3 Current maintainer: Bruce Momjian (bruce@momjian.us)
4 Last updated: Fri Feb 29 09:56:41 EST 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
33 * Check for unreferenced table files created by transactions that were
34 in-progress when the server terminated abruptly
36 http://archives.postgresql.org/pgsql-patches/2006-06/msg00096.php
38 * Set proper permissions on non-system schemas during db creation
40 Currently all schemas are owned by the super-user because they are copied
41 from the template1 database. However, since all objects are inherited
42 from the template database, it is not clear that setting schemas to the db
45 * Add function to report the time of the most recent server reload
46 * Allow statistics collector information to be pulled from the collector
47 process directly, rather than requiring the collector to write a
48 filesystem file twice a second?
49 * Allow log_min_messages to be specified on a per-module basis
51 This would allow administrators to see more detailed information from
52 specific sections of the backend, e.g. checkpoints, autovacuum, etc.
53 Another idea is to allow separate configuration files for each module,
54 or allow arbitrary SET commands to be passed to them.
56 * Simplify ability to create partitioned tables
58 This would allow creation of partitioned tables without requiring
59 creation of triggers or rules for INSERT/UPDATE/DELETE, and constraints
60 for rapid partition selection. Options could include range and hash
63 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00375.php
64 http://archives.postgresql.org/pgsql-hackers/2007-04/msg00151.php
66 * Allow auto-selection of partitioned tables for min/max() operations
67 * Allow more complex user/database default GUC settings
69 Currently ALTER USER and ALTER DATABASE support per-user and
70 per-database defaults. Consider adding per-user-and-database
71 defaults so things like search_path can be defaulted for a
72 specific user connecting to a specific database.
74 * Allow custom variable classes that can restrict who can set the values
76 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00911.php
78 * Implement the SQL standard mechanism whereby REVOKE ROLE revokes only
79 the privilege granted by the invoking role, and not those granted
82 http://archives.postgresql.org/pgsql-bugs/2007-05/msg00010.php
84 * Allow SSL authentication/encryption over unix domain sockets
86 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00924.php
90 o Allow pg_hba.conf to specify host names along with IP addresses
92 Host name lookup could occur when the postmaster reads the
93 pg_hba.conf file, or when the backend starts. Another
94 solution would be to reverse lookup the connection IP and
95 check that hostname against the host names in pg_hba.conf.
96 We could also then check that the host name maps to the IP
99 o %Allow postgresql.conf file values to be changed via an SQL
100 API, perhaps using SET GLOBAL
101 o Allow the server to be stopped/restarted via an SQL API
102 o Issue a warning if a change-on-restart-only postgresql.conf value
103 is modified and the server config files are reloaded
108 o Allow a database in tablespace t1 with tables created in
109 tablespace t2 to be used as a template for a new database created
110 with default tablespace t2
112 Currently all objects in the default database tablespace must
113 have default tablespace specifications. This is because new
114 databases are created by copying directories. If you mix default
115 tablespace tables and tablespace-specified tables in the same
116 directory, creating a new database from such a mixed directory
117 would create a new database with tables that had incorrect
118 explicit tablespaces. To fix this would require modifying
119 pg_class in the newly copied database, which we don't currently
122 o Allow reporting of which objects are in which tablespaces
124 This item is difficult because a tablespace can contain objects
125 from multiple databases. There is a server-side function that
126 returns the databases which use a specific tablespace, so this
127 requires a tool that will call that function and connect to each
128 database to find the objects in each database for that tablespace.
130 o Allow WAL replay of CREATE TABLESPACE to work when the directory
131 structure on the recovery computer is different from the original
133 o Allow per-tablespace quotas
136 * Point-In-Time Recovery (PITR)
138 o Allow a warm standby system to also allow read-only statements
141 This is useful for checking PITR recovery.
142 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00050.php
144 o %Create dump tool for write-ahead logs for use in determining
145 transaction id for point-in-time recovery
146 o Allow recovery.conf to support the same syntax as
147 postgresql.conf, including quoting
149 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00497.php
157 * Change NUMERIC to enforce the maximum precision
158 * Reduce storage space for small NUMERICs
160 http://archives.postgresql.org/pgsql-hackers/2007-02/msg01331.php
161 http://archives.postgresql.org/pgsql-patches/2007-02/msg00505.php
163 * Fix data types where equality comparison isn't intuitive, e.g. box
164 * Add support for public SYNONYMs
166 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00519.php
168 * Fix CREATE CAST on DOMAINs
170 http://archives.postgresql.org/pgsql-hackers/2006-05/msg00072.php
171 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01681.php
173 * Add support for SQL-standard GENERATED/IDENTITY columns
175 http://archives.postgresql.org/pgsql-hackers/2006-07/msg00543.php
176 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00038.php
177 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00344.php
178 http://archives.postgresql.org/pgsql-patches/2007-05/msg00076.php
180 * Improve XML support
182 http://developer.postgresql.org/index.php/XML_Support
184 * Consider placing all sequences in a single table, or create a system
188 o Allow infinite dates and intervals just like infinite timestamps
189 o Merge hardwired timezone names with the TZ database; allow either
190 kind everywhere a TZ name is currently taken
191 o Allow TIMESTAMP WITH TIME ZONE to store the original timezone
192 information, either zone name or offset from UTC [timezone]
194 If the TIMESTAMP value is stored with a time zone name, interval
195 computations should adjust based on the time zone rules.
197 o Fix SELECT '0.01 years'::interval, '0.01 months'::interval
198 o Add a GUC variable to allow output of interval values in ISO8601
200 o Have timestamp subtraction not call justify_hours()?
202 http://archives.postgresql.org/pgsql-sql/2006-10/msg00059.php
204 o Improve timestamptz subtraction to be DST-aware
206 Currently subtracting one date from another that crosses a
207 daylight savings time adjustment can return '1 day 1 hour', but
208 adding that back to the first date returns a time one hour in
209 the future. This is caused by the adjustment of '25 hours' to
210 '1 day 1 hour', and '1 day' is the same time the next day, even
211 if daylight savings adjustments are involved.
213 o Fix interval display to support values exceeding 2^31 hours
214 o Add overflow checking to timestamp and interval arithmetic
215 o Extend timezone code to allow 64-bit values so we can
216 represent years beyond 2038
218 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01363.php
220 o Use LC_TIME for localized weekday/month names, rather than
223 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00390.php
225 o Add ISO INTERVAL handling
227 http://archives.postgresql.org/pgsql-hackers/2006-01/msg00250.php
228 http://archives.postgresql.org/pgsql-bugs/2006-04/msg00248.php
230 o Support ISO INTERVAL syntax if units cannot be determined from
231 the string, and are supplied after the string
233 The SQL standard states that the units after the string
234 specify the units of the string, e.g. INTERVAL '2' MINUTE
235 should return '00:02:00'. The current behavior has the units
236 restrict the interval value to the specified unit or unit
237 range, INTERVAL '70' SECOND returns '00:00:10'.
239 For syntax that isn't uniquely ISO or PG syntax, like '1' or
240 '1:30', treat as ISO if there is a range specification clause,
241 and as PG if there no clause is present, e.g. interpret '1:30'
242 MINUTE TO SECOND as '1 minute 30 seconds', and interpret
243 '1:30' as '1 hour, 30 minutes'.
245 This makes common cases like SELECT INTERVAL '1' MONTH
246 SQL-standard results. The SQL standard supports a limited
247 number of unit combinations and doesn't support unit names in
248 the string. The PostgreSQL syntax is more flexible in the
249 range of units supported, e.g. PostgreSQL supports '1 year 1
250 hour', while the SQL standard does not.
252 o Add support for year-month syntax, INTERVAL '50-6' YEAR
254 o Interpret INTERVAL '1 year' MONTH as CAST (INTERVAL '1
255 year' AS INTERVAL MONTH), and this should return '12 months'
256 o Round or truncate values to the requested precision, e.g.
257 INTERVAL '11 months' AS YEAR should return one or zero
258 o Support precision, CREATE TABLE foo (a INTERVAL MONTH(3))
263 o Delay resolution of array expression's data type so assignment
264 coercion can be performed on empty array expressions
265 o Add support for arrays of domains
267 http://archives.postgresql.org/pgsql-patches/2007-05/msg00114.php
269 o Allow single-byte header storage for array elements
274 o Improve vacuum of large objects, like contrib/vacuumlo?
275 o Add security checking for large objects
276 o Auto-delete large objects when referencing row is deleted
278 contrib/lo offers this functionality.
280 o Allow read/write into TOAST values like large objects
282 This requires the TOAST column to be stored EXTERNAL.
284 o Add API for 64-bit large object access
286 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00781.php
290 * Add locale-aware MONEY type, and support multiple currencies
292 http://archives.postgresql.org/pgsql-general/2005-08/msg01432.php
293 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01181.php
295 * MONEY dumps in a locale-specific format making it difficult to
296 restore to a system with a different locale
297 * Allow MONEY to be easily cast to/from other numeric data types
304 * Allow INET subnet tests using non-constants to be indexed
305 * Allow to_date() and to_timestamp() accept localized month names
306 * Fix to_date()-related functions to consistently issue errors
308 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00915.php
310 * Add missing parameter handling in to_char()
312 http://archives.postgresql.org/pgsql-hackers/2005-12/msg00948.php
314 * Allow substring/replace() to get/set bit values
315 * Allow to_char() on interval values to accumulate the highest unit
318 Some special format flag would be required to request such
319 accumulation. Such functionality could also be added to EXTRACT.
320 Prevent accumulation that crosses the month/day boundary because of
321 the uneven number of days in a month.
323 o to_char(INTERVAL '1 hour 5 minutes', 'MI') => 65
324 o to_char(INTERVAL '43 hours 20 minutes', 'MI' ) => 2600
325 o to_char(INTERVAL '43 hours 20 minutes', 'WK:DD:HR:MI') => 0:1:19:20
326 o to_char(INTERVAL '3 years 5 months','MM') => 41
328 * Implement inlining of set-returning functions defined in SQL
329 * Allow SQL-language functions to return results from RETURNING queries
331 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00665.php
333 * Allow SQL-language functions to reference parameters by parameter name
335 Currently SQL-language functions can only refer to dollar parameters,
338 * Add SPI_gettypmod() to return the typemod for a TupleDesc
339 * Enforce typmod for function inputs, function results and parameters for
340 spi_prepare'd statements called from PLs
342 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01403.php
344 * Allow holdable cursors in SPI
345 * Tighten function permission checks
347 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00568.php
349 * Fix IS OF so it matches the ISO specification, and add documentation
351 http://archives.postgresql.org/pgsql-patches/2003-08/msg00060.php
352 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00060.php
354 * Add missing operators for geometric data types
356 Some geometric types do not have the full suite of geometric operators,
359 * Implement Boyer-Moore searching in strpos()
361 http://archives.postgresql.org/pgsql-patches/2007-08/msg00012.php
365 Multi-Language Support
366 ======================
368 * Add NCHAR (as distinguished from ordinary varchar),
369 * Allow locale to be set at database creation
371 Currently locale can only be set during initdb. No global tables have
372 locale-aware columns. However, the database template used during
373 database creation might have locale-aware indexes. The indexes would
374 need to be reindexed to match the new locale.
376 * Allow encoding on a per-column basis optionally using the ICU library:
378 Right now only one encoding is allowed per database. [locale]
380 http://archives.postgresql.org/pgsql-hackers/2005-03/msg00932.php
381 http://archives.postgresql.org/pgsql-patches/2005-08/msg00309.php
382 http://archives.postgresql.org/pgsql-patches/2006-03/msg00233.php
383 http://archives.postgresql.org/pgsql-hackers/2006-09/msg00662.php
385 * Add CREATE COLLATE? [locale]
386 * Support multiple simultaneous character sets, per SQL92
387 * Improve UTF8 combined character handling?
388 * Add octet_length_server() and octet_length_client()
389 * Make octet_length_client() the same as octet_length()?
390 * Fix problems with wrong runtime encoding conversion for NLS message files
391 * Add URL to more complete multi-byte regression tests
393 http://archives.postgresql.org/pgsql-hackers/2005-07/msg00272.php
395 * Fix ILIKE and regular expressions to handle case insensitivity
396 properly in multibyte encodings
398 http://archives.postgresql.org/pgsql-bugs/2005-10/msg00001.php
399 http://archives.postgresql.org/pgsql-patches/2005-11/msg00173.php
401 * Set client encoding based on the client operating system encoding
403 Currently client_encoding is set in postgresql.conf, which
404 defaults to the server encoding.
406 http://archives.postgresql.org/pgsql-hackers/2006-08/msg01696.php
413 * Automatically create rules on views so they are updateable, per SQL99
415 We can only auto-create rules for simple views. For more complex
416 cases users will still have to write rules manually.
418 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00586.php
419 http://archives.postgresql.org/pgsql-patches/2006-08/msg00255.php
421 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
422 * Allow NOTIFY in rules involving conditionals
423 * Allow VIEW/RULE recompilation when the underlying tables change
425 Another issue is whether underlying table changes should be reflected
426 in the view, e.g. should SELECT * show additional columns if they
427 are added after the view is created.
429 * Make it possible to use RETURNING together with conditional DO INSTEAD
430 rules, such as for partitioning setups
432 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00577.php
434 * Add the ability to automatically create materialized views
436 Right now materialized views require the user to create triggers on the
437 main table to keep the summary table current. SQL syntax should be able
438 to manager the triggers and summary table automatically. A more
439 sophisticated implementation would automatically retrieve from the
440 summary table when the main table is referenced, if possible.
447 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
448 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
449 * %Allow SET CONSTRAINTS to be qualified by schema/table name
450 * %Add a separate TRUNCATE permission
452 Currently only the owner can TRUNCATE a table because triggers are not
453 called, and the table is locked in exclusive mode.
455 * Allow PREPARE of cursors
456 * Allow finer control over the caching of prepared query plans
458 Currently queries prepared via the libpq API are planned on first
459 execute using the supplied parameters --- allow SQL PREPARE to do the
460 same. Also, allow control over replanning prepared queries either
461 manually or automatically when statistics for execute parameters
462 differ dramatically from those used during planning.
464 * Improve logging of prepared transactions recovered during startup
466 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00092.php
468 * Allow LISTEN/NOTIFY to store info in memory rather than tables?
470 Currently LISTEN/NOTIFY information is stored in pg_listener. Storing
471 such information in memory would improve performance.
473 * Add optional textual message to NOTIFY
475 This would allow an informational message to be added to the notify
476 message, perhaps indicating the row modified or other custom
479 * Allow multiple identical NOTIFY events to always be communicated to the
480 client, rather than sent as a single notification to the listener
481 * Add a GUC variable to warn about non-standard SQL usage in queries
482 * Add SQL-standard MERGE command, typically used to merge two tables
485 This is similar to UPDATE, then for unmatched rows, INSERT.
486 Whether concurrent access allows modifications which could cause
487 row loss is implementation independent.
489 * Add REPLACE or UPSERT command that does UPDATE, or on failure, INSERT
492 To implement this cleanly requires that the table have a unique index
493 so duplicate checking can be easily performed. It is possible to
494 do it without a unique index if we require the user to LOCK the table
497 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00501.php
498 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00536.php
500 * Add NOVICE output level for helpful messages like automatic sequence/index
502 * Add GUC to issue notice about statements that use unjoined tables
503 * Allow EXPLAIN to identify tables that were skipped because of
505 * Allow EXPLAIN output to be more easily processed by scripts, perhaps XML
506 * Enable standard_conforming_strings
507 * Make standard_conforming_strings the default in 8.5?
509 When this is done, backslash-quote should be prohibited in non-E''
510 strings because of possible confusion over how such strings treat
511 backslashes. Basically, '' is always safe for a literal single
512 quote, while \' might or might not be based on the backslash
515 * Simplify dropping roles that have objects in several databases
516 * Allow COMMENT ON to accept an expression rather than just a string
517 * Allow the count returned by SELECT, etc to be represented as an int64
518 to allow a higher range of values
519 * Add SQL99 WITH clause to SELECT
520 * Add SQL:2003 WITH RECURSIVE (hierarchical) queries to SELECT
521 * Add DEFAULT .. AS OWNER so permission checks are done as the table
524 This would be useful for SERIAL nextval() calls and CHECK constraints.
526 * Allow DISTINCT to work in multiple-argument aggregate calls
527 * Add column to pg_stat_activity that shows the progress of long-running
528 commands like CREATE INDEX and VACUUM
529 * Implement SQL:2003 window functions
530 * Improve failure message when DROP DATABASE is used on a database that
531 has prepared transactions
532 * Allow INSERT/UPDATE ... RETURNING inside a SELECT 'FROM' clause
534 http://archives.postgresql.org/pgsql-general/2006-09/msg00803.php
535 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00693.php
537 * Increase locking when DROPing objects so dependent objects cannot
538 get dropped while the DROP operation is happening
540 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00937.php
542 * -Allow AS in "SELECT col AS label" to be optional in certain cases
546 o Allow CREATE TABLE AS to determine column lengths for complex
547 expressions like SELECT col1 || col2
549 o Have WITH CONSTRAINTS also create constraint indexes
551 http://archives.postgresql.org/pgsql-patches/2007-04/msg00149.php
554 o Allow UPDATE tab SET ROW (col, ...) = (SELECT...)
556 http://archives.postgresql.org/pgsql-hackers/2006-07/msg01306.php
557 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00865.php
559 o Research self-referential UPDATEs that see inconsistent row versions
560 in read-committed mode
562 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00507.php
563 http://archives.postgresql.org/pgsql-hackers/2007-06/msg00016.php
565 o Allow GLOBAL temporary tables to exist as empty by default in
568 http://archives.postgresql.org/pgsql-hackers/2007-07/msg00006.php
573 o %Have ALTER TABLE RENAME rename SERIAL sequence names
574 o Have ALTER SEQUENCE RENAME rename the sequence name stored
575 in the sequence table
577 http://archives.postgresql.org/pgsql-bugs/2007-09/msg00092.php
578 http://archives.postgresql.org/pgsql-bugs/2007-10/msg00007.php
580 o Add ALTER DOMAIN to modify the underlying data type
581 o %Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
583 http://archives.postgresql.org/pgsql-patches/2006-02/msg00168.php
585 o %Allow ALTER TABLE to change constraint deferrability and actions
586 o Add missing object types for ALTER ... SET SCHEMA
587 o Allow ALTER TABLESPACE to move to different directories
588 o Allow databases to be moved to different tablespaces
589 o Allow moving system tables to other tablespaces, where possible
591 Currently non-global system tables must be in the default database
592 tablespace. Global system tables can never be moved.
594 o Prevent parent tables from altering or dropping constraints
595 like CHECK that are inherited by child tables unless CASCADE
597 o %Prevent child tables from altering or dropping constraints
598 like CHECK that were inherited from the parent table
599 o Have ALTER INDEX update the name of a constraint using that index
600 o Add ALTER TABLE RENAME CONSTRAINT, update index name also
601 o Allow column display reordering by recording a display,
602 storage, and permanent id for every column?
604 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00782.php
610 o Automatically maintain clustering on a table
612 This might require some background daemon to maintain clustering
613 during periods of low usage. It might also require tables to be only
614 partially filled for easier reorganization. Another idea would
615 be to create a merged heap/index data file so an index lookup would
616 automatically access the heap data too. A third idea would be to
617 store heap rows in hashed groups, perhaps using a user-supplied
619 http://archives.postgresql.org/pgsql-performance/2004-08/msg00349.php
621 o %Add default clustering to system tables
623 To do this, determine the ideal cluster index for each system
624 table and set the cluster setting during initdb.
626 o %Add VERBOSE option to report tables as they are processed,
632 o Allow COPY to report error lines and continue
634 This requires the use of a savepoint before each COPY line is
635 processed, with ROLLBACK on COPY failure.
637 o Allow COPY on a newly-created table to skip WAL logging
639 On crash recovery, the table involved in the COPY would
640 be removed or have its heap and index files truncated. One
641 issue is that no other backend should be able to add to
642 the table at the same time, which is something that is
643 currently allowed. This currently is done if the table is
644 created inside the same transaction block as the COPY because
645 no other backends can see the table.
650 o Allow column-level privileges
651 o %Allow GRANT/REVOKE permissions to be applied to all schema objects
654 The proposed syntax is:
655 GRANT SELECT ON ALL TABLES IN public TO phpuser;
656 GRANT SELECT ON NEW TABLES IN public TO phpuser;
658 o Allow GRANT/REVOKE permissions to be inherited by objects based on
661 o Allow SERIAL sequences to inherit permissions from the base table?
666 o Prevent DROP TABLE from dropping a row referenced by its own open
672 o Allow INSERT/UPDATE of the system-generated oid value for a row
673 o In rules, allow VALUES() to contain a mixture of 'old' and 'new'
679 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
683 * Referential Integrity
685 o Add MATCH PARTIAL referential integrity
686 o Change foreign key constraint for array -> element to mean element
688 o Fix problem when cascading referential triggers make changes on
689 cascaded tables, seeing the tables in an intermediate state
691 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00174.php
692 http://archives.postgresql.org/pgsql-hackers/2005-09/msg00174.php
694 o Allow DEFERRABLE and end-of-statement UNIQUE constraints?
696 This would allow UPDATE tab SET col = col + 1 to work if col has
697 a unique index. Currently, uniqueness checks are done while the
698 command is being executed, rather than at the end of the statement
701 http://people.planetpostgresql.org/greg/index.php?/archives/2006/06/10.html
702 http://archives.postgresql.org/pgsql-hackers/2006-09/msg01458.php
705 * Server-Side Languages
708 o Fix RENAME to work on variables other than OLD/NEW
710 http://archives.postgresql.org/pgsql-hackers/2002-03/msg00591.php
711 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01615.php
712 http://archives.postgresql.org/pgsql-hackers/2007-01/msg01587.php
714 o Allow function parameters to be passed by name,
715 get_employee_salary(12345 AS emp_id, 2001 AS tax_year)
716 o Allow handling of %TYPE arrays, e.g. tab.col%TYPE[]
717 o Allow listing of record column names, and access to
718 record columns via variables, e.g. columns := r.(*),
721 http://archives.postgresql.org/pgsql-patches/2005-07/msg00458.php
722 http://archives.postgresql.org/pgsql-patches/2006-05/msg00302.php
723 http://archives.postgresql.org/pgsql-patches/2006-06/msg00031.php
725 o Add support for SCROLL cursors
726 o Add support for WITH HOLD cursors
727 o Allow row and record variables to be set to NULL constants,
728 and allow NULL tests on such variables
730 Because a row is not scalar, do not allow assignment
731 from NULL-valued scalars.
733 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00070.php
736 o Add table function support to pltcl, plpythonu
737 o Add support for polymorphic arguments and return types to
738 languages other than PL/PgSQL
739 o Add capability to create and call PROCEDURES
740 o Add support for OUT and INOUT parameters to languages other
742 o Add PL/PythonU tracebacks
744 http://archives.postgresql.org/pgsql-patches/2006-02/msg00288.php
746 o Allow data to be passed in native language formats, rather
749 http://archives.postgresql.org/pgsql-hackers/2007-05/msg00289.php
755 * Have pg_ctl look at PGHOST in case it is a socket directory?
756 * Allow pg_ctl to work properly with configuration files located outside
759 pg_ctl can not read the pid file because it isn't located in the
760 config directory but in the PGDATA directory. The solution is to
761 allow pg_ctl to read and understand postgresql.conf to find the
762 data_directory value.
765 o Have psql show current values for a sequence
766 o Move psql backslash database information into the backend, use
767 mnemonic commands? [psql]
769 This would allow non-psql clients to pull the same information out
770 of the database as psql.
772 o Fix psql's \d commands more consistent
774 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
775 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00014.php
777 o Consistently display privilege information for all objects in psql
778 o Add auto-expanded mode so expanded output is used if the row
779 length is wider than the screen width.
781 Consider using auto-expanded mode for backslash commands like \df+.
783 o Prevent tab completion of SET TRANSACTION from querying the
784 database and therefore preventing the transaction isolation
785 level from being set.
787 Currently SET <tab> causes a database lookup to check all
788 supported session variables. This query causes problems
789 because setting the transaction isolation level must be the
790 first statement of a transaction.
792 o Add a \set variable to control whether \s displays line numbers
794 Another option is to add \# which lists line numbers, and
795 allows command execution.
797 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00255.php
801 o %Add dumping of comments on index columns and composite type columns
802 o %Add full object name to the tag field. eg. for operators we need
803 '=(integer, integer)', instead of just '='.
804 o Add pg_dumpall custom format dumps?
805 o Allow selection of individual object(s) of all types, not just
807 o In a selective dump, allow dumping of an object and all its
809 o Add options like pg_restore -l and -L to pg_dump
810 o Stop dumping CASCADE on DROP TYPE commands in clean mode
811 o Allow pg_dump --clean to drop roles that own objects or have
813 o Change pg_dump so that a comment on the dumped database is
814 applied to the loaded database, even if the database has a
815 different name. This will require new backend syntax, perhaps
816 COMMENT ON CURRENT DATABASE.
817 o Remove unnecessary function pointer abstractions in pg_dump source
824 Document differences between ecpg and the SQL standard and
825 information about the Informix-compatibility module.
827 o Solve cardinality > 1 for input descriptors / variables?
828 o Add a semantic check level, e.g. check if a table really exists
829 o fix handling of DB attributes that are arrays
830 o Use backend PREPARE/EXECUTE facility for ecpg where possible
832 o Fix nested C comments
833 o %sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
834 o Make SET CONNECTION thread-aware, non-standard?
835 o Allow multidimensional arrays
836 o Add internationalized message strings
837 o Implement COPY FROM STDIN
841 o Add PQescapeIdentifierConn()
842 o Prevent PQfnumber() from lowercasing unquoted the column name
844 PQfnumber() should never have been doing lowercasing, but
845 historically it has so we need a way to prevent it
847 o Allow statement results to be automatically batched to the client
849 Currently all statement results are transferred to the libpq
850 client before libpq makes the results available to the
851 application. This feature would allow the application to make
852 use of the first result rows while the rest are transferred, or
853 held on the server waiting for them to be requested by libpq.
854 One complexity is that a statement like SELECT 1/col could error
855 out mid-way through the result set.
857 o Consider disallowing multiple queries in PQexec() as an
858 additional barrier to SQL injection attacks
860 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00184.php
862 o Add PQexecf() that allows complex parameter substitution
864 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01803.php
869 * Add deferred trigger queue file
871 Right now all deferred trigger information is stored in backend
872 memory. This could exhaust memory for very large trigger queues.
873 This item involves dumping large queues into files.
875 * Allow triggers to be disabled in only the current session.
877 This is currently possible by starting a multi-statement transaction,
878 modifying the system tables, performing the desired SQL, restoring the
879 system tables, and committing the transaction. ALTER TABLE ...
880 TRIGGER requires a table lock so it is not ideal for this usage.
882 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
884 If the dump is known to be valid, allow foreign keys to be added
885 without revalidating the data.
887 * Allow statement-level triggers to access modified rows
888 * Support triggers on columns
890 http://archives.postgresql.org/pgsql-patches/2005-07/msg00107.php
892 * Allow AFTER triggers on system tables
894 System tables are modified in many places in the backend without going
895 through the executor and therefore not causing triggers to fire. To
896 complete this item, the functions that modify system tables will have
899 * Tighten trigger permission checks
901 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00564.php
903 * Allow BEFORE INSERT triggers on views
905 http://archives.postgresql.org/pgsql-general/2007-02/msg01466.php
907 * Add ability to trigger on TRUNCATE
909 http://archives.postgresql.org/pgsql-sql/2008-01/msg00050.php
916 * Add UNIQUE capability to non-btree indexes
917 * Prevent index uniqueness checks when UPDATE does not modify the column
919 Uniqueness (index) checks are done when updating a column even if the
920 column is not modified by the UPDATE.
922 * Allow the creation of on-disk bitmap indexes which can be quickly
923 combined with other bitmap indexes
925 Such indexes could be more compact if there are only a few distinct values.
926 Such indexes can also be compressed. Keeping such indexes updated can be
929 http://archives.postgresql.org/pgsql-patches/2005-07/msg00512.php
930 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01107.php
932 * Allow accurate statistics to be collected on indexes with more than
933 one column or expression indexes, perhaps using per-index statistics
935 http://archives.postgresql.org/pgsql-performance/2006-10/msg00222.php
936 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01131.php
938 * Consider compressing indexes by storing key values duplicated in
939 several rows as a single index entry
940 * Add REINDEX CONCURRENTLY, like CREATE INDEX CONCURRENTLY
942 This is difficult because you must upgrade to an exclusive table lock
943 to replace the existing index file. CREATE INDEX CONCURRENTLY does not
944 have this complication. This would allow index compaction without
947 * Allow multiple indexes to be created concurrently, ideally via a
948 single heap scan, and have a restore of a pg_dump somehow use it
950 http://archives.postgresql.org/pgsql-general/2007-05/msg01274.php
955 o Allow inherited tables to inherit indexes, UNIQUE constraints,
956 and primary/foreign keys
957 o Honor UNIQUE INDEX on base column in INSERTs/UPDATEs
958 on inherited table, e.g. INSERT INTO inherit_table
959 (unique_index_col) VALUES (dup) should fail
961 The main difficulty with this item is the problem of
962 creating an index that can span multiple tables.
964 o Allow SELECT ... FOR UPDATE on inherited tables
969 o Add more GIST index support for geometric data types
970 o Allow GIST indexes to create certain complex index types, like
971 digital trees (see Aoki)
975 http://archives.postgresql.org/pgsql-hackers/2007-09/msg00051.php
977 o Pack hash index buckets onto disk pages more efficiently
979 Currently only one hash bucket can be stored on a page. Ideally
980 several hash buckets could be stored on a single page and greater
981 granularity used for the hash algorithm.
983 http://archives.postgresql.org/pgsql-hackers/2004-06/msg00168.php
985 o Consider sorting hash buckets so entries can be found using a
986 binary search, rather than a linear scan
987 o In hash indexes, consider storing the hash value with or instead
989 o Add WAL logging for crash recovery
990 o Allow multi-column hash indexes
991 o During index creation, pre-sort the tuples to improve build speed
993 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01199.php
1000 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
1002 Ideally this requires a separate test program that can be run
1003 at initdb time or optionally later. Consider O_SYNC when
1006 * Add program to test if fsync has a delay compared to non-fsync
1014 We could use a fixed row count and a +/- count to follow MVCC
1015 visibility rules, or a single cached value could be used and
1016 invalidated if anyone modifies the table. Another idea is to
1017 get a count directly from a unique index, but for this to be
1018 faster than a sequential scan it must avoid access to the heap
1019 to obtain tuple visibility information.
1021 * Provide a way to calculate an "estimated COUNT(*)"
1023 Perhaps by using the optimizer's cardinality estimates or random
1026 http://archives.postgresql.org/pgsql-hackers/2005-11/msg00943.php
1028 * Allow data to be pulled directly from indexes
1030 Currently indexes do not have enough tuple visibility information
1031 to allow data to be pulled from the index without also accessing
1032 the heap. One way to allow this is to set a bit on index tuples
1033 to indicate if a tuple is currently visible to all transactions
1034 when the first valid heap lookup happens. This bit would have to
1035 be cleared when a heap tuple is expired.
1037 Another idea is to maintain a bitmap of heap pages where all rows
1038 are visible to all backends, and allow index lookups to reference
1039 that bitmap to avoid heap lookups, perhaps the same bitmap we might
1040 add someday to determine which heap pages need vacuuming. Frequently
1041 accessed bitmaps would have to be stored in shared memory. One 8k
1042 page of bitmaps could track 512MB of heap pages.
1044 A third idea would be for a heap scan to check if all rows are visible
1045 and if so set a per-table flag which can be checked by index scans.
1046 Any change to the table would have to clear the flag. To detect
1047 changes during the heap scan a counter could be set at the start and
1048 checked at the end --- if it is the same, the table has not been
1049 modified --- any table change would increment the counter.
1051 * Consider automatic caching of statements at various levels:
1054 o Query execute plan
1057 * Consider increasing internal areas when shared buffers is increased
1059 http://archives.postgresql.org/pgsql-hackers/2005-10/msg01419.php
1061 * Consider decreasing the amount of memory used by PrivateRefCount
1063 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00797.php
1064 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00752.php
1072 * Improve speed with indexes
1074 For large table adjustments during VACUUM FULL, it is faster to cluster
1075 or reindex rather than update the index. Also, index updates can bloat
1078 http://archives.postgresql.org/pgsql-hackers/2007-03/msg00024.php
1079 http://archives.postgresql.org/pgsql-performance/2007-05/msg00296.php
1081 * Auto-fill the free space map by scanning the buffer cache or by
1082 checking pages written by the background writer
1084 http://archives.postgresql.org/pgsql-hackers/2006-02/msg01125.php
1085 http://archives.postgresql.org/pgsql-hackers/2006-03/msg00011.php
1087 * Create a bitmap of pages that need vacuuming
1089 Instead of sequentially scanning the entire table, have the background
1090 writer or some other process record pages that have expired rows, then
1091 VACUUM can look at just those pages rather than the entire table. In
1092 the event of a system crash, the bitmap would probably be invalidated.
1093 One complexity is that index entries still have to be vacuumed, and
1094 doing this without an index scan (by using the heap values to find the
1095 index entry) might be slow and unreliable, especially for user-defined
1098 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01188.php
1099 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00121.php
1101 * Allow FSM to return free space toward the beginning of the heap file,
1102 in hopes that empty pages at the end can be truncated by VACUUM
1103 * Allow FSM page return free space based on table clustering, to assist
1104 in maintaining clustering?
1105 * Improve dead row detection during multi-statement transactions usage
1107 http://archives.postgresql.org/pgsql-patches/2007-03/msg00358.php
1111 o %Issue log message to suggest VACUUM FULL if a table is nearly
1113 o Improve control of auto-vacuum
1115 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00876.php
1117 o Prevent long-lived temporary tables from causing frozen-xid
1118 advancement starvation
1120 The problem is that autovacuum cannot vacuum them to set frozen xids;
1121 only the session that created them can do that.
1122 http://archives.postgresql.org/pgsql-general/2007-06/msg01645.php
1128 * Fix priority ordering of read and write light-weight locks (Neil)
1130 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00893.php
1131 http://archives.postgresql.org/pgsql-hackers/2004-11/msg00905.php
1133 * Fix problem when multiple subtransactions of the same outer transaction
1134 hold different types of locks, and one subtransaction aborts
1136 http://archives.postgresql.org/pgsql-hackers/2006-11/msg01011.php
1137 http://archives.postgresql.org/pgsql-hackers/2006-12/msg00001.php
1138 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00435.php
1140 * Allow UPDATEs on only non-referential integrity columns not to conflict
1141 with referential integrity locks
1143 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00073.php
1145 * Add idle_in_transaction_timeout GUC so locks are not held for long
1148 * Improve deadlock detection when a page cleaning lock conflicts
1149 with a shared buffer that is pinned
1151 http://archives.postgresql.org/pgsql-bugs/2008-01/msg00138.php
1152 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00873.php
1153 http://archives.postgresql.org/pgsql-committers/2008-01/msg00365.php
1157 Startup Time Improvements
1158 =========================
1160 * Experiment with multi-threaded backend for backend creation [thread]
1162 This would prevent the overhead associated with process creation. Most
1163 operating systems have trivial process creation time compared to
1164 database startup overhead, but a few operating systems (Win32,
1165 Solaris) might benefit from threading. Also explore the idea of
1166 a single session using multiple threads to execute a statement faster.
1172 * Eliminate need to write full pages to WAL before page modification [wal]
1174 Currently, to protect against partial disk page writes, we write
1175 full page images to WAL before they are modified so we can correct any
1176 partial page writes during recovery. These pages can also be
1177 eliminated from point-in-time archive files.
1179 o When off, write CRC to WAL and check file system blocks
1182 If CRC check fails during recovery, remember the page in case
1183 a later CRC for that page properly matches.
1185 o Write full pages during file system write and not when
1186 the page is modified in the buffer cache
1188 This allows most full page writes to happen in the background
1189 writer. It might cause problems for applying WAL on recovery
1190 into a partially-written page, but later the full page will be
1193 * Allow WAL traffic to be streamed to another server for stand-by
1195 * Reduce WAL traffic so only modified values are written rather than
1198 http://archives.postgresql.org/pgsql-hackers/2007-03/msg01589.php
1200 * Allow WAL information to recover corrupted pg_controldata
1202 http://archives.postgresql.org/pgsql-patches/2006-06/msg00025.php
1204 * Find a way to reduce rotational delay when repeatedly writing
1207 Currently fsync of WAL requires the disk platter to perform a full
1208 rotation to fsync again. One idea is to write the WAL to different
1209 offsets that might reduce the rotational delay.
1211 * Allow WAL logging to be turned off for a table, but the table
1212 might be dropped or truncated during crash recovery [walcontrol]
1214 Allow tables to bypass WAL writes and just fsync() dirty pages on
1215 commit. This should be implemented using ALTER TABLE, e.g. ALTER
1216 TABLE PERSISTENCE [ DROP | TRUNCATE | DEFAULT ]. Tables using
1217 non-default logging should not use referential integrity with
1218 default-logging tables. A table without dirty buffers during a
1219 crash could perhaps avoid the drop/truncate.
1221 * Allow WAL logging to be turned off for a table, but the table would
1222 avoid being truncated/dropped [walcontrol]
1224 To do this, only a single writer can modify the table, and writes
1225 must happen only on new pages so the new pages can be removed during
1226 crash recovery. Readers can continue accessing the table. Such
1227 tables probably cannot have indexes. One complexity is the handling
1228 of indexes on TOAST tables.
1231 Optimizer / Executor
1232 ====================
1234 * Improve selectivity functions for geometric operators
1235 * Precompile SQL functions to avoid overhead
1236 * Create utility to compute accurate random_page_cost value
1237 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
1238 * Have EXPLAIN ANALYZE issue NOTICE messages when the estimated and
1239 actual row counts differ by a specified percentage
1240 * Consider using hash buckets to do DISTINCT, rather than sorting
1242 This would be beneficial when there are few distinct values. This is
1243 already used by GROUP BY.
1245 * Log statements where the optimizer row estimates were dramatically
1246 different from the number of rows actually found?
1247 * Consider compressed annealing to search for query plans
1249 This might replace GEQO, http://sixdemonbag.org/Djinni.
1251 * Improve merge join performance by allowing mark/restore of
1254 http://archives.postgresql.org/pgsql-hackers/2007-01/msg00096.php
1258 Miscellaneous Performance
1259 =========================
1261 * Do async I/O for faster random read-ahead of data
1263 Async I/O allows multiple I/O requests to be sent to the disk with
1264 results coming back asynchronously.
1266 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00820.php
1267 http://archives.postgresql.org/pgsql-hackers/2007-12/msg00027.php
1269 * Use mmap() rather than SYSV shared memory or to write WAL files?
1271 This would remove the requirement for SYSV SHM but would introduce
1272 portability issues. Anonymous mmap (or mmap to /dev/zero) is required
1273 to prevent I/O overhead.
1275 * Consider mmap()'ing files into a backend?
1277 Doing I/O to large tables would consume a lot of address space or
1278 require frequent mapping/unmapping. Extending the file also causes
1279 mapping problems that might require mapping only individual pages,
1280 leading to thousands of mappings. Another problem is that there is no
1281 way to _prevent_ I/O to disk from the dirty shared buffers so changes
1282 could hit disk before WAL is written.
1284 * Add a script to ask system configuration questions and tune postgresql.conf
1285 * Consider ways of storing rows more compactly on disk
1287 o Reduce the row header size?
1288 o Consider reducing on-disk varlena length from four bytes to
1289 two because a heap row cannot be more than 64k in length
1291 * Consider increasing NUM_CLOG_BUFFERS
1292 * Consider having the background writer update the transaction status
1293 hint bits before writing out the page
1295 Implementing this requires the background writer to have access to system
1296 catalogs and the transaction status log.
1298 * Allow user configuration of TOAST thresholds
1300 http://archives.postgresql.org/pgsql-hackers/2007-02/msg00213.php
1302 * Allow configuration of backend priorities via the operating system
1304 Though backend priorities make priority inversion during lock
1305 waits possible, research shows that this is not a huge problem.
1307 http://archives.postgresql.org/pgsql-general/2007-02/msg00493.php
1309 * Experiment with multi-threaded backend better resource utilization
1311 This would allow a single query to make use of multiple CPU's or
1312 multiple I/O channels simultaneously. One idea is to create a
1313 background reader that can pre-fetch sequential and index scan
1314 pages needed by other backends. This could be expanded to allow
1315 concurrent reads from multiple devices in a partitioned table.
1321 * Add use of 'const' for variables in source tree
1322 * Move some things from contrib into main tree
1323 * %Remove warnings created by -Wcast-align
1324 * Move platform-specific ps status display info from ps_status.c to ports
1325 * Add optional CRC checksum to heap and index pages
1326 * Improve documentation to build only interfaces (Marc)
1327 * Remove or relicense modules that are not under the BSD license, if possible
1328 * Acquire lock on a relation before building a relcache entry for it
1329 * Allow cross-compiling by generating the zic database on the target system
1330 * Improve NLS maintenance of libpgport messages linked onto applications
1331 * Clean up casting in contrib/isn
1333 http://archives.postgresql.org/pgsql-hackers/2006-11/msg00245.php
1335 * Use UTF8 encoding for NLS messages so all server encodings can
1337 * Update Bonjour to work with newer cross-platform SDK
1339 http://archives.postgresql.org/pgsql-hackers/2006-09/msg02238.php
1340 http://archives.postgresql.org/pgsql-patches/2006-10/msg00048.php
1342 * Consider detoasting keys before sorting
1343 * Consider GnuTLS if OpenSSL license becomes a problem
1345 http://archives.postgresql.org/pgsql-patches/2006-05/msg00040.php
1346 http://archives.postgresql.org/pgsql-hackers/2006-12/msg01213.php
1348 * Consider changing documentation format from SGML to XML
1350 http://archives.postgresql.org/pgsql-docs/2006-12/msg00152.php
1352 * Consider making NAMEDATALEN more configurable in future releases
1353 * Update our code to handle 64-bit timezone files to match the zic
1354 source code, which now uses them
1355 * Have configure choose integer datetimes by default
1357 http://archives.postgresql.org/pgsql-patches/2007-05/msg00046.php
1359 * Support scoped IPv6 addresses
1361 http://archives.postgresql.org/pgsql-bugs/2007-05/msg00111.php
1363 * Consider allowing 64-bit integers and floats to be passed by value on
1366 Also change 32-bit floats (float4) to be passed by value at the same
1373 o Remove configure.in check for link failure when cause is found
1374 o Remove readdir() errno patch when runtime/mingwex/dirent.c rev
1376 o Remove psql newline patch when we find out why mingw outputs an
1378 o Allow psql to use readline once non-US code pages work with
1380 o Fix problem with shared memory on the Win32 Terminal Server
1381 o Improve signal handling
1383 http://archives.postgresql.org/pgsql-patches/2005-06/msg00027.php
1387 * Wire Protocol Changes
1389 o Allow dynamic character set handling
1390 o Add decoded type, length, precision
1392 o Update clients to use data types, typmod, schema.table.column names
1393 of result sets using new statement protocol
1399 * Add pre-parsing phase that converts non-ISO syntax to supported
1402 This could allow SQL written for other databases to run without
1405 * Allow plug-in modules to emulate features from other databases
1406 * Add features of Oracle-style packages (Pavel)
1408 A package would be a schema with session-local variables,
1409 public/private functions, and initialization functions. It
1410 is also possible to implement these capabilities
1411 in any schema and not use a separate "packages"
1414 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00384.php
1416 * Consider allowing control of upper/lower case folding of unquoted
1419 http://archives.postgresql.org/pgsql-hackers/2004-04/msg00818.php
1420 http://archives.postgresql.org/pgsql-hackers/2006-10/msg01527.php
1422 * Add autonomous transactions
1424 http://archives.postgresql.org/pgsql-hackers/2008-01/msg00893.php
1428 Features We Do _Not_ Want
1429 =========================
1431 * All backends running as threads in a single process (not wanted)
1433 This eliminates the process protection we get from the current setup.
1434 Thread creation is usually the same overhead as process creation on
1435 modern systems, so it seems unwise to use a pure threaded model.
1437 * Optimizer hints (not wanted)
1439 Optimizer hints are used to work around problems in the optimizer. We
1440 would rather have the problems reported and fixed.
1442 http://archives.postgresql.org/pgsql-hackers/2006-08/msg00506.php
1443 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00517.php
1444 http://archives.postgresql.org/pgsql-hackers/2006-10/msg00663.php
1447 Because we support postfix operators, it isn't possible to make AS
1448 optional and continue to use bison.
1449 http://archives.postgresql.org/pgsql-hackers/2003-04/msg00436.php
1451 http://archives.postgresql.org/pgsql-sql/2006-08/msg00164.php
1453 * Embedded server (not wanted)
1455 While PostgreSQL clients runs fine in limited-resource environments, the
1456 server requires multiple processes and a stable pool of resources to
1457 run reliabily and efficiently. Stripping down the PostgreSQL server
1458 to run in the same process address space as the client application
1459 would add too much complexity and failure cases.