]> granicus.if.org Git - postgresql/log
postgresql
23 years agoAdd GUC setting for Australian timezones. Uses new GUC boolean callback
Bruce Momjian [Mon, 18 Jun 2001 16:14:44 +0000 (16:14 +0000)]
Add GUC setting for Australian timezones.  Uses new GUC boolean callback
functions to clear date cache.  Allow regression tests to pass when
timezone set.

23 years agoAllow removal of system-named pg_* temp tables. Rename temp file/dir as
Bruce Momjian [Mon, 18 Jun 2001 16:13:21 +0000 (16:13 +0000)]
Allow removal of system-named pg_* temp tables.  Rename temp file/dir as
pgsql_tmp.

23 years agoImprove wording of authentication files.
Bruce Momjian [Mon, 18 Jun 2001 16:11:30 +0000 (16:11 +0000)]
Improve wording of authentication files.

23 years agoDo some restructuring to improve performance of the catcaches. Teach
Tom Lane [Mon, 18 Jun 2001 03:35:07 +0000 (03:35 +0000)]
Do some restructuring to improve performance of the catcaches.  Teach
CatalogCacheFlushRelation (formerly called SystemCacheRelationFlushed)
how to distinguish tuples it should flush from those it needn't; this
means a relcache flush event now only removes the catcache entries
it ought to, rather than zapping the caches completely as it used to.
Testing with the regression tests indicates that this considerably
improves the lifespan of catcache entries.  Also, rearrange catcache
data structures so that the limit on number of cached tuples applies
globally across all the catcaches, rather than being per-catcache.
It was a little silly to have the same size limit on both, say,
pg_attribute caches and pg_am caches (there being only four possible
rows in the latter...).  Doing LRU removal across all the caches
instead of locally in each one should reduce cache reload traffic
in the more heavily used caches and improve the efficiency of
cache memory use.

23 years agoFix *escape* handling in copy_statement_with_parameters(was my fault).
Hiroshi Inoue [Mon, 18 Jun 2001 02:16:09 +0000 (02:16 +0000)]
Fix *escape* handling in copy_statement_with_parameters(was my fault).

23 years agoKeep the list of to-be-NOTIFYed names in a plain List palloc'd in
Tom Lane [Sun, 17 Jun 2001 22:27:15 +0000 (22:27 +0000)]
Keep the list of to-be-NOTIFYed names in a plain List palloc'd in
TopTransactionContext, rather than using Dllist.  This simplifies and
speeds up the code, and eliminates a former risk of coredump when
out of memory (since the old code didn't bother to check for malloc
failure).  It also moves us one step closer to retiring Dllist...

23 years agoMake inet/cidr << and <<= operators indexable. From Alex Pilosov <alex@pilosoft...
Tom Lane [Sun, 17 Jun 2001 02:05:20 +0000 (02:05 +0000)]
Make inet/cidr << and <<= operators indexable.  From Alex Pilosov <alex@pilosoft.com>.

23 years agoTweak startup sequence so that running out of PROC array slots is
Tom Lane [Sat, 16 Jun 2001 22:58:17 +0000 (22:58 +0000)]
Tweak startup sequence so that running out of PROC array slots is
detected sooner in backend startup, and is treated as an expected error
(it gives 'Sorry, too many clients already' now).  This allows us not
to have to enforce the MaxBackends limit exactly in the postmaster.
Also, remove ProcRemove() and fold its functionality into ProcKill().
There's no good reason for a backend not to be responsible for removing
its PROC entry, and there are lots of good reasons for the postmaster
not to be touching shared-memory data structures.

23 years agoIt turns out that the relcache thinks it can distinguish different
Tom Lane [Sat, 16 Jun 2001 18:59:31 +0000 (18:59 +0000)]
It turns out that the relcache thinks it can distinguish different
rules and triggers by OID.  So, even though we have no cross-references
in the system catalogs to pg_rewrite.oid or pg_trigger.oid, we'd better
have unique indexes on them.  Put back pg_rewrite_oid_index, which I
mistakenly removed a few days ago, and add pg_trigger_oid_index.

23 years agoOops, forgot to commit doco updates for has_table_privilege.
Tom Lane [Fri, 15 Jun 2001 21:03:07 +0000 (21:03 +0000)]
Oops, forgot to commit doco updates for has_table_privilege.

23 years agofixed bug in connect.c
Michael Meskes [Fri, 15 Jun 2001 09:55:51 +0000 (09:55 +0000)]
fixed bug in connect.c

23 years agoAdd bpchar to list of string types.
D'Arcy J.M. Cain [Fri, 15 Jun 2001 02:23:18 +0000 (02:23 +0000)]
Add bpchar to list of string types.
Thanks to Steve McClure <smcclure@racemi.com> for the patch.

23 years agoAdd missing PG_SETMASK(&BlockSig) to SIGHUP_handler().
Tom Lane [Thu, 14 Jun 2001 19:59:24 +0000 (19:59 +0000)]
Add missing PG_SETMASK(&BlockSig) to SIGHUP_handler().

23 years agoRevoke public read access from pg_statistic, create new system view
Tom Lane [Thu, 14 Jun 2001 19:47:25 +0000 (19:47 +0000)]
Revoke public read access from pg_statistic, create new system view
pg_stats to provide controlled (and, hopefully, more readable) access
to statistics.  Comments on definition of pg_stats welcome.
I didn't force initdb, but the rules regress test will fail until you
do one.

23 years agoAllow remote query execution (dblink)
Bruce Momjian [Thu, 14 Jun 2001 16:49:03 +0000 (16:49 +0000)]
Allow remote query execution (dblink)

Joe Conway

23 years agoRemove more NULL lines.
Bruce Momjian [Thu, 14 Jun 2001 04:59:59 +0000 (04:59 +0000)]
Remove more NULL lines.

23 years agoFix nulls in HTML file.
Bruce Momjian [Thu, 14 Jun 2001 04:24:01 +0000 (04:24 +0000)]
Fix nulls in HTML file.

 Sergio Bruder

23 years agohas_table_privilege functions from Joe Conway (with some kibitzing from
Tom Lane [Thu, 14 Jun 2001 01:09:22 +0000 (01:09 +0000)]
has_table_privilege functions from Joe Conway (with some kibitzing from
Tom Lane).  For the moment, only the OID/name variants are provided.
I didn't force initdb, but the additions to the 'privileges' regress
test won't pass until you do one.

23 years agoDocs for:
Bruce Momjian [Wed, 13 Jun 2001 22:51:02 +0000 (22:51 +0000)]
Docs for:

> >> inet(text), cidr(text): convert a text value into inet/cidr
> >> set_masklen(inet): set masklen on the inet value

Alex Pilosov

23 years agoAllow a non-superuser database owner to vacuum all tables in his
Tom Lane [Wed, 13 Jun 2001 21:44:41 +0000 (21:44 +0000)]
Allow a non-superuser database owner to vacuum all tables in his
database, including system catalogs (but not the shared catalogs,
since they don't really belong to his database).  This is per recent
mailing list discussion.  Clean up some other code that also checks
for database ownerness by introducing a test function is_dbadmin().

23 years agoFix for duplicate oids just introduced in patch. Don't think it needs
Bruce Momjian [Wed, 13 Jun 2001 21:12:43 +0000 (21:12 +0000)]
Fix for duplicate oids just introduced in patch.  Don't think it needs
initdb unless you are using inet types.

23 years agoAttached is a patch adding following functions:
Bruce Momjian [Wed, 13 Jun 2001 21:09:00 +0000 (21:09 +0000)]
Attached is a patch adding following functions:

inet(text), cidr(text): convert a text value into inet/cidr
set_masklen(inet): set masklen on the inet value

Patch also contains regression checks for these functions.

Alex Pilosov

23 years agoFix compile failure when --enable-multibyte.
Bruce Momjian [Wed, 13 Jun 2001 21:07:12 +0000 (21:07 +0000)]
Fix compile failure when --enable-multibyte.

Marko Kreen

23 years agoUpdate TODO list.
Bruce Momjian [Wed, 13 Jun 2001 20:17:05 +0000 (20:17 +0000)]
Update TODO list.

23 years agoDon't assume free(NULL) is OK. Yes, I know ANSI C says it is.
Tom Lane [Wed, 13 Jun 2001 19:52:33 +0000 (19:52 +0000)]
Don't assume free(NULL) is OK.  Yes, I know ANSI C says it is.

23 years agoUpdate some obsolete examples.
Tom Lane [Wed, 13 Jun 2001 19:31:56 +0000 (19:31 +0000)]
Update some obsolete examples.

23 years agoRemove some dead code, simplify calling convention.
Tom Lane [Wed, 13 Jun 2001 18:56:30 +0000 (18:56 +0000)]
Remove some dead code, simplify calling convention.

23 years ago- Synced preproc.y with gram.y.
Michael Meskes [Wed, 13 Jun 2001 12:38:58 +0000 (12:38 +0000)]
- Synced preproc.y with gram.y.
- Applied bug fix by John Summerfield.

23 years agoExtend GUC concepts of parse_hook and assign_hook to all four supported
Tom Lane [Tue, 12 Jun 2001 22:54:06 +0000 (22:54 +0000)]
Extend GUC concepts of parse_hook and assign_hook to all four supported
datatypes, not only strings.  parse_hook is useless for bool, I suppose,
but it seems possibly useful for int and double to apply variable-specific
constraints that are more complex than simple range limits.  assign_hook
is definitely useful for all datatypes --- we need it right now for bool
to support date cache reset when changing Australian timezone rule setting.
Also, clean up some residual problems with the reset all/show all patch,
including memory leaks and mistaken reset of PostPortNumber.  It seems
best that RESET ALL not touch variables that don't have SUSET or
USERSET context.

23 years agoFix compile error caused by patch application.
Bruce Momjian [Tue, 12 Jun 2001 21:51:27 +0000 (21:51 +0000)]
Fix compile error caused by patch application.

23 years agoRepair problem with multi-action rules in combination with any nontrivial
Tom Lane [Tue, 12 Jun 2001 18:54:22 +0000 (18:54 +0000)]
Repair problem with multi-action rules in combination with any nontrivial
manipulation of rtable/jointree by planner.  Rewriter was generating
actions that shared rtable/jointree substructure, which caused havoc
when planner got to the later actions that it'd already mucked up.

23 years agoAdd Updatable view mention.
Bruce Momjian [Tue, 12 Jun 2001 18:15:59 +0000 (18:15 +0000)]
Add Updatable view mention.

23 years agoUpdate TODO list.
Bruce Momjian [Tue, 12 Jun 2001 18:15:50 +0000 (18:15 +0000)]
Update TODO list.

23 years agoBack out has_table_privilege patch.
Bruce Momjian [Tue, 12 Jun 2001 16:34:27 +0000 (16:34 +0000)]
Back out has_table_privilege patch.

23 years agoUpdate TODO list.
Bruce Momjian [Tue, 12 Jun 2001 16:18:11 +0000 (16:18 +0000)]
Update TODO list.

23 years agoOK -- here's take #5.
Bruce Momjian [Tue, 12 Jun 2001 15:58:34 +0000 (15:58 +0000)]
OK -- here's take #5.

It "make"s and "make check"s clean against current cvs tip.

There are now both Text and Name variants, and the regression test support
is rolled into the patch. Note that to be complete wrt Name based variants,
there are now 12 user visible versions of has_table_privilege:

has_table_privilege(Text usename, Text relname, Text priv_type)
has_table_privilege(Text usename, Name relname, Text priv_type)
has_table_privilege(Name usename, Text relname, Text priv_type)
has_table_privilege(Name usename, Name relname, Text priv_type)
has_table_privilege(Text relname, Text priv_type) /* assumes current_user */
has_table_privilege(Name relname, Text priv_type) /* assumes current_user */
has_table_privilege(Text usename, Oid reloid, Text priv_type)
has_table_privilege(Name usename, Oid reloid, Text priv_type)
has_table_privilege(Oid reloid, Text priv_type)  /* assumes current_user */
has_table_privilege(Oid usesysid, Text relname, Text priv_type)
has_table_privilege(Oid usesysid, Name relname, Text priv_type)
has_table_privilege(Oid usesysid, Oid reloid, Text priv_type)

For the Text based inputs, a new internal function, get_Name is used
(shamelessly copied from get_seq_name in sequence.c) to downcase if not
quoted, or remove quotes if quoted, and truncate. I also added a few test
cases for the downcasing, quote removal, and Name based variants to the
regression test.

Joe Conway

23 years agoI installed postgres 7.1 with --enable-odbc. I then installed
Bruce Momjian [Tue, 12 Jun 2001 15:14:47 +0000 (15:14 +0000)]
I installed postgres 7.1 with --enable-odbc.  I then installed
tclodbc (http://sourceforge.net/projects/tclodbc) and libiodbc-2.50.3
(http://www.iodbc.org/dist/libiodbc-2.50.3.tar.gz).  I could not
get either to work... postgres would not find the global odbcinst.ini
file.  I traced this to src/interfaces/odbc/gpps.c -- here are the
many things I think are wrong:

Run tclodbc and do a ``database db <DSNname>'' where ``DSNname'' is
one of the DSN's in /usr/local/etc/odbcinst.ini (or wherever the
global ini file is installed.)  The result is always the error
message that ``one of server,port,database,etc. are missing''.

Run libiodbc-2.50.3/samples/odbctest <DSNname>.  The command fails
to connect to the database and just exits.

Dave Bodenstab

23 years agoUpdate TODO list.
Bruce Momjian [Tue, 12 Jun 2001 14:50:24 +0000 (14:50 +0000)]
Update TODO list.

23 years agoClean up various to-do items associated with system indexes:
Tom Lane [Tue, 12 Jun 2001 05:55:50 +0000 (05:55 +0000)]
Clean up various to-do items associated with system indexes:
pg_database now has unique indexes on oid and on datname.
pg_shadow now has unique indexes on usename and on usesysid.
pg_am now has unique index on oid.
pg_opclass now has unique index on oid.
pg_amproc now has unique index on amid+amopclaid+amprocnum.
Remove pg_rewrite's unnecessary index on oid, delete unused RULEOID syscache.
Remove index on pg_listener and associated syscache for performance reasons
(caching rows that are certain to change before you need 'em again is
rather pointless).
Change pg_attrdef's nonunique index on adrelid into a unique index on
adrelid+adnum.

Fix various incorrect settings of pg_class.relisshared, make that the
primary reference point for whether a relation is shared or not.
IsSharedSystemRelationName() is now only consulted to initialize relisshared
during initial creation of tables and indexes.  In theory we might now
support shared user relations, though it's not clear how one would get
entries for them into pg_class &etc of multiple databases.

Fix recently reported bug that pg_attribute rows created for an index all have
the same OID.  (Proof that non-unique OID doesn't matter unless it's
actually used to do lookups ;-))

There's no need to treat pg_trigger, pg_attrdef, pg_relcheck as bootstrap
relations.  Convert them into plain system catalogs without hardwired
entries in pg_class and friends.

Unify global.bki and template1.bki into a single init script postgres.bki,
since the alleged distinction between them was misleading and pointless.
Not to mention that it didn't work for setting up indexes on shared
system relations.

Rationalize locking of pg_shadow, pg_group, pg_attrdef (no need to use
AccessExclusiveLock where ExclusiveLock or even RowExclusiveLock will do).
Also, hold locks until transaction commit where necessary.

23 years agoautoconf
Bruce Momjian [Mon, 11 Jun 2001 22:17:14 +0000 (22:17 +0000)]
autoconf

23 years agoThe attached patch enables PostgreSQL CVS to build cleanly under Cygwin
Bruce Momjian [Mon, 11 Jun 2001 22:12:48 +0000 (22:12 +0000)]
The attached patch enables PostgreSQL CVS to build cleanly under Cygwin
when built against readline 4.2.  Specifically, it handles the deprecation
of

    filename_completion_function()

with preference for

    rl_filename_completion_function()

Although, I was motivated by Cygwin support, IMO this patch is appropriate
for all platforms.  To quote from the readline source:

    #if 0
    /* Backwards compatibility (compat.c).  These will go away sometime. */
    ...
    extern READLINE_EXPORT(char, *filename_completion_function) ...
    #endif

Note that this patch is modeled after the one by Peter Eisentraut for
completion_matches():

    http://www.ca.postgresql.org/~petere/readline42.html
I tested this patch under the following environments:

    Cygwin with readline 4.1
    Cygwin with readline 4.2
    Linux with readline 2.2.1
    Linux with readline 4.2

and it behaved as expected.

Jason Tishler

23 years ago Got two patches that were found by folks on the Castor list, that we'd like to
Bruce Momjian [Mon, 11 Jun 2001 22:12:00 +0000 (22:12 +0000)]
  Got two patches that were found by folks on the Castor list, that we'd like to
submit.  These were done for the jdbc2 driver.  The first one is for support
of the Types.BIT in the PreparedStatement class.  The following lines need to be
inserted in the switch statment, at around line 530:

(Prepared statment, line 554, before the default: switch
case Types.BIT:
     if (x instanceof Boolean) {
          set(parameterIndex, ((Boolean)x).booleanValue() ? "TRUE" : "FALSE");
     } else {
          throw new PSQLException("postgresql.prep.type");
     }
     break;

The second one is dealing with blobs,

inserted in PreparedStatemant.java (After previous patch line, 558):
         case Types.BINARY:
         case Types.VARBINARY:
                              setObject(parameterIndex,x);
                              break;
and in ResultSet.java (Around line 857):
        case Types.BINARY:
        case Types.VARBINARY:
                        return getBytes(columnIndex);

Ned Wolpert <ned.wolpert@knowledgenet.com>

23 years agoAdd French psql translation. (My rendition of French anyway...)
Peter Eisentraut [Mon, 11 Jun 2001 18:23:33 +0000 (18:23 +0000)]
Add French psql translation.  (My rendition of French anyway...)
Fix typo in message.

23 years agoUpdate FAQ.
Bruce Momjian [Mon, 11 Jun 2001 16:27:52 +0000 (16:27 +0000)]
Update FAQ.

23 years agoNest macros with slightly less enthusiasm, for performance and to avoid
Tom Lane [Mon, 11 Jun 2001 05:00:56 +0000 (05:00 +0000)]
Nest macros with slightly less enthusiasm, for performance and to avoid
having non-gcc compilers spit up.

23 years agoMove temporary files into 'pg_tempfiles' subdirectory of each database
Tom Lane [Mon, 11 Jun 2001 04:12:29 +0000 (04:12 +0000)]
Move temporary files into 'pg_tempfiles' subdirectory of each database
directory (which can be made a symlink to put temp files on another disk).
Add code to delete leftover temp files during postmaster startup.
Bruce, with some kibitzing from Tom.

23 years agocontrib/intarray updates from Oleg Bartunov:
Tom Lane [Mon, 11 Jun 2001 02:31:25 +0000 (02:31 +0000)]
contrib/intarray updates from Oleg Bartunov:
it's reimplemented to use function interface version 1 and special
treating of degenerated signatures.

23 years agoUpdate EXPLAIN discussion and examples to match current sources.
Tom Lane [Mon, 11 Jun 2001 00:52:09 +0000 (00:52 +0000)]
Update EXPLAIN discussion and examples to match current sources.

23 years agoMake planner compute the number of hash buckets the same way that
Tom Lane [Mon, 11 Jun 2001 00:17:08 +0000 (00:17 +0000)]
Make planner compute the number of hash buckets the same way that
nodeHash.c will compute it (by sharing code).

23 years agoMark inet,cidr,macaddr '=' operators as mergejoinable.
Tom Lane [Sun, 10 Jun 2001 22:32:35 +0000 (22:32 +0000)]
Mark inet,cidr,macaddr '=' operators as mergejoinable.

23 years agoAdd inheritance.
Bruce Momjian [Sun, 10 Jun 2001 03:48:16 +0000 (03:48 +0000)]
Add inheritance.

23 years agoUpdate TODO list.
Bruce Momjian [Sun, 10 Jun 2001 03:48:05 +0000 (03:48 +0000)]
Update TODO list.

23 years agoUpdate or remove mentions of versions where appropriate.
Bruce Momjian [Sun, 10 Jun 2001 03:46:31 +0000 (03:46 +0000)]
Update or remove mentions of versions where appropriate.

23 years agoFix thinko in hash cost estimation: average frequency
Tom Lane [Sun, 10 Jun 2001 02:59:35 +0000 (02:59 +0000)]
Fix thinko in hash cost estimation: average frequency
should be computed from total number of distinct values in whole
relation, not # distinct values we expect to have after restriction
clauses are applied.

23 years agoUpdate TODO list.
Bruce Momjian [Sun, 10 Jun 2001 02:58:51 +0000 (02:58 +0000)]
Update TODO list.

23 years agoAdd to TODO.detail.
Bruce Momjian [Sun, 10 Jun 2001 02:57:33 +0000 (02:57 +0000)]
Add to TODO.detail.

23 years agoUpdate TODO list.
Bruce Momjian [Sun, 10 Jun 2001 01:25:01 +0000 (01:25 +0000)]
Update TODO list.

23 years agoAllow GRANT/REVOKE to/from more than one user per invocation. Command tag
Peter Eisentraut [Sat, 9 Jun 2001 23:21:55 +0000 (23:21 +0000)]
Allow GRANT/REVOKE to/from more than one user per invocation.  Command tag
for GRANT/REVOKE is now just that, not "CHANGE".

On the way, migrate some of the aclitem internal representation away from
the parser and build a real parse tree instead.  Also add some 'const'
qualifiers.

23 years agoTeach convert_to_scalar about datatypes timetz, inet, cidr, macaddr.
Tom Lane [Sat, 9 Jun 2001 22:16:18 +0000 (22:16 +0000)]
Teach convert_to_scalar about datatypes timetz, inet, cidr, macaddr.

23 years agoAdd #define for OID of type macaddr.
Tom Lane [Sat, 9 Jun 2001 21:41:42 +0000 (21:41 +0000)]
Add #define for OID of type macaddr.

23 years agoRemove RelationGetBufferWithBuffer(), which is horribly confused about
Tom Lane [Sat, 9 Jun 2001 18:16:59 +0000 (18:16 +0000)]
Remove RelationGetBufferWithBuffer(), which is horribly confused about
appropriate pin-count manipulation, and instead use ReleaseAndReadBuffer.
Make use of the fact that the passed-in buffer (if there is one) must
be pinned to avoid grabbing the bufmgr spinlock when we are able to
return this same buffer.  Eliminate unnecessary 'previous tuple' and
'next tuple' fields of HeapScanDesc and IndexScanDesc, thereby removing
a whole lot of bookkeeping from heap_getnext() and related routines.

23 years agoUpdate FAQ.
Bruce Momjian [Sat, 9 Jun 2001 14:40:41 +0000 (14:40 +0000)]
Update FAQ.

23 years agoFix a few missed NOIND usages.
Tom Lane [Sat, 9 Jun 2001 02:19:07 +0000 (02:19 +0000)]
Fix a few missed NOIND usages.

23 years agoFinish German translation, edit (original) messages a bit.
Peter Eisentraut [Fri, 8 Jun 2001 23:53:48 +0000 (23:53 +0000)]
Finish German translation, edit (original) messages a bit.

23 years agoPreliminary code cleanup in elog(). Split out some code into utility
Peter Eisentraut [Fri, 8 Jun 2001 21:16:49 +0000 (21:16 +0000)]
Preliminary code cleanup in elog().  Split out some code into utility
functions, remove indent support, make sure all strings are marked
translatable.

23 years agoDocument and work around ANT bug that prevents directory deletion.
Bruce Momjian [Thu, 7 Jun 2001 20:24:54 +0000 (20:24 +0000)]
Document and work around ANT bug that prevents directory deletion.

23 years agoAdd to TODO.detail.
Bruce Momjian [Thu, 7 Jun 2001 20:06:16 +0000 (20:06 +0000)]
Add to TODO.detail.

23 years agoUpdate TODO list.
Bruce Momjian [Thu, 7 Jun 2001 20:05:56 +0000 (20:05 +0000)]
Update TODO list.

23 years agoreset all: command line and .conf options change defaults
Bruce Momjian [Thu, 7 Jun 2001 04:50:57 +0000 (04:50 +0000)]
reset all: command line and .conf options change defaults
on RESET ALL those are restored.

show all: GUC + non-GUC.

SHOW ALL, RESET ALL

Marko Kreen

23 years agoI have ported another Python Postgres interface to Windows (PgSQL). It seems
Bruce Momjian [Thu, 7 Jun 2001 00:10:18 +0000 (00:10 +0000)]
I have ported another Python Postgres interface to Windows (PgSQL). It seems
that not many people actually use libpq on Win32; I have found another bug. Some
functions that are defined in libpq-fe.h aren't exported in the DLL version of
the library. I have added them to src/interfaces/libpq/libpqdll.def. The new
complete file is attached.

Gerhard H?ring

23 years agoThis adds unary plus capability. No grammar changes, per Tom's request.
Bruce Momjian [Thu, 7 Jun 2001 00:09:32 +0000 (00:09 +0000)]
This adds unary plus capability.  No grammar changes, per Tom's request.

Marko Kreen

23 years agoget_seq_name should truncate name to NAMEDATALEN, so that this works:
Tom Lane [Wed, 6 Jun 2001 22:03:48 +0000 (22:03 +0000)]
get_seq_name should truncate name to NAMEDATALEN, so that this works:
create sequence a1234567890123456789012345678901234567890;
select nextval('a1234567890123456789012345678901234567890');

23 years agoBe a little smarter about deciding how many most-common values to save.
Tom Lane [Wed, 6 Jun 2001 21:29:17 +0000 (21:29 +0000)]
Be a little smarter about deciding how many most-common values to save.

23 years agoAdd large object finalization cleanup to the proper java file.
Bruce Momjian [Wed, 6 Jun 2001 21:16:09 +0000 (21:16 +0000)]
Add large object finalization cleanup to the proper java file.

23 years agoRemove large object finalize code. Compile error.
Bruce Momjian [Wed, 6 Jun 2001 21:04:20 +0000 (21:04 +0000)]
Remove large object finalize code.  Compile error.

23 years agopgindent fix for asterisk indented too much in return type, for Tom.
Bruce Momjian [Wed, 6 Jun 2001 20:51:31 +0000 (20:51 +0000)]
pgindent fix for asterisk indented too much in return type, for Tom.

23 years agoPatch from Ian Lance Taylor fixing multiple cursor arguments
Jan Wieck [Wed, 6 Jun 2001 18:54:41 +0000 (18:54 +0000)]
Patch from Ian Lance Taylor fixing multiple cursor arguments
and buffer zero termination.

Jan

23 years agoAdjust WAL code so that checkpoints truncate the xlog at the previous
Tom Lane [Wed, 6 Jun 2001 17:07:46 +0000 (17:07 +0000)]
Adjust WAL code so that checkpoints truncate the xlog at the previous
checkpoint's redo pointer, not its undo pointer, per discussion in
pghackers a few days ago.  No point in hanging onto undo information
until we have the ability to do something with it --- and this solves
a rather large problem with log space for long-running transactions.
Also, change all calls of write() to detect the case where write
returned a count less than requested, but failed to set errno.
Presume that this situation indicates ENOSPC, and give the appropriate
error message, rather than a random message associated with the previous
value of errno.

23 years agoUpdate TODO list.
Bruce Momjian [Wed, 6 Jun 2001 04:53:50 +0000 (04:53 +0000)]
Update TODO list.

23 years agoCorrect permissions-checking bugs associated with ancient decision to
Tom Lane [Tue, 5 Jun 2001 19:34:56 +0000 (19:34 +0000)]
Correct permissions-checking bugs associated with ancient decision to
copy PUBLIC access rights into each newly created ACL entry.  Instead
treat each ACL entry as independent flags.  Also clean up some ugliness
in acl.h API.

23 years agoImprove planning of OR indexscan plans: for quals like
Tom Lane [Tue, 5 Jun 2001 17:13:52 +0000 (17:13 +0000)]
Improve planning of OR indexscan plans: for quals like
WHERE (a = 1 or a = 2) and b = 42
and an index on (a,b), include the clause b = 42 in the indexquals
generated for each arm of the OR clause.  Essentially this is an index-
driven conversion from CNF to DNF.  Implementation is a bit klugy, but
better than not exploiting the extra quals at all ...

23 years agoFurther work on making use of new statistics in planner. Adjust APIs
Tom Lane [Tue, 5 Jun 2001 05:26:05 +0000 (05:26 +0000)]
Further work on making use of new statistics in planner.  Adjust APIs
of costsize.c routines to pass Query root, so that costsize can figure
more things out by itself and not be so dependent on its callers to tell
it everything it needs to know.  Use selectivity of hash or merge clause
to estimate number of tuples processed internally in these joins
(this is more useful than it would've been before, since eqjoinsel is
somewhat more accurate than before).

23 years agoThis patch adds support for %TYPE in CREATE FUNCTION argument and return
Bruce Momjian [Mon, 4 Jun 2001 23:27:23 +0000 (23:27 +0000)]
This patch adds support for %TYPE in CREATE FUNCTION argument and return
types.  This version has an elog() to remind the user the type
resolution is not dynamic.

Ian Lance Taylor

23 years agoUpdate with new email on ftp location of Karel patch.
Bruce Momjian [Mon, 4 Jun 2001 20:52:51 +0000 (20:52 +0000)]
Update with new email on ftp location of Karel patch.

23 years agoAdd Ross's email to schema.
Bruce Momjian [Mon, 4 Jun 2001 20:29:08 +0000 (20:29 +0000)]
Add Ross's email to schema.

23 years agoAdd schema mention.
Bruce Momjian [Mon, 4 Jun 2001 20:28:21 +0000 (20:28 +0000)]
Add schema mention.

23 years agoUpdate TODO list.
Bruce Momjian [Mon, 4 Jun 2001 20:27:52 +0000 (20:27 +0000)]
Update TODO list.

23 years agoAdd domain to TODO.detail
Bruce Momjian [Mon, 4 Jun 2001 20:13:01 +0000 (20:13 +0000)]
Add domain to TODO.detail

23 years agoUpdate TODO list.
Bruce Momjian [Mon, 4 Jun 2001 20:12:22 +0000 (20:12 +0000)]
Update TODO list.

23 years agoprotected void finalize() {
Bruce Momjian [Mon, 4 Jun 2001 20:10:41 +0000 (20:10 +0000)]
protected void finalize() {
        close();
}

in LargeObject.java so that the db resources are released when it is
garbage collected or am I missing something?

Philip Crotwell

23 years agoGive error message, rather than coredump, for utility statements in
Tom Lane [Mon, 4 Jun 2001 16:17:30 +0000 (16:17 +0000)]
Give error message, rather than coredump, for utility statements in
conditional rules (rules with WHERE clauses).  We cannot support these
since there's noplace to hang a condition on a utility statement.
We caught the other case (attempt to attach a condition at rewrite time)
awhile ago, but this one escaped notice until now.

23 years agoMark many strings in backend not covered by elog for translation. Also,
Peter Eisentraut [Sun, 3 Jun 2001 14:53:56 +0000 (14:53 +0000)]
Mark many strings in backend not covered by elog for translation.  Also,
make strings in xlog.c look more like English and less like binary noise.

23 years agoAccept and output '-Infinity' as well as 'Infinity', per long-ago
Tom Lane [Sat, 2 Jun 2001 20:18:30 +0000 (20:18 +0000)]
Accept and output '-Infinity' as well as 'Infinity', per long-ago
suggestion from Ross Reedstrom.  Still needs work to make those symbols
convert to actual IEEE infinities (on machines where such things exist).

23 years agoUpdate TODO list.
Bruce Momjian [Sat, 2 Jun 2001 19:11:42 +0000 (19:11 +0000)]
Update TODO list.

23 years agoTweak sorting so that nulls appear at the front of a descending sort
Tom Lane [Sat, 2 Jun 2001 19:01:53 +0000 (19:01 +0000)]
Tweak sorting so that nulls appear at the front of a descending sort
(vs. at the end of a normal sort).  This ensures that explicit sorts
yield the same ordering as a btree index scan.  To be really sure that
that equivalence holds, we use the btree entries in pg_amop to decide
whether we are looking at a '<' or '>' operator.  For a sort operator
that has no btree association, we put the nulls at the front if the
operator is named '>' ... pretty grotty, but it does the right thing in
simple ASC and DESC cases, and at least there's no possibility of getting
a different answer depending on the plan type chosen.

23 years agoNative Language Support (NLS)
Peter Eisentraut [Sat, 2 Jun 2001 18:25:18 +0000 (18:25 +0000)]
Native Language Support (NLS)

Use --enable-nls to turn it on; see installation instructions for details.
See developer's guide how to make use of it in programs and how to add
translations.

psql sources have been almost fully prepared and an incomplete German
translation has been provided.  In the backend, only elog() calls are
currently translatable, and the provided German translation file is more
of a placeholder.

23 years agoParanoia about unordered comparisons in IEEE float math. If we are
Tom Lane [Sat, 2 Jun 2001 17:12:12 +0000 (17:12 +0000)]
Paranoia about unordered comparisons in IEEE float math.  If we are
given values that compare as unordered, make sure we reply that they
are equal, which is better than giving an arbitrary answer --- at least
it doesn't depend on which one is passed as which arg.

23 years agodllist.c is included in the frontend libpq interface via symlink.
Jan Wieck [Sat, 2 Jun 2001 15:16:55 +0000 (15:16 +0000)]
dllist.c is included in the frontend libpq interface via symlink.
There is no elog() available. Used fprintf(stderr, ...) and exit
instead.

Jan

23 years agoThe following patch for JDBC fixes an issue with jdbc running on a
Bruce Momjian [Fri, 1 Jun 2001 20:57:58 +0000 (20:57 +0000)]
The following patch for JDBC fixes an issue with jdbc running on a
non-multibyte database loosing 8bit characters.  This patch will cause
the jdbc driver to ignore the encoding reported by the database when
multibyte isn't enabled and use the JVM default in that case.

Barry Lind

23 years agoCheck for malloc failure.
Tom Lane [Fri, 1 Jun 2001 20:29:43 +0000 (20:29 +0000)]
Check for malloc failure.