]> granicus.if.org Git - postgresql/log
postgresql
22 years agoRemove the definition for set_name_needs_quotes() on the assumption that
Thomas G. Lockhart [Sun, 21 Apr 2002 21:37:03 +0000 (21:37 +0000)]
Remove the definition for set_name_needs_quotes() on the assumption that
 it is now obsolete. Need some regression test cases to prove otherwise...

22 years agoInitialize or set a couple of variables to suppress compiler warnings.
Thomas G. Lockhart [Sun, 21 Apr 2002 21:35:17 +0000 (21:35 +0000)]
Initialize or set a couple of variables to suppress compiler warnings.
 These were for cases protected by elog(ERROR) exits, but may as well
 keep the compiler happy. Not sure why they don't show up on my gcc-2.96.x
 version of the compiler.

22 years agoUpdate horology-no-DST-before-1970.out (tested on HPUX), plus an extrapolated horolog...
Tom Lane [Sun, 21 Apr 2002 21:06:21 +0000 (21:06 +0000)]
Update horology-no-DST-before-1970.out (tested on HPUX), plus an extrapolated horology-solaris-1947.out
(needs verification).

22 years agoProvide new option "--enable-integer-datetimes". Defaults to disabled for
Thomas G. Lockhart [Sun, 21 Apr 2002 19:56:30 +0000 (19:56 +0000)]
Provide new option "--enable-integer-datetimes". Defaults to disabled for
 now, but we may want otherwise on systems which support it.

22 years agoSupport alternate storage scheme of 64-bit integer for date/time types.
Thomas G. Lockhart [Sun, 21 Apr 2002 19:52:18 +0000 (19:52 +0000)]
Support alternate storage scheme of 64-bit integer for date/time types.
 Use "--enable-integer-datetimes" in configuration to use this rather
 than the original float8 storage. I would recommend the integer-based
 storage for any platform on which it is available. We perhaps should
 make this the default for the production release.
Change timezone(timestamptz) results to return timestamp rather than
 a character string. Formerly, we didn't have a way to represent
 timestamps with an explicit time zone other than freezing the info into
 a string. Now, we can reasonably omit the explicit time zone from the
 result and return a timestamp with values appropriate for the specified
 time zone. Much cleaner, and if you need the time zone in the result
 you can put it into a character string pretty easily anyway.
Allow fractional seconds in date/time types even for dates prior to 1BC.
Limit timestamp data types to 6 decimal places of precision. Just right
 for a micro-second storage of int8 date/time types, and reduces the
 number of places ad-hoc rounding was occuring for the float8-based types.
Use lookup tables for precision/rounding calculations for timestamp and
 interval types.  Formerly used pow() to calculate the desired value but
 with a more limited range there is no reason to not type in a lookup
 table. Should be *much* better performance, though formerly there were
 some optimizations to help minimize the number of times pow() was called.
Define a HAVE_INT64_TIMESTAMP variable. Based on the configure option
 "--enable-integer-datetimes" and the existing internal INT64_IS_BUSTED.
Add explicit date/interval operators and functions for addition and
 subtraction. Formerly relied on implicit type promotion from date to
 timestamp with time zone.
Change timezone conversion functions for the timetz type from "timetz()"
 to "timezone()". This is consistant with other time zone coersion
 functions for other types.
Bump the catalog version to 200204201.
Fix up regression tests to reflect changes in fractional seconds
 representation for date/times in BC eras.
All regression tests pass on my Linux box.

22 years agoInclude stdio.h and stdlib.h for completeness.
Thomas G. Lockhart [Sun, 21 Apr 2002 19:51:44 +0000 (19:51 +0000)]
Include stdio.h and stdlib.h for completeness.

22 years agoMake WITHOUT TIME ZONE the default for TIMESTAMP and TIME data types.
Thomas G. Lockhart [Sun, 21 Apr 2002 19:21:49 +0000 (19:21 +0000)]
Make WITHOUT TIME ZONE the default for TIMESTAMP and TIME data types.
 This is a big change from past behavior, but the last release was
 designed to handle this correctly for dump/restore upgrades.
Fix up handling of SET value arguments. Allow lists for most options at
 least at the parser level; multiple values may be rejected at the
 command processor of course.
Allow more variations on values for SET commands, including integer and
 float values where formerly stringy fields were required.
Check precision specification for date/time fields against the true
 precision range allowed by the data types. Especially useful with the
 new int8-based storage for these types, where precision is fixed and
 predictable.
Stub out a basic CREATE ASSERTION per SQL9x. Does not do anything (yet) but
 should be augmented as appropriate.
Minor fixups in braces and tabbing.

22 years agoAllow more choices for style of value in various SET commands. Formerly,
Thomas G. Lockhart [Sun, 21 Apr 2002 19:12:46 +0000 (19:12 +0000)]
Allow more choices for style of value in various SET commands. Formerly,
 most required a stringy syntax in the parser; now integers and floats
 can (or should) be handled. There is at least one cheesy error message
 mentioning sending mail to me if there are problems; should be changed
 prior to release.
Allow lists of values from the parser in more cases. If multiple arguments
 were not allowed previously, they probably are not allowed now, but at
 least the data structures being passed around are more consistant across
 more cases.

22 years agoAdd fields in the control file to check for whether the backend was
Thomas G. Lockhart [Sun, 21 Apr 2002 19:08:02 +0000 (19:08 +0000)]
Add fields in the control file to check for whether the backend was
 compiled for integer date/time storage and to check the length of
 storage for the locale fields in the same data structure.
Slightly reword some of the error messages to be more accurate on
 possible recovery options (e.g. recompile *or* re-initdb).
Bump version number on this file.

22 years agoAugment the date/time examples in the User's Guide to reflect the newer
Thomas G. Lockhart [Sun, 21 Apr 2002 19:02:39 +0000 (19:02 +0000)]
Augment the date/time examples in the User's Guide to reflect the newer
 capabilities of specifying time zones as intervals per SQL9x.
Put refentrytitle contents on the same line as the tag.
 Otherwise, leading whitespace is propagated into the product, which
 (at least) messes up the ToC layout.
Remove (some) docinfo tags containing dates. Best to omit if the dates
 are not accurate; maybe use CVS dates instead or leave them out.

22 years agoMinor wording change for readability.
Thomas G. Lockhart [Sun, 21 Apr 2002 18:58:00 +0000 (18:58 +0000)]
Minor wording change for readability.

22 years agoMake pg_dump independent of FUNC_MAX_ARGS.
Peter Eisentraut [Sun, 21 Apr 2002 05:21:18 +0000 (05:21 +0000)]
Make pg_dump independent of FUNC_MAX_ARGS.

22 years agoFix typo.
Tom Lane [Sun, 21 Apr 2002 01:03:33 +0000 (01:03 +0000)]
Fix typo.

22 years agoRestructure AclItem representation so that we can have more than eight
Tom Lane [Sun, 21 Apr 2002 00:26:44 +0000 (00:26 +0000)]
Restructure AclItem representation so that we can have more than eight
different privilege bits (might as well make use of the space we were
wasting on padding).  EXECUTE and USAGE bits for procedures, languages
now are separate privileges instead of being overlaid on SELECT.  Add
privileges for namespaces and databases.  The GRANT and REVOKE commands
work for these object types, but we don't actually enforce the privileges
yet...

22 years agoRemove --enable-syslog option
Tatsuo Ishii [Sun, 21 Apr 2002 00:22:52 +0000 (00:22 +0000)]
Remove --enable-syslog option

22 years agopq_getstring doesn't go through pq_getbyte anymore, for better performance.
Peter Eisentraut [Sat, 20 Apr 2002 23:35:43 +0000 (23:35 +0000)]
pq_getstring doesn't go through pq_getbyte anymore, for better performance.

22 years agoScanner performance improvements
Peter Eisentraut [Sat, 20 Apr 2002 21:56:15 +0000 (21:56 +0000)]
Scanner performance improvements

Use flex flags -CF.  Pass the to-be-scanned string around as StringInfo
type, to avoid querying the length repeatedly.  Clean up some code and
remove lex-compatibility cruft.  Escape backslash sequences inline.  Use
flex-provided yy_scan_buffer() function to set up input, rather than using
myinput().

22 years agoAdd missing include.
Peter Eisentraut [Sat, 20 Apr 2002 03:46:16 +0000 (03:46 +0000)]
Add missing include.

22 years agoChange naming rule for ON SELECT rules of views: they're all just
Tom Lane [Fri, 19 Apr 2002 23:13:54 +0000 (23:13 +0000)]
Change naming rule for ON SELECT rules of views: they're all just
_RETURN now, since there's no need to keep 'em unique anymore.

22 years agopg_trigger's index on tgrelid is replaced by a unique index on
Tom Lane [Fri, 19 Apr 2002 16:36:08 +0000 (16:36 +0000)]
pg_trigger's index on tgrelid is replaced by a unique index on
(tgrelid, tgname).  This provides an additional check on trigger name
uniqueness per-table (which was already enforced by the code anyway).
With this change, RelationBuildTriggers will read the triggers in
order by tgname, since it's scanning using this index.  Since a
predictable trigger ordering has been requested for some time, document
this behavior as a feature.  Also document that rules fire in name
order, since yesterday's changes to pg_rewrite indexing cause that too.

22 years agoMake PUBLIC an unreserved word (in fact, not a keyword at all),
Tom Lane [Thu, 18 Apr 2002 21:16:16 +0000 (21:16 +0000)]
Make PUBLIC an unreserved word (in fact, not a keyword at all),
per previous discussion.

22 years agoRule names are now unique per-relation, rather than unique globally.
Tom Lane [Thu, 18 Apr 2002 20:01:11 +0000 (20:01 +0000)]
Rule names are now unique per-relation, rather than unique globally.
DROP RULE and COMMENT ON RULE syntax adds an 'ON tablename' clause,
similar to TRIGGER syntaxes.  To allow loading of existing pg_dump
files containing COMMENT ON RULE, the COMMENT code will still accept
the old syntax --- but only if the target rulename is unique across
the whole database.

22 years agoSuggest another possible solution for funcs using temp tables:
Bruce Momjian [Thu, 18 Apr 2002 17:26:03 +0000 (17:26 +0000)]
Suggest another possible solution for funcs using temp tables:

<    without using PL/PgSQL EXECUTE, needs cached plan invalidation
>    without using PL/PgSQL EXECUTE, needs cache prevention/invalidation

22 years agoMore neutral:
Bruce Momjian [Thu, 18 Apr 2002 17:24:47 +0000 (17:24 +0000)]
More neutral:

>  o Abort all, some or, no SET changes made in an aborted transaction

22 years agoUpdate:
Bruce Momjian [Thu, 18 Apr 2002 14:32:36 +0000 (14:32 +0000)]
Update:

>  o Abort all or commit all SET changes made in an aborted transaction

22 years agoCorrect documention of lo_unlink return value as int, not Oid.
Bruce Momjian [Thu, 18 Apr 2002 14:28:14 +0000 (14:28 +0000)]
Correct documention of lo_unlink return value as int, not Oid.

22 years agoUpdate:
Bruce Momjian [Thu, 18 Apr 2002 05:07:38 +0000 (05:07 +0000)]
Update:

>  o Allow Java server-side programming, http://pljava.sourceforge.net
>    [java]

22 years agoAnt 1.4.1 now requires for jdbc. Mention in HISTORY release notes.
Bruce Momjian [Thu, 18 Apr 2002 04:58:13 +0000 (04:58 +0000)]
Ant 1.4.1 now requires for jdbc.  Mention in HISTORY release notes.

22 years agoAdd how to get database/user listing information.
Bruce Momjian [Thu, 18 Apr 2002 04:45:11 +0000 (04:45 +0000)]
Add how to get database/user listing information.

22 years agoAdd user:
Bruce Momjian [Thu, 18 Apr 2002 04:21:43 +0000 (04:21 +0000)]
Add user:

> * Rod is Rod Taylor <rbt@zort.ca>

22 years agoAdd to DROP COLUMN.
Bruce Momjian [Thu, 18 Apr 2002 04:17:41 +0000 (04:17 +0000)]
Add to DROP COLUMN.

22 years agoAdd to PREPARE archive.
Bruce Momjian [Thu, 18 Apr 2002 04:13:00 +0000 (04:13 +0000)]
Add to PREPARE archive.

22 years agoUpdate for Neil:
Bruce Momjian [Thu, 18 Apr 2002 04:08:22 +0000 (04:08 +0000)]
Update for Neil:

> * Cache most recent query plan(s) (Neil) [prepare]
> * Neil is Neil Conway <nconway@klamath.dyndns.org>

22 years agoUpdate for Oleg:
Bruce Momjian [Thu, 18 Apr 2002 04:07:04 +0000 (04:07 +0000)]
Update for Oleg:

>   using a sequential scan for highest/lowest values (Oleg)

22 years agoAdd:
Bruce Momjian [Thu, 18 Apr 2002 04:02:10 +0000 (04:02 +0000)]
Add:

Add to DROP COLUMN description.

22 years agoUpdate:
Bruce Momjian [Thu, 18 Apr 2002 03:59:46 +0000 (03:59 +0000)]
Update:

>  o Abort SET changes made in aborted transactions (?)

22 years agoBack out python change, needs delay.
Bruce Momjian [Thu, 18 Apr 2002 03:38:30 +0000 (03:38 +0000)]
Back out python change, needs delay.

22 years agoChange docs to do 20! rather than larger.
Bruce Momjian [Thu, 18 Apr 2002 03:37:13 +0000 (03:37 +0000)]
Change docs to do 20! rather than larger.

22 years agoUpdated:
Bruce Momjian [Thu, 18 Apr 2002 03:10:29 +0000 (03:10 +0000)]
Updated:

> * Add BETWEEN ASYMMETRIC/SYMMETRIC (Christopher)
> * Christopher is Christopher Kings-Lynne <chriskl@familyhealth.com.au>

22 years agoAdd:
Bruce Momjian [Thu, 18 Apr 2002 03:07:16 +0000 (03:07 +0000)]
Add:

> * Add BSD-licensed qsort() for Solaris

22 years agoAdd:
Bruce Momjian [Thu, 18 Apr 2002 03:04:04 +0000 (03:04 +0000)]
Add:

>  o Allow SHOW to output as a query result, like EXPLAIN

22 years agoAdd mention of -DLINUX_PROFILE to profile mention.
Bruce Momjian [Thu, 18 Apr 2002 02:48:19 +0000 (02:48 +0000)]
Add mention of -DLINUX_PROFILE to profile mention.

22 years agoAdd:
Bruce Momjian [Thu, 18 Apr 2002 01:41:23 +0000 (01:41 +0000)]
Add:

> * Try flex flags -Cf and -CF to see if performance improves

22 years agoAdd:
Bruce Momjian [Wed, 17 Apr 2002 21:52:57 +0000 (21:52 +0000)]
Add:

> * Add utility to compute accurate random_page_cost value

22 years agoOpclasses live in namespaces. I also took the opportunity to create
Tom Lane [Wed, 17 Apr 2002 20:57:57 +0000 (20:57 +0000)]
Opclasses live in namespaces.  I also took the opportunity to create
an 'opclass owner' column in pg_opclass.  Nothing is done with it at
present, but since there are plans to invent a CREATE OPERATOR CLASS
command soon, we'll probably want DROP OPERATOR CLASS too, which
suggests that a notion of ownership would be a good idea.

22 years agoAdd:
Bruce Momjian [Wed, 17 Apr 2002 20:35:23 +0000 (20:35 +0000)]
Add:

> * Allow easy display of usernames in a group

22 years agoAdd SQL92 document name.
Bruce Momjian [Wed, 17 Apr 2002 05:12:39 +0000 (05:12 +0000)]
Add SQL92 document name.

22 years agoUpdate to point directly to ANSI store.
Bruce Momjian [Wed, 17 Apr 2002 05:10:09 +0000 (05:10 +0000)]
Update to point directly to ANSI store.

22 years agoAdd mention of standards documents.
Bruce Momjian [Wed, 17 Apr 2002 05:00:01 +0000 (05:00 +0000)]
Add mention of standards documents.

22 years agoAdd steps for typical patch.
Bruce Momjian [Wed, 17 Apr 2002 02:10:22 +0000 (02:10 +0000)]
Add steps for typical patch.

22 years agoOperators live in namespaces. CREATE/DROP/COMMENT ON OPERATOR take
Tom Lane [Tue, 16 Apr 2002 23:08:12 +0000 (23:08 +0000)]
Operators live in namespaces.  CREATE/DROP/COMMENT ON OPERATOR take
qualified operator names directly, for example CREATE OPERATOR myschema.+
( ... ).  To qualify an operator name in an expression you need to write
OPERATOR(myschema.+) (thanks to Peter for suggesting an escape hatch).
I also took advantage of having to reformat pg_operator to fix something
that'd been bugging me for a while: mergejoinable operators should have
explicit links to the associated cross-data-type comparison operators,
rather than hardwiring an assumption that they are named < and >.

22 years agoAdded some rudimentary table and column tests
Dave Cramer [Tue, 16 Apr 2002 15:25:17 +0000 (15:25 +0000)]
Added some rudimentary table and column tests
added a setup/teardown to create and drop the connection, and table

22 years agofixed getColumns as per Panu Outinen's email. At this point have only repaired the...
Dave Cramer [Tue, 16 Apr 2002 13:28:44 +0000 (13:28 +0000)]
fixed getColumns as per Panu Outinen's email. At this point have only repaired the bug, haven't made it caseInsensitive

22 years agoFixed typo in preproc/type.h
Michael Meskes [Tue, 16 Apr 2002 07:03:56 +0000 (07:03 +0000)]
Fixed typo in preproc/type.h

22 years agoThe patch I sent to -patches a little while ago wasn't applied: it
Bruce Momjian [Mon, 15 Apr 2002 23:47:12 +0000 (23:47 +0000)]
The patch I sent to -patches a little while ago wasn't applied: it
was in the thread "make BufferGetBlockNumber() a macro". Tom
objected to the original patch, so I prepared a new one which
doesn't change BufferGetBlockNumber() into a macro, it just
cleans up some comments and fixes an assertion. The patch
is attached.

Neil Conway

22 years agoThe attached patch corrects an inaccuracy in src/backend/catalog/README
Bruce Momjian [Mon, 15 Apr 2002 23:46:13 +0000 (23:46 +0000)]
The attached patch corrects an inaccuracy in src/backend/catalog/README
and fixes a few spelling mistakes in src/bakckend/lmgr/README.

Neil Conway

22 years agoCATALOG VERSION UPDATED:
Bruce Momjian [Mon, 15 Apr 2002 23:45:07 +0000 (23:45 +0000)]
CATALOG VERSION UPDATED:

The indexes on most system catalogs are named with the suffix "_index";
not so with TOAST table indexes, which use "_idx". This trivial patch
changes TOAST table index names to use the "_index" suffix for
consistency.

Neil Conway

22 years agoDisable VACUUM from being called from a function because function memory
Bruce Momjian [Mon, 15 Apr 2002 23:39:42 +0000 (23:39 +0000)]
Disable VACUUM from being called from a function because function memory
would be cleared by vacuum;  fix idea from Tom Lane.

22 years agoFix for NOTIFY when NAMEDATALEN is nonstandard in server. Fix idea from
Bruce Momjian [Mon, 15 Apr 2002 23:35:51 +0000 (23:35 +0000)]
Fix for NOTIFY when NAMEDATALEN is nonstandard in server.  Fix idea from
Tom Lane to move string storage to end of structure but keep pointer in
the same location.

22 years agoFix for EINTR returns from Win9X socket operations:
Bruce Momjian [Mon, 15 Apr 2002 23:34:17 +0000 (23:34 +0000)]
Fix for EINTR returns from Win9X socket operations:

In summary, if a software writer implements timer events or other events
 which generate a signal with a timing fast enough to occur while libpq
is inside connect(), then connect returns -EINTR.  The code following
the connect call does not handle this and generates an error message.
The sum result is that the pg_connect() fails.  If the timer or other
event is right on the window of the connect() completion time, the
pg_connect() may appear to work sporadically.  If the event is too slow,
 pg_connect() will appear to always work and if the event is too fast,
pg_connect() will always fail.

David Ford

22 years agoAdjust rules for search_path so that pg_catalog is never implicitly
Tom Lane [Mon, 15 Apr 2002 22:33:21 +0000 (22:33 +0000)]
Adjust rules for search_path so that pg_catalog is never implicitly
selected as the creation target namespace; to make that happen, you
must explicitly set search_path that way.  This makes initdb a hair
more complex but seems like a good safety feature.

22 years agoUpdate as completed:
Bruce Momjian [Mon, 15 Apr 2002 20:52:12 +0000 (20:52 +0000)]
Update as completed:

> * -Allow oid to act as a foreign key

22 years agoFix text_substr bug intrduced in 7.3 development
Tatsuo Ishii [Mon, 15 Apr 2002 07:54:37 +0000 (07:54 +0000)]
Fix text_substr bug intrduced in 7.3 development
using Joe Conway's patches (submitted at pgsql-patches on 2002/04/08)
 + small fix.

22 years agoFix oversight in recent change of representation for JOIN alias
Tom Lane [Mon, 15 Apr 2002 06:05:49 +0000 (06:05 +0000)]
Fix oversight in recent change of representation for JOIN alias
variables: JOIN/ON should allow references to contained JOINs.
Per bug report from Barry Lind.

22 years agoThe contents of command.c, creatinh.c, define.c, remove.c and rename.c
Tom Lane [Mon, 15 Apr 2002 05:22:04 +0000 (05:22 +0000)]
The contents of command.c, creatinh.c, define.c, remove.c and rename.c
have been divided according to the type of object manipulated - so ALTER
TABLE code is in tablecmds.c, aggregate commands in aggregatecmds.c and
so on.

A few common support routines remain in define.c (prototypes in
src/include/commands/defrem.h).

No code has been changed except for includes to reflect the new files.
The prototypes for aggregatecmds.c, functioncmds.c, operatorcmds.c,
and typecmds.c remain in src/include/commands/defrem.h.

From John Gray <jgray@azuli.co.uk>

22 years ago1) Fix a bug about reporting varchar info thanks to Aceto.
Hiroshi Inoue [Mon, 15 Apr 2002 02:46:00 +0000 (02:46 +0000)]
1) Fix a bug about reporting varchar info thanks to Aceto.
2) Introcuced 3 drivers.
3) The version is now 7.02.0001.

22 years agoAdd bitmap email to archive.
Bruce Momjian [Sun, 14 Apr 2002 17:32:37 +0000 (17:32 +0000)]
Add bitmap email to archive.

22 years agoAllow detection of collateindex.pl in stylesheet directory or in path,
Peter Eisentraut [Sun, 14 Apr 2002 17:23:20 +0000 (17:23 +0000)]
Allow detection of collateindex.pl in stylesheet directory or in path,
which covers some recent installation schemes.

Add Mandrake installation layout to directories to check for stylesheets.

Allow documentation build to proceed if stylesheets were not found, in case
the stylesheets might be found through the SGML catalog mechanism.

22 years agoUpdate;
Bruce Momjian [Sun, 14 Apr 2002 17:20:35 +0000 (17:20 +0000)]
Update;

>    without using PL/PgSQL EXECUTE, needs cached plan invalidation

22 years agoFix comment dashes.
Bruce Momjian [Sun, 14 Apr 2002 16:47:16 +0000 (16:47 +0000)]
Fix comment dashes.

22 years ago- Synced preproc.y with gram.y
Michael Meskes [Sun, 14 Apr 2002 08:59:16 +0000 (08:59 +0000)]
- Synced preproc.y with gram.y
- Fixed one bug in structure handling resulting in using sizeof indicator instead of variable.

22 years agoFix broken SGML tag.
Tatsuo Ishii [Sun, 14 Apr 2002 01:31:37 +0000 (01:31 +0000)]
Fix broken SGML tag.

22 years agoRod's patch does what it is supposed to do, but it also includes
Bruce Momjian [Sat, 13 Apr 2002 19:57:18 +0000 (19:57 +0000)]
Rod's patch does what it is supposed to do, but it also includes
some old code to add PK constraints to CREATE TABLE. That stuff
had been removed as part of my original patch for pg_dump a
little while ago.

The attached patch fixes this by removing (again :-) ) the
code in dumpTables() to perform PK creation during CREATE
TABLE. I briefly tested it locally and it fixes both of
Tom's test cases.

Please apply.

Cheers,

Neil

--
Neil Conway <neilconway@rogers.com>

22 years agoPatch against 7.2.1 sources. Uses Solaris Intimate Shared Memory
Bruce Momjian [Sat, 13 Apr 2002 19:52:51 +0000 (19:52 +0000)]
Patch against 7.2.1 sources.  Uses Solaris Intimate Shared Memory
for Solaris on SPARC.  Scott Brunza (sbrunza@sonalysts.com) gets
credit for identifying the issue, making the change, and doing
the regression tests.

Earlier testing on 7.2rc2 and 7.2 showed performance gains of
1% to 10% on pgbench, osdb-pg, and some locally developed apps.

Solaris Intimate Shared Memory is described in "SOLARIS INTERNALS
Core Kernel Components" by Jim Mauro and Richard McDougall,
Copyright 2001 Sun Microsystem, Inc.  ISBN 0-13-022496-0

P.J. "Josh" Rovero

22 years agoUpdate/correct inheritance examples.
Tom Lane [Sat, 13 Apr 2002 17:17:29 +0000 (17:17 +0000)]
Update/correct inheritance examples.

22 years agoThis is a proposed patch to doc/src/sgml/libpgtcl.sgml which documents
Bruce Momjian [Sat, 13 Apr 2002 01:42:44 +0000 (01:42 +0000)]
This is a proposed patch to doc/src/sgml/libpgtcl.sgml which documents
the libpgtcl "pg_execute" command. This was mentioned on
pgsql-interfaces on Mar 3. I am posting it here in the hope that someone
will check to see if it makes sense and is correct SGML-wise.  I did run
it through jade, but this is my first try at this sort of thing.

ljb

22 years agoAdd:
Bruce Momjian [Sat, 13 Apr 2002 01:37:54 +0000 (01:37 +0000)]
Add:

> * Change NUMERIC to enforce the maximum precision, and increase it

22 years agoMention precision of NUMERIC in docs.
Bruce Momjian [Sat, 13 Apr 2002 01:35:09 +0000 (01:35 +0000)]
Mention precision of NUMERIC in docs.

22 years agoAdd:
Bruce Momjian [Sat, 13 Apr 2002 00:20:33 +0000 (00:20 +0000)]
Add:

> * Change NUMERIC data type to use base 10,000 internally

22 years agoChecking to decide whether relations are system relations now depends
Tom Lane [Fri, 12 Apr 2002 20:38:31 +0000 (20:38 +0000)]
Checking to decide whether relations are system relations now depends
on the namespace not the name; pg_ is not a reserved prefix for table
names anymore.  From Fernando Nasser.

22 years agoTweak error message wording.
Tom Lane [Fri, 12 Apr 2002 19:11:49 +0000 (19:11 +0000)]
Tweak error message wording.

22 years agoAllow prompting of create/drop user to handle spaces.
Bruce Momjian [Fri, 12 Apr 2002 09:42:32 +0000 (09:42 +0000)]
Allow prompting of create/drop user to handle spaces.

22 years agoAllow dropdb and others to use identifiers with spaces; IFS cleanup for
Bruce Momjian [Fri, 12 Apr 2002 09:37:11 +0000 (09:37 +0000)]
Allow dropdb and others to use identifiers with spaces;  IFS cleanup for
pg_dumpall.

22 years agoUpdate new Russian FAQ.
Bruce Momjian [Fri, 12 Apr 2002 09:17:10 +0000 (09:17 +0000)]
Update new Russian FAQ.

22 years ago1) Fix a bug in declare/fetch mode.
Hiroshi Inoue [Fri, 12 Apr 2002 08:03:12 +0000 (08:03 +0000)]
1) Fix a bug in declare/fetch mode.
2) Suppress some error logs for the request to other drivers.

22 years agoAdd mention of function CREATE INDEX usage.
Bruce Momjian [Thu, 11 Apr 2002 23:20:04 +0000 (23:20 +0000)]
Add mention of function CREATE INDEX usage.

22 years agoCleanup in use of IFS.
Bruce Momjian [Thu, 11 Apr 2002 21:22:27 +0000 (21:22 +0000)]
Cleanup in use of IFS.

22 years agoUpdate obsolete syntax in example of inheritance.
Tom Lane [Thu, 11 Apr 2002 21:18:50 +0000 (21:18 +0000)]
Update obsolete syntax in example of inheritance.

22 years agoMake sure leading/trailing/multiple consecutive space/tab in db/user/group
Bruce Momjian [Thu, 11 Apr 2002 21:16:28 +0000 (21:16 +0000)]
Make sure leading/trailing/multiple consecutive space/tab in db/user/group
names are handled OK.

22 years agoRestructure representation of aggregate functions so that they have pg_proc
Tom Lane [Thu, 11 Apr 2002 20:00:18 +0000 (20:00 +0000)]
Restructure representation of aggregate functions so that they have pg_proc
entries, per pghackers discussion.  This fixes aggregates to live in
namespaces, and also simplifies/speeds up lookup in parse_func.c.
Also, add a 'proimplicit' flag to pg_proc that controls whether a type
coercion function may be invoked implicitly, or only explicitly.  The
current settings of these flags are more permissive than I would like,
but we will need to debate and refine the behavior; for now, I avoided
breaking regression tests as much as I could.

22 years agoFix pg_upgrade to handle dbnames, user/group names with spaces.
Bruce Momjian [Thu, 11 Apr 2002 19:23:36 +0000 (19:23 +0000)]
Fix pg_upgrade to handle dbnames, user/group names with spaces.

22 years agoMake sure that usesuper is always accessed through superuser(), so that the
Peter Eisentraut [Thu, 11 Apr 2002 05:32:03 +0000 (05:32 +0000)]
Make sure that usesuper is always accessed through superuser(), so that the
single-user escape path always works.

22 years agoFix problem with invalid database name assignment in while loop.
Bruce Momjian [Thu, 11 Apr 2002 04:56:21 +0000 (04:56 +0000)]
Fix problem with invalid database name assignment in while loop.
Properly warn of invalid pg_dumpall options;  minor quoting cleanups.

22 years agoReadline and Zlib now required by default. Add options --without-readline
Peter Eisentraut [Wed, 10 Apr 2002 22:47:09 +0000 (22:47 +0000)]
Readline and Zlib now required by default.  Add options --without-readline
and --without-zlib to turn them off.

22 years agoAdd make install-strip target.
Peter Eisentraut [Wed, 10 Apr 2002 16:45:25 +0000 (16:45 +0000)]
Add make install-strip target.

22 years agoPrevent an infinite loop of error reporting.
Hiroshi Inoue [Wed, 10 Apr 2002 08:18:54 +0000 (08:18 +0000)]
Prevent an infinite loop of error reporting.

22 years agoFunctions live in namespaces. Qualified function names work, eg
Tom Lane [Tue, 9 Apr 2002 20:35:55 +0000 (20:35 +0000)]
Functions live in namespaces.  Qualified function names work, eg
SELECT schema1.func2(...).  Aggregate names can be qualified at the
syntactic level, but the qualification is ignored for the moment.

22 years agoUpdate pg_upgrade for 7.3.
Bruce Momjian [Tue, 9 Apr 2002 18:07:24 +0000 (18:07 +0000)]
Update pg_upgrade for 7.3.

22 years agoPoint to developers site, not my hostname.
Bruce Momjian [Tue, 9 Apr 2002 14:53:55 +0000 (14:53 +0000)]
Point to developers site, not my hostname.

22 years agoUpdate FAQ.
Bruce Momjian [Tue, 9 Apr 2002 14:22:34 +0000 (14:22 +0000)]
Update FAQ.