]> granicus.if.org Git - postgresql/log
postgresql
19 years agoFix pg_dumpall to do the right thing with "postgres" database, per
Tom Lane [Tue, 21 Jun 2005 15:22:18 +0000 (15:22 +0000)]
Fix pg_dumpall to do the right thing with "postgres" database, per
Dave Page.  Also, cause it to emit rather than ignore any ACL and
datconfig options that may be set for these two databases.

19 years agoCause initdb to create a third standard database "postgres", which
Tom Lane [Tue, 21 Jun 2005 04:02:34 +0000 (04:02 +0000)]
Cause initdb to create a third standard database "postgres", which
unlike template0 and template1 does not have any special status in
terms of backend functionality.  However, all external utilities such
as createuser and createdb now connect to "postgres" instead of
template1, and the documentation is changed to encourage people to use
"postgres" instead of template1 as a play area.  This should fix some
longstanding gotchas involving unexpected propagation of database
objects by createdb (when you used template1 without understanding
the implications), as well as ameliorating the problem that CREATE
DATABASE is unhappy if anyone else is connected to template1.
Patch by Dave Page, minor editing by Tom Lane.  All per recent
pghackers discussions.

19 years agoFix a potential backend crash during authentication when parsing a
Neil Conway [Tue, 21 Jun 2005 01:20:09 +0000 (01:20 +0000)]
Fix a potential backend crash during authentication when parsing a
malformed ident map file.  This was introduced by the linked list
rewrite in 8.0 -- mea maxima culpa.

Per Coverity static analysis performed by EnterpriseDB.

19 years agoTrivial dead code removal: in CreateSchemaCommand(), 'owner_name' is
Neil Conway [Tue, 21 Jun 2005 00:58:15 +0000 (00:58 +0000)]
Trivial dead code removal: in CreateSchemaCommand(), 'owner_name' is
only used in one branch of an if statement, so we can move its
declaration to that block. This also avoids an unnecessary syscache
lookup.

Per Coverity static analysis performed by EnterpriseDB.

19 years agoTrivial dead code removal: in _complete_from_query(), 'text' cannot be
Neil Conway [Tue, 21 Jun 2005 00:48:33 +0000 (00:48 +0000)]
Trivial dead code removal: in _complete_from_query(), 'text' cannot be
NULL (e.g. due to the preceding strlen()). Therefore we needn't recheck
this before initializing 'e_text'.

Per Coverity static analysis performed by EnterpriseDB.

19 years agoTrivial dead code removal: in makeObjectName(), name1 must be non-NULL
Neil Conway [Tue, 21 Jun 2005 00:35:05 +0000 (00:35 +0000)]
Trivial dead code removal: in makeObjectName(), name1 must be non-NULL
(due to the preceding strlen(), for example), so we needn't recheck this
before invoking pg_mbcliplen().

Per Coverity static analysis performed by EnterpriseDB.

19 years agoexec_eval_datum leaks memory when dealing with ROW or REC values.
Tom Lane [Mon, 20 Jun 2005 22:51:29 +0000 (22:51 +0000)]
exec_eval_datum leaks memory when dealing with ROW or REC values.
It never leaked memory before PG 8.0, so none of the callers are
expecting this.  Cleanest fix seems to be to make it allocate the needed
memory in estate->eval_econtext, where it will be cleaned up by
the next exec_eval_cleanup.  Per report from Bill Rugolsky.

19 years agoRemove read_file/write_file tests. These were originally intended to
Tom Lane [Mon, 20 Jun 2005 21:14:01 +0000 (21:14 +0000)]
Remove read_file/write_file tests.  These were originally intended to
*fail*, to test that plpython didn't allow untrusted operations.
When we changed plpython to plpythonu because python didn't actually have
a secure sandbox mode, someone (probably me :-() misinterpreted the tests
as checking whether Python's file I/O works.  Which is a stupid thing for
us to be testing.  Remove it so we don't clutter the filesystem with
random temporary files.

19 years agoplpgsql's exec_assign_value() freed the old value of a variable before
Tom Lane [Mon, 20 Jun 2005 20:44:44 +0000 (20:44 +0000)]
plpgsql's exec_assign_value() freed the old value of a variable before
copying/converting the new value, which meant that it failed badly on
"var := var" if var is of pass-by-reference type.  Fix this and a similar
hazard in exec_move_row(); not sure that the latter can manifest before
8.0, but patch it all the way back anyway.  Per report from Dave Chapeskie.

19 years agoAvoid WAL-logging individual tuple insertions during CREATE TABLE AS
Tom Lane [Mon, 20 Jun 2005 18:37:02 +0000 (18:37 +0000)]
Avoid WAL-logging individual tuple insertions during CREATE TABLE AS
(a/k/a SELECT INTO).  Instead, flush and fsync the whole relation before
committing.  We do still need the WAL log when PITR is active, however.
Simon Riggs and Tom Lane.

19 years agofix founded hole in recovery after crash, add vacuum_delay_point()
Teodor Sigaev [Mon, 20 Jun 2005 15:22:38 +0000 (15:22 +0000)]
fix founded hole in recovery after crash, add vacuum_delay_point()

19 years agoRemove statement that GiST indexes aren't WAL-logged.
Tom Lane [Mon, 20 Jun 2005 14:53:32 +0000 (14:53 +0000)]
Remove statement that GiST indexes aren't WAL-logged.

19 years agoUpdate text that stated GiST indexes aren't WAL-logged.
Tom Lane [Mon, 20 Jun 2005 13:52:17 +0000 (13:52 +0000)]
Update text that stated GiST indexes aren't WAL-logged.

19 years ago1. full functional WAL for GiST
Teodor Sigaev [Mon, 20 Jun 2005 10:29:37 +0000 (10:29 +0000)]
1. full functional WAL for GiST
2. improve vacuum for gist
   - use FSM
   - full vacuum:
      - reforms parent tuple if it's needed
        ( tuples was deleted on child page or parent tuple remains invalid
          after crash recovery )
      - truncate index file if possible
3. fixes bugs and mistakes

19 years agoCosmetic improvements to the timezone code: remove the use of the
Neil Conway [Mon, 20 Jun 2005 08:00:51 +0000 (08:00 +0000)]
Cosmetic improvements to the timezone code: remove the use of the
'register' qualifier, make some function declarations more consistent,
and so on.

19 years agoChange shell syntax that seems not to work right on FreeBSD 6-CURRENT
Tom Lane [Mon, 20 Jun 2005 02:26:50 +0000 (02:26 +0000)]
Change shell syntax that seems not to work right on FreeBSD 6-CURRENT
buildfarm machines.

19 years agoAdd lock file contents printout to "can't happen" case to help
Tom Lane [Mon, 20 Jun 2005 02:17:30 +0000 (02:17 +0000)]
Add lock file contents printout to "can't happen" case to help
investigate buildfarm failures.

19 years agoTrivial code clarity improvement to UpdateStats(); no functional change.
Neil Conway [Mon, 20 Jun 2005 02:07:47 +0000 (02:07 +0000)]
Trivial code clarity improvement to UpdateStats(); no functional change.

19 years agoFix bogus assumption that sizeof() produces an int-sized result.
Tom Lane [Mon, 20 Jun 2005 00:32:22 +0000 (00:32 +0000)]
Fix bogus assumption that sizeof() produces an int-sized result.

19 years agoMinor documentation update for RAISE expression patch.
Neil Conway [Sun, 19 Jun 2005 23:39:05 +0000 (23:39 +0000)]
Minor documentation update for RAISE expression patch.

19 years agoAvoid unnecessary palloc overhead in _bt_first(). The temporary
Tom Lane [Sun, 19 Jun 2005 22:41:00 +0000 (22:41 +0000)]
Avoid unnecessary palloc overhead in _bt_first().  The temporary
scankeys arrays that it needs can never have more than INDEX_MAX_KEYS
entries, so it's reasonable to just allocate them as fixed-size local
arrays, and save the cost of palloc/pfree.  Not a huge savings, but
a cycle saved is a cycle earned ...

19 years agoNeed #include <time.h> on some platforms.
Tom Lane [Sun, 19 Jun 2005 22:34:56 +0000 (22:34 +0000)]
Need #include <time.h> on some platforms.

19 years agoSimplify uses of readdir() by creating a function ReadDir() that
Tom Lane [Sun, 19 Jun 2005 21:34:03 +0000 (21:34 +0000)]
Simplify uses of readdir() by creating a function ReadDir() that
includes error checking and an appropriate ereport(ERROR) message.
This gets rid of rather tedious and error-prone manipulation of errno,
as well as a Windows-specific bug workaround, at more than a dozen
call sites.  After an idea in a recent patch by Heikki Linnakangas.

19 years agoArrange to fsync two-phase-commit state files only during checkpoints;
Tom Lane [Sun, 19 Jun 2005 20:00:39 +0000 (20:00 +0000)]
Arrange to fsync two-phase-commit state files only during checkpoints;
given reasonably short lifespans for prepared transactions, this should
mean that only a small minority of state files ever need to be fsynced
at all.  Per discussion with Heikki Linnakangas.

19 years agoAdd comment about \# in bcc makefiles.
Bruce Momjian [Sun, 19 Jun 2005 13:14:20 +0000 (13:14 +0000)]
Add comment about \# in bcc makefiles.

19 years agoCleanup for Win32 compile.
Bruce Momjian [Sun, 19 Jun 2005 13:10:56 +0000 (13:10 +0000)]
Cleanup for Win32 compile.

Andreas Pflug

19 years agoFix typo in comment.
Bruce Momjian [Sun, 19 Jun 2005 01:06:12 +0000 (01:06 +0000)]
Fix typo in comment.

Michael Fuhr

19 years agoWhen using C-string lookup keys in a dynahash.c hash table, use strncpy()
Tom Lane [Sat, 18 Jun 2005 20:51:30 +0000 (20:51 +0000)]
When using C-string lookup keys in a dynahash.c hash table, use strncpy()
not memcpy() to copy the offered key into the hash table during HASH_ENTER.
This avoids possible core dump if the passed key is located very near the
end of memory.  Per report from Stefan Kaltenbrunner.

19 years agoAdd a time-of-preparation column to the pg_prepared_xacts view, per an
Tom Lane [Sat, 18 Jun 2005 19:33:42 +0000 (19:33 +0000)]
Add a time-of-preparation column to the pg_prepared_xacts view, per an
old suggestion by Oliver Jowett.  Also, add a transaction column to the
pg_locks view to show the xid of each transaction holding or awaiting
locks; this allows prepared transactions to be properly associated with
the locks they own.  There was already a column named 'transaction',
and I chose to rename it to 'transactionid' --- since this column is
new in the current devel cycle there should be no backwards compatibility
issue to worry about.

19 years agoDept. of second thoughts: regular COMMIT deletes deletable files before
Tom Lane [Sat, 18 Jun 2005 05:21:09 +0000 (05:21 +0000)]
Dept. of second thoughts: regular COMMIT deletes deletable files before
releasing locks, so COMMIT PREPARED should too.

19 years agoRemove 2phase commit TODO.detail.
Bruce Momjian [Sat, 18 Jun 2005 01:51:03 +0000 (01:51 +0000)]
Remove 2phase commit TODO.detail.

19 years agoRemove TODO.detail.
Bruce Momjian [Sat, 18 Jun 2005 01:45:58 +0000 (01:45 +0000)]
Remove TODO.detail.
< * -Add two-phase commit [2phase]
> * -Add two-phase commit

19 years agoDone:
Bruce Momjian [Sat, 18 Jun 2005 01:45:03 +0000 (01:45 +0000)]
Done:

< * Add two-phase commit [2phase]
> * -Add two-phase commit [2phase]

19 years agoTwo-phase commit. Original patch by Heikki Linnakangas, with additional
Tom Lane [Fri, 17 Jun 2005 22:32:51 +0000 (22:32 +0000)]
Two-phase commit.  Original patch by Heikki Linnakangas, with additional
hacking by Alvaro Herrera and Tom Lane.

19 years agoRe-order item.
Bruce Momjian [Fri, 17 Jun 2005 19:20:27 +0000 (19:20 +0000)]
Re-order item.

> * Auto-fill the free space map by scanning the buffer cache or by
>   checking pages written by the background writer
< * Auto-fill the free space map by scanning the buffer cache or by
<   checking pages written by the background writer

19 years agoMove item:
Bruce Momjian [Fri, 17 Jun 2005 13:47:24 +0000 (13:47 +0000)]
Move item:

* Auto-fill the free space map by scanning the buffer cache or by
  checking pages written by the background writer

19 years agoUpdate NetBSD 'options' description.
Bruce Momjian [Fri, 17 Jun 2005 13:12:01 +0000 (13:12 +0000)]
Update NetBSD 'options' description.

Kris Jurka

19 years agoImprove wording of read-only GUC variables; shows -> reports.
Bruce Momjian [Fri, 17 Jun 2005 03:25:51 +0000 (03:25 +0000)]
Improve wording of read-only GUC variables;  shows -> reports.

19 years agoAdd:
Bruce Momjian [Fri, 17 Jun 2005 02:20:43 +0000 (02:20 +0000)]
Add:

>
> * Create a bitmap of pages that need vacuuming
>
>   Instead of sequentially scanning the entire table, have the background
>   writer or some other process record pages that have expired rows, then
>   VACUUM can look at just those pages rather than the entire table.  In
>   the event of a system crash, the bitmap would probably be invalidated.

19 years agoUpdate GUC description for kerberos usernames.
Bruce Momjian [Thu, 16 Jun 2005 20:47:20 +0000 (20:47 +0000)]
Update GUC description for kerberos usernames.

19 years agoAdd fsync() define for Win32 to cover cases other than wal_sync_method
Bruce Momjian [Thu, 16 Jun 2005 17:53:54 +0000 (17:53 +0000)]
Add fsync() define for Win32 to cover cases other than wal_sync_method
where we need fsync().

19 years agoAdd \x hex support to ecpg strings. This just passes them to the backend.
Bruce Momjian [Thu, 16 Jun 2005 01:43:48 +0000 (01:43 +0000)]
Add \x hex support to ecpg strings.  This just passes them to the backend.

19 years agoMove SYMMETRIC/ASYMMETRIC to reserved words to avoid shift/reduce conflicts.
Bruce Momjian [Wed, 15 Jun 2005 19:44:05 +0000 (19:44 +0000)]
Move SYMMETRIC/ASYMMETRIC to reserved words to avoid shift/reduce conflicts.

19 years agoImprove comment wording.
Bruce Momjian [Wed, 15 Jun 2005 16:28:06 +0000 (16:28 +0000)]
Improve comment wording.

19 years agoImprove hash method for bitmapsets: some examination of actual outputs
Tom Lane [Wed, 15 Jun 2005 16:24:07 +0000 (16:24 +0000)]
Improve hash method for bitmapsets: some examination of actual outputs
shows that adding a circular shift between words greatly improves the
distribution of hash outputs.

19 years agoFix display of database name during autovacuum.
Bruce Momjian [Wed, 15 Jun 2005 13:55:23 +0000 (13:55 +0000)]
Fix display of database name during autovacuum.

 Cosimo Streppone

19 years agoUpdate catalog version for recent function additions.
Bruce Momjian [Wed, 15 Jun 2005 12:56:35 +0000 (12:56 +0000)]
Update catalog version for recent function additions.

19 years agoChange the implementation of hash join to attempt to avoid unnecessary
Neil Conway [Wed, 15 Jun 2005 07:27:44 +0000 (07:27 +0000)]
Change the implementation of hash join to attempt to avoid unnecessary
work if either of the join relations are empty. The logic is:

(1) if the inner relation's startup cost is less than the outer
    relation's startup cost and this is not an outer join, read
    a single tuple from the inner relation via ExecHash()
      - if NULL, we're done

(2) read a single tuple from the outer relation
      - if NULL, we're done

(3) build the hash table on the inner relation
      - if hash table is empty and this is not an outer join,
        we're done

(4) otherwise, do hash join as usual

The implementation uses the new MultiExecProcNode API, per a
suggestion from Tom: invoking ExecHash() now produces the first
tuple from the Hash node's child node, whereas MultiExecHash()
builds the hash table.

I had to put in a bit of a kludge to get the row count returned
for EXPLAIN ANALYZE to be correct: since ExecHash() is invoked to
return a tuple, and then MultiExecHash() is invoked, we would
return one too many tuples to EXPLAIN ANALYZE. I hacked around
this by just manually detecting this situation and subtracting 1
from the EXPLAIN ANALYZE row count.

19 years agoMinor SGML markup cleanup.
Neil Conway [Wed, 15 Jun 2005 06:29:25 +0000 (06:29 +0000)]
Minor SGML markup cleanup.

19 years agoRemove old *.backup files when we do pg_stop_backup(). This
Bruce Momjian [Wed, 15 Jun 2005 01:36:08 +0000 (01:36 +0000)]
Remove old *.backup files when we do pg_stop_backup().  This
prevents a large number of *.backup files from existing in pg_xlog/

19 years agoMention we now support BETWEEN SYMMETRIC.
Bruce Momjian [Wed, 15 Jun 2005 01:28:43 +0000 (01:28 +0000)]
Mention we now support BETWEEN SYMMETRIC.

Christopher Kings-Lynne

19 years ago> Here's a patch I added against plperl, originally against beta5, now
Bruce Momjian [Wed, 15 Jun 2005 00:35:16 +0000 (00:35 +0000)]
> Here's a patch I added against plperl, originally against beta5, now
> against rc1. It simply checks with GetDatabaseEncoding() if the current
> database is in UTF-8, and if so, sets the UTF-8 flag on the arguments
> that are passed to perl. This means that it isn't necessary to
> utf8::upgrade() every string, as perl has no way of knowing offhand
> that a string is UTF-8 -- but postgres does, because the database
> encoding is specified, so it makes sense to turn the flag on. You
> should also be able to properly manipulate UTF-8 strings now from
> plperl as opposed to plperlu, because otherwise you'd have to use
> encoding 'utf8' which was not allowed. It could also eliminate some
> unexpected bugs if you assume that perl knows the string is unicode.
It
> is enabled only for perl 5.6 and higher, so earlier versions will not
> be affected.
>
> I have been assured by crab that the patch is quite harmless and will
> not break anything. It would be great to see it in 8 final! :-)

David Kamholz

19 years agoThis patch makes it possible to use the full set of timezones when doing
Bruce Momjian [Wed, 15 Jun 2005 00:34:11 +0000 (00:34 +0000)]
This patch makes it possible to use the full set of timezones when doing
"AT TIME ZONE", and not just the shorlist previously available. For
example:

SELECT CURRENT_TIMESTAMP AT TIME ZONE 'Europe/London';

works fine now. It will also obey whatever DST rules were in effect at
just that date, which the previous implementation did not.

It also supports the AT TIME ZONE on the timetz datatype. The whole
handling of DST is a bit bogus there, so I chose to make it use whatever
DST rules are in effect at the time of executig the query. not sure if
anybody is actuallyi *using* timetz though, it seems pretty
unpredictable just because of this...

Magnus Hagander

19 years agoSupport 3 and 4-byte unicode characters.
Bruce Momjian [Wed, 15 Jun 2005 00:15:08 +0000 (00:15 +0000)]
Support 3 and 4-byte unicode characters.

John Hansen

19 years ago>> Do you agree that using a hashtable for it in general is a good idea
Bruce Momjian [Wed, 15 Jun 2005 00:09:26 +0000 (00:09 +0000)]
>> Do you agree that using a hashtable for it in general is a good idea
>> assuming this sideeffect is removed, though?
>
>I have no problem with the hashtable, only with preloading it with
>everything.  What I'd like to see is that the table inherited at fork()
>contains just the data for the default timezone.  (At least in the
>normal case where that setting hasn't been changed since postmaster
>start.)

Here's a patch doing this. Changes score_timezone not to use pg_tzset(),
and thus not loading all the zones in the cache. The actual timezone
being picked will be set using set_global_timezone() which in turn calls
pg_tzset() and loads it in the cache.

Magnus Hagander

19 years agoNow \d show tablespace of indices per discussion.
Bruce Momjian [Tue, 14 Jun 2005 23:59:31 +0000 (23:59 +0000)]
Now \d show tablespace of indices per discussion.

test=# \d e
       Table "public.e"
 Column |  Type   | Modifiers
--------+---------+-----------
 i      | integer | not null
 j      | integer | not null
 k      | integer |
Indexes:
    "e_pkey" PRIMARY KEY, btree (i, j), tablespace "haha"
    "ei" btree (i)
    "ej" btree (j), tablespace "haha"
    "ek" btree (k)
Tablespace: "haha"

Qingqing Zhou

19 years agoAdd BETWEEN SYMMETRIC.
Bruce Momjian [Tue, 14 Jun 2005 23:47:39 +0000 (23:47 +0000)]
Add BETWEEN SYMMETRIC.

Pavel Stehule

19 years agoreorder:
Bruce Momjian [Tue, 14 Jun 2005 23:44:47 +0000 (23:44 +0000)]
reorder:
< * -Add BETWEEN ASYMMETRIC/SYMMETRIC
> * -Add BETWEEN SYMMETRIC/ASYMMETRIC

19 years agoDone:
Bruce Momjian [Tue, 14 Jun 2005 23:44:27 +0000 (23:44 +0000)]
Done:

> * -Add BETWEEN ASYMMETRIC/SYMMETRIC

19 years agoUse '&nbsp;' escapes for leading spaces in table values, for use in
Bruce Momjian [Tue, 14 Jun 2005 22:15:57 +0000 (22:15 +0000)]
Use '&nbsp;' escapes for leading spaces in table values, for use in
EXPLAIN output.

Jean-Paul Argudo

19 years agoSimplify shared-memory lock data structures as per recent discussion:
Tom Lane [Tue, 14 Jun 2005 22:15:33 +0000 (22:15 +0000)]
Simplify shared-memory lock data structures as per recent discussion:
it is sufficient to track whether a backend holds a lock or not, and
store information about transaction vs. session locks only in the
inside-the-backend LocalLockTable.  Since there can now be but one
PROCLOCK per lock per backend, LockCountMyLocks() is no longer needed,
thus eliminating some O(N^2) behavior when a backend holds many locks.
Also simplify the LockAcquire/LockRelease API by passing just a
'sessionLock' boolean instead of a transaction ID.  The previous API
was designed with the idea that per-transaction lock holding would be
important for subtransactions, but now that we have subtransactions we
know that this is unwanted.  While at it, add an 'isTempObject' parameter
to LockAcquire to indicate whether the lock is being taken on a temp
table.  This is not used just yet, but will be needed shortly for
two-phase commit.

19 years agoAdd pg_postmaster_start_time() function.
Bruce Momjian [Tue, 14 Jun 2005 21:04:42 +0000 (21:04 +0000)]
Add pg_postmaster_start_time() function.

Euler Taveira de Oliveira
Matthias Schmidt

19 years agoDone:
Bruce Momjian [Tue, 14 Jun 2005 20:43:02 +0000 (20:43 +0000)]
Done:

>  o -Have SHOW ALL show descriptions for server-side variables

19 years agoHave SHOW ALL include variable descriptions.
Bruce Momjian [Tue, 14 Jun 2005 20:42:53 +0000 (20:42 +0000)]
Have SHOW ALL include variable descriptions.

Matthias Schmidt

19 years agoUpdate:
Bruce Momjian [Tue, 14 Jun 2005 18:24:06 +0000 (18:24 +0000)]
Update:

<  o Have SHOW ALL and pg_settings show descriptions for server-side variables
>  o Have SHOW ALL show descriptions for server-side variables

19 years agoMove item into proper section:
Bruce Momjian [Tue, 14 Jun 2005 18:10:42 +0000 (18:10 +0000)]
Move item into proper section:

        o Have SHOW ALL and pg_settings show descriptions for server-side variables

19 years agoAttached is a makefile I hacked up to build pg_config under MSVC - the
Bruce Momjian [Tue, 14 Jun 2005 17:50:38 +0000 (17:50 +0000)]
Attached is a makefile I hacked up to build pg_config under MSVC - the
reason is that it's required (more or less) in order to build the latest
DBD::Pg code and I was testing that out under MSVC.

Andrew Dunstan

19 years agoAdd GUC krb_server_hostname so the server hostname can be specified as
Bruce Momjian [Tue, 14 Jun 2005 17:43:14 +0000 (17:43 +0000)]
Add GUC krb_server_hostname so the server hostname can be specified as
part of service principal.  If not set, any service principal matching
an entry in the keytab can be used.

NEW KERBEROS MATCHING BEHAVIOR FOR 8.1.

Todd Kover

19 years agoDone:
Bruce Momjian [Tue, 14 Jun 2005 15:11:11 +0000 (15:11 +0000)]
Done:

>  o -Allow PL/PgSQL's RAISE function to take expressions

19 years agoThe random selection in function linear() could deliver a value equal to max
Tom Lane [Tue, 14 Jun 2005 14:21:16 +0000 (14:21 +0000)]
The random selection in function linear() could deliver a value equal to max
if geqo_rand() returns exactly 1.0, resulting in failure due to indexing
off the end of the pool array.  Also, since this is using inexact float math,
it seems wise to guard against roundoff error producing values slightly
outside the expected range.  Per report from bug@zedware.org.

19 years agoWAL for GiST. It work for online backup and so on, but on
Teodor Sigaev [Tue, 14 Jun 2005 11:45:14 +0000 (11:45 +0000)]
WAL for GiST. It work for online backup and so on, but on
recovery after crash (power loss etc) it may say that it can't restore
index and index should be reindexed.

Some refactoring code.

19 years agoAllow the parameters to PL/PgSQL's RAISE statement to be expressions,
Neil Conway [Tue, 14 Jun 2005 06:43:15 +0000 (06:43 +0000)]
Allow the parameters to PL/PgSQL's RAISE statement to be expressions,
instead of just scalar variables. Add regression tests and update the
documentation. Along the way, remove some redundant error checking
code from exec_stmt_perform().

Original patch from Pavel Stehule, reworked by Neil Conway.

19 years agoTeach planner to optionally ignore index columns that have an equality
Tom Lane [Tue, 14 Jun 2005 04:04:30 +0000 (04:04 +0000)]
Teach planner to optionally ignore index columns that have an equality
constraint while determining whether the index sort order matches the
query's ORDER BY.  This for example allows an index on (x,y) to match
... WHERE x = 42 ORDER BY y;
It only works for btree indexes, but since those are the only ones we
currently have that are ordered at all, that's good enough for now.
Per popular demand.

19 years agoDone:
Bruce Momjian [Tue, 14 Jun 2005 03:20:10 +0000 (03:20 +0000)]
Done:

> * -Use index to restrict rows returned by multi-key index when used with

19 years agoAdd -L option to psql to log sessions.
Bruce Momjian [Tue, 14 Jun 2005 02:57:45 +0000 (02:57 +0000)]
Add -L option to psql to log sessions.

Lorne Sunley

19 years agoCleanup for "#option dump" in PL/PgSQL: don't print empty ELSE blocks,
Neil Conway [Tue, 14 Jun 2005 00:10:02 +0000 (00:10 +0000)]
Cleanup for "#option dump" in PL/PgSQL: don't print empty ELSE blocks,
fix two grammatical errors, and print the INTO target of EXECUTE INTO
if one is specified.

19 years agoChange the planner to allow indexscan qualification clauses to use
Tom Lane [Mon, 13 Jun 2005 23:14:49 +0000 (23:14 +0000)]
Change the planner to allow indexscan qualification clauses to use
nonconsecutive columns of a multicolumn index, as per discussion around
mid-May (pghackers thread "Best way to scan on-disk bitmaps").  This
turns out to require only minimal changes in btree, and so far as I can
see none at all in GiST.  btcostestimate did need some work, but its
original assumption that index selectivity == heap selectivity was
quite bogus even before this.

19 years agoPer discussion on -hackers, this patch changes psql's "expanded" output
Neil Conway [Mon, 13 Jun 2005 06:36:22 +0000 (06:36 +0000)]
Per discussion on -hackers, this patch changes psql's "expanded" output
mode to only affect the presentation of normal query results, not the
output of psql slash commands. Documentation updated. I also made
some unrelated minor psql cleanup. Per suggestion from Stuart Cooper.

19 years agoSGML cleanup: consistently use "endterm" in <xref>s that link to the
Neil Conway [Mon, 13 Jun 2005 02:40:08 +0000 (02:40 +0000)]
SGML cleanup: consistently use "endterm" in <xref>s that link to the
reference page for SQL commands, so that the link text is italicized.

19 years agoAdjust lo_open() so that specifying INV_READ without INV_WRITE creates
Tom Lane [Mon, 13 Jun 2005 02:26:53 +0000 (02:26 +0000)]
Adjust lo_open() so that specifying INV_READ without INV_WRITE creates
a descriptor that uses the current transaction snapshot, rather than
SnapshotNow as it did before (and still does if INV_WRITE is set).
This means pg_dump will now dump a consistent snapshot of large object
contents, as it never could do before.  Also, add a lo_create() function
that is similar to lo_creat() but allows the desired OID of the large
object to be specified.  This will simplify pg_restore considerably
(but I'll fix that in a separate commit).

19 years agoDocument the --enable-integer-datetimes configure option where the other
Tom Lane [Sun, 12 Jun 2005 15:51:50 +0000 (15:51 +0000)]
Document the --enable-integer-datetimes configure option where the other
configure options are documented.  Per Michael Glaesemann.

19 years agoThis patch removes some old code from libpq that implements a URI-like
Neil Conway [Sun, 12 Jun 2005 00:07:07 +0000 (00:07 +0000)]
This patch removes some old code from libpq that implements a URI-like
syntax for database connection parameters. It has been inside an
#ifdef NOT_USED block since 2001 or so and is marked as "broken", so
I don't think it is likely to be rehabilitated any time soon.

19 years agolibpq was not consistently checking for memory allocation failures. This
Neil Conway [Sun, 12 Jun 2005 00:00:21 +0000 (00:00 +0000)]
libpq was not consistently checking for memory allocation failures. This
patch adds missing checks to the call sites of malloc(), strdup(),
PQmakeEmptyPGresult(), pqResultAlloc(), and pqResultStrdup(), and updates
the documentation. Per original report from Volkan Yazici about
PQmakeEmptyPGresult() not checking for malloc() failure.

19 years agoSeparate predicate-testing code out of indxpath.c, making it a module
Tom Lane [Fri, 10 Jun 2005 22:25:37 +0000 (22:25 +0000)]
Separate predicate-testing code out of indxpath.c, making it a module
in its own right.  As proposed by Simon Riggs, but with some editorializing
of my own.

19 years agoAdjust comment about %t and %s to cover %m as well. Some trivial
Tom Lane [Fri, 10 Jun 2005 20:48:54 +0000 (20:48 +0000)]
Adjust comment about %t and %s to cover %m as well.  Some trivial
code beautification too.

19 years agoFix bug in MIC -> EUC_JP conversion. Per Atsushi Ogawa.
Tatsuo Ishii [Fri, 10 Jun 2005 16:43:56 +0000 (16:43 +0000)]
Fix bug in MIC -> EUC_JP conversion. Per Atsushi Ogawa.

19 years agoAdd space before :DBNAME for proper \set expansion.
Bruce Momjian [Fri, 10 Jun 2005 16:31:48 +0000 (16:31 +0000)]
Add space before :DBNAME for proper \set expansion.

Andreas Seltenreich

19 years agoMinor cleanup of documentation for recently-added HISTFILE psql variable.
Neil Conway [Fri, 10 Jun 2005 16:26:59 +0000 (16:26 +0000)]
Minor cleanup of documentation for recently-added HISTFILE psql variable.

19 years agoImplement two new special variables in PL/PgSQL: SQLSTATE and SQLERRM.
Neil Conway [Fri, 10 Jun 2005 16:23:11 +0000 (16:23 +0000)]
Implement two new special variables in PL/PgSQL: SQLSTATE and SQLERRM.
These contain the SQLSTATE and error message of the current exception,
respectively. They are scope-local variables that are only defined
in exception handlers (so attempting to reference them outside an
exception handler is an error). Update the regression tests and the
documentation.

Also, do some minor related cleanup: export an unpack_sql_state()
function from the backend and use it to unpack a SQLSTATE into a
string, and add a free_var() function to pl_exec.c

Original patch from Pavel Stehule, review by Neil Conway.

19 years agoUse saveHistory so we get a nice error message on failure.
Bruce Momjian [Fri, 10 Jun 2005 15:40:41 +0000 (15:40 +0000)]
Use saveHistory so we get a nice error message on failure.

Andreas Seltenreich

19 years agothe following patch makes the filename used to store the readline
Bruce Momjian [Fri, 10 Jun 2005 15:34:26 +0000 (15:34 +0000)]
the following patch makes the filename used to store the readline
history customizable through a variable named HISTFILE, analogous to
psql's already implemented HISTCONTROL and HISTSIZE variables, and
bash's HISTFILE-Variable.

The motivation was to be able to get psql to maintain separate
histories for separate databases.  This is now easily achievable
through a line like the following in ~/.psqlrc:

\set HISTFILE ~/.psql_history-:DBNAME

Andreas Seltenreich

19 years agoMore Win32 tilde code comments.
Bruce Momjian [Fri, 10 Jun 2005 14:49:31 +0000 (14:49 +0000)]
More Win32 tilde code comments.

19 years agoAdd comment on tilde expansion.
Bruce Momjian [Fri, 10 Jun 2005 14:41:32 +0000 (14:41 +0000)]
Add comment on tilde expansion.

19 years agoRemove unneeded variable test, per Tom.
Bruce Momjian [Fri, 10 Jun 2005 04:01:36 +0000 (04:01 +0000)]
Remove unneeded variable test, per Tom.

19 years agoQuick hack to allow the outer query's tuple_fraction to be passed down
Tom Lane [Fri, 10 Jun 2005 03:32:25 +0000 (03:32 +0000)]
Quick hack to allow the outer query's tuple_fraction to be passed down
to a subquery if the outer query is simple enough that the LIMIT can
be reflected directly to the subquery.  This didn't use to be very
interesting, because a subquery that couldn't have been flattened into
the upper query was usually not going to be very responsive to
tuple_fraction anyway.  But with new code that allows UNION ALL subqueries
to pay attention to tuple_fraction, this is useful to do.  In particular
this lets the optimization occur when the UNION ALL is directly inside
a view.

19 years agoAdd the "PGPASSFILE" environment variable to specify to the password
Bruce Momjian [Fri, 10 Jun 2005 03:02:30 +0000 (03:02 +0000)]
Add the "PGPASSFILE" environment variable to specify to the password
file.

Andrew Dunstan

19 years agoIf a LIMIT is applied to a UNION ALL query, plan each UNION arm as
Tom Lane [Fri, 10 Jun 2005 02:21:05 +0000 (02:21 +0000)]
If a LIMIT is applied to a UNION ALL query, plan each UNION arm as
if the limit were directly applied to it.  This does not actually
add a LIMIT plan node to the generated subqueries --- that would be
useless overhead --- but it does cause the planner to prefer fast-
start plans when the limit is small.  After an idea from Phil Endecott.

19 years agoRevise searching of subplan target lists to use something more efficient
Tom Lane [Fri, 10 Jun 2005 00:28:54 +0000 (00:28 +0000)]
Revise searching of subplan target lists to use something more efficient
than tlist_member calls.  Building a large join tlist is still O(N^2),
but with a much smaller constant factor than before.

19 years agoThis patch shows the full path name when doing a \s in psql,
Bruce Momjian [Thu, 9 Jun 2005 23:28:10 +0000 (23:28 +0000)]
This patch shows the full path name when doing a \s in psql,
if you have previously issued a \cd command.

Greg Sabino Mullane

19 years agoFree buffer allocated via malloc (process is short-lived, but fix it anyway).
Bruce Momjian [Thu, 9 Jun 2005 22:36:27 +0000 (22:36 +0000)]
Free buffer allocated via malloc (process is short-lived, but fix it anyway).