4 Current maintainer: Bruce Momjian (pgman@candle.pha.pa.us)
5 Last updated: Tue Apr 25 10:30:50 EDT 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?
53 * Allow log_min_messages to be specified on a per-module basis
55 This would allow administrators to see more detailed information from
56 specific sections of the backend, e.g. checkpoints, autovacuum, etc.
58 * -Re-enable the GUC full_page_writes in 8.2 when reliability issues have
62 * Improve replication solutions
66 You can use any of the master/slave replication servers to use a
67 standby server for data warehousing. To allow read/write queries to
68 multiple servers, you need multi-master replication like pgcluster.
70 o Allow replication over unreliable or non-persistent links
75 o -Add "include file" functionality in postgresql.conf
76 o %Allow commenting of variables in postgresql.conf to restore them
79 Currently, if a variable is commented out, it keeps the
80 previous uncommented value until a server restarted.
82 o %Allow per-database permissions to be set via GRANT
84 Allow database connection checks based on GRANT rules in
85 addition to the existing access checks in pg_hba.conf.
87 o Allow pg_hba.conf to specify host names along with IP addresses
89 Host name lookup could occur when the postmaster reads the
90 pg_hba.conf file, or when the backend starts. Another
91 solution would be to reverse lookup the connection IP and
92 check that hostname against the host names in pg_hba.conf.
93 We could also then check that the host name maps to the IP
96 o %Allow postgresql.conf file values to be changed via an SQL
97 API, perhaps using SET GLOBAL
98 o Allow the server to be stopped/restarted via an SQL API
99 o -Issue a warning if a change-on-restart-only postgresql.conf value
100 is modified and the server config files are reloaded
101 o Mark change-on-restart-only values in postgresql.conf
106 * Allow a database in tablespace t1 with tables created in
107 tablespace t2 to be used as a template for a new database created
108 with default tablespace t2
110 All objects in the default database tablespace must have default
111 tablespace specifications. This is because new databases are
112 created by copying directories. If you mix default tablespace
113 tables and tablespace-specified tables in the same directory,
114 creating a new database from such a mixed directory would create a
115 new database with tables that had incorrect explicit tablespaces.
116 To fix this would require modifying pg_class in the newly copied
117 database, which we don't currently do.
119 * Allow reporting of which objects are in which tablespaces
121 This item is difficult because a tablespace can contain objects
122 from multiple databases. There is a server-side function that
123 returns the databases which use a specific tablespace, so this
124 requires a tool that will call that function and connect to each
125 database to find the objects in each database for that tablespace.
127 o %Add a GUC variable to control the tablespace for temporary objects
130 It could start with a random tablespace from a supplied list and
131 cycle through the list.
133 o Allow WAL replay of CREATE TABLESPACE to work when the directory
134 structure on the recovery computer is different from the original
136 o Allow per-tablespace quotas
139 * Point-In-Time Recovery (PITR)
141 o Allow point-in-time recovery to archive partially filled
142 write-ahead logs [pitr]
144 Currently only full WAL files are archived. This means that the
145 most recent transactions aren't available for recovery in case
146 of a disk failure. This could be triggered by a user command or
149 o Automatically force archiving of partially-filled WAL files when
150 pg_stop_backup() is called or the server is stopped
152 Doing this will allow administrators to know more easily when
153 the archive contains all the files needed for point-in-time
156 o %Create dump tool for write-ahead logs for use in determining
157 transaction id for point-in-time recovery
158 o Allow a warm standby system to also allow read-only statements
161 This is useful for checking PITR recovery.
163 o Allow the PITR process to be debugged and data examined
164 o Add reporting of the current WAL file, perhaps as part of
165 partial log file archiving
171 * Allow server log information to be output as INSERT statements
173 This would allow server log information to be easily loaded into
174 a database for analysis.
176 * %Add ability to monitor the use of temporary sort files
177 * Allow server logs to be remotely read and removed using SQL commands
178 * Allow protocol-level BIND parameter values to be logged
184 * Improve the MONEY data type
186 Change the MONEY data type to use DECIMAL internally, with special
187 locale-aware output formatting.
189 * Change NUMERIC to enforce the maximum precision
190 * Add NUMERIC division operator that doesn't round?
192 Currently NUMERIC _rounds_ the result to the specified precision.
193 This means division can return a result that multiplied by the
194 divisor is greater than the dividend, e.g. this returns a value > 10:
196 SELECT (10::numeric(2,0) / 6::numeric(2,0))::numeric(2,0) * 6;
198 The positive modulus result returned by NUMERICs might be considered
199 inaccurate, in one sense.
201 * %Disallow changing default expression of a SERIAL column
202 * %Disallow ALTER SEQUENCE changes for SERIAL sequences because pg_dump
203 does not dump the changes
204 * Fix data types where equality comparison isn't intuitive, e.g. box
205 * -Zero umasked bits in conversion from INET cast to CIDR
206 * -Prevent INET cast to CIDR from dropping netmask, SELECT '1.1.1.1'::inet::cidr
207 * -Allow INET + INT8 to increment the host part of the address or
208 throw an error on overflow
209 * -Add 'tid != tid ' operator for use in corruption recovery
210 * Allow user-defined types to specify a type modifier at table creation
216 o Allow infinite dates and intervals just like infinite timestamps
217 o Merge hardwired timezone names with the TZ database; allow either
218 kind everywhere a TZ name is currently taken
219 o Allow customization of the known set of TZ names (generalize the
220 present australian_timezones hack)
221 o Allow TIMESTAMP WITH TIME ZONE to store the original timezone
222 information, either zone name or offset from UTC [timezone]
224 If the TIMESTAMP value is stored with a time zone name, interval
225 computations should adjust based on the time zone rules.
227 o Fix SELECT '0.01 years'::interval, '0.01 months'::interval
228 o Add a GUC variable to allow output of interval values in ISO8601
230 o Improve timestamptz subtraction to be DST-aware
232 Currently, subtracting one date from another that crosses a
233 daylight savings time adjustment can return '1 day 1 hour', but
234 adding that back to the first date returns a time one hour in
235 the future. This is caused by the adjustment of '25 hours' to
236 '1 day 1 hour', and '1 day' is the same time the next day, even
237 if daylight savings adjustments are involved.
239 o Fix interval display to support values exceeding 2^31 hours
240 o Add overflow checking to timestamp and interval arithmetic
241 o Add ISO INTERVAL handling
242 o -Add support for day-time syntax, INTERVAL '1 2:03:04' DAY TO
244 o Support ISO INTERVAL syntax if units cannot be determined from
245 the string, and are supplied after the string
247 The SQL standard states that the units after the string
248 specify the units of the string, e.g. INTERVAL '2' MINUTE
249 should return '00:02:00'. The current behavior has the units
250 restrict the interval value to the specified unit or unit
251 range, INTERVAL '70' SECOND returns '00:00:10'.
253 For syntax that isn't uniquely ISO or PG syntax, like '1' or
254 '1:30', treat as ISO if there is a range specification clause,
255 and as PG if there no clause is present, e.g. interpret '1:30'
256 MINUTE TO SECOND as '1 minute 30 seconds', and interpret
257 '1:30' as '1 hour, 30 minutes'.
259 This makes common cases like SELECT INTERVAL '1' MONTH
260 SQL-standard results. The SQL standard supports a limited
261 number of unit combinations and doesn't support unit names in
262 the string. The PostgreSQL syntax is more flexible in the
263 range of units supported, e.g. PostgreSQL supports '1 year 1
264 hour', while the SQL standard does not.
266 o Add support for year-month syntax, INTERVAL '50-6' YEAR TO MONTH
267 o Interpret INTERVAL '1 year' MONTH as CAST (INTERVAL '1 year' AS
268 INTERVAL MONTH), and this should return '12 months'
269 o Round or truncate values to the requested precision, e.g.
270 INTERVAL '11 months' AS YEAR should return one or zero
271 o Support precision, CREATE TABLE foo (a INTERVAL MONTH(3))
276 o -Allow NULLs in arrays
277 o Delay resolution of array expression's data type so assignment
278 coercion can be performed on empty array expressions
283 o Improve vacuum of large objects, like /contrib/vacuumlo?
284 o Add security checking for large objects
285 o Auto-delete large objects when referencing row is deleted
287 /contrib/lo offers this functionality.
289 o Allow read/write into TOAST values like large objects
291 This requires the TOAST column to be stored EXTERNAL.
297 * Allow INET subnet tests using non-constants to be indexed
298 * -Add transaction_timestamp(), statement_timestamp(), clock_timestamp()
301 Current CURRENT_TIMESTAMP returns the start time of the current
302 transaction, and gettimeofday() returns the wallclock time. This will
303 make time reporting more consistent and will allow reporting of
304 the statement start time.
306 * %Add pg_get_acldef(), pg_get_typedefault(), pg_get_attrdef(),
307 pg_get_tabledef(), pg_get_domaindef(), pg_get_functiondef()
308 * -Allow to_char() to print localized month names
309 * Allow to_date() and to_timestamp() accept localized month names
310 * Add missing parameter handling in to_char()
312 http://archives.postgresql.org/pgsql-hackers/2005-12/msg00948.php
314 * Allow functions to have a schema search path specified at creation time
315 * Allow substring/replace() to get/set bit values
316 * Allow to_char() on interval values to accumulate the highest unit
319 Some special format flag would be required to request such
320 accumulation. Such functionality could also be added to EXTRACT.
321 Prevent accumulation that crosses the month/day boundary because of
322 the uneven number of days in a month.
324 o to_char(INTERVAL '1 hour 5 minutes', 'MI') => 65
325 o to_char(INTERVAL '43 hours 20 minutes', 'MI' ) => 2600
326 o to_char(INTERVAL '43 hours 20 minutes', 'WK:DD:HR:MI') => 0:1:19:20
327 o to_char(INTERVAL '3 years 5 months','MM') => 41
329 * -Add sleep() function, remove from regress.c
330 * -Allow user-defined functions retuning a domain value to enforce domain
332 * Add SPI_gettypmod() to return the typemod for a TupleDesc
335 Multi-Language Support
336 ======================
338 * Add NCHAR (as distinguished from ordinary varchar),
339 * Allow locale to be set at database creation
341 Currently locale can only be set during initdb. No global tables have
342 locale-aware columns. However, the database template used during
343 database creation might have locale-aware indexes. The indexes would
344 need to be reindexed to match the new locale.
346 * Allow encoding on a per-column basis
348 Right now only one encoding is allowed per database.
350 * Support multiple simultaneous character sets, per SQL92
351 * Improve UTF8 combined character handling?
352 * Add octet_length_server() and octet_length_client()
353 * Make octet_length_client() the same as octet_length()?
354 * Fix problems with wrong runtime encoding conversion for NLS message files
360 * Automatically create rules on views so they are updateable, per SQL99
362 We can only auto-create rules for simple views. For more complex
363 cases users will still have to write rules manually.
365 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
366 * Allow NOTIFY in rules involving conditionals
367 * Allow VIEW/RULE recompilation when the underlying tables change
369 Another issue is whether underlying table changes should be reflected
370 in the view, e.g. should SELECT * show additional columns if they
371 are added after the view is created.
377 * Change LIMIT/OFFSET and FETCH/MOVE to use int8
378 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
379 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
380 * %Allow SET CONSTRAINTS to be qualified by schema/table name
381 * -Allow TRUNCATE ... CASCADE/RESTRICT
383 This is like DELETE CASCADE, but truncates.
385 * %Add a separate TRUNCATE permission
387 Currently only the owner can TRUNCATE a table because triggers are not
388 called, and the table is locked in exclusive mode.
390 * Allow PREPARE of cursors
391 * -Allow PREPARE to automatically determine parameter types based on the SQL
393 * Allow finer control over the caching of prepared query plans
395 Currently, queries prepared via the libpq API are planned on first
396 execute using the supplied parameters --- allow SQL PREPARE to do the
397 same. Also, allow control over replanning prepared queries either
398 manually or automatically when statistics for execute parameters
399 differ dramatically from those used during planning.
401 * Allow LISTEN/NOTIFY to store info in memory rather than tables?
403 Currently LISTEN/NOTIFY information is stored in pg_listener. Storing
404 such information in memory would improve performance.
406 * Add optional textual message to NOTIFY
408 This would allow an informational message to be added to the notify
409 message, perhaps indicating the row modified or other custom
412 * Add a GUC variable to warn about non-standard SQL usage in queries
413 * Add SQL-standard MERGE command, typically used to merge two tables
416 This is similar to UPDATE, then for unmatched rows, INSERT.
417 Whether concurrent access allows modifications which could cause
418 row loss is implementation independent.
420 * Add REPLACE or UPSERT command that does UPDATE, or on failure, INSERT
423 To implement this cleanly requires that the table have a unique index
424 so duplicate checking can be easily performed. It is possible to
425 do it without a unique index if we require the user to LOCK the table
428 * Add NOVICE output level for helpful messages like automatic sequence/index
430 * -Add COMMENT ON for all cluster global objects (roles, databases
432 * -Make row-wise comparisons work per SQL spec
434 Right now, '(a, b) < (1, 2)' is processed as 'a < 1 and b < 2', but
435 the SQL standard requires it to be processed as a column-by-column
436 comparison, so the proper comparison is '(a < 1) OR (a = 1 AND b < 2)'.
438 * Add RESET SESSION command to reset all session state
440 This would include resetting of all variables (RESET ALL), dropping of
441 temporary tables, removing any NOTIFYs, cursors, open transactions,
442 prepared queries, currval()s, etc. This could be used for connection
443 pooling. We could also change RESET ALL to have this functionality.
444 The difficult of this features is allowing RESET ALL to not affect
445 changes made by the interface driver for its internal use. One idea
446 is for this to be a protocol-only feature. Another approach is to
447 notify the protocol when a RESET SESSION command is used.
449 * Add GUC to issue notice about statements that use unjoined tables
450 * Allow EXPLAIN to identify tables that were skipped because of
452 * Allow EXPLAIN output to be more easily processed by scripts
453 * -Enable escape_string_warning and standard_conforming_strings
454 * Make standard_conforming_strings the default in 8.3?
456 When this is done, backslash-quote should be prohibited in non-E''
457 strings because of possible confusion over how such strings treat
458 backslashes. Basically, '' is always safe for a literal single
459 quote, while \' might or might not be based on the backslash
462 * Simplify dropping roles that have objects in several databases
463 * Allow COMMENT ON to accept an expression rather than just a string
464 * Allow the count returned by SELECT, etc to be to represent as an int64
465 to allow a higher range of values
466 * Make CLUSTER preserve recently-dead tuples per MVCC requirements
467 * Add SQL99 WITH clause to SELECT
468 * Add SQL99 WITH RECURSIVE to SELECT
473 o Allow CREATE TABLE AS to determine column lengths for complex
474 expressions like SELECT col1 || col2
475 o Use more reliable method for CREATE DATABASE to get a consistent
477 o -Add ON COMMIT capability to CREATE TABLE AS ... SELECT
481 o Allow UPDATE to handle complex aggregates [update]?
482 o -Allow an alias to be provided for the target table in
484 o Allow UPDATE tab SET ROW (col, ...) = (...) for updating multiple
490 o %Have ALTER TABLE RENAME rename SERIAL sequence names
491 o Add ALTER DOMAIN to modify the underlying data type
492 o %Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
493 o %Allow ALTER TABLE to change constraint deferrability and actions
494 o Add missing object types for ALTER ... SET SCHEMA
495 o Allow ALTER TABLESPACE to move to different directories
496 o Allow databases to be moved to different tablespaces
497 o Allow moving system tables to other tablespaces, where possible
499 Currently non-global system tables must be in the default database
500 tablespace. Global system tables can never be moved.
502 o Prevent parent tables from altering or dropping constraints
503 like CHECK that are inherited by child tables unless CASCADE
505 o %Prevent child tables from altering or dropping constraints
506 like CHECK that were inherited from the parent table
507 o Have ALTER INDEX update the name of a constraint using that index
508 o Add ALTER TABLE RENAME CONSTRAINT, update index name also
513 o Automatically maintain clustering on a table
515 This might require some background daemon to maintain clustering
516 during periods of low usage. It might also require tables to be only
517 partially filled for easier reorganization. Another idea would
518 be to create a merged heap/index data file so an index lookup would
519 automatically access the heap data too. A third idea would be to
520 store heap rows in hashed groups, perhaps using a user-supplied
523 o %Add default clustering to system tables
525 To do this, determine the ideal cluster index for each system
526 table and set the cluster setting during initdb.
531 o Allow COPY to report error lines and continue
533 This requires the use of a savepoint before each COPY line is
534 processed, with ROLLBACK on COPY failure.
536 o -Have COPY return the number of rows loaded/unloaded?
537 o Allow COPY on a newly-created table to skip WAL logging
539 On crash recovery, the table involved in the COPY would
540 be removed or have its heap and index files truncated. One
541 issue is that no other backend should be able to add to
542 the table at the same time, which is something that is
545 o Allow COPY to output from views
547 Another idea would be to allow actual SELECT statements in a COPY.
552 o Allow column-level privileges
553 o %Allow GRANT/REVOKE permissions to be applied to all schema objects
556 The proposed syntax is:
557 GRANT SELECT ON ALL TABLES IN public TO phpuser;
558 GRANT SELECT ON NEW TABLES IN public TO phpuser;
560 * Allow GRANT/REVOKE permissions to be inherited by objects based on
563 * Allow SERIAL sequences to inherit permissions from the base table?
568 o Allow UPDATE/DELETE WHERE CURRENT OF cursor
570 This requires using the row ctid to map cursor rows back to the
571 original heap row. This become more complicated if WITH HOLD cursors
572 are to be supported because WITH HOLD cursors have a copy of the row
573 and no FOR UPDATE lock.
575 o Prevent DROP TABLE from dropping a row referenced by its own open
578 o -Allow pooled connections to list all open WITH HOLD cursors
580 Because WITH HOLD cursors exist outside transactions, this allows
581 them to be listed so they can be closed.
586 o Allow INSERT/UPDATE of the system-generated oid value for a row
587 o Allow INSERT INTO tab (col1, ..) VALUES (val1, ..), (val2, ..)
588 o Allow INSERT/UPDATE ... RETURNING new.col or old.col
590 This is useful for returning the auto-generated key for an INSERT.
591 One complication is how to handle rules that run as part of
597 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
599 o Add SET PATH for schemas?
601 This is basically the same as SET search_path.
604 * Server-Side Languages
606 o Fix PL/pgSQL RENAME to work on variables other than OLD/NEW
607 o Allow function parameters to be passed by name,
608 get_employee_salary(emp_id => 12345, tax_year => 2001)
609 o Add Oracle-style packages
610 o Add table function support to pltcl, plpython
611 o Add capability to create and call PROCEDURES
612 o Allow PL/pgSQL to handle %TYPE arrays, e.g. tab.col%TYPE[]
613 o Allow function argument names to be statements from PL/PgSQL
614 o Add MOVE to PL/pgSQL
615 o Add support for polymorphic arguments and return types to
616 languages other than PL/PgSQL
617 o Add support for OUT and INOUT parameters to languages other
619 o Add single-step debugging of PL/PgSQL functions
620 o Allow PL/PgSQL to support WITH HOLD cursors
626 * -Have initdb set the input DateStyle (MDY or DMY) based on locale
627 * Have pg_ctl look at PGHOST in case it is a socket directory?
628 * Allow pg_ctl to work properly with configuration files located outside
631 pg_ctl can not read the pid file because it isn't located in the
632 config directory but in the PGDATA directory. The solution is to
633 allow pg_ctl to read and understand postgresql.conf to find the
634 data_directory value.
639 o Have psql show current values for a sequence
640 o Move psql backslash database information into the backend, use
641 mnemonic commands? [psql]
643 This would allow non-psql clients to pull the same information out
644 of the database as psql.
646 o Fix psql's \d commands more consistent
648 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
649 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
651 o Allow psql \pset boolean variables to set to fixed values, rather
653 o Consistently display privilege information for all objects in psql
654 o -Improve psql's handling of multi-line statements
656 Currently, while \e saves a single statement as one entry, interactive
657 statements are saved one line at a time. Ideally all statements
658 would be saved like \e does.
660 o -Allow multi-line column values to align in the proper columns
662 If the second output column value is 'a\nb', the 'b' should appear
663 in the second display column, rather than the first column as it
666 o Display IN, INOUT, and OUT parameters in \df+
668 It probably requires psql to output newlines in the proper
669 column, which is already on the TODO list.
671 o Add auto-expanded mode so expanded output is used if the row
672 length is wider than the screen width.
674 Consider using auto-expanded mode for backslash commands like \df+.
676 o Prevent tab completion of SET TRANSACTION from querying the
677 database and therefore preventing the transaction isolation
678 level from being set.
680 Currently, SET <tab> causes a database lookup to check all
681 supported session variables. This query causes problems
682 because setting the transaction isolation level must be the
683 first statement of a transaction.
688 o %Allow pg_dump to use multiple -t and -n switches [pg_dump]
689 o %Add dumping of comments on index columns and composite type columns
690 o %Add full object name to the tag field. eg. for operators we need
691 '=(integer, integer)', instead of just '='.
692 o Add pg_dumpall custom format dumps?
693 o %Add CSV output format
694 o -Update pg_dump and psql to use the new COPY libpq API (Christopher)
695 o Remove unnecessary function pointer abstractions in pg_dump source
697 o Allow selection of individual object(s) of all types, not just
699 o In a selective dump, allow dumping of an object and all its
701 o Add options like pg_restore -l and -L to pg_dump
702 o Stop dumping CASCADE on DROP TYPE commands in clean mode
703 o Allow pg_dump --clean to drop roles that own objects or have
705 o Add -f to pg_dumpall
712 Document differences between ecpg and the SQL standard and
713 information about the Informix-compatibility module.
715 o Solve cardinality > 1 for input descriptors / variables?
716 o Add a semantic check level, e.g. check if a table really exists
717 o fix handling of DB attributes that are arrays
718 o Use backend PREPARE/EXECUTE facility for ecpg where possible
720 o Fix nested C comments
721 o %sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
722 o Make SET CONNECTION thread-aware, non-standard?
723 o Allow multidimensional arrays
724 o Add internationalized message strings
725 o Add COPY TO STDIN / STDOUT handling
730 o Add a function to support Parse/DescribeStatement capability
731 o Add PQescapeIdentifier()
732 o Prevent PQfnumber() from lowercasing unquoted the column name
734 PQfnumber() should never have been doing lowercasing, but
735 historically it has so we need a way to prevent it
737 o Allow statement results to be automatically batched to the client
739 Currently, all statement results are transferred to the libpq
740 client before libpq makes the results available to the
741 application. This feature would allow the application to make
742 use of the first result rows while the rest are transferred, or
743 held on the server waiting for them to be requested by libpq.
744 One complexity is that a statement like SELECT 1/col could error
745 out mid-way through the result set.
747 o Add new version of PQescapeString() that doesn't double backslashes
748 that are part of a client-only multibyte sequence
750 Single-quote is not a valid byte in any supported client-only
751 encoding. This requires using mblen() to determine if the
752 backslash is inside or outside a multi-byte sequence.
754 o Add new version of PQescapeString() that doesn't double
755 backslashes when standard_conforming_strings is true and
756 non-E strings are used
759 Referential Integrity
760 =====================
762 * Add MATCH PARTIAL referential integrity
763 * Add deferred trigger queue file
765 Right now all deferred trigger information is stored in backend
766 memory. This could exhaust memory for very large trigger queues.
767 This item involves dumping large queues into files.
769 * Change foreign key constraint for array -> element to mean element
771 * Allow DEFERRABLE UNIQUE constraints?
772 * Allow triggers to be disabled in only the current session.
774 This is currently possible by starting a multi-statement transaction,
775 modifying the system tables, performing the desired SQL, restoring the
776 system tables, and committing the transaction. ALTER TABLE ...
777 TRIGGER requires a table lock so it is not ideal for this usage.
779 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
781 If the dump is known to be valid, allow foreign keys to be added
782 without revalidating the data.
784 * Allow statement-level triggers to access modified rows
785 * Support triggers on columns (Greg Sabino Mullane)
786 * Enforce referential integrity for system tables
787 * Allow AFTER triggers on system tables
789 System tables are modified in many places in the backend without going
790 through the executor and therefore not causing triggers to fire. To
791 complete this item, the functions that modify system tables will have
798 * Flush cached query plans when the dependent objects change,
799 when the cardinality of parameters changes dramatically, or
800 when new ANALYZE statistics are available
802 A more complex solution would be to save multiple plans for different
803 cardinality and use the appropriate plan based on the EXECUTE values.
805 * Track dependencies in function bodies and recompile/invalidate
807 This is particularly important for references to temporary tables
808 in PL/PgSQL because PL/PgSQL caches query plans. The only workaround
809 in PL/PgSQL is to use EXECUTE. One complexity is that a function
810 might itself drop and recreate dependent tables, causing it to
811 invalidate its own query plan.
817 * Add pre-parsing phase that converts non-ISO syntax to supported
820 This could allow SQL written for other databases to run without
823 * Allow plug-in modules to emulate features from other databases
824 * SQL*Net listener that makes PostgreSQL appear as an Oracle database
826 * Allow statements across databases or servers with transaction
829 This can be done using dblink and two-phase commit.
831 * Add the features of packages
833 o Make private objects accessible only to objects in the same schema
834 o Allow current_schema.objname to access current schema objects
835 o Add session variables
836 o Allow nested schemas
842 * Allow inherited tables to inherit index, UNIQUE constraint, and primary
844 * UNIQUE INDEX on base column not honored on INSERTs/UPDATEs from
845 inherited table: INSERT INTO inherit_table (unique_index_col) VALUES
848 The main difficulty with this item is the problem of creating an index
849 that can span more than one table.
851 * Allow SELECT ... FOR UPDATE on inherited tables
852 * Add UNIQUE capability to non-btree indexes
853 * Prevent index uniqueness checks when UPDATE does not modify the column
855 Uniqueness (index) checks are done when updating a column even if the
856 column is not modified by the UPDATE.
858 * Allow the creation of on-disk bitmap indexes which can be quickly
859 combined with other bitmap indexes
861 Such indexes could be more compact if there are only a few distinct values.
862 Such indexes can also be compressed. Keeping such indexes updated can be
865 * Allow use of indexes to search for NULLs
867 One solution is to create a partial index on an IS NULL expression.
869 * Allow accurate statistics to be collected on indexes with more than
870 one column or expression indexes, perhaps using per-index statistics
871 * Add fillfactor to control reserved free space during index creation
872 * Allow the creation of indexes with mixed ascending/descending
875 This is possible now by creating an operator class with reversed sort
876 operators. One complexity is that NULLs would then appear at the start
877 of the result set, and this might affect certain sort types, like
880 * Allow constraint_exclusion to work for UNIONs like it does for
881 inheritance, allow it to work for UPDATE and DELETE statements, and allow
882 it to be used for all statements with little performance impact
883 * Allow CREATE INDEX to take an additional parameter for use with
885 * Consider compressing indexes by storing key values duplicated in
886 several rows as a single index entry
888 This is difficult because it requires datatype-specific knowledge.
893 o Add more GIST index support for geometric data types
894 o Allow GIST indexes to create certain complex index types, like
895 digital trees (see Aoki)
899 o Pack hash index buckets onto disk pages more efficiently
901 Currently only one hash bucket can be stored on a page. Ideally
902 several hash buckets could be stored on a single page and greater
903 granularity used for the hash algorithm.
905 o Consider sorting hash buckets so entries can be found using a
906 binary search, rather than a linear scan
908 o In hash indexes, consider storing the hash value with or instead
911 o Add WAL logging for crash recovery
912 o Allow multi-column hash indexes
918 * Improve commit_delay handling to reduce fsync()
919 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
921 Ideally this requires a separate test program that can be run
922 at initdb time or optionally later. Consider O_SYNC when
925 * %Add an option to sync() before fsync()'ing checkpoint files
926 * Add program to test if fsync has a delay compared to non-fsync
932 * Allow free-behind capability for large sequential scans, perhaps using
935 Posix_fadvise() can control both sequential/random file caching and
936 free-behind behavior, but it is unclear how the setting affects other
937 backends that also have the file open, and the feature is not supported
938 on all operating systems.
942 We could use a fixed row count and a +/- count to follow MVCC
943 visibility rules, or a single cached value could be used and
944 invalidated if anyone modifies the table. Another idea is to
945 get a count directly from a unique index, but for this to be
946 faster than a sequential scan it must avoid access to the heap
947 to obtain tuple visibility information.
949 * Add estimated_count(*) to return an estimate of COUNT(*)
951 This would use the planner ANALYZE statistics to return an estimated
954 * Allow data to be pulled directly from indexes
956 Currently indexes do not have enough tuple visibility information
957 to allow data to be pulled from the index without also accessing
958 the heap. One way to allow this is to set a bit on index tuples
959 to indicate if a tuple is currently visible to all transactions
960 when the first valid heap lookup happens. This bit would have to
961 be cleared when a heap tuple is expired.
963 Another idea is to maintain a bitmap of heap pages where all rows
964 are visible to all backends, and allow index lookups to reference
965 that bitmap to avoid heap lookups, perhaps the same bitmap we might
966 add someday to determine which heap pages need vacuuming. Frequently
967 accessed bitmaps would have to be stored in shared memory. One 8k
968 page of bitmaps could track 512MB of heap pages.
970 * Consider automatic caching of statements at various levels:
976 * Allow sequential scans to take advantage of other concurrent
977 sequential scans, also called "Synchronised Scanning"
979 One possible implementation is to start sequential scans from the lowest
980 numbered buffer in the shared cache, and when reaching the end wrap
981 around to the beginning, rather than always starting sequential scans
982 at the start of the table.
988 * Improve speed with indexes
990 For large table adjustments during VACUUM FULL, it is faster to
991 reindex rather than update the index.
993 * Reduce lock time during VACUUM FULL by moving tuples with read lock,
994 then write lock and truncate table
996 Moved tuples are invisible to other backends so they don't require a
997 write lock. However, the read lock promotion to write lock could lead
998 to deadlock situations.
1000 * Auto-fill the free space map by scanning the buffer cache or by
1001 checking pages written by the background writer
1002 * Create a bitmap of pages that need vacuuming
1004 Instead of sequentially scanning the entire table, have the background
1005 writer or some other process record pages that have expired rows, then
1006 VACUUM can look at just those pages rather than the entire table. In
1007 the event of a system crash, the bitmap would probably be invalidated.
1008 One complexity is that index entries still have to be vacuumed, and
1009 doing this without an index scan (by using the heap values to find the
1010 index entry) might be slow and unreliable, especially for user-defined
1013 * -Add system view to show free space map contents
1014 * Allow FSM to return free space toward the beginning of the heap file,
1015 in hopes that empty pages at the end can be truncated by VACUUM
1016 * Allow FSM page return free space based on table clustering, to assist
1017 in maintaining clustering?
1022 o Use free-space map information to guide refilling
1023 o %Issue log message to suggest VACUUM FULL if a table is nearly
1025 o Improve xid wraparound detection by recording per-table rather
1027 o Consider logging activity either to the logs or a system view
1033 * Fix priority ordering of read and write light-weight locks (Neil)
1036 Startup Time Improvements
1037 =========================
1039 * Experiment with multi-threaded backend for backend creation [thread]
1041 This would prevent the overhead associated with process creation. Most
1042 operating systems have trivial process creation time compared to
1043 database startup overhead, but a few operating systems (Win32,
1044 Solaris) might benefit from threading. Also explore the idea of
1045 a single session using multiple threads to execute a statement faster.
1047 * Experiment with multi-threaded backend better resource utilization
1049 This would allow a single query to make use of multiple CPU's or
1050 multiple I/O channels simultaneously. One idea is to create a
1051 background reader that can pre-fetch sequential and index scan
1052 pages needed by other backends. This could be expanded to allow
1053 concurrent reads from multiple devices in a partitioned table.
1055 * Add connection pooling
1057 It is unclear if this should be done inside the backend code or done
1058 by something external like pgpool. The passing of file descriptors to
1059 existing backends is one of the difficulties with a backend approach.
1065 * Eliminate need to write full pages to WAL before page modification [wal]
1067 Currently, to protect against partial disk page writes, we write
1068 full page images to WAL before they are modified so we can correct any
1069 partial page writes during recovery. These pages can also be
1070 eliminated from point-in-time archive files.
1072 o When off, write CRC to WAL and check file system blocks
1075 If CRC check fails during recovery, remember the page in case
1076 a later CRC for that page properly matches.
1078 o Write full pages during file system write and not when
1079 the page is modified in the buffer cache
1081 This allows most full page writes to happen in the background
1082 writer. It might cause problems for applying WAL on recovery
1083 into a partially-written page, but later the full page will be
1086 * Allow WAL traffic to be streamed to another server for stand-by
1088 * Reduce WAL traffic so only modified values are written rather than
1090 * Allow the pg_xlog directory location to be specified during initdb
1091 with a symlink back to the /data location
1092 * Allow WAL information to recover corrupted pg_controldata
1093 * Find a way to reduce rotational delay when repeatedly writing
1096 Currently fsync of WAL requires the disk platter to perform a full
1097 rotation to fsync again. One idea is to write the WAL to different
1098 offsets that might reduce the rotational delay.
1100 * Allow buffered WAL writes and fsync
1102 Instead of guaranteeing recovery of all committed transactions, this
1103 would provide improved performance by delaying WAL writes and fsync
1104 so an abrupt operating system restart might lose a few seconds of
1105 committed transactions but still be consistent. We could perhaps
1106 remove the 'fsync' parameter (which results in an an inconsistent
1107 database) in favor of this capability.
1109 * Allow WAL logging to be turned off for a table, but the table
1110 might be dropped or truncated during crash recovery [walcontrol]
1112 Allow tables to bypass WAL writes and just fsync() dirty pages on
1113 commit. This should be implemented using ALTER TABLE, e.g. ALTER
1114 TABLE PERSISTENCE [ DROP | TRUNCATE | DEFAULT ]. Tables using
1115 non-default logging should not use referential integrity with
1116 default-logging tables. A table without dirty buffers during a
1117 crash could perhaps avoid the drop/truncate.
1119 * Allow WAL logging to be turned off for a table, but the table would
1120 avoid being truncated/dropped [walcontrol]
1122 To do this, only a single writer can modify the table, and writes
1123 must happen only on new pages so the new pages can be removed during
1124 crash recovery. Readers can continue accessing the table. Such
1125 tables probably cannot have indexes. One complexity is the handling
1126 of indexes on TOAST tables.
1129 Optimizer / Executor
1130 ====================
1132 * Improve selectivity functions for geometric operators
1133 * Allow ORDER BY ... LIMIT # to select high/low value without sort or
1134 index using a sequential scan for highest/lowest values
1136 Right now, if no index exists, ORDER BY ... LIMIT # requires we sort
1137 all values to return the high/low value. Instead The idea is to do a
1138 sequential scan to find the high/low value, thus avoiding the sort.
1139 MIN/MAX already does this, but not for LIMIT > 1.
1141 * Precompile SQL functions to avoid overhead
1142 * Create utility to compute accurate random_page_cost value
1143 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
1144 * Have EXPLAIN ANALYZE highlight poor optimizer estimates
1145 * Consider using hash buckets to do DISTINCT, rather than sorting
1147 This would be beneficial when there are few distinct values. This is
1148 already used by GROUP BY.
1150 * Log statements where the optimizer row estimates were dramatically
1151 different from the number of rows actually found?
1154 Miscellaneous Performance
1155 =========================
1157 * Do async I/O for faster random read-ahead of data
1159 Async I/O allows multiple I/O requests to be sent to the disk with
1160 results coming back asynchronously.
1162 * Use mmap() rather than SYSV shared memory or to write WAL files?
1164 This would remove the requirement for SYSV SHM but would introduce
1165 portability issues. Anonymous mmap (or mmap to /dev/zero) is required
1166 to prevent I/O overhead.
1168 * Consider mmap()'ing files into a backend?
1170 Doing I/O to large tables would consume a lot of address space or
1171 require frequent mapping/unmapping. Extending the file also causes
1172 mapping problems that might require mapping only individual pages,
1173 leading to thousands of mappings. Another problem is that there is no
1174 way to _prevent_ I/O to disk from the dirty shared buffers so changes
1175 could hit disk before WAL is written.
1177 * Add a script to ask system configuration questions and tune postgresql.conf
1178 * Merge xmin/xmax/cmin/cmax back into three header fields
1180 Before subtransactions, there used to be only three fields needed to
1181 store these four values. This was possible because only the current
1182 transaction looks at the cmin/cmax values. If the current transaction
1183 created and expired the row the fields stored where xmin (same as
1184 xmax), cmin, cmax, and if the transaction was expiring a row from a
1185 another transaction, the fields stored were xmin (cmin was not
1186 needed), xmax, and cmax. Such a system worked because a transaction
1187 could only see rows from another completed transaction. However,
1188 subtransactions can see rows from outer transactions, and once the
1189 subtransaction completes, the outer transaction continues, requiring
1190 the storage of all four fields. With subtransactions, an outer
1191 transaction can create a row, a subtransaction expire it, and when the
1192 subtransaction completes, the outer transaction still has to have
1193 proper visibility of the row's cmin, for example, for cursors.
1195 One possible solution is to create a phantom cid which represents a
1196 cmin/cmax pair and is stored in local memory. Another idea is to
1197 store both cmin and cmax only in local memory.
1199 * Research storing disk pages with no alignment/padding
1205 * Add use of 'const' for variables in source tree
1206 * Rename some /contrib modules from pg* to pg_*
1207 * Move some things from /contrib into main tree
1208 * Move some /contrib modules out to their own project sites
1209 * %Remove warnings created by -Wcast-align
1210 * Move platform-specific ps status display info from ps_status.c to ports
1211 * Add optional CRC checksum to heap and index pages
1212 * Improve documentation to build only interfaces (Marc)
1213 * Remove or relicense modules that are not under the BSD license, if possible
1214 * %Remove memory/file descriptor freeing before ereport(ERROR)
1215 * Acquire lock on a relation before building a relcache entry for it
1216 * %Promote debug_query_string into a server-side function current_query()
1217 * %Allow the identifier length to be increased via a configure option
1218 * Allow cross-compiling by generating the zic database on the target system
1219 * Improve NLS maintenance of libpgport messages linked onto applications
1220 * Allow ecpg to work with MSVC and BCC
1221 * Add xpath_array() to /contrib/xml2 to return results as an array
1222 * Allow building in directories containing spaces
1224 This is probably not possible because 'gmake' and other compiler tools
1225 do not fully support quoting of paths with spaces.
1227 * -Allow installing to directories containing spaces
1229 This is possible if proper quoting is added to the makefiles for the
1230 install targets. Because PostgreSQL supports relocatable installs, it
1231 is already possible to install into a directory that doesn't contain
1232 spaces and then copy the install to a directory with spaces.
1234 * Fix sgmltools so PDFs can be generated with bookmarks
1235 * -%Clean up compiler warnings (especially with gcc version 4)
1236 * Use UTF8 encoding for NLS messages so all server encodings can
1238 * Update Bonjour to work with newer cross-platform SDK
1239 * -Remove BeOS and QNX-specific code
1240 * Split out libpq pgpass and environment documentation sections to make
1241 it easier for non-developers to find
1242 * Improve port/qsort() to handle sorts with 50% unique and 50% duplicate
1245 This involves choosing better pivot points for the quicksort.
1250 o Remove configure.in check for link failure when cause is found
1251 o Remove readdir() errno patch when runtime/mingwex/dirent.c rev
1253 o Remove psql newline patch when we find out why mingw outputs an
1255 o Allow psql to use readline once non-US code pages work with
1257 o Re-enable timezone output on log_line_prefix '%t' when a
1258 shorter timezone string is available
1259 o Fix problem with shared memory on the Win32 Terminal Server
1260 o Improve signal handling,
1261 http://archives.postgresql.org/pgsql-patches/2005-06/msg00027.php
1262 o Add long file support for binary pg_dump output
1264 While Win32 supports 64-bit files, the MinGW API does not,
1265 meaning we have to build an fseeko replacement on top of the
1266 Win32 API, and we have to make sure MinGW handles it. Another
1267 option is to wait for the MinGW project to fix it, or use the
1268 code from the LibGW32C project as a guide.
1271 o Check WSACancelBlockingCall() for interrupts (win32intr)
1274 * Wire Protocol Changes
1276 o Allow dynamic character set handling
1277 o Add decoded type, length, precision
1279 o Update clients to use data types, typmod, schema.table.column names
1280 of result sets using new statement protocol
1283 ---------------------------------------------------------------------------
1286 Developers who have claimed items are:
1287 --------------------------------------
1288 * Alvaro is Alvaro Herrera <alvherre@dcc.uchile.cl>
1289 * Andrew is Andrew Dunstan <andrew@dunslane.net>
1290 * Bruce is Bruce Momjian <pgman@candle.pha.pa.us> of Software Research Assoc.
1291 * Christopher is Christopher Kings-Lynne <chriskl@familyhealth.com.au> of
1292 Family Health Network
1293 * D'Arcy is D'Arcy J.M. Cain <darcy@druid.net> of The Cain Gang Ltd.
1294 * Fabien is Fabien Coelho <coelho@cri.ensmp.fr>
1295 * Gavin is Gavin Sherry <swm@linuxworld.com.au> of Alcove Systems Engineering
1296 * Greg is Greg Sabino Mullane <greg@turnstep.com>
1297 * Jan is Jan Wieck <JanWieck@Yahoo.com> of Afilias, Inc.
1298 * Joe is Joe Conway <mail@joeconway.com>
1299 * Karel is Karel Zak <zakkr@zf.jcu.cz>
1300 * Magnus is Magnus Hagander <mha@sollentuna.net>
1301 * Marc is Marc Fournier <scrappy@hub.org> of PostgreSQL, Inc.
1302 * Matthew T. O'Connor <matthew@zeut.net>
1303 * Michael is Michael Meskes <meskes@postgresql.org> of Credativ
1304 * Neil is Neil Conway <neilc@samurai.com>
1305 * Oleg is Oleg Bartunov <oleg@sai.msu.su>
1306 * Peter is Peter Eisentraut <peter_e@gmx.net>
1307 * Philip is Philip Warner <pjw@rhyme.com.au> of Albatross Consulting Pty. Ltd.
1308 * Rod is Rod Taylor <pg@rbt.ca>
1309 * Simon is Simon Riggs <simon@2ndquadrant.com>
1310 * Stephan is Stephan Szabo <sszabo@megazone23.bigpanda.com>
1311 * Tatsuo is Tatsuo Ishii <t-ishii@sra.co.jp> of Software Research Assoc.
1312 * Teodor is Teodor Sigaev <teodor@sigaev.ru>
1313 * Tom is Tom Lane <tgl@sss.pgh.pa.us> of Red Hat