3 Current maintainer: Bruce Momjian (bruce@momjian.us)
4 Last updated: Thu Apr 17 16:56:51 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 Bracketed items, "[]", have more detail.
14 This list contains all known PostgreSQL bugs and feature requests. If
15 you would like to work on an item, please read the Developer's FAQ
16 first. There is also a developer's wiki at
17 http://developer.postgresql.org.
23 * -Allow administrators to safely terminate individual sessions either
24 via an SQL function or SIGTERM
25 * Check for unreferenced table files created by transactions that were
26 in-progress when the server terminated abruptly
28 http://archives.postgresql.org/pgsql-patches/2006-06/msg00096.php
30 * Set proper permissions on non-system schemas during db creation
32 Currently all schemas are owned by the super-user because they are copied
33 from the template1 database. However, since all objects are inherited
34 from the template database, it is not clear that setting schemas to the db
37 * Add function to report the time of the most recent server reload
38 * Allow statistics collector information to be pulled from the collector
39 process directly, rather than requiring the collector to write a
40 filesystem file twice a second?
41 * Reduce file system activity overhead of statistics file pgstat.stat
43 http://archives.postgresql.org/pgsql-general/2007-12/msg00106.php
45 * Allow statistics last vacuum/analyze execution times to be displayed
46 without requiring stats_row_level to be enabled
48 http://archives.postgresql.org/pgsql-docs/2007-04/msg00028.php
50 * Allow log_min_messages to be specified on a per-module basis
52 This would allow administrators to see more detailed information from
53 specific sections of the backend, e.g. checkpoints, autovacuum, etc.
54 Another idea is to allow separate configuration files for each module,
55 or allow arbitrary SET commands to be passed to them.
57 * Simplify ability to create partitioned tables
59 This would allow creation of partitioned tables without requiring
60 creation of triggers or rules for INSERT/UPDATE/DELETE, and constraints
61 for rapid partition selection. Options could include range and hash
64 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00375.php
65 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00151.php
66 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00028.php
67 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00248.php
68 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00387.php
69 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00413.php
71 * Allow auto-selection of partitioned tables for min/max() operations
72 * Allow more complex user/database default GUC settings
74 Currently ALTER USER and ALTER DATABASE support per-user and
75 per-database defaults. Consider adding per-user-and-database
76 defaults so things like search_path can be defaulted for a
77 specific user connecting to a specific database.
79 * Allow custom variable classes that can restrict who can set the values
81 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00911.php
83 * Implement the SQL standard mechanism whereby REVOKE ROLE revokes only
84 the privilege granted by the invoking role, and not those granted
87 http://archives.postgresql.org/pgsql-bugs/2007-05/msg00010.php
89 * Allow SSL authentication/encryption over unix domain sockets
91 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00924.php
93 * Allow SSL key file permission checks to be optionally disabled when
94 sharing SSL keys with other applications
96 http://archives.postgresql.org/pgsql-bugs/2007-12/msg00069.php
98 * Allow client certificate names to be checked against the client
101 This is already implemented in
102 libpq/fe-secure.c::verify_peer_name_matches_certificate() but the code
105 * Configuration files
107 o Allow pg_hba.conf to specify host names along with IP addresses
109 Host name lookup could occur when the postmaster reads the
110 pg_hba.conf file, or when the backend starts. Another
111 solution would be to reverse lookup the connection IP and
112 check that hostname against the host names in pg_hba.conf.
113 We could also then check that the host name maps to the IP
116 o %Allow postgresql.conf file values to be changed via an SQL
117 API, perhaps using SET GLOBAL
118 o Allow the server to be stopped/restarted via an SQL API
119 o Issue a warning if a change-on-restart-only postgresql.conf value
120 is modified and the server config files are reloaded
121 o Consider normalizing fractions in postgresql.conf, perhaps
124 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00550.php
126 o Allow Kerberos to disable stripping of realms so we can
127 check the username@realm against multiple realms
129 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00009.php
134 o Allow a database in tablespace t1 with tables created in
135 tablespace t2 to be used as a template for a new database created
136 with default tablespace t2
138 Currently all objects in the default database tablespace must
139 have default tablespace specifications. This is because new
140 databases are created by copying directories. If you mix default
141 tablespace tables and tablespace-specified tables in the same
142 directory, creating a new database from such a mixed directory
143 would create a new database with tables that had incorrect
144 explicit tablespaces. To fix this would require modifying
145 pg_class in the newly copied database, which we don't currently
148 o Allow reporting of which objects are in which tablespaces
150 This item is difficult because a tablespace can contain objects
151 from multiple databases. There is a server-side function that
152 returns the databases which use a specific tablespace, so this
153 requires a tool that will call that function and connect to each
154 database to find the objects in each database for that tablespace.
156 o Allow WAL replay of CREATE TABLESPACE to work when the directory
157 structure on the recovery computer is different from the original
159 o Allow per-tablespace quotas
162 * Point-In-Time Recovery (PITR)
164 o Allow a warm standby system to also allow read-only statements
167 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00050.php
169 o %Create dump tool for write-ahead logs for use in determining
170 transaction id for point-in-time recovery
172 This is useful for checking PITR recovery.
174 o Allow recovery.conf to support the same syntax as
175 postgresql.conf, including quoting
177 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00497.php
179 o Fix server restart problem when the server was shutdown during
182 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00800.php
184 o Recreate pg_xlog/archive_status/ if it doesn't exist after
185 restoring from a PITR backup
187 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00487.php
193 * Change NUMERIC to enforce the maximum precision
194 * Reduce storage space for small NUMERICs
196 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01331.php
197 http://archives.postgresql.org/pgsql-patches/2007-02/msg00505.php
198 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00715.php
200 * Fix data types where equality comparison isn't intuitive, e.g. box
201 * Add support for public SYNONYMs
203 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00519.php
205 * Fix CREATE CAST on DOMAINs
207 http://archives.postgresql.org/pgsql-hackers/2006-05/msg00072.php
208 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01681.php
210 * Allow domains to be cast
212 http://archives.postgresql.org/pgsql-hackers/2003-06/msg01206.php
213 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00289.php
215 * Add support for SQL-standard GENERATED/IDENTITY columns
217 http://archives.postgresql.org/pgsql-hackers/2006-07/msg00543.php
218 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00038.php
219 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00344.php
220 http://archives.postgresql.org/pgsql-patches/2007-05/msg00076.php
221 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00604.php
223 * Improve XML support
225 http://developer.postgresql.org/index.php/XML_Support
227 * Consider placing all sequences in a single table, or create a system
230 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
232 * Allow the UUID type to accept non-standard formats
234 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01214.php
236 * Allow text search dictionary to filter out only stop words
238 http://archives.postgresql.org/pgsql-patches/2007-11/msg00081.php
240 * Consider a function-based API for '@@' full text searches
242 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00511.php
244 * Improve text search error messages
246 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00966.php
247 http://archives.postgresql.org/pgsql-hackers/2007-11/msg01146.php
249 * Consider a special data type for regular expressions
251 http://archives.postgresql.org/pgsql-hackers/2007-08/msg01067.php
253 * Reduce BIT data type overhead using short varlena headers
255 http://archives.postgresql.org/pgsql-general/2007-12/msg00273.php
257 * Allow xml arrays to be cast to other data types
259 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00981.php
260 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00231.php
261 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00471.php
263 * Simplify integer cross-data-type operators
265 http://archives.postgresql.org/pgsql-bugs/2008-01/msg00189.php
267 * Allow XML to accept more liberal DOCTYPE specifications
269 http://archives.postgresql.org/pgsql-general/2008-02/msg00347.php
274 o Allow infinite dates and intervals just like infinite timestamps
275 o Merge hardwired timezone names with the TZ database; allow either
276 kind everywhere a TZ name is currently taken
277 o Allow TIMESTAMP WITH TIME ZONE to store the original timezone
278 information, either zone name or offset from UTC [timezone]
280 If the TIMESTAMP value is stored with a time zone name, interval
281 computations should adjust based on the time zone rules.
283 o Fix SELECT '0.01 years'::interval, '0.01 months'::interval
284 o Add a GUC variable to allow output of interval values in ISO8601
286 o Have timestamp subtraction not call justify_hours()?
288 http://archives.postgresql.org/pgsql-sql/2006-10/msg00059.php
290 o Improve timestamptz subtraction to be DST-aware
292 Currently subtracting one date from another that crosses a
293 daylight savings time adjustment can return '1 day 1 hour', but
294 adding that back to the first date returns a time one hour in
295 the future. This is caused by the adjustment of '25 hours' to
296 '1 day 1 hour', and '1 day' is the same time the next day, even
297 if daylight savings adjustments are involved.
299 o Fix interval display to support values exceeding 2^31 hours
300 o Add overflow checking to timestamp and interval arithmetic
301 o Extend timezone code to allow 64-bit values so we can
302 represent years beyond 2038
304 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01363.php
306 o Use LC_TIME for localized weekday/month names, rather than
309 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00390.php
311 o Add ISO INTERVAL handling
313 http://archives.postgresql.org/pgsql-hackers/2006-01/msg00250.php
314 http://archives.postgresql.org/pgsql-bugs/2006-04/msg00248.php
316 o Support ISO INTERVAL syntax if units cannot be determined from
317 the string, and are supplied after the string
319 The SQL standard states that the units after the string
320 specify the units of the string, e.g. INTERVAL '2' MINUTE
321 should return '00:02:00'. The current behavior has the units
322 restrict the interval value to the specified unit or unit
323 range, INTERVAL '70' SECOND returns '00:00:10'.
325 For syntax that isn't uniquely ISO or PG syntax, like '1' or
326 '1:30', treat as ISO if there is a range specification clause,
327 and as PG if there no clause is present, e.g. interpret '1:30'
328 MINUTE TO SECOND as '1 minute 30 seconds', and interpret
329 '1:30' as '1 hour, 30 minutes'.
331 This makes common cases like SELECT INTERVAL '1' MONTH
332 SQL-standard results. The SQL standard supports a limited
333 number of unit combinations and doesn't support unit names in
334 the string. The PostgreSQL syntax is more flexible in the
335 range of units supported, e.g. PostgreSQL supports '1 year 1
336 hour', while the SQL standard does not.
338 o Add support for year-month syntax, INTERVAL '50-6' YEAR
340 o Interpret INTERVAL '1 year' MONTH as CAST (INTERVAL '1
341 year' AS INTERVAL MONTH), and this should return '12 months'
342 o Round or truncate values to the requested precision, e.g.
343 INTERVAL '11 months' AS YEAR should return one or zero
344 o Support precision, CREATE TABLE foo (a INTERVAL MONTH(3))
349 o Delay resolution of array expression's data type so assignment
350 coercion can be performed on empty array expressions
351 o Add support for arrays of domains
353 http://archives.postgresql.org/pgsql-patches/2007-05/msg00114.php
355 o Allow single-byte header storage for array elements
360 o Improve vacuum of large objects, like contrib/vacuumlo?
361 o Add security checking for large objects
362 o Auto-delete large objects when referencing row is deleted
364 contrib/lo offers this functionality.
366 o Allow read/write into TOAST values like large objects
368 This requires the TOAST column to be stored EXTERNAL.
370 o Add API for 64-bit large object access
372 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00781.php
376 * Add locale-aware MONEY type, and support multiple currencies
378 http://archives.postgresql.org/pgsql-general/2005-08/msg01432.php
379 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01181.php
381 * MONEY dumps in a locale-specific format making it difficult to
382 restore to a system with a different locale
383 * Allow MONEY to be easily cast to/from other numeric data types
389 * Allow INET subnet tests using non-constants to be indexed
390 * Allow to_date() and to_timestamp() accept localized month names
391 * Fix to_date()-related functions to consistently issue errors
393 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00915.php
395 * Add missing parameter handling in to_char()
397 http://archives.postgresql.org/pgsql-hackers/2005-12/msg00948.php
399 * Allow substring/replace() to get/set bit values
400 * Allow to_char() on interval values to accumulate the highest unit
403 Some special format flag would be required to request such
404 accumulation. Such functionality could also be added to EXTRACT.
405 Prevent accumulation that crosses the month/day boundary because of
406 the uneven number of days in a month.
408 o to_char(INTERVAL '1 hour 5 minutes', 'MI') => 65
409 o to_char(INTERVAL '43 hours 20 minutes', 'MI' ) => 2600
410 o to_char(INTERVAL '43 hours 20 minutes', 'WK:DD:HR:MI') => 0:1:19:20
411 o to_char(INTERVAL '3 years 5 months','MM') => 41
413 * Implement inlining of set-returning functions defined in SQL
414 * Allow SQL-language functions to return results from RETURNING queries
416 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00665.php
418 * Allow SQL-language functions to reference parameters by parameter name
420 Currently SQL-language functions can only refer to dollar parameters,
423 * Add SPI_gettypmod() to return the typemod for a TupleDesc
424 * Enforce typmod for function inputs, function results and parameters for
425 spi_prepare'd statements called from PLs
427 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01403.php
429 * Allow holdable cursors in SPI
430 * Tighten function permission checks
432 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00568.php
434 * Fix IS OF so it matches the ISO specification, and add documentation
436 http://archives.postgresql.org/pgsql-patches/2003-08/msg00060.php
437 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00060.php
439 * Add missing operators for geometric data types
441 Some geometric types do not have the full suite of geometric operators,
444 * Implement Boyer-Moore searching in strpos()
446 http://archives.postgresql.org/pgsql-patches/2007-08/msg00012.php
448 * Prevent malicious functions from being executed with the permissions
449 of unsuspecting users
451 Index functions are safe, so VACUUM and ANALYZE are safe too.
452 Triggers, CHECK and DEFAULT expressions, and rules are still vulnerable.
453 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00268.php
455 * Reduce memory usage of aggregates in set returning functions
457 http://archives.postgresql.org/pgsql-performance/2008-01/msg00031.php
459 * Add temporal versions of generate_series()
461 http://archives.postgresql.org/pgsql-hackers/2007-04/msg01180.php
463 * Add array_accum() and array_to_set() functions for arrays
465 The standards specify array_agg() and UNNEST.
466 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00464.php
468 * Fix /contrib/ltree operator
470 http://archives.postgresql.org/pgsql-bugs/2007-11/msg00044.php
472 * Fix inconsistent precedence of =, >, and < compared to <>, >=, and <=
474 http://archives.postgresql.org/pgsql-bugs/2007-12/msg00145.php
476 * Fix regular expression bug when using complex back-references
478 http://archives.postgresql.org/pgsql-bugs/2007-10/msg00000.php
480 * Have /contrib/dblink reuse unnamed connections
482 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00895.php
484 * Add SQL-standard array_agg() and unnest() array functions
486 http://archives.postgresql.org/pgsql-hackers/2008-01/msg01017.php
488 * Allow calling of a procedure outside a SELECT that can control the
491 http://archives.postgresql.org/pgsql-hackers/2007-10/msg01375.php
495 Multi-Language Support
496 ======================
498 * Add NCHAR (as distinguished from ordinary varchar),
499 * Allow locale to be set at database creation
501 Currently locale can only be set during initdb. No global tables have
502 locale-aware columns. However, the database template used during
503 database creation might have locale-aware indexes. The indexes would
504 need to be reindexed to match the new locale.
506 * Allow encoding on a per-column basis optionally using the ICU library
508 Right now only one encoding is allowed per database. [locale]
510 http://archives.postgresql.org/pgsql-hackers/2005-03/msg00932.php
511 http://archives.postgresql.org/pgsql-patches/2005-08/msg00309.php
512 http://archives.postgresql.org/pgsql-patches/2006-03/msg00233.php
513 http://archives.postgresql.org/pgsql-hackers/2006-09/msg00662.php
514 http://wiki.postgresql.org/wiki/Todo:Collate
515 http://wiki.postgresql.org/wiki/Todo:ICU
517 * Add CREATE COLLATE? [locale]
518 * Support multiple simultaneous character sets, per SQL92
519 * Improve UTF8 combined character handling?
520 * Add octet_length_server() and octet_length_client()
521 * Make octet_length_client() the same as octet_length()?
522 * Fix problems with wrong runtime encoding conversion for NLS message files
523 * Add URL to more complete multi-byte regression tests
525 http://archives.postgresql.org/pgsql-hackers/2005-07/msg00272.php
527 * Fix ILIKE and regular expressions to handle case insensitivity
528 properly in multibyte encodings
530 http://archives.postgresql.org/pgsql-bugs/2005-10/msg00001.php
531 http://archives.postgresql.org/pgsql-patches/2005-11/msg00173.php
533 * Set client encoding based on the client operating system encoding
535 Currently client_encoding is set in postgresql.conf, which
536 defaults to the server encoding.
537 http://archives.postgresql.org/pgsql-hackers/2006-08/msg01696.php
539 * Change memory allocation for multi-byte functions so memory is
540 allocated inside conversion functions
542 Currently we preallocate memory based on worst-case usage.
549 * Automatically create rules on views so they are updateable, per SQL99
551 We can only auto-create rules for simple views. For more complex
552 cases users will still have to write rules manually.
554 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00586.php
555 http://archives.postgresql.org/pgsql-patches/2006-08/msg00255.php
557 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
558 * Allow VIEW/RULE recompilation when the underlying tables change
560 Another issue is whether underlying table changes should be reflected
561 in the view, e.g. should SELECT * show additional columns if they
562 are added after the view is created.
564 * Make it possible to use RETURNING together with conditional DO INSTEAD
565 rules, such as for partitioning setups
567 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00577.php
569 * Add the ability to automatically create materialized views
571 Right now materialized views require the user to create triggers on the
572 main table to keep the summary table current. SQL syntax should be able
573 to manager the triggers and summary table automatically. A more
574 sophisticated implementation would automatically retrieve from the
575 summary table when the main table is referenced, if possible.
582 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
583 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
584 * %Allow SET CONSTRAINTS to be qualified by schema/table name
585 * %Add a separate TRUNCATE permission
587 Currently only the owner can TRUNCATE a table because triggers are not
588 called, and the table is locked in exclusive mode.
590 * Allow PREPARE of cursors
591 * Allow finer control over the caching of prepared query plans
593 Currently queries prepared via the libpq API are planned on first
594 execute using the supplied parameters --- allow SQL PREPARE to do the
595 same. Also, allow control over replanning prepared queries either
596 manually or automatically when statistics for execute parameters
597 differ dramatically from those used during planning.
599 * Improve logging of prepared transactions recovered during startup
601 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00092.php
603 * Add a GUC variable to warn about non-standard SQL usage in queries
604 * Add SQL-standard MERGE command, typically used to merge two tables
607 This is similar to UPDATE, then for unmatched rows, INSERT.
608 Whether concurrent access allows modifications which could cause
609 row loss is implementation independent.
611 * Add REPLACE or UPSERT command that does UPDATE, or on failure, INSERT
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
616 do it without a unique index if we require the user to LOCK the table
619 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00501.php
620 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00536.php
622 * Add NOVICE output level for helpful messages like automatic sequence/index
624 * Add GUC to issue notice about statements that use unjoined tables
625 * Allow EXPLAIN to identify tables that were skipped because of
627 * Allow EXPLAIN output to be more easily processed by scripts, perhaps XML
628 * Enable standard_conforming_strings
629 * Make standard_conforming_strings the default in 8.5?
631 When this is done, backslash-quote should be prohibited in non-E''
632 strings because of possible confusion over how such strings treat
633 backslashes. Basically, '' is always safe for a literal single
634 quote, while \' might or might not be based on the backslash
637 * Simplify dropping roles that have objects in several databases
638 * Allow COMMENT ON to accept an expression rather than just a string
639 * Allow the count returned by SELECT, etc to be represented as an int64
640 to allow a higher range of values
641 * Add SQL99 WITH clause to SELECT
642 * Add SQL:2003 WITH RECURSIVE (hierarchical) queries to SELECT
644 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01375.php
645 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00642.php
646 http://archives.postgresql.org/pgsql-patches/2007-03/msg00139.php
647 http://archives.postgresql.org/pgsql-hackers/2007-11/msg01334.php
648 http://archives.postgresql.org/pgsql-patches/2008-01/msg00105.php
649 http://archives.postgresql.org/pgsql-patches/2008-03/msg00327.php
651 * Add DEFAULT .. AS OWNER so permission checks are done as the table
654 This would be useful for SERIAL nextval() calls and CHECK constraints.
656 * Allow DISTINCT to work in multiple-argument aggregate calls
657 * Add column to pg_stat_activity that shows the progress of long-running
658 commands like CREATE INDEX and VACUUM
659 * Implement SQL:2003 window functions
660 * Improve failure message when DROP DATABASE is used on a database that
661 has prepared transactions
662 * Allow INSERT/UPDATE ... RETURNING inside a SELECT 'FROM' clause
664 http://archives.postgresql.org/pgsql-general/2006-09/msg00803.php
665 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00693.php
667 * Increase locking when DROPing objects so dependent objects cannot
668 get dropped while the DROP operation is happening
670 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00937.php
672 * -Allow AS in "SELECT col AS label" to be optional in certain cases
674 * Allow INSERT ... DELETE ... RETURNING, namely allow the DELETE ...
675 RETURNING to supply values to the INSERT
676 http://archives.postgresql.org/pgsql-hackers/2008-02/thrd2.php#00979
678 * Add comments on system tables/columns using the information in
681 Ideally the information would be pulled from the SGML file
684 * Improve reporting of UNION type mismatches
686 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00944.php
687 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00597.php
692 o Allow CREATE TABLE AS to determine column lengths for complex
693 expressions like SELECT col1 || col2
695 o Have WITH CONSTRAINTS also create constraint indexes
697 http://archives.postgresql.org/pgsql-patches/2007-04/msg00149.php
699 o Have CONSTRAINT cname NOT NULL record the contraint name
701 Right now pg_attribute.attnotnull records the NOT NULL status
702 of the column, but does not record the contraint name
704 o Prevent concurrent CREATE TABLE table1 from sometimes returning
705 a cryptic error message
707 http://archives.postgresql.org/pgsql-bugs/2007-10/msg00169.php
711 o Allow UPDATE tab SET ROW (col, ...) = (SELECT...)
713 http://archives.postgresql.org/pgsql-hackers/2006-07/msg01306.php
714 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00865.php
715 http://archives.postgresql.org/pgsql-patches/2007-04/msg00315.php
716 http://archives.postgresql.org/pgsql-patches/2008-03/msg00237.php
718 o Research self-referential UPDATEs that see inconsistent row versions
719 in read-committed mode
721 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00507.php
722 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00016.php
724 o Allow GLOBAL temporary tables to exist as empty by default in
727 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00006.php
732 o Have ALTER TABLE RENAME rename SERIAL sequence names
734 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
736 o Have ALTER SEQUENCE RENAME rename the sequence name stored
737 in the sequence table
739 http://archives.postgresql.org/pgsql-bugs/2007-09/msg00092.php
740 http://archives.postgresql.org/pgsql-bugs/2007-10/msg00007.php
741 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
743 o Add ALTER DOMAIN to modify the underlying data type
744 o %Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
746 http://archives.postgresql.org/pgsql-patches/2006-02/msg00168.php
748 o %Allow ALTER TABLE to change constraint deferrability and actions
749 o Add missing object types for ALTER ... SET SCHEMA
750 o Allow ALTER TABLESPACE to move to different directories
751 o Allow databases to be moved to different tablespaces
752 o Allow moving system tables to other tablespaces, where possible
754 Currently non-global system tables must be in the default database
755 tablespace. Global system tables can never be moved.
757 o Prevent parent tables from altering or dropping constraints
758 like CHECK that are inherited by child tables unless CASCADE
760 o %Prevent child tables from altering or dropping constraints
761 like CHECK that were inherited from the parent table
762 o Have ALTER INDEX update the name of a constraint using that index
763 o Add ALTER TABLE RENAME CONSTRAINT, update index name also
764 o Allow column display reordering by recording a display,
765 storage, and permanent id for every column?
767 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00782.php
773 o Automatically maintain clustering on a table
775 This might require some background daemon to maintain clustering
776 during periods of low usage. It might also require tables to be only
777 partially filled for easier reorganization. Another idea would
778 be to create a merged heap/index data file so an index lookup would
779 automatically access the heap data too. A third idea would be to
780 store heap rows in hashed groups, perhaps using a user-supplied
782 http://archives.postgresql.org/pgsql-performance/2004-08/msg00349.php
784 o %Add default clustering to system tables
786 To do this, determine the ideal cluster index for each system
787 table and set the cluster setting during initdb.
789 o %Add VERBOSE option to report tables as they are processed,
795 o Allow COPY to report error lines and continue
797 This requires the use of a savepoint before each COPY line is
798 processed, with ROLLBACK on COPY failure.
799 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00572.php
801 o Allow COPY on a newly-created table to skip WAL logging
803 On crash recovery, the table involved in the COPY would
804 be removed or have its heap and index files truncated. One
805 issue is that no other backend should be able to add to
806 the table at the same time, which is something that is
807 currently allowed. This currently is done if the table is
808 created inside the same transaction block as the COPY because
809 no other backends can see the table.
811 o Consider using a ring buffer for COPY FROM
813 http://archives.postgresql.org/pgsql-patches/2008-02/msg00140.php
814 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01080.php
816 o Allow COPY FROM to create index entries in bulk
818 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00811.php
820 o Allow COPY in CSV mode to control whether a quoted zero-length
821 string is treated as NULL
823 Currently this is always treated as a zero-length string,
824 which generates an error when loading into an integer column
825 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00905.php
831 o Allow column-level privileges
832 o %Allow GRANT/REVOKE permissions to be applied to all schema objects
835 The proposed syntax is:
836 GRANT SELECT ON ALL TABLES IN public TO phpuser;
837 GRANT SELECT ON NEW TABLES IN public TO phpuser;
839 o Allow GRANT/REVOKE permissions to be inherited by objects based on
842 o Allow SERIAL sequences to inherit permissions from the base table?
847 o Prevent DROP TABLE from dropping a row referenced by its own open
853 o Allow INSERT/UPDATE of the system-generated oid value for a row
854 o In rules, allow VALUES() to contain a mixture of 'old' and 'new'
860 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
866 o Allow LISTEN/NOTIFY to store info in memory rather than tables?
868 Currently LISTEN/NOTIFY information is stored in pg_listener.
869 Storing such information in memory would improve performance.
871 o Add optional textual message to NOTIFY
873 This would allow an informational message to be added to the notify
874 message, perhaps indicating the row modified or other custom
877 o Allow multiple identical NOTIFY events to always be communicated
878 to the client, rather than sent as a single notification to the
881 http://archives.postgresql.org/pgsql-general/2008-01/msg00057.php
883 o Allow NOTIFY in rules involving conditionals
884 o Improve LISTEN concurrency
886 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01106.php
890 Referential Integrity
891 =====================
893 * Add MATCH PARTIAL referential integrity
894 * Change foreign key constraint for array -> element to mean element
896 * Fix problem when cascading referential triggers make changes on
897 cascaded tables, seeing the tables in an intermediate state
899 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00174.php
900 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00174.php
902 * Allow DEFERRABLE and end-of-statement UNIQUE constraints?
904 This would allow UPDATE tab SET col = col + 1 to work if col has
905 a unique index. Currently, uniqueness checks are done while the
906 command is being executed, rather than at the end of the statement
908 http://people.planetpostgresql.org/greg/index.php?/archives/2006/06/10.html
909 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01458.php
911 * Optimize referential integrity checks
913 http://archives.postgresql.org/pgsql-performance/2005-10/msg00458.php
914 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00744.php
917 Server-Side Languages
918 =====================
921 o Fix RENAME to work on variables other than OLD/NEW
923 http://archives.postgresql.org/pgsql-hackers/2002-03/msg00591.php
924 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01615.php
925 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01587.php
927 o Allow function parameters to be passed by name,
928 get_employee_salary(12345 AS emp_id, 2001 AS tax_year)
929 o Allow handling of %TYPE arrays, e.g. tab.col%TYPE[]
930 o Allow listing of record column names, and access to
931 record columns via variables, e.g. columns := r.(*),
934 http://archives.postgresql.org/pgsql-patches/2005-07/msg00458.php
935 http://archives.postgresql.org/pgsql-patches/2006-05/msg00302.php
936 http://archives.postgresql.org/pgsql-patches/2006-06/msg00031.php
938 o Add support for SCROLL cursors
939 o Add support for WITH HOLD cursors
940 o Allow row and record variables to be set to NULL constants,
941 and allow NULL tests on such variables
943 Because a row is not scalar, do not allow assignment
944 from NULL-valued scalars.
946 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00070.php
948 o Review handling of MOVE and FETCH
950 http://archives.postgresql.org/pgsql-patches/2007-04/msg00527.php
952 o Improve logic of determining if an identifier is a a
953 variable or column name
955 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00436.php
957 o Consider keeping seperate cached copies when search_path changes
959 http://archives.postgresql.org/pgsql-hackers/2008-01/msg01009.php
961 o Add CASE capability to language (already in SQL)
963 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00696.php
969 o Add table function support to pltcl, plpythonu
970 o Add support for polymorphic arguments and return types to
971 languages other than PL/PgSQL
972 o Add capability to create and call PROCEDURES
973 o Add support for OUT and INOUT parameters to languages other
975 o Add PL/PythonU tracebacks
977 http://archives.postgresql.org/pgsql-patches/2006-02/msg00288.php
979 o Allow data to be passed in native language formats, rather
982 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00289.php
984 o Add ability to obfuscate function bodies
986 http://archives.postgresql.org/pgsql-patches/2008-01/msg00125.php
993 * Have pg_ctl look at PGHOST in case it is a socket directory?
994 * Allow pg_ctl to work properly with configuration files located outside
997 pg_ctl can not read the pid file because it isn't located in the
998 config directory but in the PGDATA directory. The solution is to
999 allow pg_ctl to read and understand postgresql.conf to find the
1000 data_directory value.
1002 * Add a function like pg_get_indexdef() that report more detailed index
1005 http://archives.postgresql.org/pgsql-bugs/2007-12/msg00166.php
1008 o Have psql show current values for a sequence
1009 o Move psql backslash database information into the backend, use
1010 mnemonic commands? [psql]
1012 This would allow non-psql clients to pull the same information out
1013 of the database as psql.
1015 o Make psql's \d commands more consistent
1017 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
1018 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
1020 o Consistently display privilege information for all objects in psql
1021 o Add auto-expanded mode so expanded output is used if the row
1022 length is wider than the screen width.
1024 Consider using auto-expanded mode for backslash commands like \df+.
1026 o Prevent tab completion of SET TRANSACTION from querying the
1027 database and therefore preventing the transaction isolation
1028 level from being set.
1030 Currently SET <tab> causes a database lookup to check all
1031 supported session variables. This query causes problems
1032 because setting the transaction isolation level must be the
1033 first statement of a transaction.
1035 o Add a \set variable to control whether \s displays line numbers
1037 Another option is to add \# which lists line numbers, and
1038 allows command execution.
1040 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00255.php
1042 o Prevent escape string warnings when object names have
1045 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00227.php
1047 o Have \d show foreign keys that reference a table's primary key
1049 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00424.php
1051 o Have \d show child tables that inherit from the specified parent
1052 o Have \l+ show database size, if permissions allow
1054 Ideally it will not generate an error for invalid permissions
1056 o Include the symbolic SQLSTATE name in verbose error reports
1058 http://archives.postgresql.org/pgsql-general/2007-09/msg00438.php
1060 o Improve display if enums
1062 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00826.php
1067 * pg_dump / pg_restore
1068 o %Add dumping of comments on index columns and composite type columns
1069 o %Add full object name to the tag field. eg. for operators we need
1070 '=(integer, integer)', instead of just '='.
1071 o Add pg_dumpall custom format dumps?
1072 o Allow selection of individual object(s) of all types, not just
1074 o In a selective dump, allow dumping of an object and all its
1076 o Add options like pg_restore -l and -L to pg_dump
1077 o Stop dumping CASCADE on DROP TYPE commands in clean mode
1078 o Allow pg_dump --clean to drop roles that own objects or have
1080 o Change pg_dump so that a comment on the dumped database is
1081 applied to the loaded database, even if the database has a
1082 different name. This will require new backend syntax, perhaps
1083 COMMENT ON CURRENT DATABASE.
1084 o Remove unnecessary function pointer abstractions in pg_dump source
1086 o Allow pg_dump to utilize multiple CPUs and I/O channels by dumping
1087 multiple objects simultaneously
1089 The difficulty with this is getting multiple dump processes to
1090 produce a single dump output file. It also would require
1091 several sessions to share the same snapshot.
1092 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00205.php
1094 o Allow pg_restore to utilize multiple CPUs and I/O channels by
1095 restoring multiple objects simultaneously
1097 This might require a pg_restore flag to indicate how many
1098 simultaneous operations should be performed. Only pg_dump's
1099 -Fc format has the necessary dependency information.
1100 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00963.php
1102 o To better utilize resources, allow pg_restore to check foreign
1103 keys simultaneously, where possible
1104 o Allow pg_restore to create all indexes of a table
1105 concurrently, via a single heap scan
1107 This requires a pg_dump -Fc file because that format contains
1108 the required dependency information.
1109 http://archives.postgresql.org/pgsql-general/2007-05/msg01274.php
1111 o Allow pg_restore to load different parts of the COPY data
1113 o Prevent pg_dump/pg_restore from being affected by
1116 Using psql to restore a pg_dump dump is also affected.
1118 o Remove pre-7.3 pg_dump code that assumes pg_depend does not exit
1119 o Allow pre/data/post files when schema and data are dumped
1120 separately, for performance reasons
1122 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00205.php
1129 Document differences between ecpg and the SQL standard and
1130 information about the Informix-compatibility module.
1132 o Solve cardinality > 1 for input descriptors / variables?
1133 o Add a semantic check level, e.g. check if a table really exists
1134 o fix handling of DB attributes that are arrays
1135 o Use backend PREPARE/EXECUTE facility for ecpg where possible
1137 o Fix nested C comments
1138 o %sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
1139 o Make SET CONNECTION thread-aware, non-standard?
1140 o Allow multidimensional arrays
1141 o Add internationalized message strings
1142 o Implement COPY FROM STDIN
1146 o Add PQescapeIdentifierConn()
1147 o Prevent PQfnumber() from lowercasing unquoted the column name
1149 PQfnumber() should never have been doing lowercasing, but
1150 historically it has so we need a way to prevent it
1152 o Allow statement results to be automatically batched to the client
1154 Currently all statement results are transferred to the libpq
1155 client before libpq makes the results available to the
1156 application. This feature would allow the application to make
1157 use of the first result rows while the rest are transferred, or
1158 held on the server waiting for them to be requested by libpq.
1159 One complexity is that a statement like SELECT 1/col could error
1160 out mid-way through the result set.
1162 o Consider disallowing multiple queries in PQexec() as an
1163 additional barrier to SQL injection attacks
1165 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00184.php
1167 o Add PQexecf() that allows complex parameter substitution
1169 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01803.php
1171 o Add SQLSTATE severity to PGconn return status
1173 http://archives.postgresql.org/pgsql-interfaces/2007-11/msg00015.php
1179 * Add deferred trigger queue file
1181 Right now all deferred trigger information is stored in backend
1182 memory. This could exhaust memory for very large trigger queues.
1183 This item involves dumping large queues into files.
1185 * Allow triggers to be disabled in only the current session.
1187 This is currently possible by starting a multi-statement transaction,
1188 modifying the system tables, performing the desired SQL, restoring the
1189 system tables, and committing the transaction. ALTER TABLE ...
1190 TRIGGER requires a table lock so it is not ideal for this usage.
1192 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
1194 If the dump is known to be valid, allow foreign keys to be added
1195 without revalidating the data.
1197 * Allow statement-level triggers to access modified rows
1198 * Support triggers on columns
1200 http://archives.postgresql.org/pgsql-patches/2005-07/msg00107.php
1202 * Allow AFTER triggers on system tables
1204 System tables are modified in many places in the backend without going
1205 through the executor and therefore not causing triggers to fire. To
1206 complete this item, the functions that modify system tables will have
1209 * Tighten trigger permission checks
1211 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00564.php
1213 * Allow BEFORE INSERT triggers on views
1215 http://archives.postgresql.org/pgsql-general/2007-02/msg01466.php
1217 * -Add ability to trigger on TRUNCATE
1224 * Add UNIQUE capability to non-btree indexes
1225 * Prevent index uniqueness checks when UPDATE does not modify the column
1227 Uniqueness (index) checks are done when updating a column even if the
1228 column is not modified by the UPDATE.
1230 * Allow the creation of on-disk bitmap indexes which can be quickly
1231 combined with other bitmap indexes
1233 Such indexes could be more compact if there are only a few distinct values.
1234 Such indexes can also be compressed. Keeping such indexes updated can be
1237 http://archives.postgresql.org/pgsql-patches/2005-07/msg00512.php
1238 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01107.php
1239 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00265.php
1240 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01214.php
1241 http://archives.postgresql.org/pgsql-patches/2007-05/msg00013.php
1242 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00741.php
1244 * Allow accurate statistics to be collected on indexes with more than
1245 one column or expression indexes, perhaps using per-index statistics
1247 http://archives.postgresql.org/pgsql-performance/2006-10/msg00222.php
1248 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01131.php
1250 * Consider increasing the number of default statistics target, and
1251 reduce statistics target overhead
1253 Also consider having a larger statistics target for indexed columns
1254 and expression indexes
1255 http://archives.postgresql.org/pgsql-general/2007-05/msg01228.php
1256 http://archives.postgresql.org/pgsql-general/2007-06/msg00542.php
1257 http://archives.postgresql.org/pgsql-hackers/2008-01/msg01066.php
1259 * Consider compressing indexes by storing key values duplicated in
1260 several rows as a single index entry
1262 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00341.php
1263 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01264.php
1264 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00465.php
1265 http://archives.postgresql.org/pgsql-patches/2007-03/msg00163.php
1266 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00014.php
1267 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00487.php
1269 * Add REINDEX CONCURRENTLY, like CREATE INDEX CONCURRENTLY
1271 This is difficult because you must upgrade to an exclusive table lock
1272 to replace the existing index file. CREATE INDEX CONCURRENTLY does not
1273 have this complication. This would allow index compaction without
1275 http://archives.postgresql.org/pgsql-performance/2007-08/msg00289.php
1277 * Allow multiple indexes to be created concurrently, ideally via a
1278 single heap scan, and have pg_restore use it
1280 * Consider sorting entries before inserting into btree index
1282 http://archives.postgresql.org/pgsql-general/2008-01/msg01010.php
1284 * Allow index scans to return matching index keys
1286 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01079.php
1291 o Allow inherited tables to inherit indexes, UNIQUE constraints,
1292 and primary/foreign keys
1293 o Honor UNIQUE INDEX on base column in INSERTs/UPDATEs
1294 on inherited table, e.g. INSERT INTO inherit_table
1295 (unique_index_col) VALUES (dup) should fail
1297 The main difficulty with this item is the problem of
1298 creating an index that can span multiple tables.
1300 o Allow SELECT ... FOR UPDATE on inherited tables
1301 o Require all CHECK constraints to be inherited
1303 http://archives.postgresql.org/pgsql-bugs/2007-04/msg00026.php
1305 o Add checks to prevent a CREATE RULE views on inherited tables
1307 http://archives.postgresql.org/pgsql-general/2008-02/msg01420.php
1308 http://archives.postgresql.org/pgsql-general/2008-03/msg00077.php
1313 o Add more GIST index support for geometric data types
1314 o Allow GIST indexes to create certain complex index types, like
1315 digital trees (see Aoki)
1320 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00051.php
1322 o Pack hash index buckets onto disk pages more efficiently
1324 Currently only one hash bucket can be stored on a page. Ideally
1325 several hash buckets could be stored on a single page and greater
1326 granularity used for the hash algorithm.
1328 http://archives.postgresql.org/pgsql-hackers/2004-06/msg00168.php
1330 o Consider sorting hash buckets so entries can be found using a
1331 binary search, rather than a linear scan
1332 o In hash indexes, consider storing the hash value with or instead
1334 o Add WAL logging for crash recovery
1335 o Allow multi-column hash indexes
1336 o -During index creation, pre-sort the tuples to improve build speed
1343 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
1345 Ideally this requires a separate test program that can be run
1346 at initdb time or optionally later. Consider O_SYNC when
1349 * Add program to test if fsync has a delay compared to non-fsync
1350 * Consider sorting writes during checkpoint
1352 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00541.php
1361 We could use a fixed row count and a +/- count to follow MVCC
1362 visibility rules, or a single cached value could be used and
1363 invalidated if anyone modifies the table. Another idea is to
1364 get a count directly from a unique index, but for this to be
1365 faster than a sequential scan it must avoid access to the heap
1366 to obtain tuple visibility information.
1368 * Provide a way to calculate an "estimated COUNT(*)"
1370 Perhaps by using the optimizer's cardinality estimates or random
1373 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00943.php
1375 * Allow data to be pulled directly from indexes
1377 Currently indexes do not have enough tuple visibility information
1378 to allow data to be pulled from the index without also accessing
1379 the heap. One way to allow this is to set a bit on index tuples
1380 to indicate if a tuple is currently visible to all transactions
1381 when the first valid heap lookup happens. This bit would have to
1382 be cleared when a heap tuple is expired.
1384 Another idea is to maintain a bitmap of heap pages where all rows
1385 are visible to all backends, and allow index lookups to reference
1386 that bitmap to avoid heap lookups, perhaps the same bitmap we might
1387 add someday to determine which heap pages need vacuuming. Frequently
1388 accessed bitmaps would have to be stored in shared memory. One 8k
1389 page of bitmaps could track 512MB of heap pages.
1391 A third idea would be for a heap scan to check if all rows are visible
1392 and if so set a per-table flag which can be checked by index scans.
1393 Any change to the table would have to clear the flag. To detect
1394 changes during the heap scan a counter could be set at the start and
1395 checked at the end --- if it is the same, the table has not been
1396 modified --- any table change would increment the counter.
1398 http://archives.postgresql.org/pgsql-patches/2007-10/msg00166.php
1399 http://archives.postgresql.org/pgsql-patches/2008-01/msg00049.php
1401 * Consider automatic caching of statements at various levels:
1404 o Query execute plan
1406 http://archives.postgresql.org/pgsql-hackers/2008-04/msg00823.php
1408 * Consider increasing internal areas when shared buffers is increased
1410 http://archives.postgresql.org/pgsql-hackers/2005-10/msg01419.php
1412 * Consider decreasing the amount of memory used by PrivateRefCount
1414 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00797.php
1415 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00752.php
1417 * Consider allowing higher priority queries to have referenced buffer
1418 cache pages stay in memory longer
1420 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00562.php
1427 * Improve speed with indexes
1429 For large table adjustments during VACUUM FULL, it is faster to cluster
1430 or reindex rather than update the index. Also, index updates can bloat
1433 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00024.php
1434 http://archives.postgresql.org/pgsql-performance/2007-05/msg00296.php
1435 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00307.php
1437 * Auto-fill the free space map by scanning the buffer cache or by
1438 checking pages written by the background writer
1440 http://archives.postgresql.org/pgsql-hackers/2006-02/msg01125.php
1441 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00011.php
1443 * Create a bitmap of pages that need vacuuming
1445 Instead of sequentially scanning the entire table, have the background
1446 writer or some other process record pages that have expired rows, then
1447 VACUUM can look at just those pages rather than the entire table. In
1448 the event of a system crash, the bitmap would probably be invalidated.
1449 One complexity is that index entries still have to be vacuumed, and
1450 doing this without an index scan (by using the heap values to find the
1451 index entry) might be slow and unreliable, especially for user-defined
1454 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01188.php
1455 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00121.php
1456 http://archives.postgresql.org/pgsql-patches/2007-03/msg00508.php
1457 http://archives.postgresql.org/pgsql-patches/2007-04/msg00347.php
1458 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00156.php
1459 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00546.php
1460 http://archives.postgresql.org/pgsql-hackers/2008-04/msg00416.php
1462 * Allow FSM to return free space toward the beginning of the heap file,
1463 in hopes that empty pages at the end can be truncated by VACUUM
1464 * Allow FSM page return free space based on table clustering, to assist
1465 in maintaining clustering?
1466 * Improve dead row detection during multi-statement transactions usage
1468 http://archives.postgresql.org/pgsql-patches/2007-03/msg00358.php
1470 * Consider a more compact data representation for dead tuples
1472 http://archives.postgresql.org/pgsql-patches/2007-05/msg00143.php
1477 o %Issue log message to suggest VACUUM FULL if a table is nearly
1479 o Improve control of auto-vacuum
1481 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00876.php
1483 o Prevent long-lived temporary tables from causing frozen-xid
1484 advancement starvation
1486 The problem is that autovacuum cannot vacuum them to set frozen xids;
1487 only the session that created them can do that.
1488 http://archives.postgresql.org/pgsql-general/2007-06/msg01645.php
1490 o Store per-table autovacuum settings in pg_class.reloptions.
1492 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01440.php
1493 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00724.php
1495 o Prevent autovacuum from running if an old transaction is still
1496 running from the last vacuum
1498 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00899.php
1505 * Fix priority ordering of read and write light-weight locks (Neil)
1507 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00893.php
1508 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00905.php
1510 * Fix problem when multiple subtransactions of the same outer transaction
1511 hold different types of locks, and one subtransaction aborts
1513 http://archives.postgresql.org/pgsql-hackers/2006-11/msg01011.php
1514 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00001.php
1515 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00435.php
1516 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00773.php
1518 * Allow UPDATEs on only non-referential integrity columns not to conflict
1519 with referential integrity locks
1521 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00073.php
1523 * Add idle_in_transaction_timeout GUC so locks are not held for long
1526 * Improve deadlock detection when a page cleaning lock conflicts
1527 with a shared buffer that is pinned
1529 http://archives.postgresql.org/pgsql-bugs/2008-01/msg00138.php
1530 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00873.php
1531 http://archives.postgresql.org/pgsql-committers/2008-01/msg00365.php
1533 * Detect deadlocks involving LockBufferForCleanup()
1535 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00873.php
1539 Startup Time Improvements
1540 =========================
1542 * Experiment with multi-threaded backend for backend creation [thread]
1544 This would prevent the overhead associated with process creation. Most
1545 operating systems have trivial process creation time compared to
1546 database startup overhead, but a few operating systems (Win32,
1547 Solaris) might benefit from threading. Also explore the idea of
1548 a single session using multiple threads to execute a statement faster.
1555 * Eliminate need to write full pages to WAL before page modification [wal]
1557 Currently, to protect against partial disk page writes, we write
1558 full page images to WAL before they are modified so we can correct any
1559 partial page writes during recovery. These pages can also be
1560 eliminated from point-in-time archive files.
1562 o When off, write CRC to WAL and check file system blocks
1565 If CRC check fails during recovery, remember the page in case
1566 a later CRC for that page properly matches.
1568 o Write full pages during file system write and not when
1569 the page is modified in the buffer cache
1571 This allows most full page writes to happen in the background
1572 writer. It might cause problems for applying WAL on recovery
1573 into a partially-written page, but later the full page will be
1576 * Allow WAL traffic to be streamed to another server for stand-by
1578 * Reduce WAL traffic so only modified values are written rather than
1581 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01589.php
1583 * Allow WAL information to recover corrupted pg_controldata
1585 http://archives.postgresql.org/pgsql-patches/2006-06/msg00025.php
1587 * Find a way to reduce rotational delay when repeatedly writing
1590 Currently fsync of WAL requires the disk platter to perform a full
1591 rotation to fsync again. One idea is to write the WAL to different
1592 offsets that might reduce the rotational delay.
1594 * Allow WAL logging to be turned off for a table, but the table
1595 might be dropped or truncated during crash recovery [walcontrol]
1597 Allow tables to bypass WAL writes and just fsync() dirty pages on
1598 commit. This should be implemented using ALTER TABLE, e.g. ALTER
1599 TABLE PERSISTENCE [ DROP | TRUNCATE | DEFAULT ]. Tables using
1600 non-default logging should not use referential integrity with
1601 default-logging tables. A table without dirty buffers during a
1602 crash could perhaps avoid the drop/truncate.
1604 * Allow WAL logging to be turned off for a table, but the table would
1605 avoid being truncated/dropped [walcontrol]
1607 To do this, only a single writer can modify the table, and writes
1608 must happen only on new pages so the new pages can be removed during
1609 crash recovery. Readers can continue accessing the table. Such
1610 tables probably cannot have indexes. One complexity is the handling
1611 of indexes on TOAST tables.
1613 * Speed WAL recovery by allowing more than one page to be prefetched
1615 This should be done utilizing the same infrastructure used for
1616 prefetching in general to avoid introducing complex error-prone code
1618 http://archives.postgresql.org/pgsql-general/2007-12/msg00683.php
1619 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00497.php
1620 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01279.php
1622 * Improve WAL concurrency by increasing lock granularity
1624 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00556.php
1626 * Be more aggressive about creating WAL files
1628 http://archives.postgresql.org/pgsql-hackers/2007-10/msg01325.php
1629 http://archives.postgresql.org/pgsql-hackers/2004-07/msg01075.php
1630 http://archives.postgresql.org/pgsql-hackers/2005-04/msg00556.php
1632 * Have resource managers report the duration of their status changes
1634 http://archives.postgresql.org/pgsql-hackers/2007-10/msg01468.php
1636 * Move pgfoundry's xlogdump to /contrib and have it rely more closely
1637 on the WAL backend code
1639 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00035.php
1644 Optimizer / Executor
1645 ====================
1647 * Improve selectivity functions for geometric operators
1648 * Precompile SQL functions to avoid overhead
1649 * Create utility to compute accurate random_page_cost value
1650 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
1651 * Have EXPLAIN ANALYZE issue NOTICE messages when the estimated and
1652 actual row counts differ by a specified percentage
1653 * Improve how ANALYZE computes in-doubt tuples
1655 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00771.php
1657 * Consider using hash buckets to do DISTINCT, rather than sorting
1659 This would be beneficial when there are few distinct values. This is
1660 already used by GROUP BY.
1662 * Log statements where the optimizer row estimates were dramatically
1663 different from the number of rows actually found?
1664 * Consider compressed annealing to search for query plans
1666 This might replace GEQO, http://sixdemonbag.org/Djinni.
1668 * Improve merge join performance by allowing mark/restore of
1671 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00096.php
1673 * Consider using a hash for joining to a large IN (VALUES ...) list
1675 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00450.php
1682 * Consider having the background writer update the transaction status
1683 hint bits before writing out the page
1685 Implementing this requires the background writer to have access to system
1686 catalogs and the transaction status log.
1688 * Consider adding buffers the background writer finds reusable to the
1691 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00781.php
1693 * Automatically tune bgwriter_delay based on activity rather then using a
1696 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00781.php
1698 * Consider wither increasing BM_MAX_USAGE_COUNT improves performance
1700 http://archives.postgresql.org/pgsql-hackers/2007-06/msg01007.php
1702 * Test to see if calling PreallocXlogFiles() from the background writer
1703 will help with WAL segment creation latency
1705 http://archives.postgresql.org/pgsql-patches/2007-06/msg00340.php
1709 Miscellaneous Performance
1710 =========================
1712 * Do async I/O for faster random read-ahead of data
1714 Async I/O allows multiple I/O requests to be sent to the disk with
1715 results coming back asynchronously.
1717 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00820.php
1718 http://archives.postgresql.org/pgsql-performance/2007-09/msg00255.php
1719 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00027.php
1720 http://archives.postgresql.org/pgsql-patches/2008-01/msg00170.php
1722 * Use mmap() rather than SYSV shared memory or to write WAL files?
1724 This would remove the requirement for SYSV SHM but would introduce
1725 portability issues. Anonymous mmap (or mmap to /dev/zero) is required
1726 to prevent I/O overhead.
1728 * Consider mmap()'ing files into a backend?
1730 Doing I/O to large tables would consume a lot of address space or
1731 require frequent mapping/unmapping. Extending the file also causes
1732 mapping problems that might require mapping only individual pages,
1733 leading to thousands of mappings. Another problem is that there is no
1734 way to _prevent_ I/O to disk from the dirty shared buffers so changes
1735 could hit disk before WAL is written.
1737 * Add a script to ask system configuration questions and tune postgresql.conf
1738 * Consider ways of storing rows more compactly on disk
1740 o Reduce the row header size?
1741 o Consider reducing on-disk varlena length from four bytes to
1742 two because a heap row cannot be more than 64k in length
1744 * Consider increasing NUM_CLOG_BUFFERS
1746 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00030.php
1747 http://archives.postgresql.org/pgsql-performance/2007-08/msg00024.php
1749 * Consider transaction start/end performance improvements
1751 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00948.php
1752 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00361.php
1754 * Allow user configuration of TOAST thresholds
1756 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00213.php
1757 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00082.php
1759 * Allow configuration of backend priorities via the operating system
1761 Though backend priorities make priority inversion during lock
1762 waits possible, research shows that this is not a huge problem.
1764 http://archives.postgresql.org/pgsql-general/2007-02/msg00493.php
1766 * Experiment with multi-threaded backend better I/O utilization
1768 This would allow a single query to make use of multiple I/O channels
1769 simultaneously. One idea is to create a background reader that can
1770 pre-fetch sequential and index scan pages needed by other backends.
1771 This could be expanded to allow concurrent reads from multiple devices
1772 in a partitioned table.
1774 * Experiment with multi-threaded backend better CPU utilization
1776 This would allow several CPUs to be used for a single query, such as
1777 for sorting or query execution.
1779 * Consider increasing the minimum allowed number of shared buffers
1781 http://archives.postgresql.org/pgsql-bugs/2008-02/msg00157.php
1783 * Expire published xmin for read-only and idle transactions
1785 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00343.php
1787 * Consider if CommandCounterIncrement() can avoid its
1788 AcceptInvalidationMessages() call
1790 http://archives.postgresql.org/pgsql-committers/2007-11/msg00585.php
1792 * Improve performance of shared invalidation queue for multiple CPUs
1794 http://archives.postgresql.org/pgsql-performance/2008-01/msg00023.php
1796 * Consider Cartesian joins when both relations are needed to form an
1797 indexscan qualification for a third relation
1799 http://archives.postgresql.org/pgsql-performance/2007-12/msg00090.php
1801 * Consider not storing a NULL bitmap on disk if all the NULLs are
1804 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00624.php
1805 http://archives.postgresql.org/pgsql-patches/2007-12/msg00109.php
1807 * Sort large UPDATE/DELETEs so it is done in heap order
1809 http://archives.postgresql.org/pgsql-hackers/2008-01/msg01119.php
1811 * -Avoid tuple some tuple copying in sort routines
1812 * SMP scalability improvements
1814 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00439.php
1815 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00206.php
1816 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00361.php
1818 * Research reducing deTOASTing in more places
1820 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00895.php
1822 * Consider being smarter about memory and external files used during
1825 http://archives.postgresql.org/pgsql-hackers/2007-11/msg01101.php
1826 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00045.php
1828 * Allow one transaction to see tuples using the snapshot of another
1831 This would assist multiple backends in working together.
1832 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00400.php
1838 * Add use of 'const' for variables in source tree
1839 * Move some things from contrib into main tree
1840 * %Remove warnings created by -Wcast-align
1841 * Move platform-specific ps status display info from ps_status.c to ports
1842 * Add optional CRC checksum to heap and index pages
1843 * Improve documentation to build only interfaces (Marc)
1844 * Remove or relicense modules that are not under the BSD license, if possible
1845 * Acquire lock on a relation before building a relcache entry for it
1846 * Allow cross-compiling by generating the zic database on the target system
1847 * Improve NLS maintenance of libpgport messages linked onto applications
1848 * Clean up casting in contrib/isn
1850 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00245.php
1852 * Use UTF8 encoding for NLS messages so all server encodings can
1854 * Update Bonjour to work with newer cross-platform SDK
1856 http://archives.postgresql.org/pgsql-hackers/2006-09/msg02238.php
1857 http://archives.postgresql.org/pgsql-patches/2006-10/msg00048.php
1859 * Consider detoasting keys before sorting
1860 * Consider GnuTLS if OpenSSL license becomes a problem
1862 http://archives.postgresql.org/pgsql-patches/2006-05/msg00040.php
1863 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01213.php
1865 * Consider changing documentation format from SGML to XML
1867 http://archives.postgresql.org/pgsql-docs/2006-12/msg00152.php
1869 * Consider making NAMEDATALEN more configurable in future releases
1870 * Update our code to handle 64-bit timezone files to match the zic
1871 source code, which now uses them
1872 * Have configure choose integer datetimes by default
1874 http://archives.postgresql.org/pgsql-patches/2007-05/msg00046.php
1876 * Support scoped IPv6 addresses
1878 http://archives.postgresql.org/pgsql-bugs/2007-05/msg00111.php
1880 * Consider allowing 64-bit integers and floats to be passed by value on
1883 Also change 32-bit floats (float4) to be passed by value at the same
1886 * Research use of signals and sleep wake ups
1888 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00003.php
1890 * Add automated check for invalid C++ source code constructs
1892 http://archives.postgresql.org/pgsql-patches/2007-07/msg00056.php
1894 * Consider simplifying how memory context resets handle child contexts
1896 http://archives.postgresql.org/pgsql-patches/2007-08/msg00067.php
1898 * Remove use of MAKE_PTR and MAKE_OFFSET macros
1900 http://archives.postgresql.org/pgsql-general/2007-08/msg01510.php
1902 * Convert single quotes to apostrophes in the PDF documentation
1904 http://archives.postgresql.org/pgsql-docs/2007-12/msg00059.php
1906 * Create three versions of libpgport to simplify client code
1908 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00154.php
1910 * Remove old-style routines for manipulating tuples
1912 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00851.php
1914 * Improve detection of shared memory segments being used by other
1917 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00656.php
1919 * Implement the non-threaded Avahi service discovery protocol
1921 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00939.php
1922 http://archives.postgresql.org/pgsql-patches/2008-02/msg00097.php
1923 http://archives.postgresql.org/pgsql-hackers/2008-03/msg01211.php
1924 http://archives.postgresql.org/pgsql-patches/2008-04/msg00001.php
1928 o Remove configure.in check for link failure when cause is found
1929 o Remove readdir() errno patch when runtime/mingwex/dirent.c rev
1931 o Remove psql newline patch when we find out why mingw outputs an
1933 o Allow psql to use readline once non-US code pages work with
1935 o Fix problem with shared memory on the Win32 Terminal Server
1936 o Diagnose problem where shared memory can sometimes not be
1937 attached by postmaster children
1939 http://archives.postgresql.org/pgsql-general/2007-08/msg01377.php
1941 o Improve signal handling
1943 http://archives.postgresql.org/pgsql-patches/2005-06/msg00027.php
1945 o Convert MSVC build system to remove most batch files
1947 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00961.php
1949 o Prevent SSL from sending network packets to avoid interference
1950 with Win32 signal emulation
1952 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00455.php
1954 o Support pgxs when using MSVC
1956 o Fix MSVC NLS support, like for to_char()
1958 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00485.php
1959 http://archives.postgresql.org/pgsql-patches/2008-02/msg00038.php
1961 o Find a correct rint() substitute on Windows
1963 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00808.php
1967 * Wire Protocol Changes
1969 o Allow dynamic character set handling
1970 o Add decoded type, length, precision
1972 o Update clients to use data types, typmod, schema.table.column names
1973 of result sets using new statement protocol
1979 * Add pre-parsing phase that converts non-ISO syntax to supported
1982 This could allow SQL written for other databases to run without
1985 * Allow plug-in modules to emulate features from other databases
1986 * Add features of Oracle-style packages (Pavel)
1988 A package would be a schema with session-local variables,
1989 public/private functions, and initialization functions. It
1990 is also possible to implement these capabilities
1991 in any schema and not use a separate "packages"
1994 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00384.php
1996 * Consider allowing control of upper/lower case folding of unquoted
1999 http://archives.postgresql.org/pgsql-hackers/2004-04/msg00818.php
2000 http://archives.postgresql.org/pgsql-hackers/2006-10/msg01527.php
2001 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00849.php
2003 * Add autonomous transactions
2005 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00893.php
2009 Features We Do _Not_ Want
2010 =========================
2012 * All backends running as threads in a single process (not wanted)
2014 This eliminates the process protection we get from the current setup.
2015 Thread creation is usually the same overhead as process creation on
2016 modern systems, so it seems unwise to use a pure threaded model.
2018 * Optimizer hints (not wanted)
2020 Optimizer hints are used to work around problems in the optimizer. We
2021 would rather have the problems reported and fixed.
2023 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00506.php
2024 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00517.php
2025 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00663.php
2028 Because we support postfix operators, it isn't possible to make AS
2029 optional and continue to use bison.
2030 http://archives.postgresql.org/pgsql-hackers/2003-04/msg00436.php
2032 http://archives.postgresql.org/pgsql-sql/2006-08/msg00164.php
2034 * Embedded server (not wanted)
2036 While PostgreSQL clients runs fine in limited-resource environments, the
2037 server requires multiple processes and a stable pool of resources to
2038 run reliabily and efficiently. Stripping down the PostgreSQL server
2039 to run in the same process address space as the client application
2040 would add too much complexity and failure cases.