSQL is to use EXECUTE. One complexity is that a function
760 might itself drop and recreate dependent tables, causing it to
761 invalidate its own query plan.
767 * Add pre-parsing phase that converts non-ISO syntax to supported
770 This could allow SQL written for other databases to run without
773 * Allow plug-in modules to emulate features from other databases
774 * SQL*Net listener that makes PostgreSQL appear as an Oracle database
776 * Allow statements across databases or servers with transaction
779 This can be done using dblink and two-phase commit.
781 * Add the features of packages
783 o Make private objects accessible only to objects in the same schema
784 o Allow current_schema.objname to access current schema objects
785 o Add session variables
786 o Allow nested schemas
792 * Allow inherited tables to inherit index, UNIQUE constraint, and primary
794 * UNIQUE INDEX on base column not honored on INSERTs/UPDATEs from
795 inherited table: INSERT INTO inherit_table (unique_index_col) VALUES
798 The main difficulty with this item is the problem of creating an index
799 that can span more than one table.
801 * Allow SELECT ... FOR UPDATE on inherited tables
802 * Add UNIQUE capability to non-btree indexes
803 * Prevent index uniqueness checks when UPDATE does not modify the column
805 Uniqueness (index) checks are done when updating a column even if the
806 column is not modified by the UPDATE.
808 * Allow the creation of on-disk bitmap indexes which can be quickly
809 combined with other bitmap indexes
811 Such indexes could be more compact if there are only a few distinct values.
812 Such indexes can also be compressed. Keeping such indexes updated can be
815 * Allow use of indexes to search for NULLs
817 One solution is to create a partial index on an IS NULL expression.
819 * Allow accurate statistics to be collected on indexes with more than
820 one column or expression indexes, perhaps using per-index statistics
821 * Add fillfactor to control reserved free space during index creation
822 * Allow the creation of indexes with mixed ascending/descending specifiers
823 * Allow constraint_exclusion to work for UNIONs like it does for
824 inheritance, allow it to work for UPDATE and DELETE statements, and allow
825 it to be used for all statements with little performance impact
826 * Allow CREATE INDEX to take an additional parameter for use with
828 * Consider compressing indexes by storing key values duplicated in
829 several rows as a single index entry
831 This is difficult because it requires datatype-specific knowledge.
836 o Add more GIST index support for geometric data types
837 o Allow GIST indexes to create certain complex index types, like
838 digital trees (see Aoki)
842 o Pack hash index buckets onto disk pages more efficiently
844 Currently only one hash bucket can be stored on a page. Ideally
845 several hash buckets could be stored on a single page and greater
846 granularity used for the hash algorithm.
848 o Consider sorting hash buckets so entries can be found using a
849 binary search, rather than a linear scan
851 o In hash indexes, consider storing the hash value with or instead
854 o Add WAL logging for crash recovery
855 o Allow multi-column hash indexes
861 * Improve commit_delay handling to reduce fsync()
862 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
864 Ideally this requires a separate test program that can be run
865 at initdb time or optionally later. Consider O_SYNC when
868 * %Add an option to sync() before fsync()'ing checkpoint files
869 * Add program to test if fsync has a delay compared to non-fsync
875 * Allow free-behind capability for large sequential scans, perhaps using
878 Posix_fadvise() can control both sequential/random file caching and
879 free-behind behavior, but it is unclear how the setting affects other
880 backends that also have the file open, and the feature is not supported
881 on all operating systems.
885 We could use a fixed row count and a +/- count to follow MVCC
886 visibility rules, or a single cached value could be used and
887 invalidated if anyone modifies the table. Another idea is to
888 get a count directly from a unique index, but for this to be
889 faster than a sequential scan it must avoid access to the heap
890 to obtain tuple visibility information.
892 * Add estimated_count(*) to return an estimate of COUNT(*)
894 This would use the planner ANALYZE statistics to return an estimated
897 * Allow data to be pulled directly from indexes
899 Currently indexes do not have enough tuple visibility information
900 to allow data to be pulled from the index without also accessing
901 the heap. One way to allow this is to set a bit on index tuples
902 to indicate if a tuple is currently visible to all transactions
903 when the first valid heap lookup happens. This bit would have to
904 be cleared when a heap tuple is expired.
906 Another idea is to maintain a bitmap of heap pages where all rows
907 are visible to all backends, and allow index lookups to reference
908 that bitmap to avoid heap lookups, perhaps the same bitmap we might
909 add someday to determine which heap pages need vacuuming. Frequently
910 accessed bitmaps would have to be stored in shared memory. One 8k
911 page of bitmaps could track 512MB of heap pages.
913 * Consider automatic caching of statements at various levels:
919 * Allow sequential scans to take advantage of other concurrent
920 sequential scans, also called "Synchronised Scanning"
922 One possible implementation is to start sequential scans from the lowest
923 numbered buffer in the shared cache, and when reaching the end wrap
924 around to the beginning, rather than always starting sequential scans
925 at the start of the table.
931 * Improve speed with indexes
933 For large table adjustments during VACUUM FULL, it is faster to
934 reindex rather than update the index.
936 * Reduce lock time during VACUUM FULL by moving tuples with read lock,
937 then write lock and truncate table
939 Moved tuples are invisible to other backends so they don't require a
940 write lock. However, the read lock promotion to write lock could lead
941 to deadlock situations.
943 * Auto-fill the free space map by scanning the buffer cache or by
944 checking pages written by the background writer
945 * Create a bitmap of pages that need vacuuming
947 Instead of sequentially scanning the entire table, have the background
948 writer or some other process record pages that have expired rows, then
949 VACUUM can look at just those pages rather than the entire table. In
950 the event of a system crash, the bitmap would probably be invalidated.
951 One complexity is that index entries still have to be vacuumed, and
952 doing this without an index scan (by using the heap values to find the
953 index entry) might be slow and unreliable, especially for user-defined
956 * -Add system view to show free space map contents
957 * Allow FSM to return free space toward the beginning of the heap file,
958 in hopes that empty pages at the end can be truncated by VACUUM
959 * Allow FSM page return free space based on table clustering, to assist
960 in maintaining clustering?
965 o Use free-space map information to guide refilling
966 o %Issue log message to suggest VACUUM FULL if a table is nearly
968 o Improve xid wraparound detection by recording per-table rather
970 o Consider logging activity either to the logs or a system view
976 * Fix priority ordering of read and write light-weight locks (Neil)
979 Startup Time Improvements
980 =========================
982 * Experiment with multi-threaded backend [thread]
984 This would prevent the overhead associated with process creation. Most
985 operating systems have trivial process creation time compared to
986 database startup overhead, but a few operating systems (Win32,
987 Solaris) might benefit from threading. Also explore the idea of
988 a single session using multiple threads to execute a statement faster.
990 * Add connection pooling
992 It is unclear if this should be done inside the backend code or done
993 by something external like pgpool. The passing of file descriptors to
994 existing backends is one of the difficulties with a backend approach.
1000 * Eliminate need to write full pages to WAL before page modification [wal]
1002 Currently, to protect against partial disk page writes, we write
1003 full page images to WAL before they are modified so we can correct any
1004 partial page writes during recovery. These pages can also be
1005 eliminated from point-in-time archive files.
1007 o When off, write CRC to WAL and check file system blocks
1010 If CRC check fails during recovery, remember the page in case
1011 a later CRC for that page properly matches.
1013 o Write full pages during file system write and not when
1014 the page is modified in the buffer cache
1016 This allows most full page writes to happen in the background
1017 writer. It might cause problems for applying WAL on recovery
1018 into a partially-written page, but later the full page will be
1021 * Allow WAL traffic to be streamed to another server for stand-by
1023 * Reduce WAL traffic so only modified values are written rather than
1025 * Allow the pg_xlog directory location to be specified during initdb
1026 with a symlink back to the /data location
1027 * Allow WAL information to recover corrupted pg_controldata
1028 * Find a way to reduce rotational delay when repeatedly writing
1031 Currently fsync of WAL requires the disk platter to perform a full
1032 rotation to fsync again. One idea is to write the WAL to different
1033 offsets that might reduce the rotational delay.
1035 * Allow buffered WAL writes and fsync
1037 Instead of guaranteeing recovery of all committed transactions, this
1038 would provide improved performance by delaying WAL writes and fsync
1039 so an abrupt operating system restart might lose a few seconds of
1040 committed transactions but still be consistent. We could perhaps
1041 remove the 'fsync' parameter (which results in an an inconsistent
1042 database) in favor of this capability.
1044 * Allow WAL logging to be turned off for a table, but the table
1045 might be dropped or truncated during crash recovery [walcontrol]
1047 Allow tables to bypass WAL writes and just fsync() dirty pages on
1048 commit. This should be implemented using ALTER TABLE, e.g. ALTER
1049 TABLE PERSISTENCE [ DROP | TRUNCATE | DEFAULT ]. Tables using
1050 non-default logging should not use referential integrity with
1051 default-logging tables. A table without dirty buffers during a
1052 crash could perhaps avoid the drop/truncate.
1054 * Allow WAL logging to be turned off for a table, but the table would
1055 avoid being truncated/dropped [walcontrol]
1057 To do this, only a single writer can modify the table, and writes
1058 must happen only on new pages so the new pages can be removed during
1059 crash recovery. Readers can continue accessing the table. Such
1060 tables probably cannot have indexes. One complexity is the handling
1061 of indexes on TOAST tables.
1064 Optimizer / Executor
1065 ====================
1067 * Improve selectivity functions for geometric operators
1068 * Allow ORDER BY ... LIMIT # to select high/low value without sort or
1069 index using a sequential scan for highest/lowest values
1071 Right now, if no index exists, ORDER BY ... LIMIT # requires we sort
1072 all values to return the high/low value. Instead The idea is to do a
1073 sequential scan to find the high/low value, thus avoiding the sort.
1074 MIN/MAX already does this, but not for LIMIT > 1.
1076 * Precompile SQL functions to avoid overhead
1077 * Create utility to compute accurate random_page_cost value
1078 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
1079 * Have EXPLAIN ANALYZE highlight poor optimizer estimates
1080 * Consider using hash buckets to do DISTINCT, rather than sorting
1082 This would be beneficial when there are few distinct values. This is
1083 already used by GROUP BY.
1085 * Log statements where the optimizer row estimates were dramatically
1086 different from the number of rows actually found?
1089 Miscellaneous Performance
1090 =========================
1092 * Do async I/O for faster random read-ahead of data
1094 Async I/O allows multiple I/O requests to be sent to the disk with
1095 results coming back asynchronously.
1097 * Use mmap() rather than SYSV shared memory or to write WAL files?
1099 This would remove the requirement for SYSV SHM but would introduce
1100 portability issues. Anonymous mmap (or mmap to /dev/zero) is required
1101 to prevent I/O overhead.
1103 * Consider mmap()'ing files into a backend?
1105 Doing I/O to large tables would consume a lot of address space or
1106 require frequent mapping/unmapping. Extending the file also causes
1107 mapping problems that might require mapping only individual pages,
1108 leading to thousands of mappings. Another problem is that there is no
1109 way to _prevent_ I/O to disk from the dirty shared buffers so changes
1110 could hit disk before WAL is written.
1112 * Add a script to ask system configuration questions and tune postgresql.conf
1113 * Merge xmin/xmax/cmin/cmax back into three header fields
1115 Before subtransactions, there used to be only three fields needed to
1116 store these four values. This was possible because only the current
1117 transaction looks at the cmin/cmax values. If the current transaction
1118 created and expired the row the fields stored where xmin (same as
1119 xmax), cmin, cmax, and if the transaction was expiring a row from a
1120 another transaction, the fields stored were xmin (cmin was not
1121 needed), xmax, and cmax. Such a system worked because a transaction
1122 could only see rows from another completed transaction. However,
1123 subtransactions can see rows from outer transactions, and once the
1124 subtransaction completes, the outer transaction continues, requiring
1125 the storage of all four fields. With subtransactions, an outer
1126 transaction can create a row, a subtransaction expire it, and when the
1127 subtransaction completes, the outer transaction still has to have
1128 proper visibility of the row's cmin, for example, for cursors.
1130 One possible solution is to create a phantom cid which represents a
1131 cmin/cmax pair and is stored in local memory. Another idea is to
1132 store both cmin and cmax only in local memory.
1134 * Research storing disk pages with no alignment/padding
1140 * Add use of 'const' for variables in source tree
1141 * Rename some /contrib modules from pg* to pg_*
1142 * Move some things from /contrib into main tree
1143 * Move some /contrib modules out to their own project sites
1144 * %Remove warnings created by -Wcast-align
1145 * Move platform-specific ps status display info from ps_status.c to ports
1146 * Add optional CRC checksum to heap and index pages
1147 * Improve documentation to build only interfaces (Marc)
1148 * Remove or relicense modules that are not under the BSD license, if possible
1149 * %Remove memory/file descriptor freeing before ereport(ERROR)
1150 * Acquire lock on a relation before building a relcache entry for it
1151 * %Promote debug_query_string into a server-side function current_query()
1152 * %Allow the identifier length to be increased via a configure option
1153 * Allow cross-compiling by generating the zic database on the target system
1154 * Improve NLS maintenance of libpgport messages linked onto applications
1155 * Allow ecpg to work with MSVC and BCC
1156 * Add xpath_array() to /contrib/xml2 to return results as an array
1157 * Allow building in directories containing spaces
1159 This is probably not possible because 'gmake' and other compiler tools
1160 do not fully support quoting of paths with spaces.
1162 * -Allow installing to directories containing spaces
1164 This is possible if proper quoting is added to the makefiles for the
1165 install targets. Because PostgreSQL supports relocatable installs, it
1166 is already possible to install into a directory that doesn't contain
1167 spaces and then copy the install to a directory with spaces.
1169 * Fix sgmltools so PDFs can be generated with bookmarks
1170 * -%Clean up compiler warnings (especially with gcc version 4)
1171 * Use UTF8 encoding for NLS messages so all server encodings can
1173 * Update Bonjour to work with newer cross-platform SDK
1174 * -Remove BeOS and QNX-specific code
1175 * Split out libpq pgpass and environment documentation sections to make
1176 it easier for non-developers to find
1177 * Improve port/qsort() to handle sorts with 50% unique and 50% duplicate
1180 This involves choosing better pivot points for the quicksort.
1185 o Remove configure.in check for link failure when cause is found
1186 o Remove readdir() errno patch when runtime/mingwex/dirent.c rev
1188 o Remove psql newline patch when we find out why mingw outputs an
1190 o Allow psql to use readline once non-US code pages work with
1192 o Re-enable timezone output on log_line_prefix '%t' when a
1193 shorter timezone string is available
1194 o Fix problem with shared memory on the Win32 Terminal Server
1195 o Improve signal handling,
1196 http://archives.postgresql.org/pgsql-patches/2005-06/msg00027.php
1197 o Add long file support for binary pg_dump output
1199 While Win32 supports 64-bit files, the MinGW API does not,
1200 meaning we have to build an fseeko replacement on top of the
1201 Win32 API, and we have to make sure MinGW handles it. Another
1202 option is to wait for the MinGW project to fix it, or use the
1203 code from the LibGW32C project as a guide.
1206 * Wire Protocol Changes
1208 o Allow dynamic character set handling
1209 o Add decoded type, length, precision
1211 o Update clients to use data types, typmod, schema.table.column names
1212 of result sets using new statement protocol
1215 ---------------------------------------------------------------------------
1218 Developers who have claimed items are:
1219 --------------------------------------
1220 * Alvaro is Alvaro Herrera <alvherre@dcc.uchile.cl>
1221 * Andrew is Andrew Dunstan <andrew@dunslane.net>
1222 * Bruce is Bruce Momjian <pgman@candle.pha.pa.us> of Software Research Assoc.
1223 * Christopher is Christopher Kings-Lynne <chriskl@familyhealth.com.au> of
1224 Family Health Network
1225 * D'Arcy is D'Arcy J.M. Cain <darcy@druid.net> of The Cain Gang Ltd.
1226 * Fabien is Fabien Coelho <coelho@cri.ensmp.fr>
1227 * Gavin is Gavin Sherry <swm@linuxworld.com.au> of Alcove Systems Engineering
1228 * Greg is Greg Sabino Mullane <greg@turnstep.com>
1229 * Jan is Jan Wieck <JanWieck@Yahoo.com> of Afilias, Inc.
1230 * Joe is Joe Conway <mail@joeconway.com>
1231 * Karel is Karel Zak <zakkr@zf.jcu.cz>
1232 * Magnus is Magnus Hagander <mha@sollentuna.net>
1233 * Marc is Marc Fournier <scrappy@hub.org> of PostgreSQL, Inc.
1234 * Matthew T. O'Connor <matthew@zeut.net>
1235 * Michael is Michael Meskes <meskes@postgresql.org> of Credativ
1236 * Neil is Neil Conway <neilc@samurai.com>
1237 * Oleg is Oleg Bartunov <oleg@sai.msu.su>
1238 * Peter is Peter Eisentraut <peter_e@gmx.net>
1239 * Philip is Philip Warner <pjw@rhyme.com.au> of Albatross Consulting Pty. Ltd.
1240 * Rod is Rod Taylor <pg@rbt.ca>
1241 * Simon is Simon Riggs <simon@2ndquadrant.com>
1242 * Stephan is Stephan Szabo <sszabo@megazone23.bigpanda.com>
1243 * Tatsuo is Tatsuo Ishii <t-ishii@sra.co.jp> of Software Research Assoc.
1244 * Teodor is Teodor Sigaev <teodor@sigaev.ru>
1245 * Tom is Tom Lane <tgl@sss.pgh.pa.us> of Red Hat