3 Current maintainer: Bruce Momjian (bruce@momjian.us)
4 Last updated: Thu Apr 24 12:14:19 EDT 2008
6 The most recent version of this document can be viewed at
7 http://www.postgresql.org/docs/faqs.TODO.html.
9 #A hyphen, "-", marks changes that will appear in the upcoming 8.3 release.#
10 #A percent sign, "%", marks items that are easier to implement.#
12 This list contains all known PostgreSQL bugs and feature requests. If
13 you would like to work on an item, please read the Developer's FAQ
14 first. There is also a developer's wiki at
15 http://developer.postgresql.org.
21 * -Allow administrators to safely terminate individual sessions either
22 via an SQL function or SIGTERM
23 * Check for unreferenced table files created by transactions that were
24 in-progress when the server terminated abruptly
26 http://archives.postgresql.org/pgsql-patches/2006-06/msg00096.php
28 * Set proper permissions on non-system schemas during db creation
30 Currently all schemas are owned by the super-user because they are copied
31 from the template1 database. However, since all objects are inherited
32 from the template database, it is not clear that setting schemas to the db
35 * Add function to report the time of the most recent server reload
36 * Allow statistics collector information to be pulled from the collector
37 process directly, rather than requiring the collector to write a
38 filesystem file twice a second?
39 * Reduce file system activity overhead of statistics file pgstat.stat
41 http://archives.postgresql.org/pgsql-general/2007-12/msg00106.php
43 * Allow statistics last vacuum/analyze execution times to be displayed
44 without requiring stats_row_level to be enabled
46 http://archives.postgresql.org/pgsql-docs/2007-04/msg00028.php
48 * Allow log_min_messages to be specified on a per-module basis
50 This would allow administrators to see more detailed information from
51 specific sections of the backend, e.g. checkpoints, autovacuum, etc.
52 Another idea is to allow separate configuration files for each module,
53 or allow arbitrary SET commands to be passed to them.
55 * Simplify ability to create partitioned tables
57 This would allow creation of partitioned tables without requiring
58 creation of triggers or rules for INSERT/UPDATE/DELETE, and constraints
59 for rapid partition selection. Options could include range and hash
62 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00375.php
63 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00151.php
64 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00028.php
65 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00248.php
66 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00387.php
67 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00413.php
69 * Allow auto-selection of partitioned tables for min/max() operations
70 * Allow more complex user/database default GUC settings
72 Currently ALTER USER and ALTER DATABASE support per-user and
73 per-database defaults. Consider adding per-user-and-database
74 defaults so things like search_path can be defaulted for a
75 specific user connecting to a specific database.
77 * Allow custom variable classes that can restrict who can set the values
79 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00911.php
81 * Implement the SQL standard mechanism whereby REVOKE ROLE revokes only
82 the privilege granted by the invoking role, and not those granted
85 http://archives.postgresql.org/pgsql-bugs/2007-05/msg00010.php
87 * Allow SSL authentication/encryption over unix domain sockets
89 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00924.php
91 * Allow SSL key file permission checks to be optionally disabled when
92 sharing SSL keys with other applications
94 http://archives.postgresql.org/pgsql-bugs/2007-12/msg00069.php
96 * Allow client certificate names to be checked against the client
99 This is already implemented in
100 libpq/fe-secure.c::verify_peer_name_matches_certificate() but the code
103 * Configuration files
105 o Allow pg_hba.conf to specify host names along with IP addresses
107 Host name lookup could occur when the postmaster reads the
108 pg_hba.conf file, or when the backend starts. Another
109 solution would be to reverse lookup the connection IP and
110 check that hostname against the host names in pg_hba.conf.
111 We could also then check that the host name maps to the IP
114 o %Allow postgresql.conf file values to be changed via an SQL
115 API, perhaps using SET GLOBAL
116 o Allow the server to be stopped/restarted via an SQL API
117 o Issue a warning if a change-on-restart-only postgresql.conf value
118 is modified and the server config files are reloaded
119 o Consider normalizing fractions in postgresql.conf, perhaps
122 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00550.php
124 o Allow Kerberos to disable stripping of realms so we can
125 check the username@realm against multiple realms
127 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00009.php
132 o Allow a database in tablespace t1 with tables created in
133 tablespace t2 to be used as a template for a new database created
134 with default tablespace t2
136 Currently all objects in the default database tablespace must
137 have default tablespace specifications. This is because new
138 databases are created by copying directories. If you mix default
139 tablespace tables and tablespace-specified tables in the same
140 directory, creating a new database from such a mixed directory
141 would create a new database with tables that had incorrect
142 explicit tablespaces. To fix this would require modifying
143 pg_class in the newly copied database, which we don't currently
146 o Allow reporting of which objects are in which tablespaces
148 This item is difficult because a tablespace can contain objects
149 from multiple databases. There is a server-side function that
150 returns the databases which use a specific tablespace, so this
151 requires a tool that will call that function and connect to each
152 database to find the objects in each database for that tablespace.
154 o Allow WAL replay of CREATE TABLESPACE to work when the directory
155 structure on the recovery computer is different from the original
157 o Allow per-tablespace quotas
160 * Point-In-Time Recovery (PITR)
162 o Allow a warm standby system to also allow read-only statements
164 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00050.php
166 o %Create dump tool for write-ahead logs for use in determining
167 transaction id for point-in-time recovery
169 This is useful for checking PITR recovery.
171 o Allow recovery.conf to support the same syntax as
172 postgresql.conf, including quoting
174 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00497.php
176 o Fix server restart problem when the server was shutdown during
179 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00800.php
181 o Recreate pg_xlog/archive_status/ if it doesn't exist after
182 restoring from a PITR backup
184 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00487.php
190 * Change NUMERIC to enforce the maximum precision
191 * Reduce storage space for small NUMERICs
193 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01331.php
194 http://archives.postgresql.org/pgsql-patches/2007-02/msg00505.php
195 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00715.php
197 * Fix data types where equality comparison isn't intuitive, e.g. box
198 * Add support for public SYNONYMs
200 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00519.php
202 * Fix CREATE CAST on DOMAINs
204 http://archives.postgresql.org/pgsql-hackers/2006-05/msg00072.php
205 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01681.php
207 * Allow domains to be cast
209 http://archives.postgresql.org/pgsql-hackers/2003-06/msg01206.php
210 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00289.php
212 * Add support for SQL-standard GENERATED/IDENTITY columns
214 http://archives.postgresql.org/pgsql-hackers/2006-07/msg00543.php
215 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00038.php
216 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00344.php
217 http://archives.postgresql.org/pgsql-patches/2007-05/msg00076.php
218 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00604.php
220 * Improve XML support
222 http://developer.postgresql.org/index.php/XML_Support
224 * Consider placing all sequences in a single table, or create a system
227 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
229 * Allow the UUID type to accept non-standard formats
231 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01214.php
233 * Allow text search dictionary to filter out only stop words
235 http://archives.postgresql.org/pgsql-patches/2007-11/msg00081.php
237 * Consider a function-based API for '@@' full text searches
239 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00511.php
241 * Improve text search error messages
243 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00966.php
244 http://archives.postgresql.org/pgsql-hackers/2007-11/msg01146.php
246 * Consider a special data type for regular expressions
248 http://archives.postgresql.org/pgsql-hackers/2007-08/msg01067.php
250 * Reduce BIT data type overhead using short varlena headers
252 http://archives.postgresql.org/pgsql-general/2007-12/msg00273.php
254 * Allow xml arrays to be cast to other data types
256 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00981.php
257 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00231.php
258 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00471.php
260 * Simplify integer cross-data-type operators
262 http://archives.postgresql.org/pgsql-bugs/2008-01/msg00189.php
264 * Allow XML to accept more liberal DOCTYPE specifications
266 http://archives.postgresql.org/pgsql-general/2008-02/msg00347.php
268 * Allow adding enumerated values to an existing enumerated data
272 o Allow infinite dates and intervals just like infinite timestamps
273 o Merge hardwired timezone names with the TZ database; allow either
274 kind everywhere a TZ name is currently taken
275 o Allow TIMESTAMP WITH TIME ZONE to store the original timezone
276 information, either zone name or offset from UTC
278 If the TIMESTAMP value is stored with a time zone name, interval
279 computations should adjust based on the time zone rules.
280 http://archives.postgresql.org/pgsql-hackers/2004-10/msg00705.php
282 o Fix SELECT '0.01 years'::interval, '0.01 months'::interval
283 o Add a GUC variable to allow output of interval values in ISO8601
285 o Have timestamp subtraction not call justify_hours()?
287 http://archives.postgresql.org/pgsql-sql/2006-10/msg00059.php
289 o Improve timestamptz subtraction to be DST-aware
291 Currently subtracting one date from another that crosses a
292 daylight savings time adjustment can return '1 day 1 hour', but
293 adding that back to the first date returns a time one hour in
294 the future. This is caused by the adjustment of '25 hours' to
295 '1 day 1 hour', and '1 day' is the same time the next day, even
296 if daylight savings adjustments are involved.
298 o Fix interval display to support values exceeding 2^31 hours
299 o Add overflow checking to timestamp and interval arithmetic
300 o Extend timezone code to allow 64-bit values so we can
301 represent years beyond 2038
303 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01363.php
305 o Use LC_TIME for localized weekday/month names, rather than
308 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00390.php
310 o Add ISO INTERVAL handling
312 http://archives.postgresql.org/pgsql-hackers/2006-01/msg00250.php
313 http://archives.postgresql.org/pgsql-bugs/2006-04/msg00248.php
315 o Support ISO INTERVAL syntax if units cannot be determined from
316 the string, and are supplied after the string
318 The SQL standard states that the units after the string
319 specify the units of the string, e.g. INTERVAL '2' MINUTE
320 should return '00:02:00'. The current behavior has the units
321 restrict the interval value to the specified unit or unit
322 range, INTERVAL '70' SECOND returns '00:00:10'.
324 For syntax that isn't uniquely ISO or PG syntax, like '1' or
325 '1:30', treat as ISO if there is a range specification clause,
326 and as PG if there no clause is present, e.g. interpret '1:30'
327 MINUTE TO SECOND as '1 minute 30 seconds', and interpret
328 '1:30' as '1 hour, 30 minutes'.
330 This makes common cases like SELECT INTERVAL '1' MONTH
331 SQL-standard results. The SQL standard supports a limited
332 number of unit combinations and doesn't support unit names in
333 the string. The PostgreSQL syntax is more flexible in the
334 range of units supported, e.g. PostgreSQL supports '1 year 1
335 hour', while the SQL standard does not.
337 o Add support for year-month syntax, INTERVAL '50-6' YEAR
339 o Interpret INTERVAL '1 year' MONTH as CAST (INTERVAL '1
340 year' AS INTERVAL MONTH), and this should return '12 months'
341 o Round or truncate values to the requested precision, e.g.
342 INTERVAL '11 months' AS YEAR should return one or zero
343 o Support precision, CREATE TABLE foo (a INTERVAL MONTH(3))
348 o Delay resolution of array expression's data type so assignment
349 coercion can be performed on empty array expressions
350 o Add support for arrays of domains
352 http://archives.postgresql.org/pgsql-patches/2007-05/msg00114.php
354 o Allow single-byte header storage for array elements
359 o Improve vacuum of large objects, like contrib/vacuumlo?
360 o Add security checking for large objects
361 o Auto-delete large objects when referencing row is deleted
363 contrib/lo offers this functionality.
365 o Allow read/write into TOAST values like large objects
367 This requires the TOAST column to be stored EXTERNAL.
369 o Add API for 64-bit large object access
371 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00781.php
375 * Add locale-aware MONEY type, and support multiple currencies
377 http://archives.postgresql.org/pgsql-general/2005-08/msg01432.php
378 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01181.php
380 * MONEY dumps in a locale-specific format making it difficult to
381 restore to a system with a different locale
382 * Allow MONEY to be easily cast to/from other numeric data types
388 * Allow INET subnet tests using non-constants to be indexed
389 * Allow to_date() and to_timestamp() accept localized month names
390 * Fix to_date()-related functions to consistently issue errors
392 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00915.php
394 * Add missing parameter handling in to_char()
396 http://archives.postgresql.org/pgsql-hackers/2005-12/msg00948.php
398 * Allow substring/replace() to get/set bit values
399 * Allow to_char() on interval values to accumulate the highest unit
402 Some special format flag would be required to request such
403 accumulation. Such functionality could also be added to EXTRACT.
404 Prevent accumulation that crosses the month/day boundary because of
405 the uneven number of days in a month.
407 o to_char(INTERVAL '1 hour 5 minutes', 'MI') => 65
408 o to_char(INTERVAL '43 hours 20 minutes', 'MI' ) => 2600
409 o to_char(INTERVAL '43 hours 20 minutes', 'WK:DD:HR:MI') => 0:1:19:20
410 o to_char(INTERVAL '3 years 5 months','MM') => 41
412 * Implement inlining of set-returning functions defined in SQL
413 * Allow SQL-language functions to return results from RETURNING queries
415 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00665.php
417 * Allow SQL-language functions to reference parameters by parameter name
419 Currently SQL-language functions can only refer to dollar parameters,
422 * Add SPI_gettypmod() to return the typemod for a TupleDesc
423 * Enforce typmod for function inputs, function results and parameters for
424 spi_prepare'd statements called from PLs
426 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01403.php
428 * Allow holdable cursors in SPI
429 * Tighten function permission checks
431 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00568.php
433 * Fix IS OF so it matches the ISO specification, and add documentation
435 http://archives.postgresql.org/pgsql-patches/2003-08/msg00060.php
436 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00060.php
438 * Add missing operators for geometric data types
440 Some geometric types do not have the full suite of geometric operators,
443 * Implement Boyer-Moore searching in strpos()
445 http://archives.postgresql.org/pgsql-patches/2007-08/msg00012.php
447 * Prevent malicious functions from being executed with the permissions
448 of unsuspecting users
450 Index functions are safe, so VACUUM and ANALYZE are safe too.
451 Triggers, CHECK and DEFAULT expressions, and rules are still vulnerable.
452 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00268.php
454 * Reduce memory usage of aggregates in set returning functions
456 http://archives.postgresql.org/pgsql-performance/2008-01/msg00031.php
458 * Add temporal versions of generate_series()
460 http://archives.postgresql.org/pgsql-hackers/2007-04/msg01180.php
462 * Add array_accum() and array_to_set() functions for arrays
464 The standards specify array_agg() and UNNEST.
465 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00464.php
467 * Fix /contrib/ltree operator
469 http://archives.postgresql.org/pgsql-bugs/2007-11/msg00044.php
471 * Fix inconsistent precedence of =, >, and < compared to <>, >=, and <=
473 http://archives.postgresql.org/pgsql-bugs/2007-12/msg00145.php
475 * Fix regular expression bug when using complex back-references
477 http://archives.postgresql.org/pgsql-bugs/2007-10/msg00000.php
479 * Have /contrib/dblink reuse unnamed connections
481 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00895.php
483 * Add SQL-standard array_agg() and unnest() array functions
485 http://archives.postgresql.org/pgsql-hackers/2008-01/msg01017.php
487 * Allow calling of a procedure outside a SELECT that can control the
490 http://archives.postgresql.org/pgsql-hackers/2007-10/msg01375.php
494 Multi-Language Support
495 ======================
497 * Add NCHAR (as distinguished from ordinary varchar),
498 * Allow locale to be set at database creation
500 Currently locale can only be set during initdb. No global tables have
501 locale-aware columns. However, the database template used during
502 database creation might have locale-aware indexes. The indexes would
503 need to be reindexed to match the new locale.
505 * Allow encoding on a per-column basis optionally using the ICU library;
508 Right now only one encoding is allowed per database.
510 http://archives.postgresql.org/pgsql-hackers/2005-03/msg00932.php
511 http://archives.postgresql.org/pgsql-patches/2005-08/msg00039.php
512 http://archives.postgresql.org/pgsql-patches/2005-08/msg00309.php
513 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00110.php
514 http://archives.postgresql.org/pgsql-patches/2005-09/msg00020.php
515 http://archives.postgresql.org/pgsql-hackers/2005-12/msg01121.php
516 http://archives.postgresql.org/pgsql-hackers/2006-01/msg00767.php
517 http://archives.postgresql.org/pgsql-patches/2006-03/msg00233.php
518 http://archives.postgresql.org/pgsql-hackers/2006-09/msg00662.php
519 http://wiki.postgresql.org/wiki/Todo:Collate
520 http://wiki.postgresql.org/wiki/Todo:ICU
522 * Support multiple simultaneous character sets, per SQL92
523 * Improve UTF8 combined character handling?
524 * Add octet_length_server() and octet_length_client()
525 * Make octet_length_client() the same as octet_length()?
526 * Fix problems with wrong runtime encoding conversion for NLS message files
527 * Add URL to more complete multi-byte regression tests
529 http://archives.postgresql.org/pgsql-hackers/2005-07/msg00272.php
531 * Fix ILIKE and regular expressions to handle case insensitivity
532 properly in multibyte encodings
534 http://archives.postgresql.org/pgsql-bugs/2005-10/msg00001.php
535 http://archives.postgresql.org/pgsql-patches/2005-11/msg00173.php
537 * Set client encoding based on the client operating system encoding
539 Currently client_encoding is set in postgresql.conf, which
540 defaults to the server encoding.
541 http://archives.postgresql.org/pgsql-hackers/2006-08/msg01696.php
543 * Change memory allocation for multi-byte functions so memory is
544 allocated inside conversion functions
546 Currently we preallocate memory based on worst-case usage.
553 * Automatically create rules on views so they are updateable, per SQL99
555 We can only auto-create rules for simple views. For more complex
556 cases users will still have to write rules manually.
558 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00586.php
559 http://archives.postgresql.org/pgsql-patches/2006-08/msg00255.php
561 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
562 * Allow VIEW/RULE recompilation when the underlying tables change
564 Another issue is whether underlying table changes should be reflected
565 in the view, e.g. should SELECT * show additional columns if they
566 are added after the view is created.
568 * Make it possible to use RETURNING together with conditional DO INSTEAD
569 rules, such as for partitioning setups
571 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00577.php
573 * Add the ability to automatically create materialized views
575 Right now materialized views require the user to create triggers on the
576 main table to keep the summary table current. SQL syntax should be able
577 to manager the triggers and summary table automatically. A more
578 sophisticated implementation would automatically retrieve from the
579 summary table when the main table is referenced, if possible.
586 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
587 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
588 * %Allow SET CONSTRAINTS to be qualified by schema/table name
589 * %Add a separate TRUNCATE permission
591 Currently only the owner can TRUNCATE a table because triggers are not
592 called, and the table is locked in exclusive mode.
594 * Allow PREPARE of cursors
595 * Allow finer control over the caching of prepared query plans
597 Currently queries prepared via the libpq API are planned on first
598 execute using the supplied parameters --- allow SQL PREPARE to do the
599 same. Also, allow control over replanning prepared queries either
600 manually or automatically when statistics for execute parameters
601 differ dramatically from those used during planning.
603 * Improve logging of prepared transactions recovered during startup
605 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00092.php
607 * Add a GUC variable to warn about non-standard SQL usage in queries
608 * Add SQL-standard MERGE/REPLACE/UPSERT command
610 MERGE is typically used to merge two tables. REPLACE or UPSERT
611 command does UPDATE, or on failure, INSERT. This is similar to UPDATE,
612 then for unmatched rows, INSERT. Whether concurrent access allows
613 modifications which could cause row loss is implementation independent.
614 To implement this cleanly requires that the table have a unique index
615 so duplicate checking can be easily performed. It is possible to do it
616 without a unique index if we require the user to LOCK the table before
619 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00501.php
620 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00536.php
621 http://archives.postgresql.org/pgsql-hackers/2008-04/msg01475.php
623 * Add NOVICE output level for helpful messages like automatic sequence/index
625 * Add GUC to issue notice about statements that use unjoined tables
626 * Allow EXPLAIN to identify tables that were skipped because of
628 * Allow EXPLAIN output to be more easily processed by scripts, perhaps XML
629 * Enable standard_conforming_strings
630 * Make standard_conforming_strings the default in 8.5?
632 When this is done, backslash-quote should be prohibited in non-E''
633 strings because of possible confusion over how such strings treat
634 backslashes. Basically, '' is always safe for a literal single
635 quote, while \' might or might not be based on the backslash
638 * Simplify dropping roles that have objects in several databases
639 * Allow COMMENT ON to accept an expression rather than just a string
640 * Allow the count returned by SELECT, etc to be represented as an int64
641 to allow a higher range of values
642 * Add SQL99 WITH clause to SELECT
643 * Add SQL:2003 WITH RECURSIVE (hierarchical) queries to SELECT
645 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01375.php
646 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00642.php
647 http://archives.postgresql.org/pgsql-patches/2007-03/msg00139.php
648 http://archives.postgresql.org/pgsql-hackers/2007-11/msg01334.php
649 http://archives.postgresql.org/pgsql-patches/2008-01/msg00105.php
650 http://archives.postgresql.org/pgsql-patches/2008-03/msg00327.php
652 * Add DEFAULT .. AS OWNER so permission checks are done as the table
655 This would be useful for SERIAL nextval() calls and CHECK constraints.
657 * Allow DISTINCT to work in multiple-argument aggregate calls
658 * Add column to pg_stat_activity that shows the progress of long-running
659 commands like CREATE INDEX and VACUUM
660 * Implement SQL:2003 window functions
661 * Improve failure message when DROP DATABASE is used on a database that
662 has prepared transactions
663 * Allow INSERT/UPDATE ... RETURNING inside a SELECT 'FROM' clause
665 http://archives.postgresql.org/pgsql-general/2006-09/msg00803.php
666 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00693.php
668 * Increase locking when DROPing objects so dependent objects cannot
669 get dropped while the DROP operation is happening
671 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00937.php
673 * -Allow AS in "SELECT col AS label" to be optional in certain cases
675 * Allow INSERT ... DELETE ... RETURNING, namely allow the DELETE ...
676 RETURNING to supply values to the INSERT
677 http://archives.postgresql.org/pgsql-hackers/2008-02/thrd2.php#00979
679 * Add comments on system tables/columns using the information in
682 Ideally the information would be pulled from the SGML file
685 * Improve reporting of UNION type mismatches
687 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00944.php
688 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00597.php
693 o Allow CREATE TABLE AS to determine column lengths for complex
694 expressions like SELECT col1 || col2
696 o Have WITH CONSTRAINTS also create constraint indexes
698 http://archives.postgresql.org/pgsql-patches/2007-04/msg00149.php
700 o Have CONSTRAINT cname NOT NULL record the contraint name
702 Right now pg_attribute.attnotnull records the NOT NULL status
703 of the column, but does not record the contraint name
705 o Prevent concurrent CREATE TABLE table1 from sometimes returning
706 a cryptic error message
708 http://archives.postgresql.org/pgsql-bugs/2007-10/msg00169.php
712 o Allow UPDATE tab SET ROW (col, ...) = (SELECT...)
714 http://archives.postgresql.org/pgsql-hackers/2006-07/msg01306.php
715 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00865.php
716 http://archives.postgresql.org/pgsql-patches/2007-04/msg00315.php
717 http://archives.postgresql.org/pgsql-patches/2008-03/msg00237.php
719 o Research self-referential UPDATEs that see inconsistent row versions
720 in read-committed mode
722 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00507.php
723 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00016.php
725 o Allow GLOBAL temporary tables to exist as empty by default in
728 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00006.php
733 o Have ALTER TABLE RENAME rename SERIAL sequence names
735 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
737 o Have ALTER SEQUENCE RENAME rename the sequence name stored
738 in the sequence table
740 http://archives.postgresql.org/pgsql-bugs/2007-09/msg00092.php
741 http://archives.postgresql.org/pgsql-bugs/2007-10/msg00007.php
742 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
744 o Add ALTER DOMAIN to modify the underlying data type
745 o %Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
747 http://archives.postgresql.org/pgsql-patches/2006-02/msg00168.php
749 o %Allow ALTER TABLE to change constraint deferrability and actions
750 o Add missing object types for ALTER ... SET SCHEMA
751 o Allow ALTER TABLESPACE to move to different directories
752 o Allow databases to be moved to different tablespaces
753 o Allow moving system tables to other tablespaces, where possible
755 Currently non-global system tables must be in the default database
756 tablespace. Global system tables can never be moved.
758 o Prevent parent tables from altering or dropping constraints
759 like CHECK that are inherited by child tables unless CASCADE
761 o %Prevent child tables from altering or dropping constraints
762 like CHECK that were inherited from the parent table
763 o Have ALTER INDEX update the name of a constraint using that index
764 o Add ALTER TABLE RENAME CONSTRAINT, update index name also
765 o Allow column display reordering by recording a display,
766 storage, and permanent id for every column?
768 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00782.php
774 o Automatically maintain clustering on a table
776 This might require some background daemon to maintain clustering
777 during periods of low usage. It might also require tables to be only
778 partially filled for easier reorganization. Another idea would
779 be to create a merged heap/index data file so an index lookup would
780 automatically access the heap data too. A third idea would be to
781 store heap rows in hashed groups, perhaps using a user-supplied
783 http://archives.postgresql.org/pgsql-performance/2004-08/msg00349.php
785 o %Add default clustering to system tables
787 To do this, determine the ideal cluster index for each system
788 table and set the cluster setting during initdb.
790 o %Add VERBOSE option to report tables as they are processed,
796 o Allow COPY to report error lines and continue
798 This requires the use of a savepoint before each COPY line is
799 processed, with ROLLBACK on COPY failure.
800 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00572.php
802 o Allow COPY on a newly-created table to skip WAL logging
804 On crash recovery, the table involved in the COPY would
805 be removed or have its heap and index files truncated. One
806 issue is that no other backend should be able to add to
807 the table at the same time, which is something that is
808 currently allowed. This currently is done if the table is
809 created inside the same transaction block as the COPY because
810 no other backends can see the table.
812 o Consider using a ring buffer for COPY FROM
814 http://archives.postgresql.org/pgsql-patches/2008-02/msg00140.php
815 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01080.php
817 o Allow COPY FROM to create index entries in bulk
819 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00811.php
821 o Allow COPY in CSV mode to control whether a quoted zero-length
822 string is treated as NULL
824 Currently this is always treated as a zero-length string,
825 which generates an error when loading into an integer column
826 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00905.php
832 o Allow column-level privileges
833 o %Allow GRANT/REVOKE permissions to be applied to all schema objects
836 The proposed syntax is:
837 GRANT SELECT ON ALL TABLES IN public TO phpuser;
838 GRANT SELECT ON NEW TABLES IN public TO phpuser;
840 o Allow GRANT/REVOKE permissions to be inherited by objects based on
843 o Allow SERIAL sequences to inherit permissions from the base table?
848 o Prevent DROP TABLE from dropping a row referenced by its own open
854 o Allow INSERT/UPDATE of the system-generated oid value for a row
855 o In rules, allow VALUES() to contain a mixture of 'old' and 'new'
861 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
867 o Allow LISTEN/NOTIFY to store info in memory rather than tables?
869 Currently LISTEN/NOTIFY information is stored in pg_listener.
870 Storing such information in memory would improve performance.
872 o Add optional textual message to NOTIFY
874 This would allow an informational message to be added to the notify
875 message, perhaps indicating the row modified or other custom
878 o Allow multiple identical NOTIFY events to always be communicated
879 to the client, rather than sent as a single notification to the
882 http://archives.postgresql.org/pgsql-general/2008-01/msg00057.php
884 o Allow NOTIFY in rules involving conditionals
885 o Improve LISTEN concurrency
887 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01106.php
891 Referential Integrity
892 =====================
894 * Add MATCH PARTIAL referential integrity
895 * Change foreign key constraint for array -> element to mean element
897 * Fix problem when cascading referential triggers make changes on
898 cascaded tables, seeing the tables in an intermediate state
900 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00174.php
901 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00174.php
903 * Allow DEFERRABLE and end-of-statement UNIQUE constraints?
905 This would allow UPDATE tab SET col = col + 1 to work if col has
906 a unique index. Currently, uniqueness checks are done while the
907 command is being executed, rather than at the end of the statement
909 http://people.planetpostgresql.org/greg/index.php?/archives/2006/06/10.html
910 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01458.php
912 * Optimize referential integrity checks
914 http://archives.postgresql.org/pgsql-performance/2005-10/msg00458.php
915 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00744.php
918 Server-Side Languages
919 =====================
922 o Fix RENAME to work on variables other than OLD/NEW
924 http://archives.postgresql.org/pgsql-hackers/2002-03/msg00591.php
925 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01615.php
926 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01587.php
928 o Allow function parameters to be passed by name,
929 get_employee_salary(12345 AS emp_id, 2001 AS tax_year)
930 o Allow handling of %TYPE arrays, e.g. tab.col%TYPE[]
931 o Allow listing of record column names, and access to
932 record columns via variables, e.g. columns := r.(*),
935 http://archives.postgresql.org/pgsql-patches/2005-07/msg00458.php
936 http://archives.postgresql.org/pgsql-patches/2006-05/msg00302.php
937 http://archives.postgresql.org/pgsql-patches/2006-06/msg00031.php
939 o Add support for SCROLL cursors
940 o Add support for WITH HOLD cursors
941 o Allow row and record variables to be set to NULL constants,
942 and allow NULL tests on such variables
944 Because a row is not scalar, do not allow assignment
945 from NULL-valued scalars.
947 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00070.php
949 o Review handling of MOVE and FETCH
951 http://archives.postgresql.org/pgsql-patches/2007-04/msg00527.php
953 o Improve logic of determining if an identifier is a a
954 variable or column name
956 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00436.php
958 o Consider keeping seperate cached copies when search_path changes
960 http://archives.postgresql.org/pgsql-hackers/2008-01/msg01009.php
962 o Add CASE capability to language (already in SQL)
964 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00696.php
970 o Add table function support to pltcl, plpythonu
971 o Add support for polymorphic arguments and return types to
972 languages other than PL/PgSQL
973 o Add capability to create and call PROCEDURES
974 o Add support for OUT and INOUT parameters to languages other
976 o Add PL/PythonU tracebacks
978 http://archives.postgresql.org/pgsql-patches/2006-02/msg00288.php
980 o Allow data to be passed in native language formats, rather
983 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00289.php
985 o Add ability to obfuscate function bodies
987 http://archives.postgresql.org/pgsql-patches/2008-01/msg00125.php
994 * Have pg_ctl look at PGHOST in case it is a socket directory?
995 * Allow pg_ctl to work properly with configuration files located outside
998 pg_ctl can not read the pid file because it isn't located in the
999 config directory but in the PGDATA directory. The solution is to
1000 allow pg_ctl to read and understand postgresql.conf to find the
1001 data_directory value.
1003 * Add a function like pg_get_indexdef() that report more detailed index
1006 http://archives.postgresql.org/pgsql-bugs/2007-12/msg00166.php
1009 o Have psql show current values for a sequence
1010 o Move psql backslash database information into the backend, use
1013 This would allow non-psql clients to pull the same information out
1014 of the database as psql.
1015 http://archives.postgresql.org/pgsql-hackers/2004-01/msg00191.php
1017 o Make psql's \d commands more consistent
1019 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
1020 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
1022 o Consistently display privilege information for all objects in psql
1023 o Add auto-expanded mode so expanded output is used if the row
1024 length is wider than the screen width.
1026 Consider using auto-expanded mode for backslash commands like \df+.
1028 o Prevent tab completion of SET TRANSACTION from querying the
1029 database and therefore preventing the transaction isolation
1030 level from being set.
1032 Currently SET <tab> causes a database lookup to check all
1033 supported session variables. This query causes problems
1034 because setting the transaction isolation level must be the
1035 first statement of a transaction.
1037 o Add a \set variable to control whether \s displays line numbers
1039 Another option is to add \# which lists line numbers, and
1040 allows command execution.
1042 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00255.php
1044 o Prevent escape string warnings when object names have
1047 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00227.php
1049 o Have \d show foreign keys that reference a table's primary key
1051 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00424.php
1053 o Have \d show child tables that inherit from the specified parent
1054 o Have \l+ show database size, if permissions allow
1056 Ideally it will not generate an error for invalid permissions
1058 o Include the symbolic SQLSTATE name in verbose error reports
1060 http://archives.postgresql.org/pgsql-general/2007-09/msg00438.php
1062 o Improve display if enums
1064 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00826.php
1066 o Add prompt escape to display the client and server versions
1069 * pg_dump / pg_restore
1070 o %Add dumping of comments on index columns and composite type columns
1071 o %Add full object name to the tag field. eg. for operators we need
1072 '=(integer, integer)', instead of just '='.
1073 o Add pg_dumpall custom format dumps?
1074 o Allow selection of individual object(s) of all types, not just
1076 o In a selective dump, allow dumping of an object and all its
1078 o Add options like pg_restore -l and -L to pg_dump
1079 o Stop dumping CASCADE on DROP TYPE commands in clean mode
1080 o Allow pg_dump --clean to drop roles that own objects or have
1082 o Change pg_dump so that a comment on the dumped database is
1083 applied to the loaded database, even if the database has a
1084 different name. This will require new backend syntax, perhaps
1085 COMMENT ON CURRENT DATABASE.
1086 o Remove unnecessary function pointer abstractions in pg_dump source
1088 o Allow pg_dump to utilize multiple CPUs and I/O channels by dumping
1089 multiple objects simultaneously
1091 The difficulty with this is getting multiple dump processes to
1092 produce a single dump output file. It also would require
1093 several sessions to share the same snapshot.
1094 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00205.php
1096 o Allow pg_restore to utilize multiple CPUs and I/O channels by
1097 restoring multiple objects simultaneously
1099 This might require a pg_restore flag to indicate how many
1100 simultaneous operations should be performed. Only pg_dump's
1101 -Fc format has the necessary dependency information.
1102 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00963.php
1104 o To better utilize resources, allow pg_restore to check foreign
1105 keys simultaneously, where possible
1106 o Allow pg_restore to create all indexes of a table
1107 concurrently, via a single heap scan
1109 This requires a pg_dump -Fc file because that format contains
1110 the required dependency information.
1111 http://archives.postgresql.org/pgsql-general/2007-05/msg01274.php
1113 o Allow pg_restore to load different parts of the COPY data
1115 o Prevent pg_dump/pg_restore from being affected by
1118 Using psql to restore a pg_dump dump is also affected.
1120 o Remove pre-7.3 pg_dump code that assumes pg_depend does not exit
1121 o Allow pre/data/post files when schema and data are dumped
1122 separately, for performance reasons
1124 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00205.php
1131 Document differences between ecpg and the SQL standard and
1132 information about the Informix-compatibility module.
1134 o Solve cardinality > 1 for input descriptors / variables?
1135 o Add a semantic check level, e.g. check if a table really exists
1136 o fix handling of DB attributes that are arrays
1137 o Use backend PREPARE/EXECUTE facility for ecpg where possible
1139 o Fix nested C comments
1140 o %sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
1141 o Make SET CONNECTION thread-aware, non-standard?
1142 o Allow multidimensional arrays
1143 o Add internationalized message strings
1144 o Implement COPY FROM STDIN
1148 o Add PQescapeIdentifierConn()
1149 o Prevent PQfnumber() from lowercasing unquoted the column name
1151 PQfnumber() should never have been doing lowercasing, but
1152 historically it has so we need a way to prevent it
1154 o Allow statement results to be automatically batched to the client
1156 Currently all statement results are transferred to the libpq
1157 client before libpq makes the results available to the
1158 application. This feature would allow the application to make
1159 use of the first result rows while the rest are transferred, or
1160 held on the server waiting for them to be requested by libpq.
1161 One complexity is that a statement like SELECT 1/col could error
1162 out mid-way through the result set.
1164 o Consider disallowing multiple queries in PQexec() as an
1165 additional barrier to SQL injection attacks
1167 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00184.php
1169 o Add PQexecf() that allows complex parameter substitution
1171 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01803.php
1173 o Add SQLSTATE severity to PGconn return status
1175 http://archives.postgresql.org/pgsql-interfaces/2007-11/msg00015.php
1181 * Add deferred trigger queue file
1183 Right now all deferred trigger information is stored in backend
1184 memory. This could exhaust memory for very large trigger queues.
1185 This item involves dumping large queues into files.
1187 * Allow triggers to be disabled in only the current session.
1189 This is currently possible by starting a multi-statement transaction,
1190 modifying the system tables, performing the desired SQL, restoring the
1191 system tables, and committing the transaction. ALTER TABLE ...
1192 TRIGGER requires a table lock so it is not ideal for this usage.
1194 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
1196 If the dump is known to be valid, allow foreign keys to be added
1197 without revalidating the data.
1199 * Allow statement-level triggers to access modified rows
1200 * Support triggers on columns
1202 http://archives.postgresql.org/pgsql-patches/2005-07/msg00107.php
1204 * Allow AFTER triggers on system tables
1206 System tables are modified in many places in the backend without going
1207 through the executor and therefore not causing triggers to fire. To
1208 complete this item, the functions that modify system tables will have
1211 * Tighten trigger permission checks
1213 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00564.php
1215 * Allow BEFORE INSERT triggers on views
1217 http://archives.postgresql.org/pgsql-general/2007-02/msg01466.php
1219 * -Add ability to trigger on TRUNCATE
1226 * Add UNIQUE capability to non-btree indexes
1227 * Prevent index uniqueness checks when UPDATE does not modify the column
1229 Uniqueness (index) checks are done when updating a column even if the
1230 column is not modified by the UPDATE.
1232 * Allow the creation of on-disk bitmap indexes which can be quickly
1233 combined with other bitmap indexes
1235 Such indexes could be more compact if there are only a few distinct values.
1236 Such indexes can also be compressed. Keeping such indexes updated can be
1239 http://archives.postgresql.org/pgsql-patches/2005-07/msg00512.php
1240 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01107.php
1241 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00265.php
1242 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01214.php
1243 http://archives.postgresql.org/pgsql-patches/2007-05/msg00013.php
1244 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00741.php
1246 * Allow accurate statistics to be collected on indexes with more than
1247 one column or expression indexes, perhaps using per-index statistics
1249 http://archives.postgresql.org/pgsql-performance/2006-10/msg00222.php
1250 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01131.php
1252 * Consider increasing the number of default statistics target, and
1253 reduce statistics target overhead
1255 Also consider having a larger statistics target for indexed columns
1256 and expression indexes
1257 http://archives.postgresql.org/pgsql-general/2007-05/msg01228.php
1258 http://archives.postgresql.org/pgsql-general/2007-06/msg00542.php
1259 http://archives.postgresql.org/pgsql-hackers/2008-01/msg01066.php
1261 * Consider smaller indexes that record a range of values per heap page,
1262 rather than having one index entry for every heap row
1264 This is useful if the heap is clustered by the indexed values.
1265 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00341.php
1266 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01264.php
1267 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00465.php
1268 http://archives.postgresql.org/pgsql-patches/2007-03/msg00163.php
1269 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00014.php
1270 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00487.php
1271 http://archives.postgresql.org/pgsql-hackers/2008-04/msg01589.php
1273 * Add REINDEX CONCURRENTLY, like CREATE INDEX CONCURRENTLY
1275 This is difficult because you must upgrade to an exclusive table lock
1276 to replace the existing index file. CREATE INDEX CONCURRENTLY does not
1277 have this complication. This would allow index compaction without
1279 http://archives.postgresql.org/pgsql-performance/2007-08/msg00289.php
1281 * Allow multiple indexes to be created concurrently, ideally via a
1282 single heap scan, and have pg_restore use it
1284 * Consider sorting entries before inserting into btree index
1286 http://archives.postgresql.org/pgsql-general/2008-01/msg01010.php
1292 o Allow inherited tables to inherit indexes, UNIQUE constraints,
1293 and primary/foreign keys
1294 o Honor UNIQUE INDEX on base column in INSERTs/UPDATEs
1295 on inherited table, e.g. INSERT INTO inherit_table
1296 (unique_index_col) VALUES (dup) should fail
1298 The main difficulty with this item is the problem of
1299 creating an index that can span multiple tables.
1301 o Allow SELECT ... FOR UPDATE on inherited tables
1302 o Require all CHECK constraints to be inherited
1304 http://archives.postgresql.org/pgsql-bugs/2007-04/msg00026.php
1306 o Add checks to prevent a CREATE RULE views on inherited tables
1308 http://archives.postgresql.org/pgsql-general/2008-02/msg01420.php
1309 http://archives.postgresql.org/pgsql-general/2008-03/msg00077.php
1314 o Add more GIST index support for geometric data types
1315 o Allow GIST indexes to create certain complex index types, like
1316 digital trees (see Aoki)
1321 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00051.php
1323 o Pack hash index buckets onto disk pages more efficiently
1325 Currently only one hash bucket can be stored on a page. Ideally
1326 several hash buckets could be stored on a single page and greater
1327 granularity used for the hash algorithm.
1329 http://archives.postgresql.org/pgsql-hackers/2004-06/msg00168.php
1331 o Consider sorting hash buckets so entries can be found using a
1332 binary search, rather than a linear scan
1333 o In hash indexes, consider storing the hash value with or instead
1335 o Add WAL logging for crash recovery
1336 o Allow multi-column hash indexes
1337 o -During index creation, pre-sort the tuples to improve build speed
1344 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
1346 Ideally this requires a separate test program that can be run
1347 at initdb time or optionally later. Consider O_SYNC when
1350 * Add program to test if fsync has a delay compared to non-fsync
1351 * Consider sorting writes during checkpoint
1353 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00541.php
1362 We could use a fixed row count and a +/- count to follow MVCC
1363 visibility rules, or a single cached value could be used and
1364 invalidated if anyone modifies the table. Another idea is to
1365 get a count directly from a unique index, but for this to be
1366 faster than a sequential scan it must avoid access to the heap
1367 to obtain tuple visibility information.
1369 * Provide a way to calculate an "estimated COUNT(*)"
1371 Perhaps by using the optimizer's cardinality estimates or random
1374 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00943.php
1376 * Allow data to be pulled directly from indexes
1378 Currently indexes do not have enough tuple visibility information
1379 to allow data to be pulled from the index without also accessing
1380 the heap. One way to allow this is to set a bit on index tuples
1381 to indicate if a tuple is currently visible to all transactions
1382 when the first valid heap lookup happens. This bit would have to
1383 be cleared when a heap tuple is expired.
1385 Another idea is to maintain a bitmap of heap pages where all rows
1386 are visible to all backends, and allow index lookups to reference
1387 that bitmap to avoid heap lookups, perhaps the same bitmap we might
1388 add someday to determine which heap pages need vacuuming. Frequently
1389 accessed bitmaps would have to be stored in shared memory. One 8k
1390 page of bitmaps could track 512MB of heap pages.
1392 A third idea would be for a heap scan to check if all rows are visible
1393 and if so set a per-table flag which can be checked by index scans.
1394 Any change to the table would have to clear the flag. To detect
1395 changes during the heap scan a counter could be set at the start and
1396 checked at the end --- if it is the same, the table has not been
1397 modified --- any table change would increment the counter.
1399 http://archives.postgresql.org/pgsql-patches/2007-10/msg00166.php
1400 http://archives.postgresql.org/pgsql-patches/2008-01/msg00049.php
1402 * Consider automatic caching of statements at various levels:
1405 o Query execute plan
1407 http://archives.postgresql.org/pgsql-hackers/2008-04/msg00823.php
1409 * Consider increasing internal areas when shared buffers is increased
1411 http://archives.postgresql.org/pgsql-hackers/2005-10/msg01419.php
1413 * Consider decreasing the amount of memory used by PrivateRefCount
1415 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00797.php
1416 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00752.php
1418 * Consider allowing higher priority queries to have referenced buffer
1419 cache pages stay in memory longer
1421 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00562.php
1428 * Improve speed with indexes
1430 For large table adjustments during VACUUM FULL, it is faster to cluster
1431 or reindex rather than update the index. Also, index updates can bloat
1434 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00024.php
1435 http://archives.postgresql.org/pgsql-performance/2007-05/msg00296.php
1436 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00307.php
1438 * Auto-fill the free space map by scanning the buffer cache or by
1439 checking pages written by the background writer
1441 http://archives.postgresql.org/pgsql-hackers/2006-02/msg01125.php
1442 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00011.php
1444 * Create a bitmap of pages that need vacuuming
1446 Instead of sequentially scanning the entire table, have the background
1447 writer or some other process record pages that have expired rows, then
1448 VACUUM can look at just those pages rather than the entire table. In
1449 the event of a system crash, the bitmap would probably be invalidated.
1450 One complexity is that index entries still have to be vacuumed, and
1451 doing this without an index scan (by using the heap values to find the
1452 index entry) might be slow and unreliable, especially for user-defined
1455 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01188.php
1456 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00121.php
1457 http://archives.postgresql.org/pgsql-patches/2007-03/msg00508.php
1458 http://archives.postgresql.org/pgsql-patches/2007-04/msg00347.php
1459 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00156.php
1460 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00546.php
1461 http://archives.postgresql.org/pgsql-hackers/2008-04/msg00416.php
1463 * Allow FSM to return free space toward the beginning of the heap file,
1464 in hopes that empty pages at the end can be truncated by VACUUM
1465 * Allow FSM page return free space based on table clustering, to assist
1466 in maintaining clustering?
1467 * Improve dead row detection during multi-statement transactions usage
1469 http://archives.postgresql.org/pgsql-patches/2007-03/msg00358.php
1471 * Consider a more compact data representation for dead tuples
1473 http://archives.postgresql.org/pgsql-patches/2007-05/msg00143.php
1478 o %Issue log message to suggest VACUUM FULL if a table is nearly
1480 o Improve control of auto-vacuum
1482 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00876.php
1484 o Prevent long-lived temporary tables from causing frozen-xid
1485 advancement starvation
1487 The problem is that autovacuum cannot vacuum them to set frozen xids;
1488 only the session that created them can do that.
1489 http://archives.postgresql.org/pgsql-general/2007-06/msg01645.php
1491 o Store per-table autovacuum settings in pg_class.reloptions.
1493 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01440.php
1494 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00724.php
1496 o Prevent autovacuum from running if an old transaction is still
1497 running from the last vacuum
1499 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00899.php
1506 * Fix priority ordering of read and write light-weight locks (Neil)
1508 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00893.php
1509 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00905.php
1511 * Fix problem when multiple subtransactions of the same outer transaction
1512 hold different types of locks, and one subtransaction aborts
1514 http://archives.postgresql.org/pgsql-hackers/2006-11/msg01011.php
1515 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00001.php
1516 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00435.php
1517 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00773.php
1519 * Allow UPDATEs on only non-referential integrity columns not to conflict
1520 with referential integrity locks
1522 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00073.php
1524 * Add idle_in_transaction_timeout GUC so locks are not held for long
1527 * Improve deadlock detection when a page cleaning lock conflicts
1528 with a shared buffer that is pinned
1530 http://archives.postgresql.org/pgsql-bugs/2008-01/msg00138.php
1531 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00873.php
1532 http://archives.postgresql.org/pgsql-committers/2008-01/msg00365.php
1534 * Detect deadlocks involving LockBufferForCleanup()
1536 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00873.php
1540 Startup Time Improvements
1541 =========================
1543 * Experiment with multi-threaded backend for backend creation
1545 This would prevent the overhead associated with process creation. Most
1546 operating systems have trivial process creation time compared to
1547 database startup overhead, but a few operating systems (Win32,
1548 Solaris) might benefit from threading. Also explore the idea of
1549 a single session using multiple threads to execute a statement faster.
1556 * Eliminate need to write full pages to WAL before page modification
1558 Currently, to protect against partial disk page writes, we write
1559 full page images to WAL before they are modified so we can correct any
1560 partial page writes during recovery. These pages can also be
1561 eliminated from point-in-time archive files.
1562 http://archives.postgresql.org/pgsql-hackers/2002-06/msg00655.php
1564 o When off, write CRC to WAL and check file system blocks
1567 If CRC check fails during recovery, remember the page in case
1568 a later CRC for that page properly matches.
1570 o Write full pages during file system write and not when
1571 the page is modified in the buffer cache
1573 This allows most full page writes to happen in the background
1574 writer. It might cause problems for applying WAL on recovery
1575 into a partially-written page, but later the full page will be
1578 * Allow WAL traffic to be streamed to another server for stand-by
1580 * Reduce WAL traffic so only modified values are written rather than
1583 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01589.php
1585 * Allow WAL information to recover corrupted pg_controldata
1587 http://archives.postgresql.org/pgsql-patches/2006-06/msg00025.php
1589 * Find a way to reduce rotational delay when repeatedly writing
1592 Currently fsync of WAL requires the disk platter to perform a full
1593 rotation to fsync again. One idea is to write the WAL to different
1594 offsets that might reduce the rotational delay.
1595 http://archives.postgresql.org/pgsql-hackers/2002-11/msg00483.php
1597 * Allow WAL logging to be turned off for a table, but the table
1598 might be dropped or truncated during crash recovery
1600 Allow tables to bypass WAL writes and just fsync() dirty pages on
1601 commit. This should be implemented using ALTER TABLE, e.g. ALTER
1602 TABLE PERSISTENCE [ DROP | TRUNCATE | DEFAULT ]. Tables using
1603 non-default logging should not use referential integrity with
1604 default-logging tables. A table without dirty buffers during a
1605 crash could perhaps avoid the drop/truncate.
1606 http://archives.postgresql.org/pgsql-hackers/2005-12/msg01016.php
1608 * Allow WAL logging to be turned off for a table, but the table would
1609 avoid being truncated/dropped
1611 To do this, only a single writer can modify the table, and writes
1612 must happen only on new pages so the new pages can be removed during
1613 crash recovery. Readers can continue accessing the table. Such
1614 tables probably cannot have indexes. One complexity is the handling
1615 of indexes on TOAST tables.
1616 http://archives.postgresql.org/pgsql-hackers/2005-12/msg01016.php
1618 * Speed WAL recovery by allowing more than one page to be prefetched
1620 This should be done utilizing the same infrastructure used for
1621 prefetching in general to avoid introducing complex error-prone code
1623 http://archives.postgresql.org/pgsql-general/2007-12/msg00683.php
1624 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00497.php
1625 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01279.php
1627 * Improve WAL concurrency by increasing lock granularity
1629 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00556.php
1631 * Be more aggressive about creating WAL files
1633 http://archives.postgresql.org/pgsql-hackers/2007-10/msg01325.php
1634 http://archives.postgresql.org/pgsql-hackers/2004-07/msg01075.php
1635 http://archives.postgresql.org/pgsql-hackers/2005-04/msg00556.php
1637 * Have resource managers report the duration of their status changes
1639 http://archives.postgresql.org/pgsql-hackers/2007-10/msg01468.php
1641 * Move pgfoundry's xlogdump to /contrib and have it rely more closely
1642 on the WAL backend code
1644 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00035.php
1649 Optimizer / Executor
1650 ====================
1652 * Improve selectivity functions for geometric operators
1653 * Precompile SQL functions to avoid overhead
1654 * Create utility to compute accurate random_page_cost value
1655 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
1656 * Have EXPLAIN ANALYZE issue NOTICE messages when the estimated and
1657 actual row counts differ by a specified percentage
1658 * Improve how ANALYZE computes in-doubt tuples
1660 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00771.php
1662 * Consider using hash buckets to do DISTINCT, rather than sorting
1664 This would be beneficial when there are few distinct values. This is
1665 already used by GROUP BY.
1667 * Log statements where the optimizer row estimates were dramatically
1668 different from the number of rows actually found?
1669 * Consider compressed annealing to search for query plans
1671 This might replace GEQO, http://sixdemonbag.org/Djinni.
1673 * Improve merge join performance by allowing mark/restore of
1676 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00096.php
1678 * Consider using a hash for joining to a large IN (VALUES ...) list
1680 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00450.php
1687 * Consider having the background writer update the transaction status
1688 hint bits before writing out the page
1690 Implementing this requires the background writer to have access to system
1691 catalogs and the transaction status log.
1693 * Consider adding buffers the background writer finds reusable to the
1696 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00781.php
1698 * Automatically tune bgwriter_delay based on activity rather then using a
1701 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00781.php
1703 * Consider wither increasing BM_MAX_USAGE_COUNT improves performance
1705 http://archives.postgresql.org/pgsql-hackers/2007-06/msg01007.php
1707 * Test to see if calling PreallocXlogFiles() from the background writer
1708 will help with WAL segment creation latency
1710 http://archives.postgresql.org/pgsql-patches/2007-06/msg00340.php
1714 Miscellaneous Performance
1715 =========================
1717 * Do async I/O for faster random read-ahead of data
1719 Async I/O allows multiple I/O requests to be sent to the disk with
1720 results coming back asynchronously.
1722 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00820.php
1723 http://archives.postgresql.org/pgsql-performance/2007-09/msg00255.php
1724 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00027.php
1725 http://archives.postgresql.org/pgsql-patches/2008-01/msg00170.php
1727 * Use mmap() rather than SYSV shared memory or to write WAL files?
1729 This would remove the requirement for SYSV SHM but would introduce
1730 portability issues. Anonymous mmap (or mmap to /dev/zero) is required
1731 to prevent I/O overhead.
1733 * Consider mmap()'ing files into a backend?
1735 Doing I/O to large tables would consume a lot of address space or
1736 require frequent mapping/unmapping. Extending the file also causes
1737 mapping problems that might require mapping only individual pages,
1738 leading to thousands of mappings. Another problem is that there is no
1739 way to _prevent_ I/O to disk from the dirty shared buffers so changes
1740 could hit disk before WAL is written.
1742 * Add a script to ask system configuration questions and tune postgresql.conf
1743 * Consider ways of storing rows more compactly on disk
1745 o Reduce the row header size?
1746 o Consider reducing on-disk varlena length from four bytes to
1747 two because a heap row cannot be more than 64k in length
1749 * Consider increasing NUM_CLOG_BUFFERS
1751 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00030.php
1752 http://archives.postgresql.org/pgsql-performance/2007-08/msg00024.php
1754 * Consider transaction start/end performance improvements
1756 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00948.php
1757 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00361.php
1759 * Allow user configuration of TOAST thresholds
1761 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00213.php
1762 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00082.php
1764 * Allow configuration of backend priorities via the operating system
1766 Though backend priorities make priority inversion during lock
1767 waits possible, research shows that this is not a huge problem.
1769 http://archives.postgresql.org/pgsql-general/2007-02/msg00493.php
1771 * Experiment with multi-threaded backend better I/O utilization
1773 This would allow a single query to make use of multiple I/O channels
1774 simultaneously. One idea is to create a background reader that can
1775 pre-fetch sequential and index scan pages needed by other backends.
1776 This could be expanded to allow concurrent reads from multiple devices
1777 in a partitioned table.
1779 * Experiment with multi-threaded backend better CPU utilization
1781 This would allow several CPUs to be used for a single query, such as
1782 for sorting or query execution.
1784 * Consider increasing the minimum allowed number of shared buffers
1786 http://archives.postgresql.org/pgsql-bugs/2008-02/msg00157.php
1788 * Expire published xmin for read-only and idle transactions
1790 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00343.php
1792 * Consider if CommandCounterIncrement() can avoid its
1793 AcceptInvalidationMessages() call
1795 http://archives.postgresql.org/pgsql-committers/2007-11/msg00585.php
1797 * Improve performance of shared invalidation queue for multiple CPUs
1799 http://archives.postgresql.org/pgsql-performance/2008-01/msg00023.php
1801 * Consider Cartesian joins when both relations are needed to form an
1802 indexscan qualification for a third relation
1804 http://archives.postgresql.org/pgsql-performance/2007-12/msg00090.php
1806 * Consider not storing a NULL bitmap on disk if all the NULLs are
1809 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00624.php
1810 http://archives.postgresql.org/pgsql-patches/2007-12/msg00109.php
1812 * Sort large UPDATE/DELETEs so it is done in heap order
1814 http://archives.postgresql.org/pgsql-hackers/2008-01/msg01119.php
1816 * -Avoid tuple some tuple copying in sort routines
1817 * SMP scalability improvements
1819 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00439.php
1820 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00206.php
1821 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00361.php
1823 * Research reducing deTOASTing in more places
1825 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00895.php
1827 * Consider being smarter about memory and external files used during
1830 http://archives.postgresql.org/pgsql-hackers/2007-11/msg01101.php
1831 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00045.php
1833 * Allow one transaction to see tuples using the snapshot of another
1836 This would assist multiple backends in working together.
1837 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00400.php
1843 * Add use of 'const' for variables in source tree
1844 * Move some things from contrib into main tree
1845 * %Remove warnings created by -Wcast-align
1846 * Move platform-specific ps status display info from ps_status.c to ports
1847 * Add optional CRC checksum to heap and index pages
1848 * Improve documentation to build only interfaces (Marc)
1849 * Remove or relicense modules that are not under the BSD license, if possible
1850 * Acquire lock on a relation before building a relcache entry for it
1851 * Allow cross-compiling by generating the zic database on the target system
1852 * Improve NLS maintenance of libpgport messages linked onto applications
1853 * Clean up casting in contrib/isn
1855 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00245.php
1857 * Use UTF8 encoding for NLS messages so all server encodings can
1859 * Update Bonjour to work with newer cross-platform SDK
1861 http://archives.postgresql.org/pgsql-hackers/2006-09/msg02238.php
1862 http://archives.postgresql.org/pgsql-patches/2006-10/msg00048.php
1864 * Consider detoasting keys before sorting
1865 * Consider GnuTLS if OpenSSL license becomes a problem
1867 http://archives.postgresql.org/pgsql-patches/2006-05/msg00040.php
1868 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01213.php
1870 * Consider changing documentation format from SGML to XML
1872 http://archives.postgresql.org/pgsql-docs/2006-12/msg00152.php
1874 * Consider making NAMEDATALEN more configurable in future releases
1875 * Update our code to handle 64-bit timezone files to match the zic
1876 source code, which now uses them
1877 * Have configure choose integer datetimes by default
1879 http://archives.postgresql.org/pgsql-patches/2007-05/msg00046.php
1881 * Support scoped IPv6 addresses
1883 http://archives.postgresql.org/pgsql-bugs/2007-05/msg00111.php
1885 * Consider allowing 64-bit integers and floats to be passed by value on
1888 Also change 32-bit floats (float4) to be passed by value at the same
1891 * Research use of signals and sleep wake ups
1893 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00003.php
1895 * Add automated check for invalid C++ source code constructs
1897 http://archives.postgresql.org/pgsql-patches/2007-07/msg00056.php
1899 * Consider simplifying how memory context resets handle child contexts
1901 http://archives.postgresql.org/pgsql-patches/2007-08/msg00067.php
1903 * Remove use of MAKE_PTR and MAKE_OFFSET macros
1905 http://archives.postgresql.org/pgsql-general/2007-08/msg01510.php
1907 * Convert single quotes to apostrophes in the PDF documentation
1909 http://archives.postgresql.org/pgsql-docs/2007-12/msg00059.php
1911 * Create three versions of libpgport to simplify client code
1913 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00154.php
1915 * Remove old-style routines for manipulating tuples
1917 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00851.php
1919 * Improve detection of shared memory segments being used by other
1922 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00656.php
1924 * Implement the non-threaded Avahi service discovery protocol
1926 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00939.php
1927 http://archives.postgresql.org/pgsql-patches/2008-02/msg00097.php
1928 http://archives.postgresql.org/pgsql-hackers/2008-03/msg01211.php
1929 http://archives.postgresql.org/pgsql-patches/2008-04/msg00001.php
1933 o Remove configure.in check for link failure when cause is found
1934 o Remove readdir() errno patch when runtime/mingwex/dirent.c rev
1936 o Remove psql newline patch when we find out why mingw outputs an
1938 o Allow psql to use readline once non-US code pages work with
1940 o Fix problem with shared memory on the Win32 Terminal Server
1941 o Diagnose problem where shared memory can sometimes not be
1942 attached by postmaster children
1944 http://archives.postgresql.org/pgsql-general/2007-08/msg01377.php
1946 o Improve signal handling
1948 http://archives.postgresql.org/pgsql-patches/2005-06/msg00027.php
1950 o Convert MSVC build system to remove most batch files
1952 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00961.php
1954 o Prevent SSL from sending network packets to avoid interference
1955 with Win32 signal emulation
1957 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00455.php
1959 o Support pgxs when using MSVC
1961 o Fix MSVC NLS support, like for to_char()
1963 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00485.php
1964 http://archives.postgresql.org/pgsql-patches/2008-02/msg00038.php
1966 o Find a correct rint() substitute on Windows
1968 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00808.php
1972 * Wire Protocol Changes
1974 o Allow dynamic character set handling
1975 o Add decoded type, length, precision
1977 o Update clients to use data types, typmod, schema.table.column names
1978 of result sets using new statement protocol
1984 * Add pre-parsing phase that converts non-ISO syntax to supported
1987 This could allow SQL written for other databases to run without
1990 * Allow plug-in modules to emulate features from other databases
1991 * Add features of Oracle-style packages (Pavel)
1993 A package would be a schema with session-local variables,
1994 public/private functions, and initialization functions. It
1995 is also possible to implement these capabilities
1996 in any schema and not use a separate "packages"
1999 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00384.php
2001 * Consider allowing control of upper/lower case folding of unquoted
2004 http://archives.postgresql.org/pgsql-hackers/2004-04/msg00818.php
2005 http://archives.postgresql.org/pgsql-hackers/2006-10/msg01527.php
2006 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00849.php
2008 * Add autonomous transactions
2010 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00893.php
2014 Features We Do _Not_ Want
2015 =========================
2017 * All backends running as threads in a single process (not wanted)
2019 This eliminates the process protection we get from the current setup.
2020 Thread creation is usually the same overhead as process creation on
2021 modern systems, so it seems unwise to use a pure threaded model.
2023 * Optimizer hints (not wanted)
2025 Optimizer hints are used to work around problems in the optimizer. We
2026 would rather have the problems reported and fixed.
2028 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00506.php
2029 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00517.php
2030 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00663.php
2033 Because we support postfix operators, it isn't possible to make AS
2034 optional and continue to use bison.
2035 http://archives.postgresql.org/pgsql-hackers/2003-04/msg00436.php
2037 http://archives.postgresql.org/pgsql-sql/2006-08/msg00164.php
2039 * Embedded server (not wanted)
2041 While PostgreSQL clients runs fine in limited-resource environments, the
2042 server requires multiple processes and a stable pool of resources to
2043 run reliabily and efficiently. Stripping down the PostgreSQL server
2044 to run in the same process address space as the client application
2045 would add too much complexity and failure cases.