4 Current maintainer: Bruce Momjian (pgman@candle.pha.pa.us)
5 Last updated: Fri Aug 26 15:41:03 EDT 2005
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 8.1 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
23 * %Remove behavior of postmaster -o after making postmaster/postgres
25 * %Allow pooled connections to list all prepared queries
27 This would allow an application inheriting a pooled connection to know
28 the queries prepared in the current session.
30 * Allow major upgrades without dump/reload, perhaps using pg_upgrade
32 * Check for unreferenced table files created by transactions that were
33 in-progress when the server terminated abruptly
34 * Allow administrators to safely terminate individual sessions either
35 via an SQL function or SIGTERM
37 Lock table corruption following SIGTERM of an individual backend
38 has been reported in 8.0. A possible cause was fixed in 8.1, but
39 it is unknown whether other problems exist. This item mostly
40 requires additional testing rather than of writing any new code.
42 * %Set proper permissions on non-system schemas during db creation
44 Currently all schemas are owned by the super-user because they are
45 copied from the template1 database.
47 * Support table partitioning that allows a single table to be stored
48 in subtables that are partitioned based on the primary key or a WHERE
52 * Improve replication solutions
56 You can use any of the master/slave replication servers to use a
57 standby server for data warehousing. To allow read/write queries to
58 multiple servers, you need multi-master replication like pgcluster.
60 o Allow replication over unreliable or non-persistent links
65 o %Add "include file" functionality in postgresql.conf
66 o %Allow commenting of variables in postgresql.conf to restore them
69 Currently, if a variable is commented out, it keeps the
70 previous uncommented value until a server restarted.
72 o %Allow pg_hba.conf settings to be controlled via SQL
74 This would add a function to load the SQL table from
75 pg_hba.conf, and one to writes its contents to the flat file.
76 The table should have a line number that is a float so rows
77 can be inserted between existing rows, e.g. row 2.5 goes
78 between row 2 and row 3.
80 o %Allow postgresql.conf file values to be changed via an SQL
81 API, perhaps using SET GLOBAL
82 o Allow the server to be stopped/restarted via an SQL API
87 * Allow a database in tablespace t1 with tables created in
88 tablespace t2 to be used as a template for a new database created
89 with default tablespace t2
91 All objects in the default database tablespace must have default
92 tablespace specifications. This is because new databases are
93 created by copying directories. If you mix default tablespace
94 tables and tablespace-specified tables in the same directory,
95 creating a new database from such a mixed directory would create a
96 new database with tables that had incorrect explicit tablespaces.
97 To fix this would require modifying pg_class in the newly copied
98 database, which we don't currently do.
100 * Allow reporting of which objects are in which tablespaces
102 This item is difficult because a tablespace can contain objects
103 from multiple databases. There is a server-side function that
104 returns the databases which use a specific tablespace, so this
105 requires a tool that will call that function and connect to each
106 database to find the objects in each database for that tablespace.
108 o %Add a GUC variable to control the tablespace for temporary objects
111 It could start with a random tablespace from a supplied list and
112 cycle through the list.
114 o Allow WAL replay of CREATE TABLESPACE to work when the directory
115 structure on the recovery computer is different from the original
117 o Allow per-tablespace quotas
120 * Point-In-Time Recovery (PITR)
122 o Allow point-in-time recovery to archive partially filled
123 write-ahead logs [pitr]
125 Currently only full WAL files are archived. This means that the
126 most recent transactions aren't available for recovery in case
127 of a disk failure. This could be triggered by a user command or
130 o Automatically force archiving of partially-filled WAL files when
131 pg_stop_backup() is called or the server is stopped
133 Doing this will allow administrators to know more easily when
134 the archive contins all the files needed for point-in-time
137 o %Create dump tool for write-ahead logs for use in determining
138 transaction id for point-in-time recovery
139 o Allow a warm standby system to also allow read-only queries
142 This is useful for checking PITR recovery.
144 o Allow the PITR process to be debugged and data examined
150 * Allow server log information to be output as INSERT statements
152 This would allow server log information to be easily loaded into
153 a database for analysis.
155 * %Add ability to monitor the use of temporary sort files
156 * Allow server logs to be remotely read and removed using SQL commands
162 * Improve the MONEY data type
164 Change the MONEY data type to use DECIMAL internally, with special
165 locale-aware output formatting.
167 * Change NUMERIC to enforce the maximum precision, and increase it
168 * Add NUMERIC division operator that doesn't round?
170 Currently NUMERIC _rounds_ the result to the specified precision.
171 This means division can return a result that multiplied by the
172 divisor is greater than the dividend, e.g. this returns a value > 10:
174 SELECT (10::numeric(2,0) / 6::numeric(2,0))::numeric(2,0) * 6;
176 The positive modulus result returned by NUMERICs might be considered
177 inaccurate, in one sense.
179 * Have sequence dependency track use of DEFAULT sequences,
181 * %Disallow changing default expression of a SERIAL column?
182 * Fix data types where equality comparison isn't intuitive, e.g. box
183 * %Prevent INET cast to CIDR if the unmasked bits are not zero, or
185 * %Prevent INET cast to CIDR from droping netmask, SELECT '1.1.1.1'::inet::cidr
186 * Allow INET + INT4 to increment the host part of the address, or
187 throw an error on overflow
188 * %Add 'tid != tid ' operator for use in corruption recovery
193 o Allow infinite dates just like infinite timestamps
194 o Add a GUC variable to allow output of interval values in ISO8601
196 o Merge hardwired timezone names with the TZ database; allow either
197 kind everywhere a TZ name is currently taken
198 o Allow customization of the known set of TZ names (generalize the
199 present australian_timezones hack)
200 o Allow TIMESTAMP WITH TIME ZONE to store the original timezone
201 information, either zone name or offset from UTC [timezone]
203 If the TIMESTAMP value is stored with a time zone name, interval
204 computations should adjust based on the time zone rules.
206 o Add ISO INTERVAL handling
207 o Add support for day-time syntax, INTERVAL '1 2:03:04' DAY TO
209 o Add support for year-month syntax, INTERVAL '50-6' YEAR TO MONTH
210 o For syntax that isn't uniquely ISO or PG syntax, like '1:30' or
211 '1', treat as ISO if there is a range specification clause,
212 and as PG if there no clause is present, e.g. interpret
213 '1:30' MINUTE TO SECOND as '1 minute 30 seconds', and
214 interpret '1:30' as '1 hour, 30 minutes'
215 o Interpret INTERVAL '1 year' MONTH as CAST (INTERVAL '1 year' AS
216 INTERVAL MONTH), and this should return '12 months'
217 o Round or truncate values to the requested precision, e.g.
218 INTERVAL '11 months' AS YEAR should return one or zero
219 o Support precision, CREATE TABLE foo (a INTERVAL MONTH(3))
224 o Allow NULLs in arrays
225 o Delay resolution of array expression's data type so assignment
226 coercion can be performed on empty array expressions
231 o Improve vacuum of large objects, like /contrib/vacuumlo?
232 o Add security checking for large objects
233 o Auto-delete large objects when referencing row is deleted
235 /contrib/lo offers this functionality.
237 o Allow read/write into TOAST values like large objects
239 This requires the TOAST column to be stored EXTERNAL.
245 * Allow INET subnet tests using non-constants to be indexed
246 * Add transaction_timestamp(), statement_timestamp(), clock_timestamp()
249 Current CURRENT_TIMESTAMP returns the start time of the current
250 transaction, and gettimeofday() returns the wallclock time. This will
251 make time reporting more consistent and will allow reporting of
252 the statement start time.
254 * %Add pg_get_acldef(), pg_get_typedefault(), and pg_get_attrdef()
255 * Allow to_char() to print localized month names
256 * Allow functions to have a schema search path specified at creation time
257 * Allow substring/replace() to get/set bit values
258 * Allow to_char() on interval values to accumulate the highest unit
261 Some special format flag would be required to request such
262 accumulation. Such functionality could also be added to EXTRACT.
263 Prevent accumulation that crosses the month/day boundary because of
264 the uneven number of days in a month.
266 o to_char(INTERVAL '1 hour 5 minutes', 'MI') => 65
267 o to_char(INTERVAL '43 hours 20 minutes', 'MI' ) => 2600
268 o to_char(INTERVAL '43 hours 20 minutes', 'WK:DD:HR:MI') => 0:1:19:20
269 o to_char(INTERVAL '3 years 5 months','MM') => 41
271 * Add sleep() function, remove from regress.c
274 Multi-Language Support
275 ======================
277 * Add NCHAR (as distinguished from ordinary varchar),
278 * Allow locale to be set at database creation
280 Currently locale can only be set during initdb. No global tables have
281 locale-aware columns. However, the database template used during
282 database creation might have locale-aware indexes. The indexes would
283 need to be reindexed to match the new locale.
285 * Allow encoding on a per-column basis
287 Right now only one encoding is allowed per database.
289 * Support multiple simultaneous character sets, per SQL92
290 * Improve UTF8 combined character handling?
291 * Add octet_length_server() and octet_length_client()
292 * Make octet_length_client() the same as octet_length()?
293 * Fix problems with wrong runtime encoding conversion for NLS message files
299 * %Automatically create rules on views so they are updateable, per SQL99
301 We can only auto-create rules for simple views. For more complex
302 cases users will still have to write rules.
304 * Add the functionality for WITH CHECK OPTION clause of CREATE VIEW
305 * Allow NOTIFY in rules involving conditionals
306 * Allow VIEW/RULE recompilation when the underlying tables change
312 * Change LIMIT/OFFSET and FETCH/MOVE to use int8
313 * Add CORRESPONDING BY to UNION/INTERSECT/EXCEPT
314 * Add ROLLUP, CUBE, GROUPING SETS options to GROUP BY
315 * %Allow SET CONSTRAINTS to be qualified by schema/table name
316 * %Allow TRUNCATE ... CASCADE/RESTRICT
318 This is like DELETE CASCADE, but truncates.
320 * %Add a separate TRUNCATE permission
322 Currently only the owner can TRUNCATE a table because triggers are not
323 called, and the table is locked in exclusive mode.
325 * Allow PREPARE of cursors
326 * Allow PREPARE to automatically determine parameter types based on the SQL
328 * Allow finer control over the caching of prepared query plans
330 Currently, queries prepared via the libpq API are planned on first
331 execute using the supplied parameters --- allow SQL PREPARE to do the
332 same. Also, allow control over replanning prepared queries either
333 manually or automatically when statistics for execute parameters
334 differ dramatically from those used during planning.
336 * Allow LISTEN/NOTIFY to store info in memory rather than tables?
338 Currently LISTEN/NOTIFY information is stored in pg_listener. Storing
339 such information in memory would improve performance.
341 * Add optional textual message to NOTIFY
343 This would allow an informational message to be added to the notify
344 message, perhaps indicating the row modified or other custom
347 * Add a GUC variable to warn about non-standard SQL usage in queries
348 * Add MERGE command that does UPDATE/DELETE, or on failure, INSERT (rules,
350 * Add NOVICE output level for helpful messages like automatic sequence/index
352 * %Add COMMENT ON for all cluster global objects (roles, databases
354 * %Make row-wise comparisons work per SQL spec
355 * Add RESET CONNECTION command to reset all session state
357 This would include resetting of all variables (RESET ALL), dropping of
358 temporary tables, removing any NOTIFYs, cursors, open transactions,
359 prepared queries, currval()s, etc. This could be used for connection
360 pooling. We could also change RESET ALL to have this functionality.
361 The difficult of this features is allowing RESET ALL to not affect
362 changes made by the interface driver for its internal use. One idea
363 is for this to be a protocol-only feature. Another approach is to
364 notify the protocol when a RESET CONNECTION command is used.
366 * Add GUC to issue notice about queries that use unjoined tables
367 * Allow EXPLAIN to identify tables that were skipped because of
369 * Allow EXPLAIN output to be more easily processed by scripts
374 o Allow CREATE TABLE AS to determine column lengths for complex
375 expressions like SELECT col1 || col2
377 o Use more reliable method for CREATE DATABASE to get a consistent
380 o Add ON COMMIT capability to CREATE TABLE AS ... SELECT
384 o Allow UPDATE to handle complex aggregates [update]?
385 o Allow an alias to be provided for the target table in
388 This is not SQL-spec but many DBMSs allow it.
390 o Allow UPDATE tab SET ROW (col, ...) = (...) for updating multiple
396 o %Have ALTER TABLE RENAME rename SERIAL sequence names
397 o Add ALTER DOMAIN to modify the underlying data type
398 o %Allow ALTER TABLE ... ALTER CONSTRAINT ... RENAME
399 o %Allow ALTER TABLE to change constraint deferrability and actions
400 o Add missing object types for ALTER ... SET SCHEMA
401 o Allow ALTER TABLESPACE to move to different directories
402 o Allow databases to be moved to different tablespaces
403 o Allow moving system tables to other tablespaces, where possible
405 Currently non-global system tables must be in the default database
406 tablespace. Global system tables can never be moved.
408 o %Disallow dropping of an inherited constraint
409 o %Prevent child tables from altering or dropping constraints
410 like CHECK that were inherited from the parent table
415 o Automatically maintain clustering on a table
417 This might require some background daemon to maintain clustering
418 during periods of low usage. It might also require tables to be only
419 paritally filled for easier reorganization. Another idea would
420 be to create a merged heap/index data file so an index lookup would
421 automatically access the heap data too. A third idea would be to
422 store heap rows in hashed groups, perhaps using a user-supplied
425 o %Add default clustering to system tables
427 To do this, determine the ideal cluster index for each system
428 table and set the cluster setting during initdb.
433 o Allow COPY to report error lines and continue
435 This requires the use of a savepoint before each COPY line is
436 processed, with ROLLBACK on COPY failure.
438 o %Have COPY return the number of rows loaded/unloaded?
443 o Allow column-level privileges
444 o %Allow GRANT/REVOKE permissions to be applied to all schema objects
447 The proposed syntax is:
448 GRANT SELECT ON ALL TABLES IN public TO phpuser;
449 GRANT SELECT ON NEW TABLES IN public TO phpuser;
451 * Allow GRANT/REVOKE permissions to be inherited by objects based on
457 o Allow UPDATE/DELETE WHERE CURRENT OF cursor
459 This requires using the row ctid to map cursor rows back to the
460 original heap row. This become more complicated if WITH HOLD cursors
461 are to be supported because WITH HOLD cursors have a copy of the row
462 and no FOR UPDATE lock.
464 o Prevent DROP TABLE from dropping a row referenced by its own open
467 o %Allow pooled connections to list all open WITH HOLD cursors
469 Because WITH HOLD cursors exist outside transactions, this allows
470 them to be listed so they can be closed.
475 o Allow INSERT/UPDATE of the system-generated oid value for a row
476 o Allow INSERT INTO tab (col1, ..) VALUES (val1, ..), (val2, ..)
477 o Allow INSERT/UPDATE ... RETURNING new.col or old.col
479 This is useful for returning the auto-generated key for an INSERT.
480 One complication is how to handle rules that run as part of
486 o Add SET PERFORMANCE_TIPS option to suggest INDEX, VACUUM, VACUUM
488 o Add SET PATH for schemas?
490 This is basically the same as SET search_path.
493 * Server-Side Languages
495 o Fix PL/pgSQL RENAME to work on variables other than OLD/NEW
496 o Allow function parameters to be passed by name,
497 get_employee_salary(emp_id => 12345, tax_year => 2001)
498 o Add Oracle-style packages
499 o Add table function support to pltcl, plpython
500 o Add capability to create and call PROCEDURES
501 o Allow PL/pgSQL to handle %TYPE arrays, e.g. tab.col%TYPE[]
502 o Allow function argument names to be queries from PL/PgSQL
503 o Add MOVE to PL/pgSQL
504 o Add support for polymorphic arguments and return types to
505 languages other than PL/PgSQL
506 o Add support for OUT and INOUT parameters to languages other
513 * Add a libpq function to support Parse/DescribeStatement capability
514 * Prevent libpq's PQfnumber() from lowercasing the column name?
515 * Add PQescapeIdentifier() to libpq
516 * Have initdb set the input DateStyle (MDY or DMY) based on locale?
517 * Have pg_ctl look at PGHOST in case it is a socket directory?
518 * Allow pg_ctl to work properly with configuration files located outside
521 pg_ctl can not read the pid file because it isn't located in the
522 config directory but in the PGDATA directory. The solution is to
523 allow pg_ctl to read and understand postgresql.conf to find the
524 data_directory value.
529 o Have psql show current values for a sequence
530 o Move psql backslash database information into the backend, use
531 mnemonic commands? [psql]
533 This would allow non-psql clients to pull the same information out
534 of the database as psql.
536 o Fix psql's display of schema information (Neil)
537 o Allow psql \pset boolean variables to set to fixed values, rather
539 o Consistently display privilege information for all objects in psql
540 o Improve psql's handling of multi-line queries
542 Currently, while \e saves a single query as one entry, interactive
543 queries are saved one line at a time. Ideally all queries
544 whould be saved like \e does.
546 o Allow multi-line column values to align in the proper columns
548 If the second output column value is 'a\nb', the 'b' should appear
549 in the second display column, rather than the first column as it
555 o %Have pg_dump use multi-statement transactions for INSERT dumps
556 o %Allow pg_dump to use multiple -t and -n switches [pg_dump]
557 o %Add dumping of comments on composite type columns
558 o %Add dumping of comments on index columns
559 o %Replace crude DELETE FROM method of pg_dumpall --clean for
560 cleaning of roles with separate DROP commands
561 o Stop dumping CASCADE on DROP TYPE commands in clean mode
562 o %Add full object name to the tag field. eg. for operators we need
563 '=(integer, integer)', instead of just '='.
564 o Add pg_dumpall custom format dumps?
565 o %Add CSV output format
566 o Update pg_dump and psql to use the new COPY libpq API (Christopher)
567 o Remove unnecessary function pointer abstractions in pg_dump source
575 Document differences between ecpg and the SQL standard and
576 information about the Informix-compatibility module.
578 o Solve cardinality > 1 for input descriptors / variables?
579 o Add a semantic check level, e.g. check if a table really exists
580 o fix handling of DB attributes that are arrays
581 o Use backend PREPARE/EXECUTE facility for ecpg where possible
583 o Fix nested C comments
584 o %sqlwarn[6] should be 'W' if the PRECISION or SCALE value specified
585 o Make SET CONNECTION thread-aware, non-standard?
586 o Allow multidimensional arrays
587 o Add internationalized message strings
590 Referential Integrity
591 =====================
593 * Add MATCH PARTIAL referential integrity
594 * Add deferred trigger queue file
596 Right now all deferred trigger information is stored in backend
597 memory. This could exhaust memory for very large trigger queues.
598 This item involves dumping large queues into files.
600 * Change foreign key constraint for array -> element to mean element
602 * Allow DEFERRABLE UNIQUE constraints?
603 * Allow triggers to be disabled in only the current session.
605 This is currently possible by starting a multi-statement transaction,
606 modifying the system tables, performing the desired SQL, restoring the
607 system tables, and committing the transaction. ALTER TABLE ...
608 TRIGGER requires a table lock so it is not ideal for this usage.
610 * With disabled triggers, allow pg_dump to use ALTER TABLE ADD FOREIGN KEY
612 If the dump is known to be valid, allow foreign keys to be added
613 without revalidating the data.
615 * Allow statement-level triggers to access modified rows
616 * Support triggers on columns (Greg Sabino Mullane)
617 * Enforce referential integrity for system tables
618 * Allow AFTER triggers on system tables
620 System tables are modified in many places in the backend without going
621 through the executor and therefore not causing triggers to fire. To
622 complete this item, the functions that modify system tables will have
629 * Flush cached query plans when the dependent objects change
630 * Track dependencies in function bodies and recompile/invalidate
632 This is particularly important for references to temporary tables
633 in PL/PgSQL because PL/PgSQL caches query plans. The only workaround
634 in PL/PgSQL is to use EXECUTE. One complexity is that a function
635 might itself drop and recreate dependent tables, causing it to
636 invalidate its own query plan.
642 * Add SQL99 WITH clause to SELECT
643 * Add SQL99 WITH RECURSIVE to SELECT
644 * Add pre-parsing phase that converts non-ISO syntax to supported
647 This could allow SQL written for other databases to run without
650 * Allow plug-in modules to emulate features from other databases
651 * SQL*Net listener that makes PostgreSQL appear as an Oracle database
653 * Allow queries across databases or servers with transaction
656 This can be done using dblink and two-phase commit.
658 * Add the features of packages
660 o Make private objects accessable only to objects in the same schema
661 o Allow current_schema.objname to access current schema objects
662 o Add session variables
663 o Allow nested schemas
669 * Allow inherited tables to inherit index, UNIQUE constraint, and primary
671 * UNIQUE INDEX on base column not honored on INSERTs/UPDATEs from
672 inherited table: INSERT INTO inherit_table (unique_index_col) VALUES
675 The main difficulty with this item is the problem of creating an index
676 that can span more than one table.
678 * Allow SELECT ... FOR UPDATE on inherited tables
679 * Add UNIQUE capability to non-btree indexes
680 * Prevent index uniqueness checks when UPDATE does not modify the column
682 Uniqueness (index) checks are done when updating a column even if the
683 column is not modified by the UPDATE.
685 * Allow the creation of on-disk bitmap indexes which can be quickly
686 combined with other bitmap indexes
688 Such indexes could be more compact if there are only a few distinct values.
689 Such indexes can also be compressed. Keeping such indexes updated can be
692 * Allow use of indexes to search for NULLs
694 One solution is to create a partial index on an IS NULL expression.
696 * Allow accurate statistics to be collected on indexes with more than
697 one column or expression indexes, perhaps using per-index statistics
698 * Add fillfactor to control reserved free space during index creation
699 * Allow the creation of indexes with mixed ascending/descending specifiers
700 * Allow constraint_exclusion to work for UNIONs like it does for
701 inheritance, allow it to work for UPDATE and DELETE queries, and allow
702 it to be used for all queries with little performance impact
707 o Add more GIST index support for geometric data types
708 o Allow GIST indexes to create certain complex index types, like
709 digital trees (see Aoki)
713 o Pack hash index buckets onto disk pages more efficiently
715 Currently only one hash bucket can be stored on a page. Ideally
716 several hash buckets could be stored on a single page and greater
717 granularity used for the hash algorithm.
719 o Consider sorting hash buckets so entries can be found using a
720 binary search, rather than a linear scan
722 o In hash indexes, consider storing the hash value with or instead
725 o Add WAL logging for crash recovery
726 o Allow multi-column hash indexes
732 * Improve commit_delay handling to reduce fsync()
733 * Determine optimal fdatasync/fsync, O_SYNC/O_DSYNC options
735 Ideally this requires a separate test program that can be run
736 at initdb time or optionally later.
738 * %Add an option to sync() before fsync()'ing checkpoint files
739 * Add program to test if fsync has a delay compared to non-fsync
745 * Allow free-behind capability for large sequential scans, perhaps using
748 Posix_fadvise() can control both sequential/random file caching and
749 free-behind behavior, but it is unclear how the setting affects other
750 backends that also have the file open, and the feature is not supported
751 on all operating systems.
755 We could use a fixed row count and a +/- count to follow MVCC
756 visibility rules, or a single cached value could be used and
757 invalidated if anyone modifies the table. Another idea is to
758 get a count directly from a unique index, but for this to be
759 faster than a sequential scan it must avoid access to the heap
760 to obtain tuple visibility information.
762 * Allow data to be pulled directly from indexes
764 Currently indexes do not have enough tuple visibility information
765 to allow data to be pulled from the index without also accessing
766 the heap. One way to allow this is to set a bit to index tuples
767 to indicate if a tuple is currently visible to all transactions
768 when the first valid heap lookup happens. This bit would have to
769 be cleared when a heap tuple is expired.
772 * Consider automatic caching of queries at various levels:
778 * Allow sequential scans to take advantage of other concurrent
779 sequentiqal scans, also called "Synchronised Scanning"
781 One possible implementation is to start sequential scans from the lowest
782 numbered buffer in the shared cache, and when reaching the end wrap
783 around to the beginning, rather than always starting sequential scans
784 at the start of the table.
790 * Improve speed with indexes
792 For large table adjustements during VACUUM FULL, it is faster to
793 reindex rather than update the index.
795 * Reduce lock time during VACUUM FULL by moving tuples with read lock,
796 then write lock and truncate table
798 Moved tuples are invisible to other backends so they don't require a
799 write lock. However, the read lock promotion to write lock could lead
800 to deadlock situations.
802 * Maintain a map of recently-expired rows
804 This allows vacuum to target specific pages for possible free space
805 without requiring a sequential scan.
807 * Auto-fill the free space map by scanning the buffer cache or by
808 checking pages written by the background writer
809 * Create a bitmap of pages that need vacuuming
811 Instead of sequentially scanning the entire table, have the background
812 writer or some other process record pages that have expired rows, then
813 VACUUM can look at just those pages rather than the entire table. In
814 the event of a system crash, the bitmap would probably be invalidated.
816 * %Add system view to show free space map contents
821 o Use free-space map information to guide refilling
822 o %Issue log message to suggest VACUUM FULL if a table is nearly
824 o Improve xid wraparound detection by recording per-table rather
831 * Add code to detect an SMP machine and handle spinlocks accordingly
832 from distributted.net, http://www1.distributed.net/source,
833 in client/common/cpucheck.cpp
835 On SMP machines, it is possible that locks might be released shortly,
836 while on non-SMP machines, the backend should sleep so the process
837 holding the lock can complete and release it.
839 * Research use of sched_yield() for spinlock acquisition failure
840 * Fix priority ordering of read and write light-weight locks (Neil)
843 Startup Time Improvements
844 =========================
846 * Experiment with multi-threaded backend [thread]
848 This would prevent the overhead associated with process creation. Most
849 operating systems have trivial process creation time compared to
850 database startup overhead, but a few operating systems (WIn32,
851 Solaris) might benefit from threading. Also explore the idea of
852 a single session using multiple threads to execute a query faster.
854 * Add connection pooling
856 It is unclear if this should be done inside the backend code or done
857 by something external like pgpool. The passing of file descriptors to
858 existing backends is one of the difficulties with a backend approach.
864 * Eliminate need to write full pages to WAL before page modification [wal]
866 Currently, to protect against partial disk page writes, we write
867 full page images to WAL before they are modified so we can correct any
868 partial page writes during recovery. These pages can also be
869 eliminated from point-in-time archive files.
871 o When off, write CRC to WAL and check file system blocks
874 If CRC check fails during recovery, remember the page in case
875 a later CRC for that page properly matches.
877 o Write full pages during file system write and not when
878 the page is modified in the buffer cache
880 This allows most full page writes to happen in the background
881 writer. It might cause problems for applying WAL on recovery
882 into a partially-written page, but later the full page will be
885 * Reduce WAL traffic so only modified values are written rather than
887 * Allow the pg_xlog directory location to be specified during initdb
888 with a symlink back to the /data location
889 * Allow WAL information to recover corrupted pg_controldata
890 * Find a way to reduce rotational delay when repeatedly writing
893 Currently fsync of WAL requires the disk platter to perform a full
894 rotation to fsync again. One idea is to write the WAL to different
895 offsets that might reduce the rotational delay.
897 * Allow buffered WAL writes and fsync
899 Instead of guaranteeing recovery of all committed transactions, this
900 would provide improved performance by delaying WAL writes and fsync
901 so an abrupt operating system restart might lose a few seconds of
902 committed transactions but still be consistent. We could perhaps
903 remove the 'fsync' parameter (which results in an an inconsistent
904 database) in favor of this capability.
910 * Add missing optimizer selectivities for date, r-tree, etc
911 * Allow ORDER BY ... LIMIT # to select high/low value without sort or
912 index using a sequential scan for highest/lowest values
914 Right now, if no index exists, ORDER BY ... LIMIT # requires we sort
915 all values to return the high/low value. Instead The idea is to do a
916 sequential scan to find the high/low value, thus avoiding the sort.
917 MIN/MAX already does this, but not for LIMIT > 1.
919 * Precompile SQL functions to avoid overhead
920 * Create utility to compute accurate random_page_cost value
921 * Improve ability to display optimizer analysis using OPTIMIZER_DEBUG
922 * Have EXPLAIN ANALYZE highlight poor optimizer estimates
923 * Consider using hash buckets to do DISTINCT, rather than sorting
925 This would be beneficial when there are few distinct values.
927 * Log queries where the optimizer row estimates were dramatically
928 different from the number of rows actually found?
931 Miscellaneous Performance
932 =========================
934 * Do async I/O for faster random read-ahead of data
936 Async I/O allows multiple I/O requests to be sent to the disk with
937 results coming back asynchronously.
939 * Use mmap() rather than SYSV shared memory or to write WAL files?
941 This would remove the requirement for SYSV SHM but would introduce
942 portability issues. Anonymous mmap (or mmap to /dev/zero) is required
943 to prevent I/O overhead.
945 * Consider mmap()'ing files into a backend?
947 Doing I/O to large tables would consume a lot of address space or
948 require frequent mapping/unmapping. Extending the file also causes
949 mapping problems that might require mapping only individual pages,
950 leading to thousands of mappings. Another problem is that there is no
951 way to _prevent_ I/O to disk from the dirty shared buffers so changes
952 could hit disk before WAL is written.
954 * Add a script to ask system configuration questions and tune postgresql.conf
955 * Use a phantom command counter for nested subtransactions to reduce
957 * Research storing disk pages with no alignment/padding
962 * Add use of 'const' for variables in source tree
963 * Rename some /contrib modules from pg* to pg_*
964 * Move some things from /contrib into main tree
965 * Move some /contrib modules out to their own project sites
966 * %Remove warnings created by -Wcast-align
967 * Move platform-specific ps status display info from ps_status.c to ports
968 * Add optional CRC checksum to heap and index pages
969 * Improve documentation to build only interfaces (Marc)
970 * Remove or relicense modules that are not under the BSD license, if possible
971 * %Remove memory/file descriptor freeing before ereport(ERROR)
972 * Acquire lock on a relation before building a relcache entry for it
973 * %Promote debug_query_string into a server-side function current_query()
974 * %Allow the identifier length to be increased via a configure option
975 * Remove Win32 rename/unlink looping if unnecessary
976 * Allow cross-compiling by generating the zic database on the target system
977 * Improve NLS maintenace of libpgport messages linked onto applications
978 * Allow ecpg to work with MSVC and BCC
979 * Add xpath_array() to /contrib/xml2 to return results as an array
980 * Allow building in directories containing spaces
982 This is probably not possible because 'gmake' and other compiler tools
983 do not fully support quoting of paths with spaces.
985 * Allow installing to directories containing spaces
987 This is possible if proper quoting is added to the makefiles for the
988 install targets. Because PostgreSQL supports relocatable installs, it
989 is already possible to install into a directory that doesn't contain
990 spaces and then copy the install to a directory with spaces.
992 * Fix sgmltools so PDFs can be generated with bookmarks
993 * %Clean up compiler warnings (especially with gcc version 4)
998 o Remove configure.in check for link failure when cause is found
999 o Remove readdir() errno patch when runtime/mingwex/dirent.c rev
1001 o Remove psql newline patch when we find out why mingw outputs an
1003 o Allow psql to use readline once non-US code pages work with
1005 o Re-enable timezone output on log_line_prefix '%t' when a
1006 shorter timezone string is available
1007 o Fix problem with shared memory on the Win32 Terminal Server
1008 o %Add support for Unicode
1010 To fix this, the data needs to be converted to/from UTF16/UTF8
1011 so the Win32 wcscoll() can be used, and perhaps other functions
1012 like towupper(). However, UTF8 already works with normal
1013 locales but provides no ordering or character set classes.
1016 * Wire Protocol Changes
1018 o Allow dynamic character set handling
1019 o Add decoded type, length, precision
1021 o Update clients to use data types, typmod, schema.table.column names
1022 of result sets using new query protocol
1025 ---------------------------------------------------------------------------
1028 Developers who have claimed items are:
1029 --------------------------------------
1030 * Alvaro is Alvaro Herrera <alvherre@dcc.uchile.cl>
1031 * Andrew is Andrew Dunstan <andrew@dunslane.net>
1032 * Bruce is Bruce Momjian <pgman@candle.pha.pa.us> of Software Research Assoc.
1033 * Christopher is Christopher Kings-Lynne <chriskl@familyhealth.com.au> of
1034 Family Health Network
1035 * Claudio is Claudio Natoli <claudio.natoli@memetrics.com>
1036 * D'Arcy is D'Arcy J.M. Cain <darcy@druid.net> of The Cain Gang Ltd.
1037 * Fabien is Fabien Coelho <coelho@cri.ensmp.fr>
1038 * Gavin is Gavin Sherry <swm@linuxworld.com.au> of Alcove Systems Engineering
1039 * Greg is Greg Sabino Mullane <greg@turnstep.com>
1040 * Hiroshi is Hiroshi Inoue <Inoue@tpf.co.jp>
1041 * Jan is Jan Wieck <JanWieck@Yahoo.com> of Afilias, Inc.
1042 * Joe is Joe Conway <mail@joeconway.com>
1043 * Karel is Karel Zak <zakkr@zf.jcu.cz>
1044 * Magnus is Magnus Hagander <mha@sollentuna.net>
1045 * Marc is Marc Fournier <scrappy@hub.org> of PostgreSQL, Inc.
1046 * Matthew T. O'Connor <matthew@zeut.net>
1047 * Michael is Michael Meskes <meskes@postgresql.org> of Credativ
1048 * Neil is Neil Conway <neilc@samurai.com>
1049 * Oleg is Oleg Bartunov <oleg@sai.msu.su>
1050 * Peter is Peter Eisentraut <peter_e@gmx.net>
1051 * Philip is Philip Warner <pjw@rhyme.com.au> of Albatross Consulting Pty. Ltd.
1052 * Rod is Rod Taylor <pg@rbt.ca>
1053 * Simon is Simon Riggs <simon@2ndquadrant.com>
1054 * Stephan is Stephan Szabo <sszabo@megazone23.bigpanda.com>
1055 * Tatsuo is Tatsuo Ishii <t-ishii@sra.co.jp> of Software Research Assoc.
1056 * Tom is Tom Lane <tgl@sss.pgh.pa.us> of Red Hat