3 Current maintainer: Bruce Momjian (bruce@momjian.us)
4 Last updated: Tue Mar 11 15:48:13 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
118 o Allow a database in tablespace t1 with tables created in
119 tablespace t2 to be used as a template for a new database created
120 with default tablespace t2
122 Currently all objects in the default database tablespace must
123 have default tablespace specifications. This is because new
124 databases are created by copying directories. If you mix default
125 tablespace tables and tablespace-specified tables in the same
126 directory, creating a new database from such a mixed directory
127 would create a new database with tables that had incorrect
128 explicit tablespaces. To fix this would require modifying
129 pg_class in the newly copied database, which we don't currently
132 o Allow reporting of which objects are in which tablespaces
134 This item is difficult because a tablespace can contain objects
135 from multiple databases. There is a server-side function that
136 returns the databases which use a specific tablespace, so this
137 requires a tool that will call that function and connect to each
138 database to find the objects in each database for that tablespace.
140 o Allow WAL replay of CREATE TABLESPACE to work when the directory
141 structure on the recovery computer is different from the original
143 o Allow per-tablespace quotas
146 * Point-In-Time Recovery (PITR)
148 o Allow a warm standby system to also allow read-only statements
151 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00050.php
153 o %Create dump tool for write-ahead logs for use in determining
154 transaction id for point-in-time recovery
156 This is useful for checking PITR recovery.
158 o Allow recovery.conf to support the same syntax as
159 postgresql.conf, including quoting
161 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00497.php
167 * Change NUMERIC to enforce the maximum precision
168 * Reduce storage space for small NUMERICs
170 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01331.php
171 http://archives.postgresql.org/pgsql-patches/2007-02/msg00505.php
173 * Fix data types where equality comparison isn't intuitive, e.g. box
174 * Add support for public SYNONYMs
176 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00519.php
178 * Fix CREATE CAST on DOMAINs
180 http://archives.postgresql.org/pgsql-hackers/2006-05/msg00072.php
181 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01681.php
183 * Add support for SQL-standard GENERATED/IDENTITY columns
185 http://archives.postgresql.org/pgsql-hackers/2006-07/msg00543.php
186 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00038.php
187 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00344.php
188 http://archives.postgresql.org/pgsql-patches/2007-05/msg00076.php
189 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00604.php
191 * Improve XML support
193 http://developer.postgresql.org/index.php/XML_Support
195 * Consider placing all sequences in a single table, or create a system
198 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
200 * Allow the UUID type to accept non-standard formats
202 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01214.php
204 * Allow text search dictionary to filter out only stop words
206 http://archives.postgresql.org/pgsql-patches/2007-11/msg00081.php
208 * Consider a function-based API for '@@' full text searches
210 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00511.php
215 o Allow infinite dates and intervals just like infinite timestamps
216 o Merge hardwired timezone names with the TZ database; allow either
217 kind everywhere a TZ name is currently taken
218 o Allow TIMESTAMP WITH TIME ZONE to store the original timezone
219 information, either zone name or offset from UTC [timezone]
221 If the TIMESTAMP value is stored with a time zone name, interval
222 computations should adjust based on the time zone rules.
224 o Fix SELECT '0.01 years'::interval, '0.01 months'::interval
225 o Add a GUC variable to allow output of interval values in ISO8601
227 o Have timestamp subtraction not call justify_hours()?
229 http://archives.postgresql.org/pgsql-sql/2006-10/msg00059.php
231 o Improve timestamptz subtraction to be DST-aware
233 Currently subtracting one date from another that crosses a
234 daylight savings time adjustment can return '1 day 1 hour', but
235 adding that back to the first date returns a time one hour in
236 the future. This is caused by the adjustment of '25 hours' to
237 '1 day 1 hour', and '1 day' is the same time the next day, even
238 if daylight savings adjustments are involved.
240 o Fix interval display to support values exceeding 2^31 hours
241 o Add overflow checking to timestamp and interval arithmetic
242 o Extend timezone code to allow 64-bit values so we can
243 represent years beyond 2038
245 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01363.php
247 o Use LC_TIME for localized weekday/month names, rather than
250 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00390.php
252 o Add ISO INTERVAL handling
254 http://archives.postgresql.org/pgsql-hackers/2006-01/msg00250.php
255 http://archives.postgresql.org/pgsql-bugs/2006-04/msg00248.php
257 o Support ISO INTERVAL syntax if units cannot be determined from
258 the string, and are supplied after the string
260 The SQL standard states that the units after the string
261 specify the units of the string, e.g. INTERVAL '2' MINUTE
262 should return '00:02:00'. The current behavior has the units
263 restrict the interval value to the specified unit or unit
264 range, INTERVAL '70' SECOND returns '00:00:10'.
266 For syntax that isn't uniquely ISO or PG syntax, like '1' or
267 '1:30', treat as ISO if there is a range specification clause,
268 and as PG if there no clause is present, e.g. interpret '1:30'
269 MINUTE TO SECOND as '1 minute 30 seconds', and interpret
270 '1:30' as '1 hour, 30 minutes'.
272 This makes common cases like SELECT INTERVAL '1' MONTH
273 SQL-standard results. The SQL standard supports a limited
274 number of unit combinations and doesn't support unit names in
275 the string. The PostgreSQL syntax is more flexible in the
276 range of units supported, e.g. PostgreSQL supports '1 year 1
277 hour', while the SQL standard does not.
279 o Add support for year-month syntax, INTERVAL '50-6' YEAR
281 o Interpret INTERVAL '1 year' MONTH as CAST (INTERVAL '1
282 year' AS INTERVAL MONTH), and this should return '12 months'
283 o Round or truncate values to the requested precision, e.g.
284 INTERVAL '11 months' AS YEAR should return one or zero
285 o Support precision, CREATE TABLE foo (a INTERVAL MONTH(3))
290 o Delay resolution of array expression's data type so assignment
291 coercion can be performed on empty array expressions
292 o Add support for arrays of domains
294 http://archives.postgresql.org/pgsql-patches/2007-05/msg00114.php
296 o Allow single-byte header storage for array elements
301 o Improve vacuum of large objects, like contrib/vacuumlo?
302 o Add security checking for large objects
303 o Auto-delete large objects when referencing row is deleted
305 contrib/lo offers this functionality.
307 o Allow read/write into TOAST values like large objects
309 This requires the TOAST column to be stored EXTERNAL.
311 o Add API for 64-bit large object access
313 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00781.php
317 * Add locale-aware MONEY type, and support multiple currencies
319 http://archives.postgresql.org/pgsql-general/2005-08/msg01432.php
320 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01181.php
322 * MONEY dumps in a locale-specific format making it difficult to
323 restore to a system with a different locale
324 * Allow MONEY to be easily cast to/from other numeric data types
330 * Allow INET subnet tests using non-constants to be indexed
331 * Allow to_date() and to_timestamp() accept localized month names
332 * Fix to_date()-related functions to consistently issue errors
334 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00915.php
336 * Add missing parameter handling in to_char()
338 http://archives.postgresql.org/pgsql-hackers/2005-12/msg00948.php
340 * Allow substring/replace() to get/set bit values
341 * Allow to_char() on interval values to accumulate the highest unit
344 Some special format flag would be required to request such
345 accumulation. Such functionality could also be added to EXTRACT.
346 Prevent accumulation that crosses the month/day boundary because of
347 the uneven number of days in a month.
349 o to_char(INTERVAL '1 hour 5 minutes', 'MI') => 65
350 o to_char(INTERVAL '43 hours 20 minutes', 'MI' ) => 2600
351 o to_char(INTERVAL '43 hours 20 minutes', 'WK:DD:HR:MI') => 0:1:19:20
352 o to_char(INTERVAL '3 years 5 months','MM') => 41
354 * Implement inlining of set-returning functions defined in SQL
355 * Allow SQL-language functions to return results from RETURNING queries
357 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00665.php
359 * Allow SQL-language functions to reference parameters by parameter name
361 Currently SQL-language functions can only refer to dollar parameters,
364 * Add SPI_gettypmod() to return the typemod for a TupleDesc
365 * Enforce typmod for function inputs, function results and parameters for
366 spi_prepare'd statements called from PLs
368 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01403.php
370 * Allow holdable cursors in SPI
371 * Tighten function permission checks
373 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00568.php
375 * Fix IS OF so it matches the ISO specification, and add documentation
377 http://archives.postgresql.org/pgsql-patches/2003-08/msg00060.php
378 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00060.php
380 * Add missing operators for geometric data types
382 Some geometric types do not have the full suite of geometric operators,
385 * Implement Boyer-Moore searching in strpos()
387 http://archives.postgresql.org/pgsql-patches/2007-08/msg00012.php
389 * Prevent malicious functions from being executed with the permissions
390 of unsuspecting users
392 Index functions are safe, so VACUUM and ANALYZE are safe too.
393 Triggers, CHECK and DEFAULT expressions, and rules are still vulnerable.
394 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00268.php
396 * Reduce memory usage of aggregates in set returning functions
398 http://archives.postgresql.org/pgsql-performance/2008-01/msg00031.php
400 * Add temporal versions of generate_series()
402 http://archives.postgresql.org/pgsql-hackers/2007-04/msg01180.php
406 Multi-Language Support
407 ======================
409 * Add NCHAR (as distinguished from ordinary varchar),
410 * Allow locale to be set at database creation
412 Currently locale can only be set during initdb. No global tables have
413 locale-aware columns. However, the database template used during
414 database creation might have locale-aware indexes. The indexes would
415 need to be reindexed to match the new locale.
417 * Allow encoding on a per-column basis optionally using the ICU library:
419 Right now only one encoding is allowed per database. [locale]
421 http://archives.postgresql.org/pgsql-hackers/2005-03/msg00932.php
422 http://archives.postgresql.org/pgsql-patches/2005-08/msg00309.php
423 http://archives.postgresql.org/pgsql-patches/2006-03/msg00233.php
424 http://archives.postgresql.org/pgsql-hackers/2006-09/msg00662.php
426 * Add CREATE COLLATE? [locale]
427 * Support multiple simultaneous character sets, per SQL92
428 * Improve UTF8 combined character handling?
429 * Add octet_length_server() and octet_length_client()
430 * Make octet_length_client() the same as octet_length()?
431 * Fix problems with wrong runtime encoding conversion for NLS message files
432 * Add URL to more complete multi-byte regression tests
434 http://archives.postgresql.org/pgsql-hackers/2005-07/msg00272.php
436 * Fix ILIKE and regular expressions to handle case insensitivity
437 properly in multibyte encodings
439 http://archives.postgresql.org/pgsql-bugs/2005-10/msg00001.php
440 http://archives.postgresql.org/pgsql-patches/2005-11/msg00173.php
442 * Set client encoding based on the client operating system encoding
444 Currently client_encoding is set in postgresql.conf, which
445 defaults to the server encoding.
446 http://archives.postgresql.org/pgsql-hackers/2006-08/msg01696.php
448 * Change memory allocation for multi-byte functions so memory is
449 allocated inside conversion functions
451 Currently we preallocate memory based on worst-case usage.
458 * Automatically create rules on views so they are updateable, per SQL99
460 We can only auto-create rules for simple views. For more complex
461 cases users will still have to write rules manually.
463 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00586.php
464 http://archives.postgresql.org/pgsql-patches/2006-08/msg00255.php
466 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
467 * Allow NOTIFY in rules involving conditionals
468 * Allow VIEW/RULE recompilation when the underlying tables change
470 Another issue is whether underlying table changes should be reflected
471 in the view, e.g. should SELECT * show additional columns if they
472 are added after the view is created.
474 * Make it possible to use RETURNING together with conditional DO INSTEAD
475 rules, such as for partitioning setups
477 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00577.php
479 * Add the ability to automatically create materialized views
481 Right now materialized views require the user to create triggers on the
482 main table to keep the summary table current. SQL syntax should be able
483 to manager the triggers and summary table automatically. A more
484 sophisticated implementation would automatically retrieve from the
485 summary table when the main table is referenced, if possible.
492 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
493 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
494 * %Allow SET CONSTRAINTS to be qualified by schema/table name
495 * %Add a separate TRUNCATE permission
497 Currently only the owner can TRUNCATE a table because triggers are not
498 called, and the table is locked in exclusive mode.
500 * Allow PREPARE of cursors
501 * Allow finer control over the caching of prepared query plans
503 Currently queries prepared via the libpq API are planned on first
504 execute using the supplied parameters --- allow SQL PREPARE to do the
505 same. Also, allow control over replanning prepared queries either
506 manually or automatically when statistics for execute parameters
507 differ dramatically from those used during planning.
509 * Improve logging of prepared transactions recovered during startup
511 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00092.php
513 * Allow LISTEN/NOTIFY to store info in memory rather than tables?
515 Currently LISTEN/NOTIFY information is stored in pg_listener. Storing
516 such information in memory would improve performance.
518 * Add optional textual message to NOTIFY
520 This would allow an informational message to be added to the notify
521 message, perhaps indicating the row modified or other custom
524 * Allow multiple identical NOTIFY events to always be communicated to the
525 client, rather than sent as a single notification to the listener
527 http://archives.postgresql.org/pgsql-general/2008-01/msg00057.php
529 * Add a GUC variable to warn about non-standard SQL usage in queries
530 * Add SQL-standard MERGE command, typically used to merge two tables
533 This is similar to UPDATE, then for unmatched rows, INSERT.
534 Whether concurrent access allows modifications which could cause
535 row loss is implementation independent.
537 * Add REPLACE or UPSERT command that does UPDATE, or on failure, INSERT
540 To implement this cleanly requires that the table have a unique index
541 so duplicate checking can be easily performed. It is possible to
542 do it without a unique index if we require the user to LOCK the table
545 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00501.php
546 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00536.php
548 * Add NOVICE output level for helpful messages like automatic sequence/index
550 * Add GUC to issue notice about statements that use unjoined tables
551 * Allow EXPLAIN to identify tables that were skipped because of
553 * Allow EXPLAIN output to be more easily processed by scripts, perhaps XML
554 * Enable standard_conforming_strings
555 * Make standard_conforming_strings the default in 8.5?
557 When this is done, backslash-quote should be prohibited in non-E''
558 strings because of possible confusion over how such strings treat
559 backslashes. Basically, '' is always safe for a literal single
560 quote, while \' might or might not be based on the backslash
563 * Simplify dropping roles that have objects in several databases
564 * Allow COMMENT ON to accept an expression rather than just a string
565 * Allow the count returned by SELECT, etc to be represented as an int64
566 to allow a higher range of values
567 * Add SQL99 WITH clause to SELECT
568 * Add SQL:2003 WITH RECURSIVE (hierarchical) queries to SELECT
570 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01375.php
571 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00642.php
572 http://archives.postgresql.org/pgsql-patches/2007-03/msg00139.php
574 * Add DEFAULT .. AS OWNER so permission checks are done as the table
577 This would be useful for SERIAL nextval() calls and CHECK constraints.
579 * Allow DISTINCT to work in multiple-argument aggregate calls
580 * Add column to pg_stat_activity that shows the progress of long-running
581 commands like CREATE INDEX and VACUUM
582 * Implement SQL:2003 window functions
583 * Improve failure message when DROP DATABASE is used on a database that
584 has prepared transactions
585 * Allow INSERT/UPDATE ... RETURNING inside a SELECT 'FROM' clause
587 http://archives.postgresql.org/pgsql-general/2006-09/msg00803.php
588 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00693.php
590 * Increase locking when DROPing objects so dependent objects cannot
591 get dropped while the DROP operation is happening
593 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00937.php
595 * -Allow AS in "SELECT col AS label" to be optional in certain cases
597 * Allow INSERT ... DELETE ... RETURNING, namely allow the DELETE ...
598 RETURNING to supply values to the INSERT
599 http://archives.postgresql.org/pgsql-hackers/2008-02/thrd2.php#00979
601 * Add comments on system tables/columns using the information in
604 Ideally the information would be pulled from the SGML file
610 o Allow CREATE TABLE AS to determine column lengths for complex
611 expressions like SELECT col1 || col2
613 o Have WITH CONSTRAINTS also create constraint indexes
615 http://archives.postgresql.org/pgsql-patches/2007-04/msg00149.php
617 o Have CONSTRAINT cname NOT NULL record the contraint name
619 Right now pg_attribute.attnotnull records the NOT NULL status
620 of the column, but does not record the contraint name
625 o Allow UPDATE tab SET ROW (col, ...) = (SELECT...)
627 http://archives.postgresql.org/pgsql-hackers/2006-07/msg01306.php
628 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00865.php
630 o Research self-referential UPDATEs that see inconsistent row versions
631 in read-committed mode
633 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00507.php
634 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00016.php
636 o Allow GLOBAL temporary tables to exist as empty by default in
639 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00006.php
644 o Have ALTER TABLE RENAME rename SERIAL sequence names
646 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
648 o Have ALTER SEQUENCE RENAME rename the sequence name stored
649 in the sequence table
651 http://archives.postgresql.org/pgsql-bugs/2007-09/msg00092.php
652 http://archives.postgresql.org/pgsql-bugs/2007-10/msg00007.php
653 http://archives.postgresql.org/pgsql-hackers/2008-03/msg00008.php
655 o Add ALTER DOMAIN to modify the underlying data type
656 o %Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
658 http://archives.postgresql.org/pgsql-patches/2006-02/msg00168.php
660 o %Allow ALTER TABLE to change constraint deferrability and actions
661 o Add missing object types for ALTER ... SET SCHEMA
662 o Allow ALTER TABLESPACE to move to different directories
663 o Allow databases to be moved to different tablespaces
664 o Allow moving system tables to other tablespaces, where possible
666 Currently non-global system tables must be in the default database
667 tablespace. Global system tables can never be moved.
669 o Prevent parent tables from altering or dropping constraints
670 like CHECK that are inherited by child tables unless CASCADE
672 o %Prevent child tables from altering or dropping constraints
673 like CHECK that were inherited from the parent table
674 o Have ALTER INDEX update the name of a constraint using that index
675 o Add ALTER TABLE RENAME CONSTRAINT, update index name also
676 o Allow column display reordering by recording a display,
677 storage, and permanent id for every column?
679 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00782.php
685 o Automatically maintain clustering on a table
687 This might require some background daemon to maintain clustering
688 during periods of low usage. It might also require tables to be only
689 partially filled for easier reorganization. Another idea would
690 be to create a merged heap/index data file so an index lookup would
691 automatically access the heap data too. A third idea would be to
692 store heap rows in hashed groups, perhaps using a user-supplied
694 http://archives.postgresql.org/pgsql-performance/2004-08/msg00349.php
696 o %Add default clustering to system tables
698 To do this, determine the ideal cluster index for each system
699 table and set the cluster setting during initdb.
701 o %Add VERBOSE option to report tables as they are processed,
707 o Allow COPY to report error lines and continue
709 This requires the use of a savepoint before each COPY line is
710 processed, with ROLLBACK on COPY failure.
712 o Allow COPY on a newly-created table to skip WAL logging
714 On crash recovery, the table involved in the COPY would
715 be removed or have its heap and index files truncated. One
716 issue is that no other backend should be able to add to
717 the table at the same time, which is something that is
718 currently allowed. This currently is done if the table is
719 created inside the same transaction block as the COPY because
720 no other backends can see the table.
722 o Consider using a ring buffer for COPY FROM
724 http://archives.postgresql.org/pgsql-patches/2008-02/msg00140.php
725 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01080.php
727 o Allow COPY FROM to create index entries in bulk
729 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00811.php
735 o Allow column-level privileges
736 o %Allow GRANT/REVOKE permissions to be applied to all schema objects
739 The proposed syntax is:
740 GRANT SELECT ON ALL TABLES IN public TO phpuser;
741 GRANT SELECT ON NEW TABLES IN public TO phpuser;
743 o Allow GRANT/REVOKE permissions to be inherited by objects based on
746 o Allow SERIAL sequences to inherit permissions from the base table?
751 o Prevent DROP TABLE from dropping a row referenced by its own open
757 o Allow INSERT/UPDATE of the system-generated oid value for a row
758 o In rules, allow VALUES() to contain a mixture of 'old' and 'new'
764 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
769 Referential Integrity
770 =====================
772 * Add MATCH PARTIAL referential integrity
773 * Change foreign key constraint for array -> element to mean element
775 * Fix problem when cascading referential triggers make changes on
776 cascaded tables, seeing the tables in an intermediate state
778 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00174.php
779 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00174.php
781 * Allow DEFERRABLE and end-of-statement UNIQUE constraints?
783 This would allow UPDATE tab SET col = col + 1 to work if col has
784 a unique index. Currently, uniqueness checks are done while the
785 command is being executed, rather than at the end of the statement
787 http://people.planetpostgresql.org/greg/index.php?/archives/2006/06/10.html
788 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01458.php
790 * Optimize referential integrity checks
792 http://archives.postgresql.org/pgsql-performance/2005-10/msg00458.php
793 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00744.php
796 Server-Side Languages
797 =====================
800 o Fix RENAME to work on variables other than OLD/NEW
802 http://archives.postgresql.org/pgsql-hackers/2002-03/msg00591.php
803 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01615.php
804 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01587.php
806 o Allow function parameters to be passed by name,
807 get_employee_salary(12345 AS emp_id, 2001 AS tax_year)
808 o Allow handling of %TYPE arrays, e.g. tab.col%TYPE[]
809 o Allow listing of record column names, and access to
810 record columns via variables, e.g. columns := r.(*),
813 http://archives.postgresql.org/pgsql-patches/2005-07/msg00458.php
814 http://archives.postgresql.org/pgsql-patches/2006-05/msg00302.php
815 http://archives.postgresql.org/pgsql-patches/2006-06/msg00031.php
817 o Add support for SCROLL cursors
818 o Add support for WITH HOLD cursors
819 o Allow row and record variables to be set to NULL constants,
820 and allow NULL tests on such variables
822 Because a row is not scalar, do not allow assignment
823 from NULL-valued scalars.
825 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00070.php
827 o Review handling of MOVE and FETCH
829 http://archives.postgresql.org/pgsql-patches/2007-04/msg00527.php
833 o Add table function support to pltcl, plpythonu
834 o Add support for polymorphic arguments and return types to
835 languages other than PL/PgSQL
836 o Add capability to create and call PROCEDURES
837 o Add support for OUT and INOUT parameters to languages other
839 o Add PL/PythonU tracebacks
841 http://archives.postgresql.org/pgsql-patches/2006-02/msg00288.php
843 o Allow data to be passed in native language formats, rather
846 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00289.php
853 * Have pg_ctl look at PGHOST in case it is a socket directory?
854 * Allow pg_ctl to work properly with configuration files located outside
857 pg_ctl can not read the pid file because it isn't located in the
858 config directory but in the PGDATA directory. The solution is to
859 allow pg_ctl to read and understand postgresql.conf to find the
860 data_directory value.
862 * Add a function like pg_get_indexdef() that report more detailed index
865 http://archives.postgresql.org/pgsql-bugs/2007-12/msg00166.php
867 * Prefix command-line utilities like createuser with 'pg_'
869 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00025.php
873 o Have psql show current values for a sequence
874 o Move psql backslash database information into the backend, use
875 mnemonic commands? [psql]
877 This would allow non-psql clients to pull the same information out
878 of the database as psql.
880 o Fix psql's \d commands more consistent
882 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
883 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
885 o Consistently display privilege information for all objects in psql
886 o Add auto-expanded mode so expanded output is used if the row
887 length is wider than the screen width.
889 Consider using auto-expanded mode for backslash commands like \df+.
891 o Prevent tab completion of SET TRANSACTION from querying the
892 database and therefore preventing the transaction isolation
893 level from being set.
895 Currently SET <tab> causes a database lookup to check all
896 supported session variables. This query causes problems
897 because setting the transaction isolation level must be the
898 first statement of a transaction.
900 o Add a \set variable to control whether \s displays line numbers
902 Another option is to add \# which lists line numbers, and
903 allows command execution.
905 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00255.php
907 o Prevent escape string warnings when object names have
910 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00227.php
912 o Have \d show foreign keys that reference a table's primary key
914 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00424.php
916 o Have \d show child tables that inherit from the specified parent
917 o Have \l+ show database size, if permissions allow
919 Ideally it will not generate an error for invalid permissions
923 * pg_dump / pg_restore
924 o %Add dumping of comments on index columns and composite type columns
925 o %Add full object name to the tag field. eg. for operators we need
926 '=(integer, integer)', instead of just '='.
927 o Add pg_dumpall custom format dumps?
928 o Allow selection of individual object(s) of all types, not just
930 o In a selective dump, allow dumping of an object and all its
932 o Add options like pg_restore -l and -L to pg_dump
933 o Stop dumping CASCADE on DROP TYPE commands in clean mode
934 o Allow pg_dump --clean to drop roles that own objects or have
936 o Change pg_dump so that a comment on the dumped database is
937 applied to the loaded database, even if the database has a
938 different name. This will require new backend syntax, perhaps
939 COMMENT ON CURRENT DATABASE.
940 o Remove unnecessary function pointer abstractions in pg_dump source
942 o Allow pg_dump to utilize multiple CPUs and I/O channels by dumping
943 multiple objects simultaneously
945 The difficulty with this is getting multiple dump processes to
946 produce a single dump output file. It also would require
947 several sessions to share the same snapshot.
948 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00205.php
950 o Allow pg_restore to utilize multiple CPUs and I/O channels by
951 restoring multiple objects simultaneously
953 This might require a pg_restore flag to indicate how many
954 simultaneous operations should be performed. Only pg_dump's
955 -Fc format has the necessary dependency information.
956 http://archives.postgresql.org/pgsql-hackers/2008-02/msg00963.php
958 o To better utilize resources, allow pg_restore to check foreign
959 keys simultaneously, where possible
960 o Allow pg_restore to create all indexes of a table
961 concurrently, via a single heap scan
963 This requires a pg_dump -Fc file because that format contains
964 the required dependency information.
965 http://archives.postgresql.org/pgsql-general/2007-05/msg01274.php
967 o Allow pg_restore to load different parts of the COPY data
969 o Prevent pg_dump/pg_restore from being affected by
972 Using psql to restore a pg_dump dump is also affected.
978 Document differences between ecpg and the SQL standard and
979 information about the Informix-compatibility module.
981 o Solve cardinality > 1 for input descriptors / variables?
982 o Add a semantic check level, e.g. check if a table really exists
983 o fix handling of DB attributes that are arrays
984 o Use backend PREPARE/EXECUTE facility for ecpg where possible
986 o Fix nested C comments
987 o %sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
988 o Make SET CONNECTION thread-aware, non-standard?
989 o Allow multidimensional arrays
990 o Add internationalized message strings
991 o Implement COPY FROM STDIN
995 o Add PQescapeIdentifierConn()
996 o Prevent PQfnumber() from lowercasing unquoted the column name
998 PQfnumber() should never have been doing lowercasing, but
999 historically it has so we need a way to prevent it
1001 o Allow statement results to be automatically batched to the client
1003 Currently all statement results are transferred to the libpq
1004 client before libpq makes the results available to the
1005 application. This feature would allow the application to make
1006 use of the first result rows while the rest are transferred, or
1007 held on the server waiting for them to be requested by libpq.
1008 One complexity is that a statement like SELECT 1/col could error
1009 out mid-way through the result set.
1011 o Consider disallowing multiple queries in PQexec() as an
1012 additional barrier to SQL injection attacks
1014 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00184.php
1016 o Add PQexecf() that allows complex parameter substitution
1018 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01803.php
1020 o Add SQLSTATE severity to PGconn return status
1022 http://archives.postgresql.org/pgsql-interfaces/2007-11/msg00015.php
1028 * Add deferred trigger queue file
1030 Right now all deferred trigger information is stored in backend
1031 memory. This could exhaust memory for very large trigger queues.
1032 This item involves dumping large queues into files.
1034 * Allow triggers to be disabled in only the current session.
1036 This is currently possible by starting a multi-statement transaction,
1037 modifying the system tables, performing the desired SQL, restoring the
1038 system tables, and committing the transaction. ALTER TABLE ...
1039 TRIGGER requires a table lock so it is not ideal for this usage.
1041 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
1043 If the dump is known to be valid, allow foreign keys to be added
1044 without revalidating the data.
1046 * Allow statement-level triggers to access modified rows
1047 * Support triggers on columns
1049 http://archives.postgresql.org/pgsql-patches/2005-07/msg00107.php
1051 * Allow AFTER triggers on system tables
1053 System tables are modified in many places in the backend without going
1054 through the executor and therefore not causing triggers to fire. To
1055 complete this item, the functions that modify system tables will have
1058 * Tighten trigger permission checks
1060 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00564.php
1062 * Allow BEFORE INSERT triggers on views
1064 http://archives.postgresql.org/pgsql-general/2007-02/msg01466.php
1066 * Add ability to trigger on TRUNCATE
1068 http://archives.postgresql.org/pgsql-sql/2008-01/msg00050.php
1075 * Add UNIQUE capability to non-btree indexes
1076 * Prevent index uniqueness checks when UPDATE does not modify the column
1078 Uniqueness (index) checks are done when updating a column even if the
1079 column is not modified by the UPDATE.
1081 * Allow the creation of on-disk bitmap indexes which can be quickly
1082 combined with other bitmap indexes
1084 Such indexes could be more compact if there are only a few distinct values.
1085 Such indexes can also be compressed. Keeping such indexes updated can be
1088 http://archives.postgresql.org/pgsql-patches/2005-07/msg00512.php
1089 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01107.php
1091 * Allow accurate statistics to be collected on indexes with more than
1092 one column or expression indexes, perhaps using per-index statistics
1094 http://archives.postgresql.org/pgsql-performance/2006-10/msg00222.php
1095 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01131.php
1097 * Consider increasing the number of default statistics target, and
1098 reduce statistics target overhead
1100 Also consider having a larger statistics target for indexed columns
1101 and expression indexes
1102 http://archives.postgresql.org/pgsql-general/2007-05/msg01228.php
1103 http://archives.postgresql.org/pgsql-general/2007-06/msg00542.php
1105 * Consider compressing indexes by storing key values duplicated in
1106 several rows as a single index entry
1108 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00341.php
1109 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01264.php
1110 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00465.php
1111 http://archives.postgresql.org/pgsql-patches/2007-03/msg00163.php
1113 * Add REINDEX CONCURRENTLY, like CREATE INDEX CONCURRENTLY
1115 This is difficult because you must upgrade to an exclusive table lock
1116 to replace the existing index file. CREATE INDEX CONCURRENTLY does not
1117 have this complication. This would allow index compaction without
1120 * Allow multiple indexes to be created concurrently, ideally via a
1121 single heap scan, and have pg_restore use it
1127 o Allow inherited tables to inherit indexes, UNIQUE constraints,
1128 and primary/foreign keys
1129 o Honor UNIQUE INDEX on base column in INSERTs/UPDATEs
1130 on inherited table, e.g. INSERT INTO inherit_table
1131 (unique_index_col) VALUES (dup) should fail
1133 The main difficulty with this item is the problem of
1134 creating an index that can span multiple tables.
1136 o Allow SELECT ... FOR UPDATE on inherited tables
1137 o Require all CHECK constraints to be inherited
1139 http://archives.postgresql.org/pgsql-bugs/2007-04/msg00026.php
1144 o Add more GIST index support for geometric data types
1145 o Allow GIST indexes to create certain complex index types, like
1146 digital trees (see Aoki)
1151 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00051.php
1153 o Pack hash index buckets onto disk pages more efficiently
1155 Currently only one hash bucket can be stored on a page. Ideally
1156 several hash buckets could be stored on a single page and greater
1157 granularity used for the hash algorithm.
1159 http://archives.postgresql.org/pgsql-hackers/2004-06/msg00168.php
1161 o Consider sorting hash buckets so entries can be found using a
1162 binary search, rather than a linear scan
1163 o In hash indexes, consider storing the hash value with or instead
1165 o Add WAL logging for crash recovery
1166 o Allow multi-column hash indexes
1167 o During index creation, pre-sort the tuples to improve build speed
1169 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01199.php
1176 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
1178 Ideally this requires a separate test program that can be run
1179 at initdb time or optionally later. Consider O_SYNC when
1182 * Add program to test if fsync has a delay compared to non-fsync
1191 We could use a fixed row count and a +/- count to follow MVCC
1192 visibility rules, or a single cached value could be used and
1193 invalidated if anyone modifies the table. Another idea is to
1194 get a count directly from a unique index, but for this to be
1195 faster than a sequential scan it must avoid access to the heap
1196 to obtain tuple visibility information.
1198 * Provide a way to calculate an "estimated COUNT(*)"
1200 Perhaps by using the optimizer's cardinality estimates or random
1203 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00943.php
1205 * Allow data to be pulled directly from indexes
1207 Currently indexes do not have enough tuple visibility information
1208 to allow data to be pulled from the index without also accessing
1209 the heap. One way to allow this is to set a bit on index tuples
1210 to indicate if a tuple is currently visible to all transactions
1211 when the first valid heap lookup happens. This bit would have to
1212 be cleared when a heap tuple is expired.
1214 Another idea is to maintain a bitmap of heap pages where all rows
1215 are visible to all backends, and allow index lookups to reference
1216 that bitmap to avoid heap lookups, perhaps the same bitmap we might
1217 add someday to determine which heap pages need vacuuming. Frequently
1218 accessed bitmaps would have to be stored in shared memory. One 8k
1219 page of bitmaps could track 512MB of heap pages.
1221 A third idea would be for a heap scan to check if all rows are visible
1222 and if so set a per-table flag which can be checked by index scans.
1223 Any change to the table would have to clear the flag. To detect
1224 changes during the heap scan a counter could be set at the start and
1225 checked at the end --- if it is the same, the table has not been
1226 modified --- any table change would increment the counter.
1228 * Consider automatic caching of statements at various levels:
1231 o Query execute plan
1234 * Consider increasing internal areas when shared buffers is increased
1236 http://archives.postgresql.org/pgsql-hackers/2005-10/msg01419.php
1238 * Consider decreasing the amount of memory used by PrivateRefCount
1240 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00797.php
1241 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00752.php
1243 * Consider allowing higher priority queries to have referenced buffer
1244 cache pages stay in memory longer
1246 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00562.php
1253 * Improve speed with indexes
1255 For large table adjustments during VACUUM FULL, it is faster to cluster
1256 or reindex rather than update the index. Also, index updates can bloat
1259 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00024.php
1260 http://archives.postgresql.org/pgsql-performance/2007-05/msg00296.php
1262 * Auto-fill the free space map by scanning the buffer cache or by
1263 checking pages written by the background writer
1265 http://archives.postgresql.org/pgsql-hackers/2006-02/msg01125.php
1266 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00011.php
1268 * Create a bitmap of pages that need vacuuming
1270 Instead of sequentially scanning the entire table, have the background
1271 writer or some other process record pages that have expired rows, then
1272 VACUUM can look at just those pages rather than the entire table. In
1273 the event of a system crash, the bitmap would probably be invalidated.
1274 One complexity is that index entries still have to be vacuumed, and
1275 doing this without an index scan (by using the heap values to find the
1276 index entry) might be slow and unreliable, especially for user-defined
1279 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01188.php
1280 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00121.php
1282 * Allow FSM to return free space toward the beginning of the heap file,
1283 in hopes that empty pages at the end can be truncated by VACUUM
1284 * Allow FSM page return free space based on table clustering, to assist
1285 in maintaining clustering?
1286 * Improve dead row detection during multi-statement transactions usage
1288 http://archives.postgresql.org/pgsql-patches/2007-03/msg00358.php
1290 * Consider a more compact data representation for dead tuples
1292 http://archives.postgresql.org/pgsql-patches/2007-05/msg00143.php
1297 o %Issue log message to suggest VACUUM FULL if a table is nearly
1299 o Improve control of auto-vacuum
1301 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00876.php
1303 o Prevent long-lived temporary tables from causing frozen-xid
1304 advancement starvation
1306 The problem is that autovacuum cannot vacuum them to set frozen xids;
1307 only the session that created them can do that.
1308 http://archives.postgresql.org/pgsql-general/2007-06/msg01645.php
1310 o Store per-table autovacuum settings in pg_class.reloptions.
1312 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01440.php
1313 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00724.php
1315 o Prevent autovacuum from running if an old transaction is still
1316 running from the last vacuum
1318 http://archives.postgresql.org/pgsql-hackers/2007-11/msg00899.php
1325 * Fix priority ordering of read and write light-weight locks (Neil)
1327 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00893.php
1328 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00905.php
1330 * Fix problem when multiple subtransactions of the same outer transaction
1331 hold different types of locks, and one subtransaction aborts
1333 http://archives.postgresql.org/pgsql-hackers/2006-11/msg01011.php
1334 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00001.php
1335 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00435.php
1336 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00773.php
1338 * Allow UPDATEs on only non-referential integrity columns not to conflict
1339 with referential integrity locks
1341 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00073.php
1343 * Add idle_in_transaction_timeout GUC so locks are not held for long
1346 * Improve deadlock detection when a page cleaning lock conflicts
1347 with a shared buffer that is pinned
1349 http://archives.postgresql.org/pgsql-bugs/2008-01/msg00138.php
1350 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00873.php
1351 http://archives.postgresql.org/pgsql-committers/2008-01/msg00365.php
1356 Startup Time Improvements
1357 =========================
1359 * Experiment with multi-threaded backend for backend creation [thread]
1361 This would prevent the overhead associated with process creation. Most
1362 operating systems have trivial process creation time compared to
1363 database startup overhead, but a few operating systems (Win32,
1364 Solaris) might benefit from threading. Also explore the idea of
1365 a single session using multiple threads to execute a statement faster.
1372 * Eliminate need to write full pages to WAL before page modification [wal]
1374 Currently, to protect against partial disk page writes, we write
1375 full page images to WAL before they are modified so we can correct any
1376 partial page writes during recovery. These pages can also be
1377 eliminated from point-in-time archive files.
1379 o When off, write CRC to WAL and check file system blocks
1382 If CRC check fails during recovery, remember the page in case
1383 a later CRC for that page properly matches.
1385 o Write full pages during file system write and not when
1386 the page is modified in the buffer cache
1388 This allows most full page writes to happen in the background
1389 writer. It might cause problems for applying WAL on recovery
1390 into a partially-written page, but later the full page will be
1393 * Allow WAL traffic to be streamed to another server for stand-by
1395 * Reduce WAL traffic so only modified values are written rather than
1398 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01589.php
1400 * Allow WAL information to recover corrupted pg_controldata
1402 http://archives.postgresql.org/pgsql-patches/2006-06/msg00025.php
1404 * Find a way to reduce rotational delay when repeatedly writing
1407 Currently fsync of WAL requires the disk platter to perform a full
1408 rotation to fsync again. One idea is to write the WAL to different
1409 offsets that might reduce the rotational delay.
1411 * Allow WAL logging to be turned off for a table, but the table
1412 might be dropped or truncated during crash recovery [walcontrol]
1414 Allow tables to bypass WAL writes and just fsync() dirty pages on
1415 commit. This should be implemented using ALTER TABLE, e.g. ALTER
1416 TABLE PERSISTENCE [ DROP | TRUNCATE | DEFAULT ]. Tables using
1417 non-default logging should not use referential integrity with
1418 default-logging tables. A table without dirty buffers during a
1419 crash could perhaps avoid the drop/truncate.
1421 * Allow WAL logging to be turned off for a table, but the table would
1422 avoid being truncated/dropped [walcontrol]
1424 To do this, only a single writer can modify the table, and writes
1425 must happen only on new pages so the new pages can be removed during
1426 crash recovery. Readers can continue accessing the table. Such
1427 tables probably cannot have indexes. One complexity is the handling
1428 of indexes on TOAST tables.
1430 * Speed WAL recovery by allowing more than one page to be prefetched
1432 This involves having a separate process that can be told which pages
1433 the recovery process will need in the near future.
1434 http://archives.postgresql.org/pgsql-hackers/2008-02/msg01279.php
1438 Optimizer / Executor
1439 ====================
1441 * Improve selectivity functions for geometric operators
1442 * Precompile SQL functions to avoid overhead
1443 * Create utility to compute accurate random_page_cost value
1444 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
1445 * Have EXPLAIN ANALYZE issue NOTICE messages when the estimated and
1446 actual row counts differ by a specified percentage
1447 * Consider using hash buckets to do DISTINCT, rather than sorting
1449 This would be beneficial when there are few distinct values. This is
1450 already used by GROUP BY.
1452 * Log statements where the optimizer row estimates were dramatically
1453 different from the number of rows actually found?
1454 * Consider compressed annealing to search for query plans
1456 This might replace GEQO, http://sixdemonbag.org/Djinni.
1458 * Improve merge join performance by allowing mark/restore of
1461 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00096.php
1463 * Consider using a hash for joining to a large IN (VALUES ...) list
1465 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00450.php
1469 Miscellaneous Performance
1470 =========================
1472 * Do async I/O for faster random read-ahead of data
1474 Async I/O allows multiple I/O requests to be sent to the disk with
1475 results coming back asynchronously.
1477 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00820.php
1478 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00027.php
1480 * Use mmap() rather than SYSV shared memory or to write WAL files?
1482 This would remove the requirement for SYSV SHM but would introduce
1483 portability issues. Anonymous mmap (or mmap to /dev/zero) is required
1484 to prevent I/O overhead.
1486 * Consider mmap()'ing files into a backend?
1488 Doing I/O to large tables would consume a lot of address space or
1489 require frequent mapping/unmapping. Extending the file also causes
1490 mapping problems that might require mapping only individual pages,
1491 leading to thousands of mappings. Another problem is that there is no
1492 way to _prevent_ I/O to disk from the dirty shared buffers so changes
1493 could hit disk before WAL is written.
1495 * Add a script to ask system configuration questions and tune postgresql.conf
1496 * Consider ways of storing rows more compactly on disk
1498 o Reduce the row header size?
1499 o Consider reducing on-disk varlena length from four bytes to
1500 two because a heap row cannot be more than 64k in length
1502 * Consider increasing NUM_CLOG_BUFFERS
1503 * Consider having the background writer update the transaction status
1504 hint bits before writing out the page
1506 Implementing this requires the background writer to have access to system
1507 catalogs and the transaction status log.
1509 * Consider adding buffers the BGW finds reusable to the free list
1511 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00781.php
1513 * Automatically tune bgwriter_delay based on activity rather then using a
1516 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00781.php
1518 * Allow user configuration of TOAST thresholds
1520 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00213.php
1522 * Allow configuration of backend priorities via the operating system
1524 Though backend priorities make priority inversion during lock
1525 waits possible, research shows that this is not a huge problem.
1527 http://archives.postgresql.org/pgsql-general/2007-02/msg00493.php
1529 * Experiment with multi-threaded backend better resource utilization
1531 This would allow a single query to make use of multiple CPU's or
1532 multiple I/O channels simultaneously. One idea is to create a
1533 background reader that can pre-fetch sequential and index scan
1534 pages needed by other backends. This could be expanded to allow
1535 concurrent reads from multiple devices in a partitioned table.
1537 * Consider increasing the minimum allowed number of shared buffers
1539 http://archives.postgresql.org/pgsql-bugs/2008-02/msg00157.php
1547 * Add use of 'const' for variables in source tree
1548 * Move some things from contrib into main tree
1549 * %Remove warnings created by -Wcast-align
1550 * Move platform-specific ps status display info from ps_status.c to ports
1551 * Add optional CRC checksum to heap and index pages
1552 * Improve documentation to build only interfaces (Marc)
1553 * Remove or relicense modules that are not under the BSD license, if possible
1554 * Acquire lock on a relation before building a relcache entry for it
1555 * Allow cross-compiling by generating the zic database on the target system
1556 * Improve NLS maintenance of libpgport messages linked onto applications
1557 * Clean up casting in contrib/isn
1559 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00245.php
1561 * Use UTF8 encoding for NLS messages so all server encodings can
1563 * Update Bonjour to work with newer cross-platform SDK
1565 http://archives.postgresql.org/pgsql-hackers/2006-09/msg02238.php
1566 http://archives.postgresql.org/pgsql-patches/2006-10/msg00048.php
1568 * Consider detoasting keys before sorting
1569 * Consider GnuTLS if OpenSSL license becomes a problem
1571 http://archives.postgresql.org/pgsql-patches/2006-05/msg00040.php
1572 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01213.php
1574 * Consider changing documentation format from SGML to XML
1576 http://archives.postgresql.org/pgsql-docs/2006-12/msg00152.php
1578 * Consider making NAMEDATALEN more configurable in future releases
1579 * Update our code to handle 64-bit timezone files to match the zic
1580 source code, which now uses them
1581 * Have configure choose integer datetimes by default
1583 http://archives.postgresql.org/pgsql-patches/2007-05/msg00046.php
1585 * Support scoped IPv6 addresses
1587 http://archives.postgresql.org/pgsql-bugs/2007-05/msg00111.php
1589 * Consider allowing 64-bit integers and floats to be passed by value on
1592 Also change 32-bit floats (float4) to be passed by value at the same
1599 o Remove configure.in check for link failure when cause is found
1600 o Remove readdir() errno patch when runtime/mingwex/dirent.c rev
1602 o Remove psql newline patch when we find out why mingw outputs an
1604 o Allow psql to use readline once non-US code pages work with
1606 o Fix problem with shared memory on the Win32 Terminal Server
1607 o Improve signal handling
1609 http://archives.postgresql.org/pgsql-patches/2005-06/msg00027.php
1613 * Wire Protocol Changes
1615 o Allow dynamic character set handling
1616 o Add decoded type, length, precision
1618 o Update clients to use data types, typmod, schema.table.column names
1619 of result sets using new statement protocol
1625 * Add pre-parsing phase that converts non-ISO syntax to supported
1628 This could allow SQL written for other databases to run without
1631 * Allow plug-in modules to emulate features from other databases
1632 * Add features of Oracle-style packages (Pavel)
1634 A package would be a schema with session-local variables,
1635 public/private functions, and initialization functions. It
1636 is also possible to implement these capabilities
1637 in any schema and not use a separate "packages"
1640 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00384.php
1642 * Consider allowing control of upper/lower case folding of unquoted
1645 http://archives.postgresql.org/pgsql-hackers/2004-04/msg00818.php
1646 http://archives.postgresql.org/pgsql-hackers/2006-10/msg01527.php
1648 * Add autonomous transactions
1650 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00893.php
1654 Features We Do _Not_ Want
1655 =========================
1657 * All backends running as threads in a single process (not wanted)
1659 This eliminates the process protection we get from the current setup.
1660 Thread creation is usually the same overhead as process creation on
1661 modern systems, so it seems unwise to use a pure threaded model.
1663 * Optimizer hints (not wanted)
1665 Optimizer hints are used to work around problems in the optimizer. We
1666 would rather have the problems reported and fixed.
1668 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00506.php
1669 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00517.php
1670 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00663.php
1673 Because we support postfix operators, it isn't possible to make AS
1674 optional and continue to use bison.
1675 http://archives.postgresql.org/pgsql-hackers/2003-04/msg00436.php
1677 http://archives.postgresql.org/pgsql-sql/2006-08/msg00164.php
1679 * Embedded server (not wanted)
1681 While PostgreSQL clients runs fine in limited-resource environments, the
1682 server requires multiple processes and a stable pool of resources to
1683 run reliabily and efficiently. Stripping down the PostgreSQL server
1684 to run in the same process address space as the client application
1685 would add too much complexity and failure cases.