3 Current maintainer: Bruce Momjian (bruce@momjian.us)
4 Last updated: Mon Mar 24 20:28:45 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
26 Lock table corruption following SIGTERM of an individual backend
27 has been reported in 8.0. A possible cause was fixed in 8.1, but
28 it is unknown whether other problems exist. This item mostly
29 requires additional testing rather than of writing any new code.
31 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00174.php
32 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00218.php
34 * Check for unreferenced table files created by transactions that were
35 in-progress when the server terminated abruptly
37 http://archives.postgresql.org/pgsql-patches/2006-06/msg00096.php
39 * Set proper permissions on non-system schemas during db creation
41 Currently all schemas are owned by the super-user because they are copied
42 from the template1 database. However, since all objects are inherited
43 from the template database, it is not clear that setting schemas to the db
46 * Add function to report the time of the most recent server reload
47 * Allow statistics collector information to be pulled from the collector
48 process directly, rather than requiring the collector to write a
49 filesystem file twice a second?
50 * Reduce file system activity overhead of statistics file pgstat.stat
52 http://archives.postgresql.org/pgsql-general/2007-12/msg00106.php
54 * Allow statistics last vacuum/analyze execution times to be displayed
55 without requiring stats_row_level to be enabled
57 http://archives.postgresql.org/pgsql-docs/2007-04/msg00028.php
59 * Allow log_min_messages to be specified on a per-module basis
61 This would allow administrators to see more detailed information from
62 specific sections of the backend, e.g. checkpoints, autovacuum, etc.
63 Another idea is to allow separate configuration files for each module,
64 or allow arbitrary SET commands to be passed to them.
66 * Simplify ability to create partitioned tables
68 This would allow creation of partitioned tables without requiring
69 creation of triggers or rules for INSERT/UPDATE/DELETE, and constraints
70 for rapid partition selection. Options could include range and hash
73 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00375.php
74 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00151.php
75 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00028.php
76 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00248.php
77 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00387.php
78 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00413.php
80 * Allow auto-selection of partitioned tables for min/max() operations
81 * Allow more complex user/database default GUC settings
83 Currently ALTER USER and ALTER DATABASE support per-user and
84 per-database defaults. Consider adding per-user-and-database
85 defaults so things like search_path can be defaulted for a
86 specific user connecting to a specific database.
88 * Allow custom variable classes that can restrict who can set the values
90 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00911.php
92 * Implement the SQL standard mechanism whereby REVOKE ROLE revokes only
93 the privilege granted by the invoking role, and not those granted
96 http://archives.postgresql.org/pgsql-bugs/2007-05/msg00010.php
98 * Allow SSL authentication/encryption over unix domain sockets
100 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00924.php
102 * Allow SSL key file permission checks to be optionally disabled when
103 sharing SSL keys with other applications
105 http://archives.postgresql.org/pgsql-bugs/2007-12/msg00069.php
107 * Allow client certificate names to be checked against the client
110 This is already implemented in
111 libpq/fe-secure.c::verify_peer_name_matches_certificate() but the code
114 * Configuration files
116 o Allow pg_hba.conf to specify host names along with IP addresses
118 Host name lookup could occur when the postmaster reads the
119 pg_hba.conf file, or when the backend starts. Another
120 solution would be to reverse lookup the connection IP and
121 check that hostname against the host names in pg_hba.conf.
122 We could also then check that the host name maps to the IP
125 o %Allow postgresql.conf file values to be changed via an SQL
126 API, perhaps using SET GLOBAL
127 o Allow the server to be stopped/restarted via an SQL API
128 o Issue a warning if a change-on-restart-only postgresql.conf value
129 is modified and the server config files are reloaded
130 o Consider normalizing fractions in postgresql.conf, perhaps
133 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00550.php
135 o Allow Kerberos to disable stripping of realms so we can
136 check the username@realm against multiple realms
138 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00009.php
143 o Allow a database in tablespace t1 with tables created in
144 tablespace t2 to be used as a template for a new database created
145 with default tablespace t2
147 Currently all objects in the default database tablespace must
148 have default tablespace specifications. This is because new
149 databases are created by copying directories. If you mix default
150 tablespace tables and tablespace-specified tables in the same
151 directory, creating a new database from such a mixed directory
152 would create a new database with tables that had incorrect
153 explicit tablespaces. To fix this would require modifying
154 pg_class in the newly copied database, which we don't currently
157 o Allow reporting of which objects are in which tablespaces
159 This item is difficult because a tablespace can contain objects
160 from multiple databases. There is a server-side function that
161 returns the databases which use a specific tablespace, so this
162 requires a tool that will call that function and connect to each
163 database to find the objects in each database for that tablespace.
165 o Allow WAL replay of CREATE TABLESPACE to work when the directory
166 structure on the recovery computer is different from the original
168 o Allow per-tablespace quotas
171 * Point-In-Time Recovery (PITR)
173 o Allow a warm standby system to also allow read-only statements
176 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00050.php
178 o %Create dump tool for write-ahead logs for use in determining
179 transaction id for point-in-time recovery
181 This is useful for checking PITR recovery.
183 o Allow recovery.conf to support the same syntax as
184 postgresql.conf, including quoting
186 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00497.php
188 o Fix server restart problem when the server was shutdown during
191 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00800.php
193 o Recreate pg_xlog/archive_status/ if it doesn't exist after
194 restoring from a PITR backup
196 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00487.php
202 * Change NUMERIC to enforce the maximum precision
203 * Reduce storage space for small NUMERICs
205 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01331.php
206 http://archives.postgresql.org/pgsql-patches/2007-02/msg00505.php
207 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00715.php
209 * Fix data types where equality comparison isn't intuitive, e.g. box
210 * Add support for public SYNONYMs
212 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00519.php
214 * Fix CREATE CAST on DOMAINs
216 http://archives.postgresql.org/pgsql-hackers/2006-05/msg00072.php
217 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01681.php
219 * Allow domains to be cast
221 http://archives.postgresql.org/pgsql-hackers/2003-06/msg01206.php
222 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00289.php
224 * Add support for SQL-standard GENERATED/IDENTITY columns
226 http://archives.postgresql.org/pgsql-hackers/2006-07/msg00543.php
227 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00038.php
228 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00344.php
229 http://archives.postgresql.org/pgsql-patches/2007-05/msg00076.php
230 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00604.php
232 * Improve XML support
234 http://developer.postgresql.org/index.php/XML_Support
236 * Consider placing all sequences in a single table, or create a system
239 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
241 * Allow the UUID type to accept non-standard formats
243 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01214.php
245 * Allow text search dictionary to filter out only stop words
247 http://archives.postgresql.org/pgsql-patches/2007-11/msg00081.php
249 * Consider a function-based API for '@@' full text searches
251 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00511.php
253 * Improve text search error messages
255 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00966.php
256 http://archives.postgresql.org/pgsql-hackers/2007-11/msg01146.php
258 * Consider a special data type for regular expressions
260 http://archives.postgresql.org/pgsql-hackers/2007-08/msg01067.php
262 * Reduce BIT data type overhead using short varlena headers
264 http://archives.postgresql.org/pgsql-general/2007-12/msg00273.php
266 * Allow xml arrays to be cast to other data types
268 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00981.php
269 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00231.php
270 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00471.php
276 o Allow infinite dates and intervals just like infinite timestamps
277 o Merge hardwired timezone names with the TZ database; allow either
278 kind everywhere a TZ name is currently taken
279 o Allow TIMESTAMP WITH TIME ZONE to store the original timezone
280 information, either zone name or offset from UTC [timezone]
282 If the TIMESTAMP value is stored with a time zone name, interval
283 computations should adjust based on the time zone rules.
285 o Fix SELECT '0.01 years'::interval, '0.01 months'::interval
286 o Add a GUC variable to allow output of interval values in ISO8601
288 o Have timestamp subtraction not call justify_hours()?
290 http://archives.postgresql.org/pgsql-sql/2006-10/msg00059.php
292 o Improve timestamptz subtraction to be DST-aware
294 Currently subtracting one date from another that crosses a
295 daylight savings time adjustment can return '1 day 1 hour', but
296 adding that back to the first date returns a time one hour in
297 the future. This is caused by the adjustment of '25 hours' to
298 '1 day 1 hour', and '1 day' is the same time the next day, even
299 if daylight savings adjustments are involved.
301 o Fix interval display to support values exceeding 2^31 hours
302 o Add overflow checking to timestamp and interval arithmetic
303 o Extend timezone code to allow 64-bit values so we can
304 represent years beyond 2038
306 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01363.php
308 o Use LC_TIME for localized weekday/month names, rather than
311 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00390.php
313 o Add ISO INTERVAL handling
315 http://archives.postgresql.org/pgsql-hackers/2006-01/msg00250.php
316 http://archives.postgresql.org/pgsql-bugs/2006-04/msg00248.php
318 o Support ISO INTERVAL syntax if units cannot be determined from
319 the string, and are supplied after the string
321 The SQL standard states that the units after the string
322 specify the units of the string, e.g. INTERVAL '2' MINUTE
323 should return '00:02:00'. The current behavior has the units
324 restrict the interval value to the specified unit or unit
325 range, INTERVAL '70' SECOND returns '00:00:10'.
327 For syntax that isn't uniquely ISO or PG syntax, like '1' or
328 '1:30', treat as ISO if there is a range specification clause,
329 and as PG if there no clause is present, e.g. interpret '1:30'
330 MINUTE TO SECOND as '1 minute 30 seconds', and interpret
331 '1:30' as '1 hour, 30 minutes'.
333 This makes common cases like SELECT INTERVAL '1' MONTH
334 SQL-standard results. The SQL standard supports a limited
335 number of unit combinations and doesn't support unit names in
336 the string. The PostgreSQL syntax is more flexible in the
337 range of units supported, e.g. PostgreSQL supports '1 year 1
338 hour', while the SQL standard does not.
340 o Add support for year-month syntax, INTERVAL '50-6' YEAR
342 o Interpret INTERVAL '1 year' MONTH as CAST (INTERVAL '1
343 year' AS INTERVAL MONTH), and this should return '12 months'
344 o Round or truncate values to the requested precision, e.g.
345 INTERVAL '11 months' AS YEAR should return one or zero
346 o Support precision, CREATE TABLE foo (a INTERVAL MONTH(3))
351 o Delay resolution of array expression's data type so assignment
352 coercion can be performed on empty array expressions
353 o Add support for arrays of domains
355 http://archives.postgresql.org/pgsql-patches/2007-05/msg00114.php
357 o Allow single-byte header storage for array elements
362 o Improve vacuum of large objects, like contrib/vacuumlo?
363 o Add security checking for large objects
364 o Auto-delete large objects when referencing row is deleted
366 contrib/lo offers this functionality.
368 o Allow read/write into TOAST values like large objects
370 This requires the TOAST column to be stored EXTERNAL.
372 o Add API for 64-bit large object access
374 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00781.php
378 * Add locale-aware MONEY type, and support multiple currencies
380 http://archives.postgresql.org/pgsql-general/2005-08/msg01432.php
381 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01181.php
383 * MONEY dumps in a locale-specific format making it difficult to
384 restore to a system with a different locale
385 * Allow MONEY to be easily cast to/from other numeric data types
391 * Allow INET subnet tests using non-constants to be indexed
392 * Allow to_date() and to_timestamp() accept localized month names
393 * Fix to_date()-related functions to consistently issue errors
395 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00915.php
397 * Add missing parameter handling in to_char()
399 http://archives.postgresql.org/pgsql-hackers/2005-12/msg00948.php
401 * Allow substring/replace() to get/set bit values
402 * Allow to_char() on interval values to accumulate the highest unit
405 Some special format flag would be required to request such
406 accumulation. Such functionality could also be added to EXTRACT.
407 Prevent accumulation that crosses the month/day boundary because of
408 the uneven number of days in a month.
410 o to_char(INTERVAL '1 hour 5 minutes', 'MI') => 65
411 o to_char(INTERVAL '43 hours 20 minutes', 'MI' ) => 2600
412 o to_char(INTERVAL '43 hours 20 minutes', 'WK:DD:HR:MI') => 0:1:19:20
413 o to_char(INTERVAL '3 years 5 months','MM') => 41
415 * Implement inlining of set-returning functions defined in SQL
416 * Allow SQL-language functions to return results from RETURNING queries
418 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00665.php
420 * Allow SQL-language functions to reference parameters by parameter name
422 Currently SQL-language functions can only refer to dollar parameters,
425 * Add SPI_gettypmod() to return the typemod for a TupleDesc
426 * Enforce typmod for function inputs, function results and parameters for
427 spi_prepare'd statements called from PLs
429 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01403.php
431 * Allow holdable cursors in SPI
432 * Tighten function permission checks
434 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00568.php
436 * Fix IS OF so it matches the ISO specification, and add documentation
438 http://archives.postgresql.org/pgsql-patches/2003-08/msg00060.php
439 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00060.php
441 * Add missing operators for geometric data types
443 Some geometric types do not have the full suite of geometric operators,
446 * Implement Boyer-Moore searching in strpos()
448 http://archives.postgresql.org/pgsql-patches/2007-08/msg00012.php
450 * Prevent malicious functions from being executed with the permissions
451 of unsuspecting users
453 Index functions are safe, so VACUUM and ANALYZE are safe too.
454 Triggers, CHECK and DEFAULT expressions, and rules are still vulnerable.
455 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00268.php
457 * Reduce memory usage of aggregates in set returning functions
459 http://archives.postgresql.org/pgsql-performance/2008-01/msg00031.php
461 * Add temporal versions of generate_series()
463 http://archives.postgresql.org/pgsql-hackers/2007-04/msg01180.php
465 * Add array_accum() and array_to_set() functions for arrays
467 The standards specify array_agg() and UNNEST.
468 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00464.php
470 * Fix /contrib/ltree operator
472 http://archives.postgresql.org/pgsql-bugs/2007-11/msg00044.php
474 * Fix inconsistent precedence of =, >, and < compared to <>, >=, and <=
476 http://archives.postgresql.org/pgsql-bugs/2007-12/msg00145.php
478 * Fix regular expression bug when using complex back-references
480 http://archives.postgresql.org/pgsql-bugs/2007-10/msg00000.php
482 * Have /contrib/dblink reuse unnamed connections
484 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00895.php
488 Multi-Language Support
489 ======================
491 * Add NCHAR (as distinguished from ordinary varchar),
492 * Allow locale to be set at database creation
494 Currently locale can only be set during initdb. No global tables have
495 locale-aware columns. However, the database template used during
496 database creation might have locale-aware indexes. The indexes would
497 need to be reindexed to match the new locale.
499 * Allow encoding on a per-column basis optionally using the ICU library:
501 Right now only one encoding is allowed per database. [locale]
503 http://archives.postgresql.org/pgsql-hackers/2005-03/msg00932.php
504 http://archives.postgresql.org/pgsql-patches/2005-08/msg00309.php
505 http://archives.postgresql.org/pgsql-patches/2006-03/msg00233.php
506 http://archives.postgresql.org/pgsql-hackers/2006-09/msg00662.php
508 * Add CREATE COLLATE? [locale]
509 * Support multiple simultaneous character sets, per SQL92
510 * Improve UTF8 combined character handling?
511 * Add octet_length_server() and octet_length_client()
512 * Make octet_length_client() the same as octet_length()?
513 * Fix problems with wrong runtime encoding conversion for NLS message files
514 * Add URL to more complete multi-byte regression tests
516 http://archives.postgresql.org/pgsql-hackers/2005-07/msg00272.php
518 * Fix ILIKE and regular expressions to handle case insensitivity
519 properly in multibyte encodings
521 http://archives.postgresql.org/pgsql-bugs/2005-10/msg00001.php
522 http://archives.postgresql.org/pgsql-patches/2005-11/msg00173.php
524 * Set client encoding based on the client operating system encoding
526 Currently client_encoding is set in postgresql.conf, which
527 defaults to the server encoding.
528 http://archives.postgresql.org/pgsql-hackers/2006-08/msg01696.php
530 * Change memory allocation for multi-byte functions so memory is
531 allocated inside conversion functions
533 Currently we preallocate memory based on worst-case usage.
540 * Automatically create rules on views so they are updateable, per SQL99
542 We can only auto-create rules for simple views. For more complex
543 cases users will still have to write rules manually.
545 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00586.php
546 http://archives.postgresql.org/pgsql-patches/2006-08/msg00255.php
548 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
549 * Allow NOTIFY in rules involving conditionals
550 * Allow VIEW/RULE recompilation when the underlying tables change
552 Another issue is whether underlying table changes should be reflected
553 in the view, e.g. should SELECT * show additional columns if they
554 are added after the view is created.
556 * Make it possible to use RETURNING together with conditional DO INSTEAD
557 rules, such as for partitioning setups
559 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00577.php
561 * Add the ability to automatically create materialized views
563 Right now materialized views require the user to create triggers on the
564 main table to keep the summary table current. SQL syntax should be able
565 to manager the triggers and summary table automatically. A more
566 sophisticated implementation would automatically retrieve from the
567 summary table when the main table is referenced, if possible.
574 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
575 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
576 * %Allow SET CONSTRAINTS to be qualified by schema/table name
577 * %Add a separate TRUNCATE permission
579 Currently only the owner can TRUNCATE a table because triggers are not
580 called, and the table is locked in exclusive mode.
582 * Allow PREPARE of cursors
583 * Allow finer control over the caching of prepared query plans
585 Currently queries prepared via the libpq API are planned on first
586 execute using the supplied parameters --- allow SQL PREPARE to do the
587 same. Also, allow control over replanning prepared queries either
588 manually or automatically when statistics for execute parameters
589 differ dramatically from those used during planning.
591 * Improve logging of prepared transactions recovered during startup
593 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00092.php
595 * Allow LISTEN/NOTIFY to store info in memory rather than tables?
597 Currently LISTEN/NOTIFY information is stored in pg_listener. Storing
598 such information in memory would improve performance.
600 * Add optional textual message to NOTIFY
602 This would allow an informational message to be added to the notify
603 message, perhaps indicating the row modified or other custom
606 * Allow multiple identical NOTIFY events to always be communicated to the
607 client, rather than sent as a single notification to the listener
609 http://archives.postgresql.org/pgsql-general/2008-01/msg00057.php
611 * Add a GUC variable to warn about non-standard SQL usage in queries
612 * Add SQL-standard MERGE command, typically used to merge two tables
615 This is similar to UPDATE, then for unmatched rows, INSERT.
616 Whether concurrent access allows modifications which could cause
617 row loss is implementation independent.
619 * Add REPLACE or UPSERT command that does UPDATE, or on failure, INSERT
622 To implement this cleanly requires that the table have a unique index
623 so duplicate checking can be easily performed. It is possible to
624 do it without a unique index if we require the user to LOCK the table
627 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00501.php
628 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00536.php
630 * Add NOVICE output level for helpful messages like automatic sequence/index
632 * Add GUC to issue notice about statements that use unjoined tables
633 * Allow EXPLAIN to identify tables that were skipped because of
635 * Allow EXPLAIN output to be more easily processed by scripts, perhaps XML
636 * Enable standard_conforming_strings
637 * Make standard_conforming_strings the default in 8.5?
639 When this is done, backslash-quote should be prohibited in non-E''
640 strings because of possible confusion over how such strings treat
641 backslashes. Basically, '' is always safe for a literal single
642 quote, while \' might or might not be based on the backslash
645 * Simplify dropping roles that have objects in several databases
646 * Allow COMMENT ON to accept an expression rather than just a string
647 * Allow the count returned by SELECT, etc to be represented as an int64
648 to allow a higher range of values
649 * Add SQL99 WITH clause to SELECT
650 * Add SQL:2003 WITH RECURSIVE (hierarchical) queries to SELECT
652 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01375.php
653 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00642.php
654 http://archives.postgresql.org/pgsql-patches/2007-03/msg00139.php
655 http://archives.postgresql.org/pgsql-hackers/2007-11/msg01334.php
657 * Add DEFAULT .. AS OWNER so permission checks are done as the table
660 This would be useful for SERIAL nextval() calls and CHECK constraints.
662 * Allow DISTINCT to work in multiple-argument aggregate calls
663 * Add column to pg_stat_activity that shows the progress of long-running
664 commands like CREATE INDEX and VACUUM
665 * Implement SQL:2003 window functions
666 * Improve failure message when DROP DATABASE is used on a database that
667 has prepared transactions
668 * Allow INSERT/UPDATE ... RETURNING inside a SELECT 'FROM' clause
670 http://archives.postgresql.org/pgsql-general/2006-09/msg00803.php
671 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00693.php
673 * Increase locking when DROPing objects so dependent objects cannot
674 get dropped while the DROP operation is happening
676 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00937.php
678 * -Allow AS in "SELECT col AS label" to be optional in certain cases
680 * Allow INSERT ... DELETE ... RETURNING, namely allow the DELETE ...
681 RETURNING to supply values to the INSERT
682 http://archives.postgresql.org/pgsql-hackers/2008-02/thrd2.php#00979
684 * Add comments on system tables/columns using the information in
687 Ideally the information would be pulled from the SGML file
690 * Improve reporting of UNION type mismatches
692 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00944.php
693 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00597.php
698 o Allow CREATE TABLE AS to determine column lengths for complex
699 expressions like SELECT col1 || col2
701 o Have WITH CONSTRAINTS also create constraint indexes
703 http://archives.postgresql.org/pgsql-patches/2007-04/msg00149.php
705 o Have CONSTRAINT cname NOT NULL record the contraint name
707 Right now pg_attribute.attnotnull records the NOT NULL status
708 of the column, but does not record the contraint name
710 o Prevent concurrent CREATE TABLE table1 from sometimes returning
711 a cryptic error message
713 http://archives.postgresql.org/pgsql-bugs/2007-10/msg00169.php
717 o Allow UPDATE tab SET ROW (col, ...) = (SELECT...)
719 http://archives.postgresql.org/pgsql-hackers/2006-07/msg01306.php
720 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00865.php
721 http://archives.postgresql.org/pgsql-patches/2007-04/msg00315.php
722 http://archives.postgresql.org/pgsql-patches/2008-03/msg00237.php
724 o Research self-referential UPDATEs that see inconsistent row versions
725 in read-committed mode
727 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00507.php
728 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00016.php
730 o Allow GLOBAL temporary tables to exist as empty by default in
733 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00006.php
738 o Have ALTER TABLE RENAME rename SERIAL sequence names
740 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
742 o Have ALTER SEQUENCE RENAME rename the sequence name stored
743 in the sequence table
745 http://archives.postgresql.org/pgsql-bugs/2007-09/msg00092.php
746 http://archives.postgresql.org/pgsql-bugs/2007-10/msg00007.php
747 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
749 o Add ALTER DOMAIN to modify the underlying data type
750 o %Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
752 http://archives.postgresql.org/pgsql-patches/2006-02/msg00168.php
754 o %Allow ALTER TABLE to change constraint deferrability and actions
755 o Add missing object types for ALTER ... SET SCHEMA
756 o Allow ALTER TABLESPACE to move to different directories
757 o Allow databases to be moved to different tablespaces
758 o Allow moving system tables to other tablespaces, where possible
760 Currently non-global system tables must be in the default database
761 tablespace. Global system tables can never be moved.
763 o Prevent parent tables from altering or dropping constraints
764 like CHECK that are inherited by child tables unless CASCADE
766 o %Prevent child tables from altering or dropping constraints
767 like CHECK that were inherited from the parent table
768 o Have ALTER INDEX update the name of a constraint using that index
769 o Add ALTER TABLE RENAME CONSTRAINT, update index name also
770 o Allow column display reordering by recording a display,
771 storage, and permanent id for every column?
773 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00782.php
779 o Automatically maintain clustering on a table
781 This might require some background daemon to maintain clustering
782 during periods of low usage. It might also require tables to be only
783 partially filled for easier reorganization. Another idea would
784 be to create a merged heap/index data file so an index lookup would
785 automatically access the heap data too. A third idea would be to
786 store heap rows in hashed groups, perhaps using a user-supplied
788 http://archives.postgresql.org/pgsql-performance/2004-08/msg00349.php
790 o %Add default clustering to system tables
792 To do this, determine the ideal cluster index for each system
793 table and set the cluster setting during initdb.
795 o %Add VERBOSE option to report tables as they are processed,
801 o Allow COPY to report error lines and continue
803 This requires the use of a savepoint before each COPY line is
804 processed, with ROLLBACK on COPY failure.
805 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00572.php
807 o Allow COPY on a newly-created table to skip WAL logging
809 On crash recovery, the table involved in the COPY would
810 be removed or have its heap and index files truncated. One
811 issue is that no other backend should be able to add to
812 the table at the same time, which is something that is
813 currently allowed. This currently is done if the table is
814 created inside the same transaction block as the COPY because
815 no other backends can see the table.
817 o Consider using a ring buffer for COPY FROM
819 http://archives.postgresql.org/pgsql-patches/2008-02/msg00140.php
820 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01080.php
822 o Allow COPY FROM to create index entries in bulk
824 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00811.php
826 o Allow COPY in CSV mode to control whether a quoted zero-length
827 string is treated as NULL
829 Currently this is always treated as a zero-length string,
830 which generates an error when loading into an integer column
831 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00905.php
837 o Allow column-level privileges
838 o %Allow GRANT/REVOKE permissions to be applied to all schema objects
841 The proposed syntax is:
842 GRANT SELECT ON ALL TABLES IN public TO phpuser;
843 GRANT SELECT ON NEW TABLES IN public TO phpuser;
845 o Allow GRANT/REVOKE permissions to be inherited by objects based on
848 o Allow SERIAL sequences to inherit permissions from the base table?
853 o Prevent DROP TABLE from dropping a row referenced by its own open
859 o Allow INSERT/UPDATE of the system-generated oid value for a row
860 o In rules, allow VALUES() to contain a mixture of 'old' and 'new'
866 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
871 Referential Integrity
872 =====================
874 * Add MATCH PARTIAL referential integrity
875 * Change foreign key constraint for array -> element to mean element
877 * Fix problem when cascading referential triggers make changes on
878 cascaded tables, seeing the tables in an intermediate state
880 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00174.php
881 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00174.php
883 * Allow DEFERRABLE and end-of-statement UNIQUE constraints?
885 This would allow UPDATE tab SET col = col + 1 to work if col has
886 a unique index. Currently, uniqueness checks are done while the
887 command is being executed, rather than at the end of the statement
889 http://people.planetpostgresql.org/greg/index.php?/archives/2006/06/10.html
890 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01458.php
892 * Optimize referential integrity checks
894 http://archives.postgresql.org/pgsql-performance/2005-10/msg00458.php
895 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00744.php
898 Server-Side Languages
899 =====================
902 o Fix RENAME to work on variables other than OLD/NEW
904 http://archives.postgresql.org/pgsql-hackers/2002-03/msg00591.php
905 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01615.php
906 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01587.php
908 o Allow function parameters to be passed by name,
909 get_employee_salary(12345 AS emp_id, 2001 AS tax_year)
910 o Allow handling of %TYPE arrays, e.g. tab.col%TYPE[]
911 o Allow listing of record column names, and access to
912 record columns via variables, e.g. columns := r.(*),
915 http://archives.postgresql.org/pgsql-patches/2005-07/msg00458.php
916 http://archives.postgresql.org/pgsql-patches/2006-05/msg00302.php
917 http://archives.postgresql.org/pgsql-patches/2006-06/msg00031.php
919 o Add support for SCROLL cursors
920 o Add support for WITH HOLD cursors
921 o Allow row and record variables to be set to NULL constants,
922 and allow NULL tests on such variables
924 Because a row is not scalar, do not allow assignment
925 from NULL-valued scalars.
927 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00070.php
929 o Review handling of MOVE and FETCH
931 http://archives.postgresql.org/pgsql-patches/2007-04/msg00527.php
933 o Improve logic of determining if an identifier is a a
934 variable or column name
936 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00436.php
940 o Add table function support to pltcl, plpythonu
941 o Add support for polymorphic arguments and return types to
942 languages other than PL/PgSQL
943 o Add capability to create and call PROCEDURES
944 o Add support for OUT and INOUT parameters to languages other
946 o Add PL/PythonU tracebacks
948 http://archives.postgresql.org/pgsql-patches/2006-02/msg00288.php
950 o Allow data to be passed in native language formats, rather
953 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00289.php
960 * Have pg_ctl look at PGHOST in case it is a socket directory?
961 * Allow pg_ctl to work properly with configuration files located outside
964 pg_ctl can not read the pid file because it isn't located in the
965 config directory but in the PGDATA directory. The solution is to
966 allow pg_ctl to read and understand postgresql.conf to find the
967 data_directory value.
969 * Add a function like pg_get_indexdef() that report more detailed index
972 http://archives.postgresql.org/pgsql-bugs/2007-12/msg00166.php
974 * Prefix command-line utilities like createuser with 'pg_'
976 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00025.php
980 o Have psql show current values for a sequence
981 o Move psql backslash database information into the backend, use
982 mnemonic commands? [psql]
984 This would allow non-psql clients to pull the same information out
985 of the database as psql.
987 o Fix psql's \d commands more consistent
989 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
990 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
992 o Consistently display privilege information for all objects in psql
993 o Add auto-expanded mode so expanded output is used if the row
994 length is wider than the screen width.
996 Consider using auto-expanded mode for backslash commands like \df+.
998 o Prevent tab completion of SET TRANSACTION from querying the
999 database and therefore preventing the transaction isolation
1000 level from being set.
1002 Currently SET <tab> causes a database lookup to check all
1003 supported session variables. This query causes problems
1004 because setting the transaction isolation level must be the
1005 first statement of a transaction.
1007 o Add a \set variable to control whether \s displays line numbers
1009 Another option is to add \# which lists line numbers, and
1010 allows command execution.
1012 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00255.php
1014 o Prevent escape string warnings when object names have
1017 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00227.php
1019 o Have \d show foreign keys that reference a table's primary key
1021 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00424.php
1023 o Have \d show child tables that inherit from the specified parent
1024 o Have \l+ show database size, if permissions allow
1026 Ideally it will not generate an error for invalid permissions
1028 o Include the symbolic SQLSTATE name in verbose error reports
1030 http://archives.postgresql.org/pgsql-general/2007-09/msg00438.php
1033 * pg_dump / pg_restore
1034 o %Add dumping of comments on index columns and composite type columns
1035 o %Add full object name to the tag field. eg. for operators we need
1036 '=(integer, integer)', instead of just '='.
1037 o Add pg_dumpall custom format dumps?
1038 o Allow selection of individual object(s) of all types, not just
1040 o In a selective dump, allow dumping of an object and all its
1042 o Add options like pg_restore -l and -L to pg_dump
1043 o Stop dumping CASCADE on DROP TYPE commands in clean mode
1044 o Allow pg_dump --clean to drop roles that own objects or have
1046 o Change pg_dump so that a comment on the dumped database is
1047 applied to the loaded database, even if the database has a
1048 different name. This will require new backend syntax, perhaps
1049 COMMENT ON CURRENT DATABASE.
1050 o Remove unnecessary function pointer abstractions in pg_dump source
1052 o Allow pg_dump to utilize multiple CPUs and I/O channels by dumping
1053 multiple objects simultaneously
1055 The difficulty with this is getting multiple dump processes to
1056 produce a single dump output file. It also would require
1057 several sessions to share the same snapshot.
1058 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00205.php
1060 o Allow pg_restore to utilize multiple CPUs and I/O channels by
1061 restoring multiple objects simultaneously
1063 This might require a pg_restore flag to indicate how many
1064 simultaneous operations should be performed. Only pg_dump's
1065 -Fc format has the necessary dependency information.
1066 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00963.php
1068 o To better utilize resources, allow pg_restore to check foreign
1069 keys simultaneously, where possible
1070 o Allow pg_restore to create all indexes of a table
1071 concurrently, via a single heap scan
1073 This requires a pg_dump -Fc file because that format contains
1074 the required dependency information.
1075 http://archives.postgresql.org/pgsql-general/2007-05/msg01274.php
1077 o Allow pg_restore to load different parts of the COPY data
1079 o Prevent pg_dump/pg_restore from being affected by
1082 Using psql to restore a pg_dump dump is also affected.
1084 o Remove pre-7.3 pg_dump code that assumes pg_depend does not exit
1085 o Allow pre/data/post files when schema and data are dumped
1086 separately, for performance reasons
1088 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00205.php
1095 Document differences between ecpg and the SQL standard and
1096 information about the Informix-compatibility module.
1098 o Solve cardinality > 1 for input descriptors / variables?
1099 o Add a semantic check level, e.g. check if a table really exists
1100 o fix handling of DB attributes that are arrays
1101 o Use backend PREPARE/EXECUTE facility for ecpg where possible
1103 o Fix nested C comments
1104 o %sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
1105 o Make SET CONNECTION thread-aware, non-standard?
1106 o Allow multidimensional arrays
1107 o Add internationalized message strings
1108 o Implement COPY FROM STDIN
1112 o Add PQescapeIdentifierConn()
1113 o Prevent PQfnumber() from lowercasing unquoted the column name
1115 PQfnumber() should never have been doing lowercasing, but
1116 historically it has so we need a way to prevent it
1118 o Allow statement results to be automatically batched to the client
1120 Currently all statement results are transferred to the libpq
1121 client before libpq makes the results available to the
1122 application. This feature would allow the application to make
1123 use of the first result rows while the rest are transferred, or
1124 held on the server waiting for them to be requested by libpq.
1125 One complexity is that a statement like SELECT 1/col could error
1126 out mid-way through the result set.
1128 o Consider disallowing multiple queries in PQexec() as an
1129 additional barrier to SQL injection attacks
1131 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00184.php
1133 o Add PQexecf() that allows complex parameter substitution
1135 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01803.php
1137 o Add SQLSTATE severity to PGconn return status
1139 http://archives.postgresql.org/pgsql-interfaces/2007-11/msg00015.php
1145 * Add deferred trigger queue file
1147 Right now all deferred trigger information is stored in backend
1148 memory. This could exhaust memory for very large trigger queues.
1149 This item involves dumping large queues into files.
1151 * Allow triggers to be disabled in only the current session.
1153 This is currently possible by starting a multi-statement transaction,
1154 modifying the system tables, performing the desired SQL, restoring the
1155 system tables, and committing the transaction. ALTER TABLE ...
1156 TRIGGER requires a table lock so it is not ideal for this usage.
1158 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
1160 If the dump is known to be valid, allow foreign keys to be added
1161 without revalidating the data.
1163 * Allow statement-level triggers to access modified rows
1164 * Support triggers on columns
1166 http://archives.postgresql.org/pgsql-patches/2005-07/msg00107.php
1168 * Allow AFTER triggers on system tables
1170 System tables are modified in many places in the backend without going
1171 through the executor and therefore not causing triggers to fire. To
1172 complete this item, the functions that modify system tables will have
1175 * Tighten trigger permission checks
1177 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00564.php
1179 * Allow BEFORE INSERT triggers on views
1181 http://archives.postgresql.org/pgsql-general/2007-02/msg01466.php
1183 * Add ability to trigger on TRUNCATE
1185 http://archives.postgresql.org/pgsql-sql/2008-01/msg00050.php
1192 * Add UNIQUE capability to non-btree indexes
1193 * Prevent index uniqueness checks when UPDATE does not modify the column
1195 Uniqueness (index) checks are done when updating a column even if the
1196 column is not modified by the UPDATE.
1198 * Allow the creation of on-disk bitmap indexes which can be quickly
1199 combined with other bitmap indexes
1201 Such indexes could be more compact if there are only a few distinct values.
1202 Such indexes can also be compressed. Keeping such indexes updated can be
1205 http://archives.postgresql.org/pgsql-patches/2005-07/msg00512.php
1206 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01107.php
1208 * Allow accurate statistics to be collected on indexes with more than
1209 one column or expression indexes, perhaps using per-index statistics
1211 http://archives.postgresql.org/pgsql-performance/2006-10/msg00222.php
1212 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01131.php
1214 * Consider increasing the number of default statistics target, and
1215 reduce statistics target overhead
1217 Also consider having a larger statistics target for indexed columns
1218 and expression indexes
1219 http://archives.postgresql.org/pgsql-general/2007-05/msg01228.php
1220 http://archives.postgresql.org/pgsql-general/2007-06/msg00542.php
1221 http://archives.postgresql.org/pgsql-hackers/2008-01/msg01066.php
1223 * Consider compressing indexes by storing key values duplicated in
1224 several rows as a single index entry
1226 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00341.php
1227 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01264.php
1228 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00465.php
1229 http://archives.postgresql.org/pgsql-patches/2007-03/msg00163.php
1231 * Add REINDEX CONCURRENTLY, like CREATE INDEX CONCURRENTLY
1233 This is difficult because you must upgrade to an exclusive table lock
1234 to replace the existing index file. CREATE INDEX CONCURRENTLY does not
1235 have this complication. This would allow index compaction without
1237 http://archives.postgresql.org/pgsql-performance/2007-08/msg00289.php
1239 * Allow multiple indexes to be created concurrently, ideally via a
1240 single heap scan, and have pg_restore use it
1246 o Allow inherited tables to inherit indexes, UNIQUE constraints,
1247 and primary/foreign keys
1248 o Honor UNIQUE INDEX on base column in INSERTs/UPDATEs
1249 on inherited table, e.g. INSERT INTO inherit_table
1250 (unique_index_col) VALUES (dup) should fail
1252 The main difficulty with this item is the problem of
1253 creating an index that can span multiple tables.
1255 o Allow SELECT ... FOR UPDATE on inherited tables
1256 o Require all CHECK constraints to be inherited
1258 http://archives.postgresql.org/pgsql-bugs/2007-04/msg00026.php
1260 o Add checks to prevent a CREATE RULE views on inherited tables
1262 http://archives.postgresql.org/pgsql-general/2008-02/msg01420.php
1263 http://archives.postgresql.org/pgsql-general/2008-03/msg00077.php
1268 o Add more GIST index support for geometric data types
1269 o Allow GIST indexes to create certain complex index types, like
1270 digital trees (see Aoki)
1275 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00051.php
1277 o Pack hash index buckets onto disk pages more efficiently
1279 Currently only one hash bucket can be stored on a page. Ideally
1280 several hash buckets could be stored on a single page and greater
1281 granularity used for the hash algorithm.
1283 http://archives.postgresql.org/pgsql-hackers/2004-06/msg00168.php
1285 o Consider sorting hash buckets so entries can be found using a
1286 binary search, rather than a linear scan
1287 o In hash indexes, consider storing the hash value with or instead
1289 o Add WAL logging for crash recovery
1290 o Allow multi-column hash indexes
1291 o -During index creation, pre-sort the tuples to improve build speed
1298 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
1300 Ideally this requires a separate test program that can be run
1301 at initdb time or optionally later. Consider O_SYNC when
1304 * Add program to test if fsync has a delay compared to non-fsync
1305 * Consider sorting writes during checkpoint
1307 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00541.php
1316 We could use a fixed row count and a +/- count to follow MVCC
1317 visibility rules, or a single cached value could be used and
1318 invalidated if anyone modifies the table. Another idea is to
1319 get a count directly from a unique index, but for this to be
1320 faster than a sequential scan it must avoid access to the heap
1321 to obtain tuple visibility information.
1323 * Provide a way to calculate an "estimated COUNT(*)"
1325 Perhaps by using the optimizer's cardinality estimates or random
1328 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00943.php
1330 * Allow data to be pulled directly from indexes
1332 Currently indexes do not have enough tuple visibility information
1333 to allow data to be pulled from the index without also accessing
1334 the heap. One way to allow this is to set a bit on index tuples
1335 to indicate if a tuple is currently visible to all transactions
1336 when the first valid heap lookup happens. This bit would have to
1337 be cleared when a heap tuple is expired.
1339 Another idea is to maintain a bitmap of heap pages where all rows
1340 are visible to all backends, and allow index lookups to reference
1341 that bitmap to avoid heap lookups, perhaps the same bitmap we might
1342 add someday to determine which heap pages need vacuuming. Frequently
1343 accessed bitmaps would have to be stored in shared memory. One 8k
1344 page of bitmaps could track 512MB of heap pages.
1346 A third idea would be for a heap scan to check if all rows are visible
1347 and if so set a per-table flag which can be checked by index scans.
1348 Any change to the table would have to clear the flag. To detect
1349 changes during the heap scan a counter could be set at the start and
1350 checked at the end --- if it is the same, the table has not been
1351 modified --- any table change would increment the counter.
1353 * Consider automatic caching of statements at various levels:
1356 o Query execute plan
1359 * Consider increasing internal areas when shared buffers is increased
1361 http://archives.postgresql.org/pgsql-hackers/2005-10/msg01419.php
1363 * Consider decreasing the amount of memory used by PrivateRefCount
1365 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00797.php
1366 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00752.php
1368 * Consider allowing higher priority queries to have referenced buffer
1369 cache pages stay in memory longer
1371 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00562.php
1378 * Improve speed with indexes
1380 For large table adjustments during VACUUM FULL, it is faster to cluster
1381 or reindex rather than update the index. Also, index updates can bloat
1384 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00024.php
1385 http://archives.postgresql.org/pgsql-performance/2007-05/msg00296.php
1386 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00307.php
1388 * Auto-fill the free space map by scanning the buffer cache or by
1389 checking pages written by the background writer
1391 http://archives.postgresql.org/pgsql-hackers/2006-02/msg01125.php
1392 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00011.php
1394 * Create a bitmap of pages that need vacuuming
1396 Instead of sequentially scanning the entire table, have the background
1397 writer or some other process record pages that have expired rows, then
1398 VACUUM can look at just those pages rather than the entire table. In
1399 the event of a system crash, the bitmap would probably be invalidated.
1400 One complexity is that index entries still have to be vacuumed, and
1401 doing this without an index scan (by using the heap values to find the
1402 index entry) might be slow and unreliable, especially for user-defined
1405 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01188.php
1406 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00121.php
1408 * Allow FSM to return free space toward the beginning of the heap file,
1409 in hopes that empty pages at the end can be truncated by VACUUM
1410 * Allow FSM page return free space based on table clustering, to assist
1411 in maintaining clustering?
1412 * Improve dead row detection during multi-statement transactions usage
1414 http://archives.postgresql.org/pgsql-patches/2007-03/msg00358.php
1416 * Consider a more compact data representation for dead tuples
1418 http://archives.postgresql.org/pgsql-patches/2007-05/msg00143.php
1423 o %Issue log message to suggest VACUUM FULL if a table is nearly
1425 o Improve control of auto-vacuum
1427 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00876.php
1429 o Prevent long-lived temporary tables from causing frozen-xid
1430 advancement starvation
1432 The problem is that autovacuum cannot vacuum them to set frozen xids;
1433 only the session that created them can do that.
1434 http://archives.postgresql.org/pgsql-general/2007-06/msg01645.php
1436 o Store per-table autovacuum settings in pg_class.reloptions.
1438 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01440.php
1439 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00724.php
1441 o Prevent autovacuum from running if an old transaction is still
1442 running from the last vacuum
1444 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00899.php
1451 * Fix priority ordering of read and write light-weight locks (Neil)
1453 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00893.php
1454 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00905.php
1456 * Fix problem when multiple subtransactions of the same outer transaction
1457 hold different types of locks, and one subtransaction aborts
1459 http://archives.postgresql.org/pgsql-hackers/2006-11/msg01011.php
1460 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00001.php
1461 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00435.php
1462 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00773.php
1464 * Allow UPDATEs on only non-referential integrity columns not to conflict
1465 with referential integrity locks
1467 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00073.php
1469 * Add idle_in_transaction_timeout GUC so locks are not held for long
1472 * Improve deadlock detection when a page cleaning lock conflicts
1473 with a shared buffer that is pinned
1475 http://archives.postgresql.org/pgsql-bugs/2008-01/msg00138.php
1476 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00873.php
1477 http://archives.postgresql.org/pgsql-committers/2008-01/msg00365.php
1482 Startup Time Improvements
1483 =========================
1485 * Experiment with multi-threaded backend for backend creation [thread]
1487 This would prevent the overhead associated with process creation. Most
1488 operating systems have trivial process creation time compared to
1489 database startup overhead, but a few operating systems (Win32,
1490 Solaris) might benefit from threading. Also explore the idea of
1491 a single session using multiple threads to execute a statement faster.
1498 * Eliminate need to write full pages to WAL before page modification [wal]
1500 Currently, to protect against partial disk page writes, we write
1501 full page images to WAL before they are modified so we can correct any
1502 partial page writes during recovery. These pages can also be
1503 eliminated from point-in-time archive files.
1505 o When off, write CRC to WAL and check file system blocks
1508 If CRC check fails during recovery, remember the page in case
1509 a later CRC for that page properly matches.
1511 o Write full pages during file system write and not when
1512 the page is modified in the buffer cache
1514 This allows most full page writes to happen in the background
1515 writer. It might cause problems for applying WAL on recovery
1516 into a partially-written page, but later the full page will be
1519 * Allow WAL traffic to be streamed to another server for stand-by
1521 * Reduce WAL traffic so only modified values are written rather than
1524 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01589.php
1526 * Allow WAL information to recover corrupted pg_controldata
1528 http://archives.postgresql.org/pgsql-patches/2006-06/msg00025.php
1530 * Find a way to reduce rotational delay when repeatedly writing
1533 Currently fsync of WAL requires the disk platter to perform a full
1534 rotation to fsync again. One idea is to write the WAL to different
1535 offsets that might reduce the rotational delay.
1537 * Allow WAL logging to be turned off for a table, but the table
1538 might be dropped or truncated during crash recovery [walcontrol]
1540 Allow tables to bypass WAL writes and just fsync() dirty pages on
1541 commit. This should be implemented using ALTER TABLE, e.g. ALTER
1542 TABLE PERSISTENCE [ DROP | TRUNCATE | DEFAULT ]. Tables using
1543 non-default logging should not use referential integrity with
1544 default-logging tables. A table without dirty buffers during a
1545 crash could perhaps avoid the drop/truncate.
1547 * Allow WAL logging to be turned off for a table, but the table would
1548 avoid being truncated/dropped [walcontrol]
1550 To do this, only a single writer can modify the table, and writes
1551 must happen only on new pages so the new pages can be removed during
1552 crash recovery. Readers can continue accessing the table. Such
1553 tables probably cannot have indexes. One complexity is the handling
1554 of indexes on TOAST tables.
1556 * Speed WAL recovery by allowing more than one page to be prefetched
1558 This should be done utilizing the same infrastructure used for
1559 prefetching in general to avoid introducing complex error-prone code
1561 http://archives.postgresql.org/pgsql-general/2007-12/msg00683.php
1562 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00497.php
1563 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01279.php
1565 * Improve WAL concurrency by increasing lock granularity
1567 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00556.php
1569 * Be more aggressive about creating WAL files
1571 http://archives.postgresql.org/pgsql-hackers/2007-10/msg01325.php
1573 * Have resource managers report the duration of their status changes
1575 http://archives.postgresql.org/pgsql-hackers/2007-10/msg01468.php
1579 Optimizer / Executor
1580 ====================
1582 * Improve selectivity functions for geometric operators
1583 * Precompile SQL functions to avoid overhead
1584 * Create utility to compute accurate random_page_cost value
1585 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
1586 * Have EXPLAIN ANALYZE issue NOTICE messages when the estimated and
1587 actual row counts differ by a specified percentage
1588 * Consider using hash buckets to do DISTINCT, rather than sorting
1590 This would be beneficial when there are few distinct values. This is
1591 already used by GROUP BY.
1593 * Log statements where the optimizer row estimates were dramatically
1594 different from the number of rows actually found?
1595 * Consider compressed annealing to search for query plans
1597 This might replace GEQO, http://sixdemonbag.org/Djinni.
1599 * Improve merge join performance by allowing mark/restore of
1602 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00096.php
1604 * Consider using a hash for joining to a large IN (VALUES ...) list
1606 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00450.php
1613 * Consider having the background writer update the transaction status
1614 hint bits before writing out the page
1616 Implementing this requires the background writer to have access to system
1617 catalogs and the transaction status log.
1619 * Consider adding buffers the background writer finds reusable to the
1622 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00781.php
1624 * Automatically tune bgwriter_delay based on activity rather then using a
1627 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00781.php
1629 * Consider wither increasing BM_MAX_USAGE_COUNT improves performance
1631 http://archives.postgresql.org/pgsql-hackers/2007-06/msg01007.php
1633 * Test to see if calling PreallocXlogFiles() from the background writer
1634 will help with WAL segment creation latency
1636 http://archives.postgresql.org/pgsql-patches/2007-06/msg00340.php
1640 Miscellaneous Performance
1641 =========================
1643 * Do async I/O for faster random read-ahead of data
1645 Async I/O allows multiple I/O requests to be sent to the disk with
1646 results coming back asynchronously.
1648 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00820.php
1649 http://archives.postgresql.org/pgsql-performance/2007-09/msg00255.php
1650 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00027.php
1651 http://archives.postgresql.org/pgsql-patches/2008-01/msg00170.php
1653 * Use mmap() rather than SYSV shared memory or to write WAL files?
1655 This would remove the requirement for SYSV SHM but would introduce
1656 portability issues. Anonymous mmap (or mmap to /dev/zero) is required
1657 to prevent I/O overhead.
1659 * Consider mmap()'ing files into a backend?
1661 Doing I/O to large tables would consume a lot of address space or
1662 require frequent mapping/unmapping. Extending the file also causes
1663 mapping problems that might require mapping only individual pages,
1664 leading to thousands of mappings. Another problem is that there is no
1665 way to _prevent_ I/O to disk from the dirty shared buffers so changes
1666 could hit disk before WAL is written.
1668 * Add a script to ask system configuration questions and tune postgresql.conf
1669 * Consider ways of storing rows more compactly on disk
1671 o Reduce the row header size?
1672 o Consider reducing on-disk varlena length from four bytes to
1673 two because a heap row cannot be more than 64k in length
1675 * Consider increasing NUM_CLOG_BUFFERS
1677 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00030.php
1678 http://archives.postgresql.org/pgsql-performance/2007-08/msg00024.php
1680 * Allow user configuration of TOAST thresholds
1682 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00213.php
1683 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00082.php
1685 * Allow configuration of backend priorities via the operating system
1687 Though backend priorities make priority inversion during lock
1688 waits possible, research shows that this is not a huge problem.
1690 http://archives.postgresql.org/pgsql-general/2007-02/msg00493.php
1692 * Experiment with multi-threaded backend better I/O utilization
1694 This would allow a single query to make use of multiple I/O channels
1695 simultaneously. One idea is to create a background reader that can
1696 pre-fetch sequential and index scan pages needed by other backends.
1697 This could be expanded to allow concurrent reads from multiple devices
1698 in a partitioned table.
1700 * Experiment with multi-threaded backend better CPU utilization
1702 This would allow several CPUs to be used for a single query, such as
1703 for sorting or query execution.
1705 * Consider increasing the minimum allowed number of shared buffers
1707 http://archives.postgresql.org/pgsql-bugs/2008-02/msg00157.php
1709 * Expire published xmin for read-only and idle transactions
1711 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00343.php
1713 * Consider if CommandCounterIncrement() can avoid its
1714 AcceptInvalidationMessages() call
1716 http://archives.postgresql.org/pgsql-committers/2007-11/msg00585.php
1718 * Improve performance of shared invalidation queue for multiple CPUs
1720 http://archives.postgresql.org/pgsql-performance/2008-01/msg00023.php
1722 * Consider Cartesian joins when both relations are needed to form an
1723 indexscan qualification for a third relation
1725 http://archives.postgresql.org/pgsql-performance/2007-12/msg00090.php
1727 * Consider not storing a NULL bitmap on disk if all the NULLs are
1730 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00624.php
1731 http://archives.postgresql.org/pgsql-patches/2007-12/msg00109.php
1733 * Sort large UPDATE/DELETEs so it is done in heap order
1735 http://archives.postgresql.org/pgsql-hackers/2008-01/msg01119.php
1737 * Avoid tuple some tuple copying in sort routines
1739 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01206.php
1740 http://archives.postgresql.org/pgsql-patches/2008-02/msg00176.php
1742 * SMP scalability improvements
1744 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00439.php
1745 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00206.php
1746 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00361.php
1748 * Research reducing deTOASTing in more places
1750 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00895.php
1757 * Add use of 'const' for variables in source tree
1758 * Move some things from contrib into main tree
1759 * %Remove warnings created by -Wcast-align
1760 * Move platform-specific ps status display info from ps_status.c to ports
1761 * Add optional CRC checksum to heap and index pages
1762 * Improve documentation to build only interfaces (Marc)
1763 * Remove or relicense modules that are not under the BSD license, if possible
1764 * Acquire lock on a relation before building a relcache entry for it
1765 * Allow cross-compiling by generating the zic database on the target system
1766 * Improve NLS maintenance of libpgport messages linked onto applications
1767 * Clean up casting in contrib/isn
1769 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00245.php
1771 * Use UTF8 encoding for NLS messages so all server encodings can
1773 * Update Bonjour to work with newer cross-platform SDK
1775 http://archives.postgresql.org/pgsql-hackers/2006-09/msg02238.php
1776 http://archives.postgresql.org/pgsql-patches/2006-10/msg00048.php
1778 * Consider detoasting keys before sorting
1779 * Consider GnuTLS if OpenSSL license becomes a problem
1781 http://archives.postgresql.org/pgsql-patches/2006-05/msg00040.php
1782 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01213.php
1784 * Consider changing documentation format from SGML to XML
1786 http://archives.postgresql.org/pgsql-docs/2006-12/msg00152.php
1788 * Consider making NAMEDATALEN more configurable in future releases
1789 * Update our code to handle 64-bit timezone files to match the zic
1790 source code, which now uses them
1791 * Have configure choose integer datetimes by default
1793 http://archives.postgresql.org/pgsql-patches/2007-05/msg00046.php
1795 * Support scoped IPv6 addresses
1797 http://archives.postgresql.org/pgsql-bugs/2007-05/msg00111.php
1799 * Consider allowing 64-bit integers and floats to be passed by value on
1802 Also change 32-bit floats (float4) to be passed by value at the same
1805 * Research use of signals and sleep wake ups
1807 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00003.php
1809 * Add automated check for invalid C++ source code constructs
1811 http://archives.postgresql.org/pgsql-patches/2007-07/msg00056.php
1813 * Consider simplifying how memory context resets handle child contexts
1815 http://archives.postgresql.org/pgsql-patches/2007-08/msg00067.php
1817 * Remove use of MAKE_PTR and MAKE_OFFSET macros
1819 http://archives.postgresql.org/pgsql-general/2007-08/msg01510.php
1821 * Convert single quotes to apostrophes in the PDF documentation
1823 http://archives.postgresql.org/pgsql-docs/2007-12/msg00059.php
1825 * Create three versions of libpgport to simplify client code
1827 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00154.php
1829 * Remove old-style routines for manipulating tuples
1831 http://archives.postgresql.org/pgsql-hackers/2007-10/msg00851.php
1837 o Remove configure.in check for link failure when cause is found
1838 o Remove readdir() errno patch when runtime/mingwex/dirent.c rev
1840 o Remove psql newline patch when we find out why mingw outputs an
1842 o Allow psql to use readline once non-US code pages work with
1844 o Fix problem with shared memory on the Win32 Terminal Server
1845 o Diagnose problem where shared memory can sometimes not be
1846 attached by postmaster children
1848 http://archives.postgresql.org/pgsql-general/2007-08/msg01377.php
1850 o Improve signal handling
1852 http://archives.postgresql.org/pgsql-patches/2005-06/msg00027.php
1854 o Convert MSVC build system to remove most batch files
1856 http://archives.postgresql.org/pgsql-hackers/2007-08/msg00961.php
1858 o Prevent SSL from sending network packets to avoid interference
1859 with Win32 signal emulation
1861 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00455.php
1863 o Support pgxs when using MSVC
1865 o Fix MSVC NLS support, like for to_char()
1867 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00485.php
1868 http://archives.postgresql.org/pgsql-patches/2008-02/msg00038.php
1872 * Wire Protocol Changes
1874 o Allow dynamic character set handling
1875 o Add decoded type, length, precision
1877 o Update clients to use data types, typmod, schema.table.column names
1878 of result sets using new statement protocol
1884 * Add pre-parsing phase that converts non-ISO syntax to supported
1887 This could allow SQL written for other databases to run without
1890 * Allow plug-in modules to emulate features from other databases
1891 * Add features of Oracle-style packages (Pavel)
1893 A package would be a schema with session-local variables,
1894 public/private functions, and initialization functions. It
1895 is also possible to implement these capabilities
1896 in any schema and not use a separate "packages"
1899 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00384.php
1901 * Consider allowing control of upper/lower case folding of unquoted
1904 http://archives.postgresql.org/pgsql-hackers/2004-04/msg00818.php
1905 http://archives.postgresql.org/pgsql-hackers/2006-10/msg01527.php
1907 * Add autonomous transactions
1909 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00893.php
1913 Features We Do _Not_ Want
1914 =========================
1916 * All backends running as threads in a single process (not wanted)
1918 This eliminates the process protection we get from the current setup.
1919 Thread creation is usually the same overhead as process creation on
1920 modern systems, so it seems unwise to use a pure threaded model.
1922 * Optimizer hints (not wanted)
1924 Optimizer hints are used to work around problems in the optimizer. We
1925 would rather have the problems reported and fixed.
1927 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00506.php
1928 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00517.php
1929 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00663.php
1932 Because we support postfix operators, it isn't possible to make AS
1933 optional and continue to use bison.
1934 http://archives.postgresql.org/pgsql-hackers/2003-04/msg00436.php
1936 http://archives.postgresql.org/pgsql-sql/2006-08/msg00164.php
1938 * Embedded server (not wanted)
1940 While PostgreSQL clients runs fine in limited-resource environments, the
1941 server requires multiple processes and a stable pool of resources to
1942 run reliabily and efficiently. Stripping down the PostgreSQL server
1943 to run in the same process address space as the client application
1944 would add too much complexity and failure cases.