4 Current maintainer: Bruce Momjian (bruce@momjian.us)
5 Last updated: Fri Apr 13 19:23:16 EDT 2007
7 The most recent version of this document can be viewed at
8 http://www.postgresql.org/docs/faqs.TODO.html.
10 #A hyphen, "-", marks changes that will appear in the upcoming 8.3 release.#
11 #A percent sign, "%", marks items that are easier to implement.#
13 Bracketed items, "[]", have more detail.
15 This list contains all known PostgreSQL bugs and feature requests. If
16 you would like to work on an item, please read the Developer's FAQ
17 first. There is also a developer's wiki at
18 http://developer.postgresql.org.
24 * Allow major upgrades without dump/reload, perhaps using pg_upgrade
26 * Check for unreferenced table files created by transactions that were
27 in-progress when the server terminated abruptly
29 http://archives.postgresql.org/pgsql-patches/2006-06/msg00096.php
31 * Allow administrators to safely terminate individual sessions either
32 via an SQL function or SIGTERM
34 Lock table corruption following SIGTERM of an individual backend
35 has been reported in 8.0. A possible cause was fixed in 8.1, but
36 it is unknown whether other problems exist. This item mostly
37 requires additional testing rather than of writing any new code.
38 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00174.php
40 * %Set proper permissions on non-system schemas during db creation
42 Currently all schemas are owned by the super-user because they are
43 copied from the template1 database.
45 * Support table partitioning that allows a single table to be stored
46 in subtables that are partitioned based on the primary key or a WHERE
48 * Add function to report the time of the most recent server reload
49 * Allow statistics collector information to be pulled from the collector
50 process directly, rather than requiring the collector to write a
51 filesystem file twice a second?
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 rules for INSERT/UPDATE/DELETE, and constraints for
63 rapid partition selection. Options could include range and hash
65 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00375.php
66 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00151.php
68 * Allow auto-selection of partitioned tables for min/max() operations
69 * Allow more complex user/database default GUC settings
71 Currently ALTER USER and ALTER DATABASE support per-user and
72 per-database defaults. Consider adding per-user-and-database
73 defaults so things like search_path can be defaulted for a
74 specific user connecting to a specific database.
76 * Allow custom variable classes that can restrict who can set the values
78 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00911.php
81 * Improve replication solutions
85 You can use any of the master/slave replication servers to use a
86 standby server for data warehousing. To allow read/write queries to
87 multiple servers, you need multi-master replication like pgcluster.
89 o Allow replication over unreliable or non-persistent links
94 o Allow commenting of variables in postgresql.conf to restore them
97 Currently, if a variable is commented out, it keeps the
98 previous uncommented value until a server restarted.
99 http://archives.postgresql.org/pgsql-patches/2007-03/msg00180.php
101 o Allow pg_hba.conf to specify host names along with IP addresses
103 Host name lookup could occur when the postmaster reads the
104 pg_hba.conf file, or when the backend starts. Another
105 solution would be to reverse lookup the connection IP and
106 check that hostname against the host names in pg_hba.conf.
107 We could also then check that the host name maps to the IP
110 o %Allow postgresql.conf file values to be changed via an SQL
111 API, perhaps using SET GLOBAL
112 o Allow the server to be stopped/restarted via an SQL API
113 o Issue a warning if a change-on-restart-only postgresql.conf value
114 is modified and the server config files are reloaded
115 o Mark change-on-restart-only values in postgresql.conf
120 o Allow a database in tablespace t1 with tables created in
121 tablespace t2 to be used as a template for a new database created
122 with default tablespace t2
124 All objects in the default database tablespace must have default
125 tablespace specifications. This is because new databases are
126 created by copying directories. If you mix default tablespace
127 tables and tablespace-specified tables in the same directory,
128 creating a new database from such a mixed directory would create a
129 new database with tables that had incorrect explicit tablespaces.
130 To fix this would require modifying pg_class in the newly copied
131 database, which we don't currently do.
133 o Allow reporting of which objects are in which tablespaces
135 This item is difficult because a tablespace can contain objects
136 from multiple databases. There is a server-side function that
137 returns the databases which use a specific tablespace, so this
138 requires a tool that will call that function and connect to each
139 database to find the objects in each database for that tablespace.
141 o -Add a GUC variable to control the tablespace for temporary objects
144 It could start with a random tablespace from a supplied list and
145 cycle through the list.
147 o Allow WAL replay of CREATE TABLESPACE to work when the directory
148 structure on the recovery computer is different from the original
150 o Allow per-tablespace quotas
153 * Point-In-Time Recovery (PITR)
155 o Allow a warm standby system to also allow read-only statements
158 This is useful for checking PITR recovery.
159 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00050.php
161 o %Create dump tool for write-ahead logs for use in determining
162 transaction id for point-in-time recovery
163 o Allow the PITR process to be debugged and data examined
164 o Allow recovery.conf to allow the same syntax as
165 postgresql.conf, including quoting
167 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00497.php
173 * Allow server log information to be output as INSERT statements
175 This would allow server log information to be easily loaded into
176 a database for analysis.
178 * -Add ability to monitor the use of temporary sort files
184 * -Make 64-bit version of the MONEY data type
185 * Add locale-aware MONEY type, and support multiple currencies
187 http://archives.postgresql.org/pgsql-general/2005-08/msg01432.php
188 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01181.php
190 * Change NUMERIC to enforce the maximum precision
191 * Reduce storage space for small NUMERICs
193 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01331.php
194 http://archives.postgresql.org/pgsql-patches/2007-02/msg00505.php
196 * Fix data types where equality comparison isn't intuitive, e.g. box
197 * Allow user-defined types to specify a type modifier at table creation
199 * -Allow user-defined types to accept 'typmod' parameters
201 http://archives.postgresql.org/pgsql-hackers/2005-08/msg01142.php
202 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00012.php
203 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00149.php
205 * Add support for public SYNONYMs
207 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00519.php
209 * Fix CREATE CAST on DOMAINs
211 http://archives.postgresql.org/pgsql-hackers/2006-05/msg00072.php
212 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01681.php
214 * -Add Globally/Universally Unique Identifier (GUID/UUID)
216 http://archives.postgresql.org/pgsql-patches/2006-09/msg00209.php
217 http://archives.postgresql.org/pgsql-general/2007-01/msg00853.php
219 * Add support for SQL-standard GENERATED/IDENTITY columns
221 http://archives.postgresql.org/pgsql-hackers/2006-07/msg00543.php
222 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00038.php
224 * -Support a data type with specific enumerated values (ENUM)
225 * Improve XML support
227 http://developer.postgresql.org/index.php/XML_Support
229 * Consider placing all sequences in a single table, or create a system
231 * Allow all data types to cast to and from TEXT
233 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00017.php
238 o Allow infinite dates and intervals just like infinite timestamps
239 o Merge hardwired timezone names with the TZ database; allow either
240 kind everywhere a TZ name is currently taken
241 o Allow TIMESTAMP WITH TIME ZONE to store the original timezone
242 information, either zone name or offset from UTC [timezone]
244 If the TIMESTAMP value is stored with a time zone name, interval
245 computations should adjust based on the time zone rules.
247 o Fix SELECT '0.01 years'::interval, '0.01 months'::interval
248 o Add a GUC variable to allow output of interval values in ISO8601
250 o Have timestamp subtraction not call justify_hours()?
252 http://archives.postgresql.org/pgsql-sql/2006-10/msg00059.php
254 o Improve timestamptz subtraction to be DST-aware
256 Currently subtracting one date from another that crosses a
257 daylight savings time adjustment can return '1 day 1 hour', but
258 adding that back to the first date returns a time one hour in
259 the future. This is caused by the adjustment of '25 hours' to
260 '1 day 1 hour', and '1 day' is the same time the next day, even
261 if daylight savings adjustments are involved.
263 o Fix interval display to support values exceeding 2^31 hours
264 o Add overflow checking to timestamp and interval arithmetic
265 o Extend timezone code to allow 64-bit values so we can
266 represent years beyond 2038
268 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01363.php
270 o Use LC_TIME for localized weekday/month names, rather than
273 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00390.php
275 o Add ISO INTERVAL handling
277 http://archives.postgresql.org/pgsql-hackers/2006-01/msg00250.php
278 http://archives.postgresql.org/pgsql-bugs/2006-04/msg00248.php
280 o Support ISO INTERVAL syntax if units cannot be determined from
281 the string, and are supplied after the string
283 The SQL standard states that the units after the string
284 specify the units of the string, e.g. INTERVAL '2' MINUTE
285 should return '00:02:00'. The current behavior has the units
286 restrict the interval value to the specified unit or unit
287 range, INTERVAL '70' SECOND returns '00:00:10'.
289 For syntax that isn't uniquely ISO or PG syntax, like '1' or
290 '1:30', treat as ISO if there is a range specification clause,
291 and as PG if there no clause is present, e.g. interpret '1:30'
292 MINUTE TO SECOND as '1 minute 30 seconds', and interpret
293 '1:30' as '1 hour, 30 minutes'.
295 This makes common cases like SELECT INTERVAL '1' MONTH
296 SQL-standard results. The SQL standard supports a limited
297 number of unit combinations and doesn't support unit names in
298 the string. The PostgreSQL syntax is more flexible in the
299 range of units supported, e.g. PostgreSQL supports '1 year 1
300 hour', while the SQL standard does not.
302 o Add support for year-month syntax, INTERVAL '50-6' YEAR TO MONTH
303 o Interpret INTERVAL '1 year' MONTH as CAST (INTERVAL '1 year' AS
304 INTERVAL MONTH), and this should return '12 months'
305 o Round or truncate values to the requested precision, e.g.
306 INTERVAL '11 months' AS YEAR should return one or zero
307 o Support precision, CREATE TABLE foo (a INTERVAL MONTH(3))
312 o Delay resolution of array expression's data type so assignment
313 coercion can be performed on empty array expressions
314 o Add support for arrays of domains
315 o Add support for arrays of complex types
316 o Allow single-byte header storage for array elements
321 o Improve vacuum of large objects, like /contrib/vacuumlo?
322 o Add security checking for large objects
323 o Auto-delete large objects when referencing row is deleted
325 /contrib/lo offers this functionality.
327 o Allow read/write into TOAST values like large objects
329 This requires the TOAST column to be stored EXTERNAL.
331 o Add API for 64-bit large object access
333 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00781.php
339 * Allow INET subnet tests using non-constants to be indexed
340 * Allow to_date() and to_timestamp() accept localized month names
341 * Fix to_date()-related functions to consistently issue errors
343 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00915.php
345 * Add missing parameter handling in to_char()
347 http://archives.postgresql.org/pgsql-hackers/2005-12/msg00948.php
349 * Allow functions to have a schema search path specified at creation time
350 * Allow substring/replace() to get/set bit values
351 * Allow to_char() on interval values to accumulate the highest unit
354 Some special format flag would be required to request such
355 accumulation. Such functionality could also be added to EXTRACT.
356 Prevent accumulation that crosses the month/day boundary because of
357 the uneven number of days in a month.
359 o to_char(INTERVAL '1 hour 5 minutes', 'MI') => 65
360 o to_char(INTERVAL '43 hours 20 minutes', 'MI' ) => 2600
361 o to_char(INTERVAL '43 hours 20 minutes', 'WK:DD:HR:MI') => 0:1:19:20
362 o to_char(INTERVAL '3 years 5 months','MM') => 41
364 * -Add ISO day of week format 'ID' to to_char() where Monday = 1
365 * -Add a field 'isoyear' to extract(), based on the ISO week
366 * Implement inlining of set-returning functions defined in SQL
367 * Allow SQL-language functions to return results from RETURNING queries
369 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00665.php
371 * Allow SQL-language functions to reference parameters by parameter name
373 Currently SQL-language functions can only refer to dollar parameters,
376 * Add SPI_gettypmod() to return the typemod for a TupleDesc
377 * Enforce typmod for function inputs, function results and parameters for
378 spi_prepare'd statements called from PLs
380 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01403.php
382 * Allow holdable cursors in SPI
383 * Tighten function permission checks
385 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00568.php
387 * Fix IS OF so it matches the ISO specification, and add documentation
389 http://archives.postgresql.org/pgsql-patches/2003-08/msg00060.php
390 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00060.php
392 * Add missing operators for geometric data types
394 Some geometric types do not have the full suite of geometric operators,
400 Multi-Language Support
401 ======================
403 * Add NCHAR (as distinguished from ordinary varchar),
404 * Allow locale to be set at database creation
406 Currently locale can only be set during initdb. No global tables have
407 locale-aware columns. However, the database template used during
408 database creation might have locale-aware indexes. The indexes would
409 need to be reindexed to match the new locale.
411 * Allow encoding on a per-column basis optionally using the ICU library:
413 Right now only one encoding is allowed per database. [locale]
414 http://archives.postgresql.org/pgsql-hackers/2005-03/msg00932.php
415 http://archives.postgresql.org/pgsql-patches/2005-08/msg00309.php
416 http://archives.postgresql.org/pgsql-patches/2006-03/msg00233.php
417 http://archives.postgresql.org/pgsql-hackers/2006-09/msg00662.php
419 * Add CREATE COLLATE? [locale]
420 * Support multiple simultaneous character sets, per SQL92
421 * Improve UTF8 combined character handling?
422 * Add octet_length_server() and octet_length_client()
423 * Make octet_length_client() the same as octet_length()?
424 * Fix problems with wrong runtime encoding conversion for NLS message files
425 * Add URL to more complete multi-byte regression tests
427 http://archives.postgresql.org/pgsql-hackers/2005-07/msg00272.php
429 * Fix ILIKE and regular expressions to handle case insensitivity
430 properly in multibyte encodings
432 http://archives.postgresql.org/pgsql-bugs/2005-10/msg00001.php
433 http://archives.postgresql.org/pgsql-patches/2005-11/msg00173.php
435 * Set client encoding based on the client operating system encoding
437 Currently client_encoding is set in postgresql.conf, which
438 defaults to the server encoding.
439 http://archives.postgresql.org/pgsql-hackers/2006-08/msg01696.php
441 * Fix cases where invalid byte encodings are accepted by the database,
442 but throw an error on SELECT
444 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00767.php
450 * Automatically create rules on views so they are updateable, per SQL99
452 We can only auto-create rules for simple views. For more complex
453 cases users will still have to write rules manually.
454 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00586.php
455 http://archives.postgresql.org/pgsql-patches/2006-08/msg00255.php
457 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
458 * Allow NOTIFY in rules involving conditionals
459 * Allow VIEW/RULE recompilation when the underlying tables change
461 Another issue is whether underlying table changes should be reflected
462 in the view, e.g. should SELECT * show additional columns if they
463 are added after the view is created.
469 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
470 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
471 * %Allow SET CONSTRAINTS to be qualified by schema/table name
472 * %Add a separate TRUNCATE permission
474 Currently only the owner can TRUNCATE a table because triggers are not
475 called, and the table is locked in exclusive mode.
477 * Allow PREPARE of cursors
478 * Allow finer control over the caching of prepared query plans
480 Currently queries prepared via the libpq API are planned on first
481 execute using the supplied parameters --- allow SQL PREPARE to do the
482 same. Also, allow control over replanning prepared queries either
483 manually or automatically when statistics for execute parameters
484 differ dramatically from those used during planning.
486 * Invalidate prepared queries, like INSERT, when the table definition
488 * Improve logging of prepared statements recovered during startup
490 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00092.php
492 * Allow LISTEN/NOTIFY to store info in memory rather than tables?
494 Currently LISTEN/NOTIFY information is stored in pg_listener. Storing
495 such information in memory would improve performance.
497 * Add optional textual message to NOTIFY
499 This would allow an informational message to be added to the notify
500 message, perhaps indicating the row modified or other custom
503 * Add a GUC variable to warn about non-standard SQL usage in queries
504 * Add SQL-standard MERGE command, typically used to merge two tables
507 This is similar to UPDATE, then for unmatched rows, INSERT.
508 Whether concurrent access allows modifications which could cause
509 row loss is implementation independent.
511 * Add REPLACE or UPSERT command that does UPDATE, or on failure, INSERT
514 To implement this cleanly requires that the table have a unique index
515 so duplicate checking can be easily performed. It is possible to
516 do it without a unique index if we require the user to LOCK the table
518 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00501.php
519 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00536.php
522 * Add NOVICE output level for helpful messages like automatic sequence/index
524 * -Add RESET SESSION command to reset all session state
525 * Add GUC to issue notice about statements that use unjoined tables
526 * Allow EXPLAIN to identify tables that were skipped because of
528 * Allow EXPLAIN output to be more easily processed by scripts
529 * Enable standard_conforming_strings
530 * Make standard_conforming_strings the default in 8.3?
532 When this is done, backslash-quote should be prohibited in non-E''
533 strings because of possible confusion over how such strings treat
534 backslashes. Basically, '' is always safe for a literal single
535 quote, while \' might or might not be based on the backslash
538 * Simplify dropping roles that have objects in several databases
539 * Allow COMMENT ON to accept an expression rather than just a string
540 * Allow the count returned by SELECT, etc to be to represent as an int64
541 to allow a higher range of values
542 * Add SQL99 WITH clause to SELECT
543 * Add SQL:2003 WITH RECURSIVE (hierarchical) queries to SELECT
544 * Add DEFAULT .. AS OWNER so permission checks are done as the table
547 This would be useful for SERIAL nextval() calls and CHECK constraints.
549 * Allow DISTINCT to work in multiple-argument aggregate calls
550 * Add column to pg_stat_activity that shows the progress of long-running
551 commands like CREATE INDEX and VACUUM
552 * Implement SQL:2003 window functions
553 * Improve failure message when DROP DATABASE is used on a database that
554 has prepared transactions
555 * Allow INSERT/UPDATE ... RETURNING inside a SELECT 'FROM' clause
557 http://archives.postgresql.org/pgsql-general/2006-09/msg00803.php
558 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00693.php
560 * Increase locking when DROPing objects so dependent objects cannot
561 get dropped while the DROP operation is happening
563 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00937.php
568 o Allow CREATE TABLE AS to determine column lengths for complex
569 expressions like SELECT col1 || col2
570 o Use more reliable method for CREATE DATABASE to get a consistent
572 o Fix transaction restriction checks for CREATE DATABASE and
575 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00133.php
577 o Have WITH CONSTRAINTS also create constraint indexes
578 http://archives.postgresql.org/pgsql-patches/2007-04/msg00149.php
581 o Allow UPDATE tab SET ROW (col, ...) = (SELECT...)
583 http://archives.postgresql.org/pgsql-hackers/2006-07/msg01306.php
584 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00865.php
588 o %Have ALTER TABLE RENAME rename SERIAL sequence names
589 o Add ALTER DOMAIN to modify the underlying data type
590 o %Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
592 http://archives.postgresql.org/pgsql-patches/2006-02/msg00168.php
594 o %Allow ALTER TABLE to change constraint deferrability and actions
595 o Add missing object types for ALTER ... SET SCHEMA
596 o Allow ALTER TABLESPACE to move to different directories
597 o Allow databases to be moved to different tablespaces
598 o Allow moving system tables to other tablespaces, where possible
600 Currently non-global system tables must be in the default database
601 tablespace. Global system tables can never be moved.
603 o Prevent parent tables from altering or dropping constraints
604 like CHECK that are inherited by child tables unless CASCADE
606 o %Prevent child tables from altering or dropping constraints
607 like CHECK that were inherited from the parent table
608 o Have ALTER INDEX update the name of a constraint using that index
609 o Add ALTER TABLE RENAME CONSTRAINT, update index name also
610 o Allow column display reordering by recording a display,
611 storage, and permanent id for every column?
613 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00782.php
619 o -Make CLUSTER preserve recently-dead tuples per MVCC requirements
620 o Automatically maintain clustering on a table
622 This might require some background daemon to maintain clustering
623 during periods of low usage. It might also require tables to be only
624 partially filled for easier reorganization. Another idea would
625 be to create a merged heap/index data file so an index lookup would
626 automatically access the heap data too. A third idea would be to
627 store heap rows in hashed groups, perhaps using a user-supplied
629 http://archives.postgresql.org/pgsql-performance/2004-08/msg00349.php
631 o %Add default clustering to system tables
633 To do this, determine the ideal cluster index for each system
634 table and set the cluster setting during initdb.
636 o %Add VERBOSE option to report tables as they are processed,
638 o -Add more logical syntax CLUSTER table USING index;
639 support current syntax for backward compatibility
644 o Allow COPY to report error lines and continue
646 This requires the use of a savepoint before each COPY line is
647 processed, with ROLLBACK on COPY failure.
649 o Allow COPY on a newly-created table to skip WAL logging
651 On crash recovery, the table involved in the COPY would
652 be removed or have its heap and index files truncated. One
653 issue is that no other backend should be able to add to
654 the table at the same time, which is something that is
660 o Allow column-level privileges
661 o %Allow GRANT/REVOKE permissions to be applied to all schema objects
664 The proposed syntax is:
665 GRANT SELECT ON ALL TABLES IN public TO phpuser;
666 GRANT SELECT ON NEW TABLES IN public TO phpuser;
668 o Allow GRANT/REVOKE permissions to be inherited by objects based on
671 o Allow SERIAL sequences to inherit permissions from the base table?
676 o Allow UPDATE/DELETE WHERE CURRENT OF cursor
678 This requires using the row ctid to map cursor rows back to the
679 original heap row. This become more complicated if WITH HOLD cursors
680 are to be supported because WITH HOLD cursors have a copy of the row
681 and no FOR UPDATE lock.
682 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01014.php
684 o Prevent DROP TABLE from dropping a row referenced by its own open
690 o Allow INSERT/UPDATE of the system-generated oid value for a row
691 o In rules, allow VALUES() to contain a mixture of 'old' and 'new'
697 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
699 o Add SET PATH for schemas?
701 This is basically the same as SET search_path.
704 * Referential Integrity
706 o Add MATCH PARTIAL referential integrity
707 o Change foreign key constraint for array -> element to mean element
709 o Enforce referential integrity for system tables
710 o Fix problem when cascading referential triggers make changes on
711 cascaded tables, seeing the tables in an intermediate state
713 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00174.php
714 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00174.php
716 o Allow DEFERRABLE and end-of-statement UNIQUE constraints?
718 This would allow UPDATE tab SET col = col + 1 to work if col has
719 a unique index. Currently, uniqueness checks are done while the
720 command is being executed, rather than at the end of the statement
722 http://people.planetpostgresql.org/greg/index.php?/archives/2006/06/10.html
723 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01458.php
726 * Server-Side Languages
729 o Fix RENAME to work on variables other than OLD/NEW
731 http://archives.postgresql.org/pgsql-hackers/2002-03/msg00591.php
732 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01615.php
733 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01587.php
735 o Allow function parameters to be passed by name,
736 get_employee_salary(12345 AS emp_id, 2001 AS tax_year)
737 o Add Oracle-style packages (Pavel)
739 A package would be a schema with session-local variables,
740 public/private functions, and initialization functions. It
741 is also possible to implement these capabilities
742 in all schemas and not use a separate "packages"
744 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00384.php
746 o Allow handling of %TYPE arrays, e.g. tab.col%TYPE[]
747 o Allow listing of record column names, and access to
748 record columns via variables, e.g. columns := r.(*),
751 http://archives.postgresql.org/pgsql-patches/2005-07/msg00458.php
752 http://archives.postgresql.org/pgsql-patches/2006-05/msg00302.php
753 http://archives.postgresql.org/pgsql-patches/2006-06/msg00031.php
756 o Add single-step debugging of functions
757 o Add support for WITH HOLD and SCROLL cursors
759 PL/pgSQL cursors should support the same syntax as
762 o Allow RETURN to return row or record functions
764 http://archives.postgresql.org/pgsql-patches/2005-11/msg00045.php
765 http://archives.postgresql.org/pgsql-patches/2006-08/msg00397.php
766 http://archives.postgresql.org/pgsql-hackers/2006-09/msg00388.php
768 o Fix problems with RETURN NEXT on tables with
769 dropped/added columns after function creation
771 http://archives.postgresql.org/pgsql-patches/2006-02/msg00165.php
773 o Allow row and record variables to be set to NULL constants,
774 and allow NULL tests on such variables
776 Because a row is not scalar, do not allow assignment
777 from NULL-valued scalars.
778 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00070.php
781 o Add table function support to pltcl, plpythonu
782 o Add support for polymorphic arguments and return types to
783 languages other than PL/PgSQL
784 o Add capability to create and call PROCEDURES
785 o Add support for OUT and INOUT parameters to languages other
787 o Add PL/PythonU tracebacks
789 http://archives.postgresql.org/pgsql-patches/2006-02/msg00288.php
791 o -Allow PL/PythonU to return boolean rather than 1/0
798 * Have pg_ctl look at PGHOST in case it is a socket directory?
799 * Allow pg_ctl to work properly with configuration files located outside
802 pg_ctl can not read the pid file because it isn't located in the
803 config directory but in the PGDATA directory. The solution is to
804 allow pg_ctl to read and understand postgresql.conf to find the
805 data_directory value.
807 * Make consistent use of long/short command options --- pg_ctl needs
808 long ones, pg_config doesn't have short ones, postgres doesn't have
809 enough long ones, etc.
814 o Have psql show current values for a sequence
815 o Move psql backslash database information into the backend, use
816 mnemonic commands? [psql]
818 This would allow non-psql clients to pull the same information out
819 of the database as psql.
821 o Fix psql's \d commands more consistent
823 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
824 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
826 o -Allow psql \pset boolean variables to set to fixed values, rather
828 o Consistently display privilege information for all objects in psql
829 o Add auto-expanded mode so expanded output is used if the row
830 length is wider than the screen width.
832 Consider using auto-expanded mode for backslash commands like \df+.
834 o Prevent tab completion of SET TRANSACTION from querying the
835 database and therefore preventing the transaction isolation
836 level from being set.
838 Currently SET <tab> causes a database lookup to check all
839 supported session variables. This query causes problems
840 because setting the transaction isolation level must be the
841 first statement of a transaction.
843 o Consider parsing the -c string into individual queries so each
844 is run in its own transaction
846 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00291.php
848 o Add a \set variable to control whether \s displays line numbers
850 Another option is to add \# which lists line numbers, and
851 allows command execution.
853 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00255.php
858 o %Add dumping of comments on index columns and composite type columns
859 o %Add full object name to the tag field. eg. for operators we need
860 '=(integer, integer)', instead of just '='.
861 o Add pg_dumpall custom format dumps?
862 o Remove unnecessary function pointer abstractions in pg_dump source
864 o Allow selection of individual object(s) of all types, not just
866 o In a selective dump, allow dumping of an object and all its
868 o Add options like pg_restore -l and -L to pg_dump
869 o Stop dumping CASCADE on DROP TYPE commands in clean mode
870 o Allow pg_dump --clean to drop roles that own objects or have
872 o -Add -f to pg_dumpall
879 Document differences between ecpg and the SQL standard and
880 information about the Informix-compatibility module.
882 o Solve cardinality > 1 for input descriptors / variables?
883 o Add a semantic check level, e.g. check if a table really exists
884 o fix handling of DB attributes that are arrays
885 o Use backend PREPARE/EXECUTE facility for ecpg where possible
887 o Fix nested C comments
888 o %sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
889 o Make SET CONNECTION thread-aware, non-standard?
890 o Allow multidimensional arrays
891 o Add internationalized message strings
892 o Implement COPY FROM STDIN
897 o Add PQescapeIdentifierConn()
898 o Prevent PQfnumber() from lowercasing unquoted the column name
900 PQfnumber() should never have been doing lowercasing, but
901 historically it has so we need a way to prevent it
903 o Allow statement results to be automatically batched to the client
905 Currently all statement results are transferred to the libpq
906 client before libpq makes the results available to the
907 application. This feature would allow the application to make
908 use of the first result rows while the rest are transferred, or
909 held on the server waiting for them to be requested by libpq.
910 One complexity is that a statement like SELECT 1/col could error
911 out mid-way through the result set.
912 o Fix SSL retry to avoid useless repeated connection attempts and
913 ensuing misleading error messages
914 o Consider disallowing multiple queries in PQexec() as an
915 additional barrier to SQL injection attacks
917 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00184.php
919 o Add PQexecf() that allows complex parameter substitution
921 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01803.php
926 * Add deferred trigger queue file
928 Right now all deferred trigger information is stored in backend
929 memory. This could exhaust memory for very large trigger queues.
930 This item involves dumping large queues into files.
932 * Allow triggers to be disabled in only the current session.
934 This is currently possible by starting a multi-statement transaction,
935 modifying the system tables, performing the desired SQL, restoring the
936 system tables, and committing the transaction. ALTER TABLE ...
937 TRIGGER requires a table lock so it is not ideal for this usage.
939 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
941 If the dump is known to be valid, allow foreign keys to be added
942 without revalidating the data.
944 * Allow statement-level triggers to access modified rows
945 * Support triggers on columns
947 http://archives.postgresql.org/pgsql-patches/2005-07/msg00107.php
949 * Allow AFTER triggers on system tables
951 System tables are modified in many places in the backend without going
952 through the executor and therefore not causing triggers to fire. To
953 complete this item, the functions that modify system tables will have
956 * Tighten trigger permission checks
958 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00564.php
960 * Allow BEFORE INSERT triggers on views
962 http://archives.postgresql.org/pgsql-general/2007-02/msg01466.php
969 * Flush cached query plans when the dependent objects change,
970 when the cardinality of parameters changes dramatically, or
971 when new ANALYZE statistics are available
973 A more complex solution would be to save multiple plans for different
974 cardinality and use the appropriate plan based on the EXECUTE values.
976 * Track dependencies in function bodies and recompile/invalidate
978 This is particularly important for references to temporary tables
979 in PL/PgSQL because PL/PgSQL caches query plans. The only workaround
980 in PL/PgSQL is to use EXECUTE. One complexity is that a function
981 might itself drop and recreate dependent tables, causing it to
982 invalidate its own query plan.
988 * Add UNIQUE capability to non-btree indexes
989 * Prevent index uniqueness checks when UPDATE does not modify the column
991 Uniqueness (index) checks are done when updating a column even if the
992 column is not modified by the UPDATE.
994 * Allow the creation of on-disk bitmap indexes which can be quickly
995 combined with other bitmap indexes
997 Such indexes could be more compact if there are only a few distinct values.
998 Such indexes can also be compressed. Keeping such indexes updated can be
1001 http://archives.postgresql.org/pgsql-patches/2005-07/msg00512.php
1002 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01107.php
1004 * -Allow use of indexes to search for NULLs
1005 * Allow accurate statistics to be collected on indexes with more than
1006 one column or expression indexes, perhaps using per-index statistics
1008 http://archives.postgresql.org/pgsql-performance/2006-10/msg00222.php
1009 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01131.php
1011 * -Allow the creation of indexes with mixed ascending/descending
1013 * Consider compressing indexes by storing key values duplicated in
1014 several rows as a single index entry
1016 This is difficult because it requires datatype-specific knowledge.
1018 * Add REINDEX CONCURRENTLY, like CREATE INDEX CONCURRENTLY
1020 This is difficult because you must upgrade to an exclusive table lock
1021 to replace the existing index file. CREATE INDEX CONCURRENTLY does not
1022 have this complication. This would allow index compaction without
1028 o Allow inherited tables to inherit indexes, UNIQUE constraints,
1029 and primary/foreign keys
1030 o Honor UNIQUE INDEX on base column in INSERTs/UPDATEs
1031 on inherited table, e.g. INSERT INTO inherit_table
1032 (unique_index_col) VALUES (dup) should fail
1034 The main difficulty with this item is the problem of
1035 creating an index that can span multiple tables.
1037 o Allow SELECT ... FOR UPDATE on inherited tables
1042 o Add more GIST index support for geometric data types
1043 o Allow GIST indexes to create certain complex index types, like
1044 digital trees (see Aoki)
1049 o Pack hash index buckets onto disk pages more efficiently
1051 Currently only one hash bucket can be stored on a page. Ideally
1052 several hash buckets could be stored on a single page and greater
1053 granularity used for the hash algorithm.
1055 http://archives.postgresql.org/pgsql-hackers/2004-06/msg00168.php
1057 o Consider sorting hash buckets so entries can be found using a
1058 binary search, rather than a linear scan
1059 o In hash indexes, consider storing the hash value with or instead
1061 o Add WAL logging for crash recovery
1062 o Allow multi-column hash indexes
1063 o During index creation, pre-sort the tuples to improve build speed
1065 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01199.php
1072 * Improve commit_delay handling to reduce fsync()
1073 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
1075 Ideally this requires a separate test program that can be run
1076 at initdb time or optionally later. Consider O_SYNC when
1079 * %Add an option to sync() before fsync()'ing checkpoint files
1080 * Add program to test if fsync has a delay compared to non-fsync
1081 * Reduce checkpoint performance degredation by forcing data to disk
1084 http://archives.postgresql.org/pgsql-patches/2006-12/msg00104.php
1085 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00337.php
1086 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00079.php
1092 * Allow free-behind capability for large sequential scans, perhaps using
1095 Posix_fadvise() can control both sequential/random file caching and
1096 free-behind behavior, but it is unclear how the setting affects other
1097 backends that also have the file open, and the feature is not supported
1098 on all operating systems.
1102 We could use a fixed row count and a +/- count to follow MVCC
1103 visibility rules, or a single cached value could be used and
1104 invalidated if anyone modifies the table. Another idea is to
1105 get a count directly from a unique index, but for this to be
1106 faster than a sequential scan it must avoid access to the heap
1107 to obtain tuple visibility information.
1109 * Provide a way to calculate an "estimated COUNT(*)"
1111 Perhaps by using the optimizer's cardinality estimates or random
1114 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00943.php
1116 * Allow data to be pulled directly from indexes
1118 Currently indexes do not have enough tuple visibility information
1119 to allow data to be pulled from the index without also accessing
1120 the heap. One way to allow this is to set a bit on index tuples
1121 to indicate if a tuple is currently visible to all transactions
1122 when the first valid heap lookup happens. This bit would have to
1123 be cleared when a heap tuple is expired.
1125 Another idea is to maintain a bitmap of heap pages where all rows
1126 are visible to all backends, and allow index lookups to reference
1127 that bitmap to avoid heap lookups, perhaps the same bitmap we might
1128 add someday to determine which heap pages need vacuuming. Frequently
1129 accessed bitmaps would have to be stored in shared memory. One 8k
1130 page of bitmaps could track 512MB of heap pages.
1132 * Consider automatic caching of statements at various levels:
1135 o Query execute plan
1138 * Allow sequential scans to take advantage of other concurrent
1139 sequential scans, also called "Synchronised Scanning"
1141 One possible implementation is to start sequential scans from the lowest
1142 numbered buffer in the shared cache, and when reaching the end wrap
1143 around to the beginning, rather than always starting sequential scans
1144 at the start of the table.
1146 http://archives.postgresql.org/pgsql-patches/2006-12/msg00076.php
1147 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00408.php
1148 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00784.php
1149 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00415.php
1151 * Consider increasing internal areas when shared buffers is increased
1153 http://archives.postgresql.org/pgsql-hackers/2005-10/msg01419.php
1155 * Consider decreasing the amount of memory used by PrivateRefCount
1157 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00797.php
1158 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00752.php
1165 * Improve speed with indexes
1167 For large table adjustments during VACUUM FULL, it is faster to cluster
1168 or reindex rather than update the index. Also, index updates can bloat
1171 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00024.php
1173 * Reduce lock time during VACUUM FULL by moving tuples with read lock,
1174 then write lock and truncate table
1176 Moved tuples are invisible to other backends so they don't require a
1177 write lock. However, the read lock promotion to write lock could lead
1178 to deadlock situations.
1180 * Auto-fill the free space map by scanning the buffer cache or by
1181 checking pages written by the background writer
1183 http://archives.postgresql.org/pgsql-hackers/2006-02/msg01125.php
1184 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00011.php
1186 * Create a bitmap of pages that need vacuuming
1188 Instead of sequentially scanning the entire table, have the background
1189 writer or some other process record pages that have expired rows, then
1190 VACUUM can look at just those pages rather than the entire table. In
1191 the event of a system crash, the bitmap would probably be invalidated.
1192 One complexity is that index entries still have to be vacuumed, and
1193 doing this without an index scan (by using the heap values to find the
1194 index entry) might be slow and unreliable, especially for user-defined
1197 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01188.php
1198 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00121.php
1200 * Allow FSM to return free space toward the beginning of the heap file,
1201 in hopes that empty pages at the end can be truncated by VACUUM
1202 * Allow FSM page return free space based on table clustering, to assist
1203 in maintaining clustering?
1204 * Consider shrinking expired tuples to just their headers
1206 http://archives.postgresql.org/pgsql-patches/2006-03/msg00142.php
1207 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01025.php
1209 * Allow heap reuse of UPDATEd rows if no indexed columns are changed,
1210 and old and new versions are on the same heap page?
1212 While vacuum handles DELETEs fine, updating of non-indexed columns, like
1213 counters, are difficult for VACUUM to handle efficiently. This method
1214 is possible for same-page updates because a single index row can be
1215 used to point to both old and new values.
1216 http://archives.postgresql.org/pgsql-hackers/2006-06/msg01305.php
1217 http://archives.postgresql.org/pgsql-hackers/2006-06/msg01534.php
1219 * Reuse index tuples that point to heap tuples that are not visible to
1221 * Improve dead row detection during multi-statement transactions usage
1223 http://archives.postgresql.org/pgsql-patches/2007-03/msg00358.php
1227 o Use free-space map information to guide refilling
1228 o %Issue log message to suggest VACUUM FULL if a table is nearly
1230 o Consider logging activity either to the logs or a system view
1231 o -Turn on by default
1232 o Allow multiple vacuums so large tables do not starve small
1235 http://archives.postgresql.org/pgsql-general/2007-01/msg00031.php
1237 o Improve control of auto-vacuum
1239 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00876.php
1245 * Fix priority ordering of read and write light-weight locks (Neil)
1247 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00893.php
1248 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00905.php
1250 * Fix problem when multiple subtransactions of the same outer transaction
1251 hold different types of locks, and one subtransaction aborts
1253 http://archives.postgresql.org/pgsql-hackers/2006-11/msg01011.php
1254 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00001.php
1255 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00435.php
1257 * Allow UPDATEs on only non-referential integrity columns not to conflict
1258 with referential integrity locks
1260 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00073.php
1262 * Add idle_in_transaction_timeout GUC so locks are not held for long
1267 Startup Time Improvements
1268 =========================
1270 * Experiment with multi-threaded backend for backend creation [thread]
1272 This would prevent the overhead associated with process creation. Most
1273 operating systems have trivial process creation time compared to
1274 database startup overhead, but a few operating systems (Win32,
1275 Solaris) might benefit from threading. Also explore the idea of
1276 a single session using multiple threads to execute a statement faster.
1278 * Experiment with multi-threaded backend better resource utilization
1280 This would allow a single query to make use of multiple CPU's or
1281 multiple I/O channels simultaneously. One idea is to create a
1282 background reader that can pre-fetch sequential and index scan
1283 pages needed by other backends. This could be expanded to allow
1284 concurrent reads from multiple devices in a partitioned table.
1286 * Add connection pooling
1288 It is unclear if this should be done inside the backend code or done
1289 by something external like pgpool. The passing of file descriptors to
1290 existing backends is one of the difficulties with a backend approach.
1296 * Eliminate need to write full pages to WAL before page modification [wal]
1298 Currently, to protect against partial disk page writes, we write
1299 full page images to WAL before they are modified so we can correct any
1300 partial page writes during recovery. These pages can also be
1301 eliminated from point-in-time archive files.
1303 o When off, write CRC to WAL and check file system blocks
1306 If CRC check fails during recovery, remember the page in case
1307 a later CRC for that page properly matches.
1309 o Write full pages during file system write and not when
1310 the page is modified in the buffer cache
1312 This allows most full page writes to happen in the background
1313 writer. It might cause problems for applying WAL on recovery
1314 into a partially-written page, but later the full page will be
1317 * Allow WAL traffic to be streamed to another server for stand-by
1319 * Reduce WAL traffic so only modified values are written rather than
1322 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01589.php
1324 * -Allow the pg_xlog directory location to be specified during initdb
1325 with a symlink back to the /data location
1326 * Allow WAL information to recover corrupted pg_controldata
1328 http://archives.postgresql.org/pgsql-patches/2006-06/msg00025.php
1330 * Find a way to reduce rotational delay when repeatedly writing
1333 Currently fsync of WAL requires the disk platter to perform a full
1334 rotation to fsync again. One idea is to write the WAL to different
1335 offsets that might reduce the rotational delay.
1337 * Allow buffered WAL writes and fsync
1339 Instead of guaranteeing recovery of all committed transactions, this
1340 would provide improved performance by delaying WAL writes and fsync
1341 so an abrupt operating system restart might lose a few seconds of
1342 committed transactions but still be consistent. We could perhaps
1343 remove the 'fsync' parameter (which results in an an inconsistent
1344 database) in favor of this capability.
1346 * Allow WAL logging to be turned off for a table, but the table
1347 might be dropped or truncated during crash recovery [walcontrol]
1349 Allow tables to bypass WAL writes and just fsync() dirty pages on
1350 commit. This should be implemented using ALTER TABLE, e.g. ALTER
1351 TABLE PERSISTENCE [ DROP | TRUNCATE | DEFAULT ]. Tables using
1352 non-default logging should not use referential integrity with
1353 default-logging tables. A table without dirty buffers during a
1354 crash could perhaps avoid the drop/truncate.
1356 * Allow WAL logging to be turned off for a table, but the table would
1357 avoid being truncated/dropped [walcontrol]
1359 To do this, only a single writer can modify the table, and writes
1360 must happen only on new pages so the new pages can be removed during
1361 crash recovery. Readers can continue accessing the table. Such
1362 tables probably cannot have indexes. One complexity is the handling
1363 of indexes on TOAST tables.
1366 Optimizer / Executor
1367 ====================
1369 * Improve selectivity functions for geometric operators
1370 * Allow ORDER BY ... LIMIT # to select high/low value without sort or
1371 index using a sequential scan for highest/lowest values
1373 Right now, if no index exists, ORDER BY ... LIMIT # requires we sort
1374 all values to return the high/low value. Instead The idea is to do a
1375 sequential scan to find the high/low value, thus avoiding the sort.
1376 MIN/MAX already does this, but not for LIMIT > 1.
1378 * Precompile SQL functions to avoid overhead
1379 * Create utility to compute accurate random_page_cost value
1380 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
1381 * Have EXPLAIN ANALYZE issue NOTICE messages when the estimated and
1382 actual row counts differ by a specified percentage
1383 * Consider using hash buckets to do DISTINCT, rather than sorting
1385 This would be beneficial when there are few distinct values. This is
1386 already used by GROUP BY.
1388 * Log statements where the optimizer row estimates were dramatically
1389 different from the number of rows actually found?
1390 * Consider compressed annealing to search for query plans
1392 This might replace GEQO, http://sixdemonbag.org/Djinni.
1394 * Improve merge join performance by allowing mark/restore of
1397 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00096.php
1401 Miscellaneous Performance
1402 =========================
1404 * Do async I/O for faster random read-ahead of data
1406 Async I/O allows multiple I/O requests to be sent to the disk with
1407 results coming back asynchronously.
1408 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00820.php
1410 * Use mmap() rather than SYSV shared memory or to write WAL files?
1412 This would remove the requirement for SYSV SHM but would introduce
1413 portability issues. Anonymous mmap (or mmap to /dev/zero) is required
1414 to prevent I/O overhead.
1416 * Consider mmap()'ing files into a backend?
1418 Doing I/O to large tables would consume a lot of address space or
1419 require frequent mapping/unmapping. Extending the file also causes
1420 mapping problems that might require mapping only individual pages,
1421 leading to thousands of mappings. Another problem is that there is no
1422 way to _prevent_ I/O to disk from the dirty shared buffers so changes
1423 could hit disk before WAL is written.
1425 * Add a script to ask system configuration questions and tune postgresql.conf
1426 * -Merge xmin/xmax/cmin/cmax back into three header fields
1427 * Consider ways of storing rows more compactly on disk
1429 o -Support a smaller header for short variable-length fields
1430 o Reduce the row header size?
1431 o Consider reducing on-disk varlena length from four bytes to
1432 two because a heap row cannot be more than 64k in length
1434 * Consider increasing NUM_CLOG_BUFFERS
1435 * Consider having the background writer update the transaction status
1436 hint bits before writing out the page
1437 * Allow user configuration of TOAST thresholds
1439 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00213.php
1441 * Allow configuration of backend priorities via the operating system
1443 Though backend priorities make priority inversion during lock
1444 waits possible, research shows that this is not a huge problem.
1445 http://archives.postgresql.org/pgsql-general/2007-02/msg00493.php
1447 * Consider reducing memory used for shared buffer reference count
1449 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00752.php
1456 * Add use of 'const' for variables in source tree
1457 * Move some things from /contrib into main tree
1458 * %Remove warnings created by -Wcast-align
1459 * Move platform-specific ps status display info from ps_status.c to ports
1460 * Add optional CRC checksum to heap and index pages
1461 * Improve documentation to build only interfaces (Marc)
1462 * Remove or relicense modules that are not under the BSD license, if possible
1463 * %Remove memory/file descriptor freeing before ereport(ERROR)
1464 * Acquire lock on a relation before building a relcache entry for it
1465 * %Promote debug_query_string into a server-side function current_query()
1466 * Allow cross-compiling by generating the zic database on the target system
1467 * Improve NLS maintenance of libpgport messages linked onto applications
1468 * Allow ecpg to work with MSVC and BCC
1469 * Add xpath_array() to /contrib/xml2 to return results as an array
1470 * Clean up casting in /contrib/isn
1472 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00245.php
1474 * Allow building in directories containing spaces
1476 This is probably not possible because 'gmake' and other compiler tools
1477 do not fully support quoting of paths with spaces.
1479 * Fix sgmltools so PDFs can be generated with bookmarks
1480 * Use UTF8 encoding for NLS messages so all server encodings can
1482 * Update Bonjour to work with newer cross-platform SDK
1484 http://archives.postgresql.org/pgsql-hackers/2006-09/msg02238.php
1485 http://archives.postgresql.org/pgsql-patches/2006-10/msg00048.php
1487 * Split out libpq pgpass and environment documentation sections to make
1488 it easier for non-developers to find
1489 * Consider detoasting keys before sorting
1490 * Consider GnuTLS if OpenSSL license becomes a problem
1492 http://archives.postgresql.org/pgsql-patches/2006-05/msg00040.php
1493 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01213.php
1495 * Use strlcpy() rather than our StrNCpy() macro
1497 http://archives.postgresql.org/pgsql-hackers/2006-09/msg02108.php
1499 * Consider changing documentation format from SGML to XML
1501 http://archives.postgresql.org/pgsql-docs/2006-12/msg00152.php
1503 * -Move NAMEDATALEN from postgres_ext.h to pg_config_manual.h
1504 * Consider making NAMEDATALEN more configurable in future releases
1505 * Update our code to handle 64-bit timezone files to match the zic
1506 source code, which now uses them
1511 o Remove configure.in check for link failure when cause is found
1512 o Remove readdir() errno patch when runtime/mingwex/dirent.c rev
1514 o Remove psql newline patch when we find out why mingw outputs an
1516 o Allow psql to use readline once non-US code pages work with
1518 o Re-enable timezone output on log_line_prefix '%t' when a
1519 shorter timezone string is available
1520 o Fix problem with shared memory on the Win32 Terminal Server
1521 o Improve signal handling
1523 http://archives.postgresql.org/pgsql-patches/2005-06/msg00027.php
1525 o -Add long file support for binary pg_dump output
1526 o Check WSACancelBlockingCall() for interrupts [win32intr]
1529 * Wire Protocol Changes
1531 o Allow dynamic character set handling
1532 o Add decoded type, length, precision
1534 o Update clients to use data types, typmod, schema.table.column names
1535 of result sets using new statement protocol
1541 * Add pre-parsing phase that converts non-ISO syntax to supported
1544 This could allow SQL written for other databases to run without
1547 * Allow plug-in modules to emulate features from other databases
1548 * SQL*Net listener that makes PostgreSQL appear as an Oracle database
1550 * Allow statements across databases or servers with transaction
1553 This can be done using dblink and two-phase commit.
1555 * Add the features of packages
1557 o Make private objects accessible only to objects in the same schema
1558 o Allow current_schema.objname to access current schema objects
1559 o Add session variables
1560 o Allow nested schemas
1562 * Consider allowing control of upper/lower case folding of unquoted
1565 http://archives.postgresql.org/pgsql-hackers/2004-04/msg00818.php
1566 http://archives.postgresql.org/pgsql-hackers/2006-10/msg01527.php
1569 Features We Do _Not_ Want
1570 =========================
1572 * All backends running as threads in a single process (not wanted)
1574 This eliminates the process protection we get from the current setup.
1575 Thread creation is usually the same overhead as process creation on
1576 modern systems, so it seems unwise to use a pure threaded model.
1578 * Optimizer hints (not wanted)
1580 Optimizer hints are used to work around problems in the optimizer. We
1581 would rather have the problems reported and fixed.
1582 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00506.php
1583 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00517.php
1584 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00663.php
1586 * Allow AS in "SELECT col AS label" to be optional (not wanted)
1588 Because we support postfix operators, it isn't possible to make AS
1589 optional and continue to use bison.
1590 http://archives.postgresql.org/pgsql-sql/2006-08/msg00164.php
1592 * Embedded server (not wanted)
1594 While PostgreSQL clients runs fine in limited-resource environments, the
1595 server requires multiple processes and a stable pool of resources to
1596 run reliabily and efficiently. Stripping down the PostgreSQL server
1597 to run in the same process address space as the client application
1598 would add too much complexity and failure cases.