]> granicus.if.org Git - postgresql/log
postgresql
24 years agoRemove -O2 for aix.
Bruce Momjian [Fri, 24 Mar 2000 14:39:07 +0000 (14:39 +0000)]
Remove -O2 for aix.

24 years agoMove LOCK terms down.
Bruce Momjian [Fri, 24 Mar 2000 03:57:06 +0000 (03:57 +0000)]
Move LOCK terms down.

24 years agooutfuncs.c was missing a print routine for Material plan nodes, leading
Tom Lane [Fri, 24 Mar 2000 02:58:25 +0000 (02:58 +0000)]
outfuncs.c was missing a print routine for Material plan nodes, leading
to trouble when trying to EXPLAIN VERBOSE a plan containing one.

24 years agoRename bytea functions to not have upper-case letters in their names.
Tom Lane [Fri, 24 Mar 2000 02:41:46 +0000 (02:41 +0000)]
Rename bytea functions to not have upper-case letters in their names.
Clean up grotty coding in them, too.  AFAICS from the CVS logs, these
have been broken since Postgres95, so I'm not going to insist on an
initdb to fix them now...

24 years agoFold PQsetenv working state into PGconn, rather than trying to maintain
Tom Lane [Fri, 24 Mar 2000 01:39:55 +0000 (01:39 +0000)]
Fold PQsetenv working state into PGconn, rather than trying to maintain
it in a separate object.  There's no value in keeping the state separate,
and it creates dangling-pointer problems.  Also, remove PQsetenv routines
from public API, until and unless they are redesigned to have a safer
interface.  Since they were never part of the documented API before 7.0,
it's unlikely that anyone is calling them.

24 years agoUpdate multi-byte support README
Tatsuo Ishii [Fri, 24 Mar 2000 01:37:11 +0000 (01:37 +0000)]
Update multi-byte support README

24 years agoUpdate HISTORY
Bruce Momjian [Fri, 24 Mar 2000 01:20:09 +0000 (01:20 +0000)]
Update HISTORY

24 years agoA little further tweaking of the range-query selectivity logic:
Tom Lane [Thu, 23 Mar 2000 23:35:47 +0000 (23:35 +0000)]
A little further tweaking of the range-query selectivity logic:
to avoid undue sensitivity to roundoff error, believe that a zero
or slightly negative range estimate should represent a small
positive selectivity, rather than falling back on a generic default
estimate.

24 years ago>> 5. empty define that results in an empty but terminated line ( ; )
Bruce Momjian [Thu, 23 Mar 2000 23:16:49 +0000 (23:16 +0000)]
>> 5. empty define that results in an empty but terminated line ( ; )
easy (maybe dumb) fix for 5 in attachment define.patch

greetings, Andreas

24 years agoBack out // compiler flag.
Bruce Momjian [Thu, 23 Mar 2000 22:25:36 +0000 (22:25 +0000)]
Back out // compiler flag.

24 years agoHmm, absolute pathnames for the copy makes sense. I'll whip up that
Bruce Momjian [Thu, 23 Mar 2000 21:38:58 +0000 (21:38 +0000)]
Hmm, absolute pathnames for the copy makes sense. I'll whip up that
patch in a second. Should be sufficent to just make sure the first
character is a '/', right?

Ross J. Reedstrom

24 years agoSome points for portability improvements:
Bruce Momjian [Thu, 23 Mar 2000 17:27:36 +0000 (17:27 +0000)]
Some points for portability improvements:

1. C++ style comments in C source for ecpg ( // comment )
2. compiler finds wrong include file extern.h in ecpg/lib/descriptor.c
from
include path instead of workdir (rename it ?)
3. fe-connect getsockopt takes a socklen_t as fifth arg not int (use
SOCKET_SIZE_TYPE instead)
4. char vs unsigned char in psql calls to libpq
5. empty define that results in an empty but terminated line ( ; )

Now for all but point 3 I can supply changes to the
compiler flags, to make the compiler less pedantic.
Or is someone interested in the complications ?

in the meantime can someone apply the attached patch ?

Andreas

24 years agoUpdate history/sgml.
Bruce Momjian [Thu, 23 Mar 2000 15:09:56 +0000 (15:09 +0000)]
Update history/sgml.

24 years agoPlease apply the attached patch to interfaces/libpq/libpq-fe.h. This would
Bruce Momjian [Thu, 23 Mar 2000 15:00:11 +0000 (15:00 +0000)]
Please apply the attached patch to interfaces/libpq/libpq-fe.h. This would
allow to write applications which do not have a link to postgres_ext.h in
their source directory.

Andreas Kardos

24 years ago*** empty log message ***
Michael Meskes [Thu, 23 Mar 2000 07:53:48 +0000 (07:53 +0000)]
*** empty log message ***

24 years agosubselect regress test was kind of silly; it claimed to test correlation
Tom Lane [Thu, 23 Mar 2000 07:42:13 +0000 (07:42 +0000)]
subselect regress test was kind of silly; it claimed to test correlation
cases but actually did no such thing.  Make it test some more cases than
before (including things that didn't work in 6.5).

24 years agoFloat-to-int conversion functions should return NULL when given NULL
Tom Lane [Thu, 23 Mar 2000 07:40:00 +0000 (07:40 +0000)]
Float-to-int conversion functions should return NULL when given NULL
input, not throw a gratuitous elog().

24 years agoRemove no-longer-necessary restriction against uplevel correlation vars
Tom Lane [Thu, 23 Mar 2000 07:38:30 +0000 (07:38 +0000)]
Remove no-longer-necessary restriction against uplevel correlation vars
outside WHERE clause.  Fix a couple of places that didn't handle uplevel
refs cleanly.

24 years agoHack parse_coerce so it won't try to constant-fold the dummy Const
Tom Lane [Thu, 23 Mar 2000 07:36:03 +0000 (07:36 +0000)]
Hack parse_coerce so it won't try to constant-fold the dummy Const
nodes introduced by make_subplan().  It'd be better if we used a
different node type for subplan result placeholders, but for now...

24 years agoExecSubPlan needs to be able to cope with RelabelType nodes atop the
Tom Lane [Thu, 23 Mar 2000 07:32:58 +0000 (07:32 +0000)]
ExecSubPlan needs to be able to cope with RelabelType nodes atop the
Const placeholder nodes for subplan result values.

24 years agoUpdate TODO list.
Bruce Momjian [Thu, 23 Mar 2000 06:35:58 +0000 (06:35 +0000)]
Update TODO list.

24 years agoUpdate HISTORY file for 7.0.
Bruce Momjian [Thu, 23 Mar 2000 06:30:58 +0000 (06:30 +0000)]
Update HISTORY file for 7.0.

24 years agoPrepare for 7.0 release
Tatsuo Ishii [Thu, 23 Mar 2000 02:04:07 +0000 (02:04 +0000)]
Prepare for 7.0 release

24 years agoIf we cannot get a real estimate for the selectivity of a range query,
Tom Lane [Thu, 23 Mar 2000 00:58:36 +0000 (00:58 +0000)]
If we cannot get a real estimate for the selectivity of a range query,
use a default value that's fairly small.  We were generating a result
of about 0.1, but I think 0.01 is probably better --- want to encourage
use of an indexscan in this situation.

24 years agoImprove selectivity estimation involving string constants: pay attention
Tom Lane [Thu, 23 Mar 2000 00:55:42 +0000 (00:55 +0000)]
Improve selectivity estimation involving string constants: pay attention
to more than one character, and try to do the right thing in non-ASCII
locales.

24 years agoRepair logic flaw in cost estimator: cost_nestloop() was estimating CPU
Tom Lane [Wed, 22 Mar 2000 22:08:35 +0000 (22:08 +0000)]
Repair logic flaw in cost estimator: cost_nestloop() was estimating CPU
costs using the inner path's parent->rows count as the number of tuples
processed per inner scan iteration.  This is wrong when we are using an
inner indexscan with indexquals based on join clauses, because the rows
count in a Relation node reflects the selectivity of the restriction
clauses for that rel only.  Upshot was that if join clause was very
selective, we'd drastically overestimate the true cost of the join.
Fix is to calculate correct output-rows estimate for an inner indexscan
when the IndexPath node is created and save it in the path node.
Change of path node doesn't require initdb, since path nodes don't
appear in saved rules.

24 years agoUpdate pg_dumpall again.
Bruce Momjian [Wed, 22 Mar 2000 05:25:06 +0000 (05:25 +0000)]
Update pg_dumpall again.

24 years agoFix pg_dumpall for new psql output.
Bruce Momjian [Wed, 22 Mar 2000 05:10:22 +0000 (05:10 +0000)]
Fix pg_dumpall for new psql output.

24 years agoForgot that odbc had its own copies of config.sub/config.guess.
Tom Lane [Tue, 21 Mar 2000 06:35:22 +0000 (06:35 +0000)]
Forgot that odbc had its own copies of config.sub/config.guess.
Update those to latest Autoconf sources, too.

24 years agoFix query for primary keys to reflect new DISTINCT ON () syntax.
Thomas G. Lockhart [Tue, 21 Mar 2000 06:02:22 +0000 (06:02 +0000)]
Fix query for primary keys to reflect new DISTINCT ON () syntax.
Reported by "Tibor Laszlo" <ltibor@mail.tiszanet.hu> and fix suggested by
 "Hiroshi Inoue" <Inoue@tpf.co.jp>.

24 years agoAdd syntax for BIT() and BIT VARYING(), but no underlying implementation
Thomas G. Lockhart [Tue, 21 Mar 2000 06:00:41 +0000 (06:00 +0000)]
Add syntax for BIT() and BIT VARYING(), but no underlying implementation
 is available yet.
Remove redundant call to xlateSqlType() in the character
 type handling code.

24 years agoUpdate test for new ORDER BY clause from Tom Lane.
Thomas G. Lockhart [Tue, 21 Mar 2000 05:59:14 +0000 (05:59 +0000)]
Update test for new ORDER BY clause from Tom Lane.

24 years agoRestructure planning code so that preprocessing of targetlist and quals
Tom Lane [Tue, 21 Mar 2000 05:12:12 +0000 (05:12 +0000)]
Restructure planning code so that preprocessing of targetlist and quals
to simplify constant expressions and expand SubLink nodes into SubPlans
is done in a separate routine subquery_planner() that calls union_planner().
We formerly did most of this work in query_planner(), but that's the
wrong place because it may never see the real targetlist.  Splitting
union_planner into two routines also allows us to avoid redundant work
when union_planner is invoked recursively for UNION and inheritance
cases.  Upshot is that it is now possible to do something like
select float8(count(*)) / (select count(*) from int4_tbl)  from int4_tbl
group by f1;
which has never worked before.

24 years agoCorrect typo in error message.
Tom Lane [Tue, 21 Mar 2000 04:20:45 +0000 (04:20 +0000)]
Correct typo in error message.

24 years agoThis patch corrects spelling, grammar and euphony for the psql reference
Bruce Momjian [Tue, 21 Mar 2000 01:52:12 +0000 (01:52 +0000)]
This patch corrects spelling, grammar and euphony for the psql reference
page for 7.0:

Oliver Elphick

24 years agoReverse out BYTEA type coersion.
Bruce Momjian [Mon, 20 Mar 2000 15:42:47 +0000 (15:42 +0000)]
Reverse out BYTEA type coersion.

24 years agoTurn XLOG off (do not create log file).
Vadim B. Mikheev [Mon, 20 Mar 2000 07:25:39 +0000 (07:25 +0000)]
Turn XLOG off (do not create log file).

24 years agoEmit 'this operator is deprecated' warnings for ':' and ';'.
Tom Lane [Mon, 20 Mar 2000 05:20:34 +0000 (05:20 +0000)]
Emit 'this operator is deprecated' warnings for ':' and ';'.

24 years agoConvert float8 regress test to use exp() and ln() instead of ':' and
Tom Lane [Mon, 20 Mar 2000 05:19:11 +0000 (05:19 +0000)]
Convert float8 regress test to use exp() and ln() instead of ':' and
';' operators.

24 years agoUpdate for BYTEAOID.
Bruce Momjian [Mon, 20 Mar 2000 04:26:01 +0000 (04:26 +0000)]
Update for BYTEAOID.

24 years agoSorry, I have read the misspelling 'coersion' one time too many.
Tom Lane [Mon, 20 Mar 2000 04:22:11 +0000 (04:22 +0000)]
Sorry, I have read the misspelling 'coersion' one time too many.

24 years agoAdd exp(), ln(), and some other functions to numeric-functions table.
Tom Lane [Mon, 20 Mar 2000 04:20:52 +0000 (04:20 +0000)]
Add exp(), ln(), and some other functions to numeric-functions table.

24 years agoAdd note that ':' and ';' operators are deprecated.
Tom Lane [Mon, 20 Mar 2000 04:19:47 +0000 (04:19 +0000)]
Add note that ':' and ';' operators are deprecated.

24 years agoAdd compatiblity information for bytea.
Bruce Momjian [Mon, 20 Mar 2000 04:02:47 +0000 (04:02 +0000)]
Add compatiblity information for bytea.

24 years agoAdd FORCE keyword to ColID
Hiroshi Inoue [Mon, 20 Mar 2000 00:24:37 +0000 (00:24 +0000)]
Add FORCE keyword to ColID

24 years agoMake use of pre-existing regexp match capability to eliminate redundant
Tom Lane [Sun, 19 Mar 2000 22:48:30 +0000 (22:48 +0000)]
Make use of pre-existing regexp match capability to eliminate redundant
entries in template selection rules.  Also, change alpha-dec-osf pattern
to cope with version info attached to the 'alpha' part.

24 years agoAutoconf run
Tom Lane [Sun, 19 Mar 2000 22:46:56 +0000 (22:46 +0000)]
Autoconf run

24 years agoChange configure.in to note that since we are using expr(1)'s regex match
Tom Lane [Sun, 19 Mar 2000 22:46:22 +0000 (22:46 +0000)]
Change configure.in to note that since we are using expr(1)'s regex match
command, the entries in template/.similar can really be regular
expressions.  This isn't a new feature, just an observation of what the
code already did.

24 years agocash_words_out function truncated its output by 1 character due to
Tom Lane [Sun, 19 Mar 2000 22:10:52 +0000 (22:10 +0000)]
cash_words_out function truncated its output by 1 character due to
incorrect use of StrNCpy.

24 years agoSeveral calls to StrNCpy incorrectly subtracted 1 from the length arg,
Tom Lane [Sun, 19 Mar 2000 22:10:08 +0000 (22:10 +0000)]
Several calls to StrNCpy incorrectly subtracted 1 from the length arg,
leading to postmaster accepting args 1 shorter than it had room for.

24 years agoChange MemSet and StrNCpy to evaluate their arguments only once.
Tom Lane [Sun, 19 Mar 2000 22:08:51 +0000 (22:08 +0000)]
Change MemSet and StrNCpy to evaluate their arguments only once.
Fix inadequate parenthesization in several other macros.

24 years agoSince PORTNAME is no longer used at the level of C code (it's only in
Tom Lane [Sun, 19 Mar 2000 21:59:30 +0000 (21:59 +0000)]
Since PORTNAME is no longer used at the level of C code (it's only in
Makefiles now), there's no reason for os2client to maintain its own
copy of c.h just to change #define PORTNAME.  Simplify Makefile
accordingly.  Get rid of horribly-out-of-date modified copy of c.h,
which should never have been in the distribution to start with,
since it's actually a derived file.  Now it's not needed anyway.

24 years agoProper unpack de_DE.ISO-8859-1
Bruce Momjian [Sun, 19 Mar 2000 19:52:44 +0000 (19:52 +0000)]
Proper unpack  de_DE.ISO-8859-1

24 years agoMinor code rearrangement & doc improvement in eval_const_expressions().
Tom Lane [Sun, 19 Mar 2000 18:20:38 +0000 (18:20 +0000)]
Minor code rearrangement & doc improvement in eval_const_expressions().

24 years ago*** empty log message ***
Michael Meskes [Sun, 19 Mar 2000 10:04:47 +0000 (10:04 +0000)]
*** empty log message ***

24 years agotransformExpr() did the Wrong Thing if applied to a SubLink node that
Tom Lane [Sun, 19 Mar 2000 07:13:58 +0000 (07:13 +0000)]
transformExpr() did the Wrong Thing if applied to a SubLink node that
had already been transformed.  This led to failure in examples like
UPDATE table SET fld = (SELECT ...).  Repair this, and revise the
comments to explain that transformExpr has to be robust against this
condition.  Someday we might want to fix the callers so that
transformExpr is never invoked on its own output, but that someday
is not today.

24 years agoit seems in the beta2 release DBUSERID in pg_dumpall is the _name_ of the
Bruce Momjian [Sun, 19 Mar 2000 02:19:43 +0000 (02:19 +0000)]
it seems in the beta2 release DBUSERID in pg_dumpall is the _name_ of the
user, so it doesn't need to be translated from the number to the name.

also ``create database ...'' does not take numbers for the encoding, so
the ENCODING variable does not need to be translated to a number, but left
as the text representation.  a patch is supplied to make the changes i
have found to work.  i was successful dumping and reloading my database
after these changes.
-

John M. Flinchbaugh

24 years agoFix incorrect implementation of log(x) for numeric, as well as
Tom Lane [Sun, 19 Mar 2000 01:12:18 +0000 (01:12 +0000)]
Fix incorrect implementation of log(x) for numeric, as well as
incorrect descriptions of a couple of log-related functions.
I will not force an initdb for this, but log() on a numeric won't
work until you do one...

24 years agoAnother go-round with resolution of ambiguous functions and operators.
Tom Lane [Sun, 19 Mar 2000 00:19:39 +0000 (00:19 +0000)]
Another go-round with resolution of ambiguous functions and operators.
In function parsing, try for an actual function of the given name and
input types before trying to interpret the function call as a type
coercion request, rather than after.  Before, a function that had the
same name as a type and operated on a binary-compatible type wouldn't
get invoked.  Also, cross-pollinate between func_select_candidates and
oper_select_candidates to ensure that they use as nearly the same
resolution rules as possible.  A few other minor code cleanups too.

24 years agoIn can_coerce_type, verify that a possible type-coercion function
Tom Lane [Sun, 19 Mar 2000 00:15:39 +0000 (00:15 +0000)]
In can_coerce_type, verify that a possible type-coercion function
actually returns the type it is named for.

24 years agoFixed psql -c "\slashcmd"
Peter Eisentraut [Sat, 18 Mar 2000 22:48:29 +0000 (22:48 +0000)]
Fixed psql -c "\slashcmd"

24 years agoUpdate bytea type description
Bruce Momjian [Sat, 18 Mar 2000 20:50:10 +0000 (20:50 +0000)]
Update bytea type description

24 years agoImprove error message wording in unary_op_error() --- suggest that
Tom Lane [Sat, 18 Mar 2000 19:53:54 +0000 (19:53 +0000)]
Improve error message wording in unary_op_error() --- suggest that
problem could be lack of parentheses.  This addresses cases like
X UserOp UserOp Y, which will be parsed as (X UserOp) UserOp Y,
whereas what likely was wanted was X UserOp (UserOp Y).

24 years agoClean up minor compiler warnings.
Tom Lane [Sat, 18 Mar 2000 18:57:16 +0000 (18:57 +0000)]
Clean up minor compiler warnings.

24 years agoModify lexing of multi-char operators per pghackers discussion around
Tom Lane [Sat, 18 Mar 2000 18:03:12 +0000 (18:03 +0000)]
Modify lexing of multi-char operators per pghackers discussion around
16-Mar-00: trailing + or - is not part of the operator unless the operator
also contains characters not present in SQL92-defined operators.  This
solves the 'X=-Y' problem without unduly constraining users' choice of
operator names --- in particular, no existing Postgres operator names
become invalid.

Also, remove processing of // comments, as agreed in the same thread.

24 years agoRepair typos: <xb> EOF rule should be <xh>, likewise <xq> to <xd>
Tom Lane [Sat, 18 Mar 2000 05:44:21 +0000 (05:44 +0000)]
Repair typos: <xb> EOF rule should be <xh>, likewise <xq> to <xd>

24 years agoJust noticed that the grammar actually has no provision for '+' as a
Tom Lane [Sat, 18 Mar 2000 04:32:35 +0000 (04:32 +0000)]
Just noticed that the grammar actually has no provision for '+' as a
prefix operator :-(.  Bad enough that we have no implementation of
unary plus, but at least with this fix the grammar will take it.

24 years agoAdd translation of timespan to interval.
Bruce Momjian [Sat, 18 Mar 2000 00:33:45 +0000 (00:33 +0000)]
Add translation of timespan to interval.

24 years agoPatch possible portability problem: a few places had // style comments,
Tom Lane [Fri, 17 Mar 2000 23:26:36 +0000 (23:26 +0000)]
Patch possible portability problem: a few places had // style comments,
which is not ANSI C, even though some compilers will take it...

24 years agoTypo correction (// -> /)
Tom Lane [Fri, 17 Mar 2000 23:17:23 +0000 (23:17 +0000)]
Typo correction (// -> /)

24 years agoAdd safety check on expression nesting depth. Default value is set by
Tom Lane [Fri, 17 Mar 2000 05:29:07 +0000 (05:29 +0000)]
Add safety check on expression nesting depth.  Default value is set by
a config.h #define, and the runtime value can be controlled via SET.

24 years agoFix a bunch of minor portability problems and maybe-bugs revealed by
Tom Lane [Fri, 17 Mar 2000 02:36:41 +0000 (02:36 +0000)]
Fix a bunch of minor portability problems and maybe-bugs revealed by
running gcc and HP's cc with warnings cranked way up.  Signed vs unsigned
comparisons, routines declared static and then defined not-static,
that kind of thing.  Tedious, but perhaps useful...

24 years agoUpdate config.guess and config.sub to latest versions available from
Tom Lane [Thu, 16 Mar 2000 20:41:17 +0000 (20:41 +0000)]
Update config.guess and config.sub to latest versions available from
the GNU Autoconf CVS server.

24 years agoTweak horology regress test to ensure platform-independent ordering of
Tom Lane [Thu, 16 Mar 2000 17:03:41 +0000 (17:03 +0000)]
Tweak horology regress test to ensure platform-independent ordering of
results in conversions tests.  Update horology-no-DST-before-1970.out.

24 years agoSince it has been confirmed working on the bugs list, please apply the
Bruce Momjian [Thu, 16 Mar 2000 15:35:34 +0000 (15:35 +0000)]
Since it has been confirmed working on the bugs list, please apply the
attached patch.

Andreas Kardos

24 years agolibpq++ Makefile uses -DDEBUG to turn on debugging trace to
Bruce Momjian [Thu, 16 Mar 2000 15:34:36 +0000 (15:34 +0000)]
libpq++ Makefile uses -DDEBUG to turn on debugging trace to
/tmp/trace.out.
However, elog.h uses DEBUG as a log-level flag.  As a result, tracing is
turned on even if the libpq++.so is built with DEBUG commented out in
the Makefile.

This patch changes libpq++ to use DEBUGFILE instead (which is not
defined anywhere else).

Oliver Elphick

24 years agoFix typo (introduced by xemacs markup normalization) in the syntax summary.
Thomas G. Lockhart [Thu, 16 Mar 2000 14:39:30 +0000 (14:39 +0000)]
Fix typo (introduced by xemacs markup normalization) in the syntax summary.

24 years agoSupport full POSIX-style time zone: EST+3, PST-3, etc.
Thomas G. Lockhart [Thu, 16 Mar 2000 14:38:42 +0000 (14:38 +0000)]
Support full POSIX-style time zone: EST+3, PST-3, etc.
We probably support a superset of the spec, but I don't have the spec
 to confirm this.
Update regression tests to include tests for this format.

24 years agoSupport full POSIX-style time zone: EST+3, PST-3, etc.
Thomas G. Lockhart [Thu, 16 Mar 2000 14:38:31 +0000 (14:38 +0000)]
Support full POSIX-style time zone: EST+3, PST-3, etc.
We probably support a superset of the spec, but I don't have the spec
 to confirm this.
Update regression tests to include tests for this format.
Update geometry.out with results from Linux RH 5.2 system
(for last decimal place).

24 years agoSupport full POSIX-style time zone: EST+3, PST-3, etc.
Thomas G. Lockhart [Thu, 16 Mar 2000 14:37:33 +0000 (14:37 +0000)]
Support full POSIX-style time zone: EST+3, PST-3, etc.
We probably support a superset of the spec, but I don't have the spec
 to confirm this.
Update regression tests to include tests for this format.
Fix single-space typo in printed message in regress.sh.

24 years agoFixes for char_length() change.
Tatsuo Ishii [Thu, 16 Mar 2000 11:55:43 +0000 (11:55 +0000)]
Fixes for char_length() change.

24 years agoFixes for char_length() changes.
Tatsuo Ishii [Thu, 16 Mar 2000 11:53:16 +0000 (11:53 +0000)]
Fixes for char_length() changes.

24 years agoTurns out that Mazurkiewicz's gripe about 'function inheritance' is
Tom Lane [Thu, 16 Mar 2000 06:35:07 +0000 (06:35 +0000)]
Turns out that Mazurkiewicz's gripe about 'function inheritance' is
actually a type-coercion problem.  If you have a function defined on
class A, and class B inherits from A, then the function ought to work
on class B as well --- but coerce_type didn't know that.  Now it does.

24 years agoFix for char_length() changes
Tatsuo Ishii [Thu, 16 Mar 2000 05:07:21 +0000 (05:07 +0000)]
Fix for char_length() changes

24 years agoFix some (more) problems with subselects in rules. Rewriter failed to
Tom Lane [Thu, 16 Mar 2000 03:23:18 +0000 (03:23 +0000)]
Fix some (more) problems with subselects in rules.  Rewriter failed to
mark query as having subselects if a subselect was added from a rule
WHERE condition (as opposed to a rule action).  Also, fix adjustment
of varlevelsup so that it actually has some prospect of working when
inserting an expression containing a subselect into a subquery.

24 years ago Hi,
Bruce Momjian [Thu, 16 Mar 2000 01:35:41 +0000 (01:35 +0000)]
 Hi,

 small changes in formatting.c code (better memory usage ...etc.) and
better
to_char's cache (will fastly for more to_char()s in one query).

(It is probably end of to_char() development in 7.0 cycle.)

                                                Karel

24 years agoUpdate comment obsoleted by Thomas's latest fixes.
Tom Lane [Wed, 15 Mar 2000 23:42:14 +0000 (23:42 +0000)]
Update comment obsoleted by Thomas's latest fixes.

24 years agoTweak GROUP BY so that it will still accept result-column names, but only
Tom Lane [Wed, 15 Mar 2000 23:31:19 +0000 (23:31 +0000)]
Tweak GROUP BY so that it will still accept result-column names, but only
after trying to resolve the item as an input-column name.  This allows us
to be compliant with the SQL92 spec for queries that fall within the spec,
while still accepting the same out-of-spec queries as 6.5 did.  You'll only
lose if there is an output column name that is the same as an input
column name, but doesn't refer to the same value.  7.0 will interpret
such a GROUP BY spec differently than 6.5 did.  No way around that, because
6.5 was clearly not spec compliant.

24 years ago*** empty log message ***
Michael Meskes [Wed, 15 Mar 2000 19:09:10 +0000 (19:09 +0000)]
*** empty log message ***

24 years agoFix busted TRANSLATE() code --- it coredumped due to pfree()'ing the
Tom Lane [Wed, 15 Mar 2000 17:24:18 +0000 (17:24 +0000)]
Fix busted TRANSLATE() code --- it coredumped due to pfree()'ing the
wrong pointer.

24 years agoFix " to ' on lobj docs.
Bruce Momjian [Wed, 15 Mar 2000 17:07:41 +0000 (17:07 +0000)]
Fix " to ' on lobj docs.

24 years agoRemove another incorrect UserAbortTransactionBlock() call.
Tom Lane [Wed, 15 Mar 2000 07:02:56 +0000 (07:02 +0000)]
Remove another incorrect UserAbortTransactionBlock() call.

24 years agoRemove gratuitous and incorrect begin/commit transaction calls in
Tom Lane [Wed, 15 Mar 2000 06:50:51 +0000 (06:50 +0000)]
Remove gratuitous and incorrect begin/commit transaction calls in
CREATE DB/DROP DB.  If you didn't think they were wrong, try what
happens when you compile with -DCLOBBER_FREED_MEMORY --- database
name displayed in error messages is trashed, because transaction
abort freed it.  Also, remove trailing periods in error messages,
per our prevailing style.

24 years agoFix a couple of missed changes in expected results.
Tom Lane [Wed, 15 Mar 2000 06:24:58 +0000 (06:24 +0000)]
Fix a couple of missed changes in expected results.

24 years agoAdd 'datetime' -> 'timestamp' conversion to xlateSqlFunc() to ease
Tom Lane [Wed, 15 Mar 2000 05:31:55 +0000 (05:31 +0000)]
Add 'datetime' -> 'timestamp' conversion to xlateSqlFunc() to ease
the pain of updating apps to 7.0.  Should we also translate some of
the 'datetime_foo' functions that exist in 6.* ?

24 years agoRepair unintentional damage to MULTIBYTE code.
Tom Lane [Tue, 14 Mar 2000 23:59:23 +0000 (23:59 +0000)]
Repair unintentional damage to MULTIBYTE code.

24 years agoCache fmgr lookup data for index's getnext() function in IndexScanDesc,
Tom Lane [Tue, 14 Mar 2000 23:52:01 +0000 (23:52 +0000)]
Cache fmgr lookup data for index's getnext() function in IndexScanDesc,
so that the fmgr lookup only has to happen once per index scan and not
once per tuple.  Seems to save 5% or so of CPU time for an indexscan.

24 years agoImplement column aliases on views "CREATE VIEW name (collist)".
Thomas G. Lockhart [Tue, 14 Mar 2000 23:06:59 +0000 (23:06 +0000)]
Implement column aliases on views "CREATE VIEW name (collist)".
Implement TIME WITH TIME ZONE type (timetz internal type).
Remap length() for character strings to CHAR_LENGTH() for SQL92
 and to remove the ambiguity with geometric length() functions.
Keep length() for character strings for backward compatibility.
Shrink stored views by removing internal column name list from visible rte.
Implement min(), max() for time and timetz data types.
Implement conversion of TIME to INTERVAL.
Implement abs(), mod(), fac() for the int8 data type.
Rename some math functions to generic names:
 round(), sqrt(), cbrt(), pow(), etc.
Rename NUMERIC power() function to pow().
Fix int2 factorial to calculate result in int4.
Enhance the Oracle compatibility function translate() to work with string
 arguments (from Edwin Ramirez).
Modify pg_proc system table to remove OID holes.

24 years agoWrite up new TIME WITH TIME ZONE type.
Thomas G. Lockhart [Tue, 14 Mar 2000 22:52:53 +0000 (22:52 +0000)]
Write up new TIME WITH TIME ZONE type.

24 years agoMarginal performance improvement in LockBuffer --- calculate address
Tom Lane [Tue, 14 Mar 2000 22:46:27 +0000 (22:46 +0000)]
Marginal performance improvement in LockBuffer --- calculate address
of BufferLocks[] entry just once.  Seems to save 10% or so of the
routine's runtime, which'd not be worth worrying about if it weren't
such a hotspot.