3 Current maintainer: Bruce Momjian (bruce@momjian.us)
4 Last updated: Tue Mar 11 16:16:32 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 * Allow statistics last vacuum/analyze execution times to be displayed
51 without requiring stats_row_level to be enabled
52 * Allow log_min_messages to be specified on a per-module basis
54 This would allow administrators to see more detailed information from
55 specific sections of the backend, e.g. checkpoints, autovacuum, etc.
56 Another idea is to allow separate configuration files for each module,
57 or allow arbitrary SET commands to be passed to them.
59 * Simplify ability to create partitioned tables
61 This would allow creation of partitioned tables without requiring
62 creation of triggers or rules for INSERT/UPDATE/DELETE, and constraints
63 for rapid partition selection. Options could include range and hash
66 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00375.php
67 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00151.php
69 * Allow auto-selection of partitioned tables for min/max() operations
70 * Allow more complex user/database default GUC settings
72 Currently ALTER USER and ALTER DATABASE support per-user and
73 per-database defaults. Consider adding per-user-and-database
74 defaults so things like search_path can be defaulted for a
75 specific user connecting to a specific database.
77 * Allow custom variable classes that can restrict who can set the values
79 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00911.php
81 * Implement the SQL standard mechanism whereby REVOKE ROLE revokes only
82 the privilege granted by the invoking role, and not those granted
85 http://archives.postgresql.org/pgsql-bugs/2007-05/msg00010.php
87 * Allow SSL authentication/encryption over unix domain sockets
89 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00924.php
91 * Allow client certificate names to be checked against the client
94 This is already implemented in
95 libpq/fe-secure.c::verify_peer_name_matches_certificate() but the code
100 o Allow pg_hba.conf to specify host names along with IP addresses
102 Host name lookup could occur when the postmaster reads the
103 pg_hba.conf file, or when the backend starts. Another
104 solution would be to reverse lookup the connection IP and
105 check that hostname against the host names in pg_hba.conf.
106 We could also then check that the host name maps to the IP
109 o %Allow postgresql.conf file values to be changed via an SQL
110 API, perhaps using SET GLOBAL
111 o Allow the server to be stopped/restarted via an SQL API
112 o Issue a warning if a change-on-restart-only postgresql.conf value
113 is modified and the server config files are reloaded
114 o Consider normalizing fractions in postgresql.conf, perhaps
117 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00550.php
122 o Allow a database in tablespace t1 with tables created in
123 tablespace t2 to be used as a template for a new database created
124 with default tablespace t2
126 Currently all objects in the default database tablespace must
127 have default tablespace specifications. This is because new
128 databases are created by copying directories. If you mix default
129 tablespace tables and tablespace-specified tables in the same
130 directory, creating a new database from such a mixed directory
131 would create a new database with tables that had incorrect
132 explicit tablespaces. To fix this would require modifying
133 pg_class in the newly copied database, which we don't currently
136 o Allow reporting of which objects are in which tablespaces
138 This item is difficult because a tablespace can contain objects
139 from multiple databases. There is a server-side function that
140 returns the databases which use a specific tablespace, so this
141 requires a tool that will call that function and connect to each
142 database to find the objects in each database for that tablespace.
144 o Allow WAL replay of CREATE TABLESPACE to work when the directory
145 structure on the recovery computer is different from the original
147 o Allow per-tablespace quotas
150 * Point-In-Time Recovery (PITR)
152 o Allow a warm standby system to also allow read-only statements
155 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00050.php
157 o %Create dump tool for write-ahead logs for use in determining
158 transaction id for point-in-time recovery
160 This is useful for checking PITR recovery.
162 o Allow recovery.conf to support the same syntax as
163 postgresql.conf, including quoting
165 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00497.php
171 * Change NUMERIC to enforce the maximum precision
172 * Reduce storage space for small NUMERICs
174 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01331.php
175 http://archives.postgresql.org/pgsql-patches/2007-02/msg00505.php
177 * Fix data types where equality comparison isn't intuitive, e.g. box
178 * Add support for public SYNONYMs
180 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00519.php
182 * Fix CREATE CAST on DOMAINs
184 http://archives.postgresql.org/pgsql-hackers/2006-05/msg00072.php
185 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01681.php
187 * Add support for SQL-standard GENERATED/IDENTITY columns
189 http://archives.postgresql.org/pgsql-hackers/2006-07/msg00543.php
190 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00038.php
191 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00344.php
192 http://archives.postgresql.org/pgsql-patches/2007-05/msg00076.php
193 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00604.php
195 * Improve XML support
197 http://developer.postgresql.org/index.php/XML_Support
199 * Consider placing all sequences in a single table, or create a system
202 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
204 * Allow the UUID type to accept non-standard formats
206 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01214.php
208 * Allow text search dictionary to filter out only stop words
210 http://archives.postgresql.org/pgsql-patches/2007-11/msg00081.php
212 * Consider a function-based API for '@@' full text searches
214 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00511.php
219 o Allow infinite dates and intervals just like infinite timestamps
220 o Merge hardwired timezone names with the TZ database; allow either
221 kind everywhere a TZ name is currently taken
222 o Allow TIMESTAMP WITH TIME ZONE to store the original timezone
223 information, either zone name or offset from UTC [timezone]
225 If the TIMESTAMP value is stored with a time zone name, interval
226 computations should adjust based on the time zone rules.
228 o Fix SELECT '0.01 years'::interval, '0.01 months'::interval
229 o Add a GUC variable to allow output of interval values in ISO8601
231 o Have timestamp subtraction not call justify_hours()?
233 http://archives.postgresql.org/pgsql-sql/2006-10/msg00059.php
235 o Improve timestamptz subtraction to be DST-aware
237 Currently subtracting one date from another that crosses a
238 daylight savings time adjustment can return '1 day 1 hour', but
239 adding that back to the first date returns a time one hour in
240 the future. This is caused by the adjustment of '25 hours' to
241 '1 day 1 hour', and '1 day' is the same time the next day, even
242 if daylight savings adjustments are involved.
244 o Fix interval display to support values exceeding 2^31 hours
245 o Add overflow checking to timestamp and interval arithmetic
246 o Extend timezone code to allow 64-bit values so we can
247 represent years beyond 2038
249 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01363.php
251 o Use LC_TIME for localized weekday/month names, rather than
254 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00390.php
256 o Add ISO INTERVAL handling
258 http://archives.postgresql.org/pgsql-hackers/2006-01/msg00250.php
259 http://archives.postgresql.org/pgsql-bugs/2006-04/msg00248.php
261 o Support ISO INTERVAL syntax if units cannot be determined from
262 the string, and are supplied after the string
264 The SQL standard states that the units after the string
265 specify the units of the string, e.g. INTERVAL '2' MINUTE
266 should return '00:02:00'. The current behavior has the units
267 restrict the interval value to the specified unit or unit
268 range, INTERVAL '70' SECOND returns '00:00:10'.
270 For syntax that isn't uniquely ISO or PG syntax, like '1' or
271 '1:30', treat as ISO if there is a range specification clause,
272 and as PG if there no clause is present, e.g. interpret '1:30'
273 MINUTE TO SECOND as '1 minute 30 seconds', and interpret
274 '1:30' as '1 hour, 30 minutes'.
276 This makes common cases like SELECT INTERVAL '1' MONTH
277 SQL-standard results. The SQL standard supports a limited
278 number of unit combinations and doesn't support unit names in
279 the string. The PostgreSQL syntax is more flexible in the
280 range of units supported, e.g. PostgreSQL supports '1 year 1
281 hour', while the SQL standard does not.
283 o Add support for year-month syntax, INTERVAL '50-6' YEAR
285 o Interpret INTERVAL '1 year' MONTH as CAST (INTERVAL '1
286 year' AS INTERVAL MONTH), and this should return '12 months'
287 o Round or truncate values to the requested precision, e.g.
288 INTERVAL '11 months' AS YEAR should return one or zero
289 o Support precision, CREATE TABLE foo (a INTERVAL MONTH(3))
294 o Delay resolution of array expression's data type so assignment
295 coercion can be performed on empty array expressions
296 o Add support for arrays of domains
298 http://archives.postgresql.org/pgsql-patches/2007-05/msg00114.php
300 o Allow single-byte header storage for array elements
305 o Improve vacuum of large objects, like contrib/vacuumlo?
306 o Add security checking for large objects
307 o Auto-delete large objects when referencing row is deleted
309 contrib/lo offers this functionality.
311 o Allow read/write into TOAST values like large objects
313 This requires the TOAST column to be stored EXTERNAL.
315 o Add API for 64-bit large object access
317 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00781.php
321 * Add locale-aware MONEY type, and support multiple currencies
323 http://archives.postgresql.org/pgsql-general/2005-08/msg01432.php
324 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01181.php
326 * MONEY dumps in a locale-specific format making it difficult to
327 restore to a system with a different locale
328 * Allow MONEY to be easily cast to/from other numeric data types
334 * Allow INET subnet tests using non-constants to be indexed
335 * Allow to_date() and to_timestamp() accept localized month names
336 * Fix to_date()-related functions to consistently issue errors
338 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00915.php
340 * Add missing parameter handling in to_char()
342 http://archives.postgresql.org/pgsql-hackers/2005-12/msg00948.php
344 * Allow substring/replace() to get/set bit values
345 * Allow to_char() on interval values to accumulate the highest unit
348 Some special format flag would be required to request such
349 accumulation. Such functionality could also be added to EXTRACT.
350 Prevent accumulation that crosses the month/day boundary because of
351 the uneven number of days in a month.
353 o to_char(INTERVAL '1 hour 5 minutes', 'MI') => 65
354 o to_char(INTERVAL '43 hours 20 minutes', 'MI' ) => 2600
355 o to_char(INTERVAL '43 hours 20 minutes', 'WK:DD:HR:MI') => 0:1:19:20
356 o to_char(INTERVAL '3 years 5 months','MM') => 41
358 * Implement inlining of set-returning functions defined in SQL
359 * Allow SQL-language functions to return results from RETURNING queries
361 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00665.php
363 * Allow SQL-language functions to reference parameters by parameter name
365 Currently SQL-language functions can only refer to dollar parameters,
368 * Add SPI_gettypmod() to return the typemod for a TupleDesc
369 * Enforce typmod for function inputs, function results and parameters for
370 spi_prepare'd statements called from PLs
372 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01403.php
374 * Allow holdable cursors in SPI
375 * Tighten function permission checks
377 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00568.php
379 * Fix IS OF so it matches the ISO specification, and add documentation
381 http://archives.postgresql.org/pgsql-patches/2003-08/msg00060.php
382 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00060.php
384 * Add missing operators for geometric data types
386 Some geometric types do not have the full suite of geometric operators,
389 * Implement Boyer-Moore searching in strpos()
391 http://archives.postgresql.org/pgsql-patches/2007-08/msg00012.php
393 * Prevent malicious functions from being executed with the permissions
394 of unsuspecting users
396 Index functions are safe, so VACUUM and ANALYZE are safe too.
397 Triggers, CHECK and DEFAULT expressions, and rules are still vulnerable.
398 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00268.php
400 * Reduce memory usage of aggregates in set returning functions
402 http://archives.postgresql.org/pgsql-performance/2008-01/msg00031.php
404 * Add temporal versions of generate_series()
406 http://archives.postgresql.org/pgsql-hackers/2007-04/msg01180.php
410 Multi-Language Support
411 ======================
413 * Add NCHAR (as distinguished from ordinary varchar),
414 * Allow locale to be set at database creation
416 Currently locale can only be set during initdb. No global tables have
417 locale-aware columns. However, the database template used during
418 database creation might have locale-aware indexes. The indexes would
419 need to be reindexed to match the new locale.
421 * Allow encoding on a per-column basis optionally using the ICU library:
423 Right now only one encoding is allowed per database. [locale]
425 http://archives.postgresql.org/pgsql-hackers/2005-03/msg00932.php
426 http://archives.postgresql.org/pgsql-patches/2005-08/msg00309.php
427 http://archives.postgresql.org/pgsql-patches/2006-03/msg00233.php
428 http://archives.postgresql.org/pgsql-hackers/2006-09/msg00662.php
430 * Add CREATE COLLATE? [locale]
431 * Support multiple simultaneous character sets, per SQL92
432 * Improve UTF8 combined character handling?
433 * Add octet_length_server() and octet_length_client()
434 * Make octet_length_client() the same as octet_length()?
435 * Fix problems with wrong runtime encoding conversion for NLS message files
436 * Add URL to more complete multi-byte regression tests
438 http://archives.postgresql.org/pgsql-hackers/2005-07/msg00272.php
440 * Fix ILIKE and regular expressions to handle case insensitivity
441 properly in multibyte encodings
443 http://archives.postgresql.org/pgsql-bugs/2005-10/msg00001.php
444 http://archives.postgresql.org/pgsql-patches/2005-11/msg00173.php
446 * Set client encoding based on the client operating system encoding
448 Currently client_encoding is set in postgresql.conf, which
449 defaults to the server encoding.
450 http://archives.postgresql.org/pgsql-hackers/2006-08/msg01696.php
452 * Change memory allocation for multi-byte functions so memory is
453 allocated inside conversion functions
455 Currently we preallocate memory based on worst-case usage.
462 * Automatically create rules on views so they are updateable, per SQL99
464 We can only auto-create rules for simple views. For more complex
465 cases users will still have to write rules manually.
467 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00586.php
468 http://archives.postgresql.org/pgsql-patches/2006-08/msg00255.php
470 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
471 * Allow NOTIFY in rules involving conditionals
472 * Allow VIEW/RULE recompilation when the underlying tables change
474 Another issue is whether underlying table changes should be reflected
475 in the view, e.g. should SELECT * show additional columns if they
476 are added after the view is created.
478 * Make it possible to use RETURNING together with conditional DO INSTEAD
479 rules, such as for partitioning setups
481 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00577.php
483 * Add the ability to automatically create materialized views
485 Right now materialized views require the user to create triggers on the
486 main table to keep the summary table current. SQL syntax should be able
487 to manager the triggers and summary table automatically. A more
488 sophisticated implementation would automatically retrieve from the
489 summary table when the main table is referenced, if possible.
496 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
497 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
498 * %Allow SET CONSTRAINTS to be qualified by schema/table name
499 * %Add a separate TRUNCATE permission
501 Currently only the owner can TRUNCATE a table because triggers are not
502 called, and the table is locked in exclusive mode.
504 * Allow PREPARE of cursors
505 * Allow finer control over the caching of prepared query plans
507 Currently queries prepared via the libpq API are planned on first
508 execute using the supplied parameters --- allow SQL PREPARE to do the
509 same. Also, allow control over replanning prepared queries either
510 manually or automatically when statistics for execute parameters
511 differ dramatically from those used during planning.
513 * Improve logging of prepared transactions recovered during startup
515 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00092.php
517 * Allow LISTEN/NOTIFY to store info in memory rather than tables?
519 Currently LISTEN/NOTIFY information is stored in pg_listener. Storing
520 such information in memory would improve performance.
522 * Add optional textual message to NOTIFY
524 This would allow an informational message to be added to the notify
525 message, perhaps indicating the row modified or other custom
528 * Allow multiple identical NOTIFY events to always be communicated to the
529 client, rather than sent as a single notification to the listener
531 http://archives.postgresql.org/pgsql-general/2008-01/msg00057.php
533 * Add a GUC variable to warn about non-standard SQL usage in queries
534 * Add SQL-standard MERGE command, typically used to merge two tables
537 This is similar to UPDATE, then for unmatched rows, INSERT.
538 Whether concurrent access allows modifications which could cause
539 row loss is implementation independent.
541 * Add REPLACE or UPSERT command that does UPDATE, or on failure, INSERT
544 To implement this cleanly requires that the table have a unique index
545 so duplicate checking can be easily performed. It is possible to
546 do it without a unique index if we require the user to LOCK the table
549 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00501.php
550 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00536.php
552 * Add NOVICE output level for helpful messages like automatic sequence/index
554 * Add GUC to issue notice about statements that use unjoined tables
555 * Allow EXPLAIN to identify tables that were skipped because of
557 * Allow EXPLAIN output to be more easily processed by scripts, perhaps XML
558 * Enable standard_conforming_strings
559 * Make standard_conforming_strings the default in 8.5?
561 When this is done, backslash-quote should be prohibited in non-E''
562 strings because of possible confusion over how such strings treat
563 backslashes. Basically, '' is always safe for a literal single
564 quote, while \' might or might not be based on the backslash
567 * Simplify dropping roles that have objects in several databases
568 * Allow COMMENT ON to accept an expression rather than just a string
569 * Allow the count returned by SELECT, etc to be represented as an int64
570 to allow a higher range of values
571 * Add SQL99 WITH clause to SELECT
572 * Add SQL:2003 WITH RECURSIVE (hierarchical) queries to SELECT
574 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01375.php
575 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00642.php
576 http://archives.postgresql.org/pgsql-patches/2007-03/msg00139.php
578 * Add DEFAULT .. AS OWNER so permission checks are done as the table
581 This would be useful for SERIAL nextval() calls and CHECK constraints.
583 * Allow DISTINCT to work in multiple-argument aggregate calls
584 * Add column to pg_stat_activity that shows the progress of long-running
585 commands like CREATE INDEX and VACUUM
586 * Implement SQL:2003 window functions
587 * Improve failure message when DROP DATABASE is used on a database that
588 has prepared transactions
589 * Allow INSERT/UPDATE ... RETURNING inside a SELECT 'FROM' clause
591 http://archives.postgresql.org/pgsql-general/2006-09/msg00803.php
592 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00693.php
594 * Increase locking when DROPing objects so dependent objects cannot
595 get dropped while the DROP operation is happening
597 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00937.php
599 * -Allow AS in "SELECT col AS label" to be optional in certain cases
601 * Allow INSERT ... DELETE ... RETURNING, namely allow the DELETE ...
602 RETURNING to supply values to the INSERT
603 http://archives.postgresql.org/pgsql-hackers/2008-02/thrd2.php#00979
605 * Add comments on system tables/columns using the information in
608 Ideally the information would be pulled from the SGML file
614 o Allow CREATE TABLE AS to determine column lengths for complex
615 expressions like SELECT col1 || col2
617 o Have WITH CONSTRAINTS also create constraint indexes
619 http://archives.postgresql.org/pgsql-patches/2007-04/msg00149.php
621 o Have CONSTRAINT cname NOT NULL record the contraint name
623 Right now pg_attribute.attnotnull records the NOT NULL status
624 of the column, but does not record the contraint name
629 o Allow UPDATE tab SET ROW (col, ...) = (SELECT...)
631 http://archives.postgresql.org/pgsql-hackers/2006-07/msg01306.php
632 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00865.php
634 o Research self-referential UPDATEs that see inconsistent row versions
635 in read-committed mode
637 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00507.php
638 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00016.php
640 o Allow GLOBAL temporary tables to exist as empty by default in
643 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00006.php
648 o Have ALTER TABLE RENAME rename SERIAL sequence names
650 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
652 o Have ALTER SEQUENCE RENAME rename the sequence name stored
653 in the sequence table
655 http://archives.postgresql.org/pgsql-bugs/2007-09/msg00092.php
656 http://archives.postgresql.org/pgsql-bugs/2007-10/msg00007.php
657 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
659 o Add ALTER DOMAIN to modify the underlying data type
660 o %Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
662 http://archives.postgresql.org/pgsql-patches/2006-02/msg00168.php
664 o %Allow ALTER TABLE to change constraint deferrability and actions
665 o Add missing object types for ALTER ... SET SCHEMA
666 o Allow ALTER TABLESPACE to move to different directories
667 o Allow databases to be moved to different tablespaces
668 o Allow moving system tables to other tablespaces, where possible
670 Currently non-global system tables must be in the default database
671 tablespace. Global system tables can never be moved.
673 o Prevent parent tables from altering or dropping constraints
674 like CHECK that are inherited by child tables unless CASCADE
676 o %Prevent child tables from altering or dropping constraints
677 like CHECK that were inherited from the parent table
678 o Have ALTER INDEX update the name of a constraint using that index
679 o Add ALTER TABLE RENAME CONSTRAINT, update index name also
680 o Allow column display reordering by recording a display,
681 storage, and permanent id for every column?
683 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00782.php
689 o Automatically maintain clustering on a table
691 This might require some background daemon to maintain clustering
692 during periods of low usage. It might also require tables to be only
693 partially filled for easier reorganization. Another idea would
694 be to create a merged heap/index data file so an index lookup would
695 automatically access the heap data too. A third idea would be to
696 store heap rows in hashed groups, perhaps using a user-supplied
698 http://archives.postgresql.org/pgsql-performance/2004-08/msg00349.php
700 o %Add default clustering to system tables
702 To do this, determine the ideal cluster index for each system
703 table and set the cluster setting during initdb.
705 o %Add VERBOSE option to report tables as they are processed,
711 o Allow COPY to report error lines and continue
713 This requires the use of a savepoint before each COPY line is
714 processed, with ROLLBACK on COPY failure.
716 o Allow COPY on a newly-created table to skip WAL logging
718 On crash recovery, the table involved in the COPY would
719 be removed or have its heap and index files truncated. One
720 issue is that no other backend should be able to add to
721 the table at the same time, which is something that is
722 currently allowed. This currently is done if the table is
723 created inside the same transaction block as the COPY because
724 no other backends can see the table.
726 o Consider using a ring buffer for COPY FROM
728 http://archives.postgresql.org/pgsql-patches/2008-02/msg00140.php
729 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01080.php
731 o Allow COPY FROM to create index entries in bulk
733 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00811.php
739 o Allow column-level privileges
740 o %Allow GRANT/REVOKE permissions to be applied to all schema objects
743 The proposed syntax is:
744 GRANT SELECT ON ALL TABLES IN public TO phpuser;
745 GRANT SELECT ON NEW TABLES IN public TO phpuser;
747 o Allow GRANT/REVOKE permissions to be inherited by objects based on
750 o Allow SERIAL sequences to inherit permissions from the base table?
755 o Prevent DROP TABLE from dropping a row referenced by its own open
761 o Allow INSERT/UPDATE of the system-generated oid value for a row
762 o In rules, allow VALUES() to contain a mixture of 'old' and 'new'
768 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
773 Referential Integrity
774 =====================
776 * Add MATCH PARTIAL referential integrity
777 * Change foreign key constraint for array -> element to mean element
779 * Fix problem when cascading referential triggers make changes on
780 cascaded tables, seeing the tables in an intermediate state
782 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00174.php
783 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00174.php
785 * Allow DEFERRABLE and end-of-statement UNIQUE constraints?
787 This would allow UPDATE tab SET col = col + 1 to work if col has
788 a unique index. Currently, uniqueness checks are done while the
789 command is being executed, rather than at the end of the statement
791 http://people.planetpostgresql.org/greg/index.php?/archives/2006/06/10.html
792 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01458.php
794 * Optimize referential integrity checks
796 http://archives.postgresql.org/pgsql-performance/2005-10/msg00458.php
797 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00744.php
800 Server-Side Languages
801 =====================
804 o Fix RENAME to work on variables other than OLD/NEW
806 http://archives.postgresql.org/pgsql-hackers/2002-03/msg00591.php
807 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01615.php
808 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01587.php
810 o Allow function parameters to be passed by name,
811 get_employee_salary(12345 AS emp_id, 2001 AS tax_year)
812 o Allow handling of %TYPE arrays, e.g. tab.col%TYPE[]
813 o Allow listing of record column names, and access to
814 record columns via variables, e.g. columns := r.(*),
817 http://archives.postgresql.org/pgsql-patches/2005-07/msg00458.php
818 http://archives.postgresql.org/pgsql-patches/2006-05/msg00302.php
819 http://archives.postgresql.org/pgsql-patches/2006-06/msg00031.php
821 o Add support for SCROLL cursors
822 o Add support for WITH HOLD cursors
823 o Allow row and record variables to be set to NULL constants,
824 and allow NULL tests on such variables
826 Because a row is not scalar, do not allow assignment
827 from NULL-valued scalars.
829 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00070.php
831 o Review handling of MOVE and FETCH
833 http://archives.postgresql.org/pgsql-patches/2007-04/msg00527.php
837 o Add table function support to pltcl, plpythonu
838 o Add support for polymorphic arguments and return types to
839 languages other than PL/PgSQL
840 o Add capability to create and call PROCEDURES
841 o Add support for OUT and INOUT parameters to languages other
843 o Add PL/PythonU tracebacks
845 http://archives.postgresql.org/pgsql-patches/2006-02/msg00288.php
847 o Allow data to be passed in native language formats, rather
850 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00289.php
857 * Have pg_ctl look at PGHOST in case it is a socket directory?
858 * Allow pg_ctl to work properly with configuration files located outside
861 pg_ctl can not read the pid file because it isn't located in the
862 config directory but in the PGDATA directory. The solution is to
863 allow pg_ctl to read and understand postgresql.conf to find the
864 data_directory value.
866 * Add a function like pg_get_indexdef() that report more detailed index
869 http://archives.postgresql.org/pgsql-bugs/2007-12/msg00166.php
871 * Prefix command-line utilities like createuser with 'pg_'
873 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00025.php
877 o Have psql show current values for a sequence
878 o Move psql backslash database information into the backend, use
879 mnemonic commands? [psql]
881 This would allow non-psql clients to pull the same information out
882 of the database as psql.
884 o Fix psql's \d commands more consistent
886 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
887 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
889 o Consistently display privilege information for all objects in psql
890 o Add auto-expanded mode so expanded output is used if the row
891 length is wider than the screen width.
893 Consider using auto-expanded mode for backslash commands like \df+.
895 o Prevent tab completion of SET TRANSACTION from querying the
896 database and therefore preventing the transaction isolation
897 level from being set.
899 Currently SET <tab> causes a database lookup to check all
900 supported session variables. This query causes problems
901 because setting the transaction isolation level must be the
902 first statement of a transaction.
904 o Add a \set variable to control whether \s displays line numbers
906 Another option is to add \# which lists line numbers, and
907 allows command execution.
909 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00255.php
911 o Prevent escape string warnings when object names have
914 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00227.php
916 o Have \d show foreign keys that reference a table's primary key
918 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00424.php
920 o Have \d show child tables that inherit from the specified parent
921 o Have \l+ show database size, if permissions allow
923 Ideally it will not generate an error for invalid permissions
927 * pg_dump / pg_restore
928 o %Add dumping of comments on index columns and composite type columns
929 o %Add full object name to the tag field. eg. for operators we need
930 '=(integer, integer)', instead of just '='.
931 o Add pg_dumpall custom format dumps?
932 o Allow selection of individual object(s) of all types, not just
934 o In a selective dump, allow dumping of an object and all its
936 o Add options like pg_restore -l and -L to pg_dump
937 o Stop dumping CASCADE on DROP TYPE commands in clean mode
938 o Allow pg_dump --clean to drop roles that own objects or have
940 o Change pg_dump so that a comment on the dumped database is
941 applied to the loaded database, even if the database has a
942 different name. This will require new backend syntax, perhaps
943 COMMENT ON CURRENT DATABASE.
944 o Remove unnecessary function pointer abstractions in pg_dump source
946 o Allow pg_dump to utilize multiple CPUs and I/O channels by dumping
947 multiple objects simultaneously
949 The difficulty with this is getting multiple dump processes to
950 produce a single dump output file. It also would require
951 several sessions to share the same snapshot.
952 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00205.php
954 o Allow pg_restore to utilize multiple CPUs and I/O channels by
955 restoring multiple objects simultaneously
957 This might require a pg_restore flag to indicate how many
958 simultaneous operations should be performed. Only pg_dump's
959 -Fc format has the necessary dependency information.
960 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00963.php
962 o To better utilize resources, allow pg_restore to check foreign
963 keys simultaneously, where possible
964 o Allow pg_restore to create all indexes of a table
965 concurrently, via a single heap scan
967 This requires a pg_dump -Fc file because that format contains
968 the required dependency information.
969 http://archives.postgresql.org/pgsql-general/2007-05/msg01274.php
971 o Allow pg_restore to load different parts of the COPY data
973 o Prevent pg_dump/pg_restore from being affected by
976 Using psql to restore a pg_dump dump is also affected.
982 Document differences between ecpg and the SQL standard and
983 information about the Informix-compatibility module.
985 o Solve cardinality > 1 for input descriptors / variables?
986 o Add a semantic check level, e.g. check if a table really exists
987 o fix handling of DB attributes that are arrays
988 o Use backend PREPARE/EXECUTE facility for ecpg where possible
990 o Fix nested C comments
991 o %sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
992 o Make SET CONNECTION thread-aware, non-standard?
993 o Allow multidimensional arrays
994 o Add internationalized message strings
995 o Implement COPY FROM STDIN
999 o Add PQescapeIdentifierConn()
1000 o Prevent PQfnumber() from lowercasing unquoted the column name
1002 PQfnumber() should never have been doing lowercasing, but
1003 historically it has so we need a way to prevent it
1005 o Allow statement results to be automatically batched to the client
1007 Currently all statement results are transferred to the libpq
1008 client before libpq makes the results available to the
1009 application. This feature would allow the application to make
1010 use of the first result rows while the rest are transferred, or
1011 held on the server waiting for them to be requested by libpq.
1012 One complexity is that a statement like SELECT 1/col could error
1013 out mid-way through the result set.
1015 o Consider disallowing multiple queries in PQexec() as an
1016 additional barrier to SQL injection attacks
1018 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00184.php
1020 o Add PQexecf() that allows complex parameter substitution
1022 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01803.php
1024 o Add SQLSTATE severity to PGconn return status
1026 http://archives.postgresql.org/pgsql-interfaces/2007-11/msg00015.php
1032 * Add deferred trigger queue file
1034 Right now all deferred trigger information is stored in backend
1035 memory. This could exhaust memory for very large trigger queues.
1036 This item involves dumping large queues into files.
1038 * Allow triggers to be disabled in only the current session.
1040 This is currently possible by starting a multi-statement transaction,
1041 modifying the system tables, performing the desired SQL, restoring the
1042 system tables, and committing the transaction. ALTER TABLE ...
1043 TRIGGER requires a table lock so it is not ideal for this usage.
1045 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
1047 If the dump is known to be valid, allow foreign keys to be added
1048 without revalidating the data.
1050 * Allow statement-level triggers to access modified rows
1051 * Support triggers on columns
1053 http://archives.postgresql.org/pgsql-patches/2005-07/msg00107.php
1055 * Allow AFTER triggers on system tables
1057 System tables are modified in many places in the backend without going
1058 through the executor and therefore not causing triggers to fire. To
1059 complete this item, the functions that modify system tables will have
1062 * Tighten trigger permission checks
1064 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00564.php
1066 * Allow BEFORE INSERT triggers on views
1068 http://archives.postgresql.org/pgsql-general/2007-02/msg01466.php
1070 * Add ability to trigger on TRUNCATE
1072 http://archives.postgresql.org/pgsql-sql/2008-01/msg00050.php
1079 * Add UNIQUE capability to non-btree indexes
1080 * Prevent index uniqueness checks when UPDATE does not modify the column
1082 Uniqueness (index) checks are done when updating a column even if the
1083 column is not modified by the UPDATE.
1085 * Allow the creation of on-disk bitmap indexes which can be quickly
1086 combined with other bitmap indexes
1088 Such indexes could be more compact if there are only a few distinct values.
1089 Such indexes can also be compressed. Keeping such indexes updated can be
1092 http://archives.postgresql.org/pgsql-patches/2005-07/msg00512.php
1093 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01107.php
1095 * Allow accurate statistics to be collected on indexes with more than
1096 one column or expression indexes, perhaps using per-index statistics
1098 http://archives.postgresql.org/pgsql-performance/2006-10/msg00222.php
1099 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01131.php
1101 * Consider increasing the number of default statistics target, and
1102 reduce statistics target overhead
1104 Also consider having a larger statistics target for indexed columns
1105 and expression indexes
1106 http://archives.postgresql.org/pgsql-general/2007-05/msg01228.php
1107 http://archives.postgresql.org/pgsql-general/2007-06/msg00542.php
1109 * Consider compressing indexes by storing key values duplicated in
1110 several rows as a single index entry
1112 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00341.php
1113 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01264.php
1114 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00465.php
1115 http://archives.postgresql.org/pgsql-patches/2007-03/msg00163.php
1117 * Add REINDEX CONCURRENTLY, like CREATE INDEX CONCURRENTLY
1119 This is difficult because you must upgrade to an exclusive table lock
1120 to replace the existing index file. CREATE INDEX CONCURRENTLY does not
1121 have this complication. This would allow index compaction without
1124 * Allow multiple indexes to be created concurrently, ideally via a
1125 single heap scan, and have pg_restore use it
1131 o Allow inherited tables to inherit indexes, UNIQUE constraints,
1132 and primary/foreign keys
1133 o Honor UNIQUE INDEX on base column in INSERTs/UPDATEs
1134 on inherited table, e.g. INSERT INTO inherit_table
1135 (unique_index_col) VALUES (dup) should fail
1137 The main difficulty with this item is the problem of
1138 creating an index that can span multiple tables.
1140 o Allow SELECT ... FOR UPDATE on inherited tables
1141 o Require all CHECK constraints to be inherited
1143 http://archives.postgresql.org/pgsql-bugs/2007-04/msg00026.php
1148 o Add more GIST index support for geometric data types
1149 o Allow GIST indexes to create certain complex index types, like
1150 digital trees (see Aoki)
1155 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00051.php
1157 o Pack hash index buckets onto disk pages more efficiently
1159 Currently only one hash bucket can be stored on a page. Ideally
1160 several hash buckets could be stored on a single page and greater
1161 granularity used for the hash algorithm.
1163 http://archives.postgresql.org/pgsql-hackers/2004-06/msg00168.php
1165 o Consider sorting hash buckets so entries can be found using a
1166 binary search, rather than a linear scan
1167 o In hash indexes, consider storing the hash value with or instead
1169 o Add WAL logging for crash recovery
1170 o Allow multi-column hash indexes
1171 o During index creation, pre-sort the tuples to improve build speed
1173 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01199.php
1180 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
1182 Ideally this requires a separate test program that can be run
1183 at initdb time or optionally later. Consider O_SYNC when
1186 * Add program to test if fsync has a delay compared to non-fsync
1187 * Consider sorting writes during checkpoint
1189 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00541.php
1198 We could use a fixed row count and a +/- count to follow MVCC
1199 visibility rules, or a single cached value could be used and
1200 invalidated if anyone modifies the table. Another idea is to
1201 get a count directly from a unique index, but for this to be
1202 faster than a sequential scan it must avoid access to the heap
1203 to obtain tuple visibility information.
1205 * Provide a way to calculate an "estimated COUNT(*)"
1207 Perhaps by using the optimizer's cardinality estimates or random
1210 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00943.php
1212 * Allow data to be pulled directly from indexes
1214 Currently indexes do not have enough tuple visibility information
1215 to allow data to be pulled from the index without also accessing
1216 the heap. One way to allow this is to set a bit on index tuples
1217 to indicate if a tuple is currently visible to all transactions
1218 when the first valid heap lookup happens. This bit would have to
1219 be cleared when a heap tuple is expired.
1221 Another idea is to maintain a bitmap of heap pages where all rows
1222 are visible to all backends, and allow index lookups to reference
1223 that bitmap to avoid heap lookups, perhaps the same bitmap we might
1224 add someday to determine which heap pages need vacuuming. Frequently
1225 accessed bitmaps would have to be stored in shared memory. One 8k
1226 page of bitmaps could track 512MB of heap pages.
1228 A third idea would be for a heap scan to check if all rows are visible
1229 and if so set a per-table flag which can be checked by index scans.
1230 Any change to the table would have to clear the flag. To detect
1231 changes during the heap scan a counter could be set at the start and
1232 checked at the end --- if it is the same, the table has not been
1233 modified --- any table change would increment the counter.
1235 * Consider automatic caching of statements at various levels:
1238 o Query execute plan
1241 * Consider increasing internal areas when shared buffers is increased
1243 http://archives.postgresql.org/pgsql-hackers/2005-10/msg01419.php
1245 * Consider decreasing the amount of memory used by PrivateRefCount
1247 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00797.php
1248 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00752.php
1250 * Consider allowing higher priority queries to have referenced buffer
1251 cache pages stay in memory longer
1253 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00562.php
1260 * Improve speed with indexes
1262 For large table adjustments during VACUUM FULL, it is faster to cluster
1263 or reindex rather than update the index. Also, index updates can bloat
1266 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00024.php
1267 http://archives.postgresql.org/pgsql-performance/2007-05/msg00296.php
1269 * Auto-fill the free space map by scanning the buffer cache or by
1270 checking pages written by the background writer
1272 http://archives.postgresql.org/pgsql-hackers/2006-02/msg01125.php
1273 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00011.php
1275 * Create a bitmap of pages that need vacuuming
1277 Instead of sequentially scanning the entire table, have the background
1278 writer or some other process record pages that have expired rows, then
1279 VACUUM can look at just those pages rather than the entire table. In
1280 the event of a system crash, the bitmap would probably be invalidated.
1281 One complexity is that index entries still have to be vacuumed, and
1282 doing this without an index scan (by using the heap values to find the
1283 index entry) might be slow and unreliable, especially for user-defined
1286 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01188.php
1287 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00121.php
1289 * Allow FSM to return free space toward the beginning of the heap file,
1290 in hopes that empty pages at the end can be truncated by VACUUM
1291 * Allow FSM page return free space based on table clustering, to assist
1292 in maintaining clustering?
1293 * Improve dead row detection during multi-statement transactions usage
1295 http://archives.postgresql.org/pgsql-patches/2007-03/msg00358.php
1297 * Consider a more compact data representation for dead tuples
1299 http://archives.postgresql.org/pgsql-patches/2007-05/msg00143.php
1304 o %Issue log message to suggest VACUUM FULL if a table is nearly
1306 o Improve control of auto-vacuum
1308 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00876.php
1310 o Prevent long-lived temporary tables from causing frozen-xid
1311 advancement starvation
1313 The problem is that autovacuum cannot vacuum them to set frozen xids;
1314 only the session that created them can do that.
1315 http://archives.postgresql.org/pgsql-general/2007-06/msg01645.php
1317 o Store per-table autovacuum settings in pg_class.reloptions.
1319 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01440.php
1320 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00724.php
1322 o Prevent autovacuum from running if an old transaction is still
1323 running from the last vacuum
1325 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00899.php
1332 * Fix priority ordering of read and write light-weight locks (Neil)
1334 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00893.php
1335 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00905.php
1337 * Fix problem when multiple subtransactions of the same outer transaction
1338 hold different types of locks, and one subtransaction aborts
1340 http://archives.postgresql.org/pgsql-hackers/2006-11/msg01011.php
1341 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00001.php
1342 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00435.php
1343 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00773.php
1345 * Allow UPDATEs on only non-referential integrity columns not to conflict
1346 with referential integrity locks
1348 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00073.php
1350 * Add idle_in_transaction_timeout GUC so locks are not held for long
1353 * Improve deadlock detection when a page cleaning lock conflicts
1354 with a shared buffer that is pinned
1356 http://archives.postgresql.org/pgsql-bugs/2008-01/msg00138.php
1357 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00873.php
1358 http://archives.postgresql.org/pgsql-committers/2008-01/msg00365.php
1363 Startup Time Improvements
1364 =========================
1366 * Experiment with multi-threaded backend for backend creation [thread]
1368 This would prevent the overhead associated with process creation. Most
1369 operating systems have trivial process creation time compared to
1370 database startup overhead, but a few operating systems (Win32,
1371 Solaris) might benefit from threading. Also explore the idea of
1372 a single session using multiple threads to execute a statement faster.
1379 * Eliminate need to write full pages to WAL before page modification [wal]
1381 Currently, to protect against partial disk page writes, we write
1382 full page images to WAL before they are modified so we can correct any
1383 partial page writes during recovery. These pages can also be
1384 eliminated from point-in-time archive files.
1386 o When off, write CRC to WAL and check file system blocks
1389 If CRC check fails during recovery, remember the page in case
1390 a later CRC for that page properly matches.
1392 o Write full pages during file system write and not when
1393 the page is modified in the buffer cache
1395 This allows most full page writes to happen in the background
1396 writer. It might cause problems for applying WAL on recovery
1397 into a partially-written page, but later the full page will be
1400 * Allow WAL traffic to be streamed to another server for stand-by
1402 * Reduce WAL traffic so only modified values are written rather than
1405 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01589.php
1407 * Allow WAL information to recover corrupted pg_controldata
1409 http://archives.postgresql.org/pgsql-patches/2006-06/msg00025.php
1411 * Find a way to reduce rotational delay when repeatedly writing
1414 Currently fsync of WAL requires the disk platter to perform a full
1415 rotation to fsync again. One idea is to write the WAL to different
1416 offsets that might reduce the rotational delay.
1418 * Allow WAL logging to be turned off for a table, but the table
1419 might be dropped or truncated during crash recovery [walcontrol]
1421 Allow tables to bypass WAL writes and just fsync() dirty pages on
1422 commit. This should be implemented using ALTER TABLE, e.g. ALTER
1423 TABLE PERSISTENCE [ DROP | TRUNCATE | DEFAULT ]. Tables using
1424 non-default logging should not use referential integrity with
1425 default-logging tables. A table without dirty buffers during a
1426 crash could perhaps avoid the drop/truncate.
1428 * Allow WAL logging to be turned off for a table, but the table would
1429 avoid being truncated/dropped [walcontrol]
1431 To do this, only a single writer can modify the table, and writes
1432 must happen only on new pages so the new pages can be removed during
1433 crash recovery. Readers can continue accessing the table. Such
1434 tables probably cannot have indexes. One complexity is the handling
1435 of indexes on TOAST tables.
1437 * Speed WAL recovery by allowing more than one page to be prefetched
1439 This involves having a separate process that can be told which pages
1440 the recovery process will need in the near future.
1441 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01279.php
1445 Optimizer / Executor
1446 ====================
1448 * Improve selectivity functions for geometric operators
1449 * Precompile SQL functions to avoid overhead
1450 * Create utility to compute accurate random_page_cost value
1451 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
1452 * Have EXPLAIN ANALYZE issue NOTICE messages when the estimated and
1453 actual row counts differ by a specified percentage
1454 * Consider using hash buckets to do DISTINCT, rather than sorting
1456 This would be beneficial when there are few distinct values. This is
1457 already used by GROUP BY.
1459 * Log statements where the optimizer row estimates were dramatically
1460 different from the number of rows actually found?
1461 * Consider compressed annealing to search for query plans
1463 This might replace GEQO, http://sixdemonbag.org/Djinni.
1465 * Improve merge join performance by allowing mark/restore of
1468 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00096.php
1470 * Consider using a hash for joining to a large IN (VALUES ...) list
1472 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00450.php
1476 Miscellaneous Performance
1477 =========================
1479 * Do async I/O for faster random read-ahead of data
1481 Async I/O allows multiple I/O requests to be sent to the disk with
1482 results coming back asynchronously.
1484 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00820.php
1485 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00027.php
1487 * Use mmap() rather than SYSV shared memory or to write WAL files?
1489 This would remove the requirement for SYSV SHM but would introduce
1490 portability issues. Anonymous mmap (or mmap to /dev/zero) is required
1491 to prevent I/O overhead.
1493 * Consider mmap()'ing files into a backend?
1495 Doing I/O to large tables would consume a lot of address space or
1496 require frequent mapping/unmapping. Extending the file also causes
1497 mapping problems that might require mapping only individual pages,
1498 leading to thousands of mappings. Another problem is that there is no
1499 way to _prevent_ I/O to disk from the dirty shared buffers so changes
1500 could hit disk before WAL is written.
1502 * Add a script to ask system configuration questions and tune postgresql.conf
1503 * Consider ways of storing rows more compactly on disk
1505 o Reduce the row header size?
1506 o Consider reducing on-disk varlena length from four bytes to
1507 two because a heap row cannot be more than 64k in length
1509 * Consider increasing NUM_CLOG_BUFFERS
1510 * Consider having the background writer update the transaction status
1511 hint bits before writing out the page
1513 Implementing this requires the background writer to have access to system
1514 catalogs and the transaction status log.
1516 * Consider adding buffers the BGW finds reusable to the free list
1518 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00781.php
1520 * Automatically tune bgwriter_delay based on activity rather then using a
1523 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00781.php
1525 * Allow user configuration of TOAST thresholds
1527 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00213.php
1529 * Allow configuration of backend priorities via the operating system
1531 Though backend priorities make priority inversion during lock
1532 waits possible, research shows that this is not a huge problem.
1534 http://archives.postgresql.org/pgsql-general/2007-02/msg00493.php
1536 * Experiment with multi-threaded backend better resource utilization
1538 This would allow a single query to make use of multiple CPU's or
1539 multiple I/O channels simultaneously. One idea is to create a
1540 background reader that can pre-fetch sequential and index scan
1541 pages needed by other backends. This could be expanded to allow
1542 concurrent reads from multiple devices in a partitioned table.
1544 * Consider increasing the minimum allowed number of shared buffers
1546 http://archives.postgresql.org/pgsql-bugs/2008-02/msg00157.php
1554 * Add use of 'const' for variables in source tree
1555 * Move some things from contrib into main tree
1556 * %Remove warnings created by -Wcast-align
1557 * Move platform-specific ps status display info from ps_status.c to ports
1558 * Add optional CRC checksum to heap and index pages
1559 * Improve documentation to build only interfaces (Marc)
1560 * Remove or relicense modules that are not under the BSD license, if possible
1561 * Acquire lock on a relation before building a relcache entry for it
1562 * Allow cross-compiling by generating the zic database on the target system
1563 * Improve NLS maintenance of libpgport messages linked onto applications
1564 * Clean up casting in contrib/isn
1566 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00245.php
1568 * Use UTF8 encoding for NLS messages so all server encodings can
1570 * Update Bonjour to work with newer cross-platform SDK
1572 http://archives.postgresql.org/pgsql-hackers/2006-09/msg02238.php
1573 http://archives.postgresql.org/pgsql-patches/2006-10/msg00048.php
1575 * Consider detoasting keys before sorting
1576 * Consider GnuTLS if OpenSSL license becomes a problem
1578 http://archives.postgresql.org/pgsql-patches/2006-05/msg00040.php
1579 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01213.php
1581 * Consider changing documentation format from SGML to XML
1583 http://archives.postgresql.org/pgsql-docs/2006-12/msg00152.php
1585 * Consider making NAMEDATALEN more configurable in future releases
1586 * Update our code to handle 64-bit timezone files to match the zic
1587 source code, which now uses them
1588 * Have configure choose integer datetimes by default
1590 http://archives.postgresql.org/pgsql-patches/2007-05/msg00046.php
1592 * Support scoped IPv6 addresses
1594 http://archives.postgresql.org/pgsql-bugs/2007-05/msg00111.php
1596 * Consider allowing 64-bit integers and floats to be passed by value on
1599 Also change 32-bit floats (float4) to be passed by value at the same
1606 o Remove configure.in check for link failure when cause is found
1607 o Remove readdir() errno patch when runtime/mingwex/dirent.c rev
1609 o Remove psql newline patch when we find out why mingw outputs an
1611 o Allow psql to use readline once non-US code pages work with
1613 o Fix problem with shared memory on the Win32 Terminal Server
1614 o Improve signal handling
1616 http://archives.postgresql.org/pgsql-patches/2005-06/msg00027.php
1620 * Wire Protocol Changes
1622 o Allow dynamic character set handling
1623 o Add decoded type, length, precision
1625 o Update clients to use data types, typmod, schema.table.column names
1626 of result sets using new statement protocol
1632 * Add pre-parsing phase that converts non-ISO syntax to supported
1635 This could allow SQL written for other databases to run without
1638 * Allow plug-in modules to emulate features from other databases
1639 * Add features of Oracle-style packages (Pavel)
1641 A package would be a schema with session-local variables,
1642 public/private functions, and initialization functions. It
1643 is also possible to implement these capabilities
1644 in any schema and not use a separate "packages"
1647 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00384.php
1649 * Consider allowing control of upper/lower case folding of unquoted
1652 http://archives.postgresql.org/pgsql-hackers/2004-04/msg00818.php
1653 http://archives.postgresql.org/pgsql-hackers/2006-10/msg01527.php
1655 * Add autonomous transactions
1657 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00893.php
1661 Features We Do _Not_ Want
1662 =========================
1664 * All backends running as threads in a single process (not wanted)
1666 This eliminates the process protection we get from the current setup.
1667 Thread creation is usually the same overhead as process creation on
1668 modern systems, so it seems unwise to use a pure threaded model.
1670 * Optimizer hints (not wanted)
1672 Optimizer hints are used to work around problems in the optimizer. We
1673 would rather have the problems reported and fixed.
1675 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00506.php
1676 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00517.php
1677 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00663.php
1680 Because we support postfix operators, it isn't possible to make AS
1681 optional and continue to use bison.
1682 http://archives.postgresql.org/pgsql-hackers/2003-04/msg00436.php
1684 http://archives.postgresql.org/pgsql-sql/2006-08/msg00164.php
1686 * Embedded server (not wanted)
1688 While PostgreSQL clients runs fine in limited-resource environments, the
1689 server requires multiple processes and a stable pool of resources to
1690 run reliabily and efficiently. Stripping down the PostgreSQL server
1691 to run in the same process address space as the client application
1692 would add too much complexity and failure cases.