]> granicus.if.org Git - postgresql/log
postgresql
17 years agoTranslation updates
Peter Eisentraut [Thu, 15 Nov 2007 20:38:15 +0000 (20:38 +0000)]
Translation updates

17 years agoPrevent re-use of a deleted relation's relfilenode until after the next
Tom Lane [Thu, 15 Nov 2007 20:36:40 +0000 (20:36 +0000)]
Prevent re-use of a deleted relation's relfilenode until after the next
checkpoint.  This guards against an unlikely data-loss scenario in which
we re-use the relfilenode, then crash, then replay the deletion and
recreation of the file.  Even then we'd be OK if all insertions into the
new relation had been WAL-logged ... but that's not guaranteed given all
the no-WAL-logging optimizations that have recently been added.

Patch by Heikki Linnakangas, per a discussion last month.

17 years agoUpdate config.guess and config.sub
Peter Eisentraut [Thu, 15 Nov 2007 20:21:05 +0000 (20:21 +0000)]
Update config.guess and config.sub

17 years agoAdd more detail for non-TEXT cast release item.
Bruce Momjian [Thu, 15 Nov 2007 20:10:32 +0000 (20:10 +0000)]
Add more detail for non-TEXT cast release item.

17 years agoMessage improvements
Peter Eisentraut [Thu, 15 Nov 2007 20:04:38 +0000 (20:04 +0000)]
Message improvements

17 years agoAdd to synopsis that timeout option also applies to stop mode
Peter Eisentraut [Thu, 15 Nov 2007 19:40:31 +0000 (19:40 +0000)]
Add to synopsis that timeout option also applies to stop mode

17 years agoMake help synopsis consistent
Peter Eisentraut [Thu, 15 Nov 2007 19:35:26 +0000 (19:35 +0000)]
Make help synopsis consistent

17 years agoMake synopses of -t option consistent
Peter Eisentraut [Thu, 15 Nov 2007 18:37:23 +0000 (18:37 +0000)]
Make synopses of -t option consistent

17 years agoDocument that /contrib/tsearch2 now contains a compatibility interface.
Bruce Momjian [Thu, 15 Nov 2007 17:09:40 +0000 (17:09 +0000)]
Document that /contrib/tsearch2 now contains a compatibility interface.

17 years agoAdjust example to reduce confusion between a tsvector column and
Tom Lane [Wed, 14 Nov 2007 23:48:55 +0000 (23:48 +0000)]
Adjust example to reduce confusion between a tsvector column and
an index, per Simon.

17 years agoAdd a rank/(rank+1) normalization option to ts_rank(). While the usefulness
Tom Lane [Wed, 14 Nov 2007 23:43:27 +0000 (23:43 +0000)]
Add a rank/(rank+1) normalization option to ts_rank().  While the usefulness
of this seems a bit marginal, if it's useful enough to be shown in the manual
then we probably ought to support doing it without double evaluation of the
ts_rank function.  Per my proposal earlier today.

17 years agoFix incorrect calculation of elapsed_xlogs. Itagaki Takahiro
Tom Lane [Wed, 14 Nov 2007 21:19:18 +0000 (21:19 +0000)]
Fix incorrect calculation of elapsed_xlogs.  Itagaki Takahiro

17 years agoAdd an Accept parameter to "simple" dictionaries. The default of true
Tom Lane [Wed, 14 Nov 2007 18:36:37 +0000 (18:36 +0000)]
Add an Accept parameter to "simple" dictionaries.  The default of true
gives the old behavior; selecting false allows the dictionary to be used
as a filter ahead of other dictionaries, because it will pass on rather
than accept words that aren't in its stopword list.
Jan Urbanski

17 years agoMention that GSSAPI is preferred over Kerberos authentication because
Bruce Momjian [Wed, 14 Nov 2007 14:38:30 +0000 (14:38 +0000)]
Mention that GSSAPI is preferred over Kerberos authentication because
GSSAPI is industry-standard.

17 years agoAdd note about deprecating krb5 authentication in favour of GSSAPI,
Magnus Hagander [Wed, 14 Nov 2007 14:25:55 +0000 (14:25 +0000)]
Add note about deprecating krb5 authentication in favour of GSSAPI,
per discussions (a long time ago). Documentation only, we keep full
support in the code.

17 years agoCapitalize "Migration to Version".
Bruce Momjian [Wed, 14 Nov 2007 14:06:54 +0000 (14:06 +0000)]
Capitalize "Migration to Version".

17 years agoProperly capitalize "Migrate To Version"; seems we always had that
Bruce Momjian [Wed, 14 Nov 2007 12:36:34 +0000 (12:36 +0000)]
Properly capitalize "Migrate To Version";  seems we always had that
wrong.

17 years agoMarkup cleanup.
Bruce Momjian [Wed, 14 Nov 2007 03:39:53 +0000 (03:39 +0000)]
Markup cleanup.

17 years agoReference pgfoundry instead of gborg.
Bruce Momjian [Wed, 14 Nov 2007 03:39:17 +0000 (03:39 +0000)]
Reference pgfoundry instead of gborg.

17 years agoRemove reference to gborg, instead reference pgfoundry.
Bruce Momjian [Wed, 14 Nov 2007 03:37:30 +0000 (03:37 +0000)]
Remove reference to gborg, instead reference pgfoundry.

17 years agoUpdate discussion of tsearch2 migration. I'm not entirely sure about
Tom Lane [Wed, 14 Nov 2007 03:26:24 +0000 (03:26 +0000)]
Update discussion of tsearch2 migration.  I'm not entirely sure about
the division of material between here and the tsearch2 contrib page,
but at least it's not obviously unfinished any more.

17 years agoMove contrib docs to an appendix (next door to External Projects).
Tom Lane [Wed, 14 Nov 2007 02:36:43 +0000 (02:36 +0000)]
Move contrib docs to an appendix (next door to External Projects).
Change title to Additional Supplied Modules.  Improve discussion of
how to build and install 'em.

17 years agoClean up some now-obsolete references to GBorg.
Tom Lane [Wed, 14 Nov 2007 01:58:18 +0000 (01:58 +0000)]
Clean up some now-obsolete references to GBorg.

17 years agoAdd some quick documentation for tsearch2 compatibility module.
Tom Lane [Wed, 14 Nov 2007 01:09:50 +0000 (01:09 +0000)]
Add some quick documentation for tsearch2 compatibility module.

17 years agoUpdate release notes to current CVS.
Bruce Momjian [Wed, 14 Nov 2007 00:16:27 +0000 (00:16 +0000)]
Update release notes to current CVS.

17 years agoClean up some stray references to tsearch2.
Tom Lane [Tue, 13 Nov 2007 23:36:26 +0000 (23:36 +0000)]
Clean up some stray references to tsearch2.

17 years agocontrib/tsearch2 need not be special-cased any longer.
Tom Lane [Tue, 13 Nov 2007 22:49:47 +0000 (22:49 +0000)]
contrib/tsearch2 need not be special-cased any longer.

17 years agoResurrect the code for the rewrite(ARRAY[...]) aggregate function,
Tom Lane [Tue, 13 Nov 2007 22:14:50 +0000 (22:14 +0000)]
Resurrect the code for the rewrite(ARRAY[...]) aggregate function,
and put it into contrib/tsearch2 compatibility module.

17 years agoOoops, missed one file to remove.
Tom Lane [Tue, 13 Nov 2007 21:25:25 +0000 (21:25 +0000)]
Ooops, missed one file to remove.

17 years agoUpdate description of contrib/tsearch2.
Tom Lane [Tue, 13 Nov 2007 21:10:40 +0000 (21:10 +0000)]
Update description of contrib/tsearch2.

17 years agoRe-add contrib/tsearch2 to the set of buildable contrib modules.
Tom Lane [Tue, 13 Nov 2007 21:04:45 +0000 (21:04 +0000)]
Re-add contrib/tsearch2 to the set of buildable contrib modules.

17 years agoReplace the now-incompatible-with-core contrib/tsearch2 module with a
Tom Lane [Tue, 13 Nov 2007 21:02:29 +0000 (21:02 +0000)]
Replace the now-incompatible-with-core contrib/tsearch2 module with a
compatibility package.  This supports importing dumps from past versions
using tsearch2, and provides the old names and API for most functions
that were changed.  (rewrite(ARRAY[...]) is a glaring omission, though.)

Pavel Stehule and Tom Lane

17 years agoAgain properly fix Darwin strip.
Bruce Momjian [Tue, 13 Nov 2007 18:50:54 +0000 (18:50 +0000)]
Again properly fix Darwin strip.

17 years agoAdd link to Postgres-R glossary from our documentation (has helpful
Bruce Momjian [Tue, 13 Nov 2007 18:44:20 +0000 (18:44 +0000)]
Add link to Postgres-R glossary from our documentation (has helpful
terms defined).

17 years agoModify OS/X Darin test to actually work, (no $template)
Bruce Momjian [Tue, 13 Nov 2007 18:15:01 +0000 (18:15 +0000)]
Modify OS/X Darin test to actually work, (no $template)

17 years agoFix a few contrib regression test scripts that hadn't gotten the word
Tom Lane [Tue, 13 Nov 2007 06:29:04 +0000 (06:29 +0000)]
Fix a few contrib regression test scripts that hadn't gotten the word
about best practice for including the module creation scripts: to wit
that you should suppress NOTICE messages.  This avoids creating
regression failures by adding or removing comment lines in the module
scripts.

17 years agoAdjust expected file for new CVS version label.
Bruce Momjian [Tue, 13 Nov 2007 04:27:26 +0000 (04:27 +0000)]
Adjust expected file for new CVS version label.

17 years agoAdd CVS version labels to all install/uninstall scripts.
Bruce Momjian [Tue, 13 Nov 2007 04:24:29 +0000 (04:24 +0000)]
Add CVS version labels to all install/uninstall scripts.

17 years agoFix more script breakage.
Tom Lane [Tue, 13 Nov 2007 02:17:33 +0000 (02:17 +0000)]
Fix more script breakage.

17 years agoFix insufficient search-path paranoia in SQL function definitions.
Tom Lane [Tue, 13 Nov 2007 01:55:49 +0000 (01:55 +0000)]
Fix insufficient search-path paranoia in SQL function definitions.
Remove setting of search_path in install/uninstall scripts, since unlike
other contrib modules this one does not want to let you change the
installation schema.

17 years agoAdjust script to be consistent (thanks Tom for the fix).
Bruce Momjian [Tue, 13 Nov 2007 01:19:40 +0000 (01:19 +0000)]
Adjust script to be consistent (thanks Tom for the fix).

17 years agoUndo damage from yesterday's script "cleanup".
Tom Lane [Tue, 13 Nov 2007 00:20:36 +0000 (00:20 +0000)]
Undo damage from yesterday's script "cleanup".

17 years agoI find that an out-of-the-box installation of OSSP uuid 1.6.0 installs
Tom Lane [Tue, 13 Nov 2007 00:13:19 +0000 (00:13 +0000)]
I find that an out-of-the-box installation of OSSP uuid 1.6.0 installs
itself as libuuid, not libossp-uuid which was the only case expected by
our build support.  Install a configure test to determine which name
to use (and to check that the library is present at all).

17 years agoNo master server overhead for middleware replication.
Bruce Momjian [Mon, 12 Nov 2007 18:01:44 +0000 (18:01 +0000)]
No master server overhead for middleware replication.

17 years agoChange replication chart title to "No waiting for multiple servers".
Bruce Momjian [Mon, 12 Nov 2007 17:02:14 +0000 (17:02 +0000)]
Change replication chart title to "No waiting for multiple servers".

17 years agoTrim trailing whitespace --- needed commit to update anoncvs.
Bruce Momjian [Mon, 12 Nov 2007 01:37:34 +0000 (01:37 +0000)]
Trim trailing whitespace --- needed commit to update anoncvs.

17 years agoEnsure that typmod decoration on a datatype name is validated in all cases,
Tom Lane [Sun, 11 Nov 2007 19:22:49 +0000 (19:22 +0000)]
Ensure that typmod decoration on a datatype name is validated in all cases,
even in code paths where we don't pay any subsequent attention to the typmod
value.  This seems needed in view of the fact that 8.3's generalized typmod
support will accept a lot of bogus syntax, such as "timestamp(foo)" or
"record(int, 42)" --- if we allow such things to pass without comment,
users will get confused.  Per a recent example from Greg Stark.

To implement this in a way that's not very vulnerable to future
bugs-of-omission, refactor the API of parse_type.c's TypeName lookup routines
so that typmod validation is folded into the base lookup operation.  Callers
can still choose not to receive the encoded typmod, but we'll check the
decoration anyway if it's present.

17 years agoRemove tabs from SGML files.
Bruce Momjian [Sun, 11 Nov 2007 14:23:18 +0000 (14:23 +0000)]
Remove tabs from SGML files.

17 years agoRemove /contrib/dblink/doc directory, now in SGML.
Bruce Momjian [Sun, 11 Nov 2007 05:13:09 +0000 (05:13 +0000)]
Remove /contrib/dblink/doc directory, now in SGML.

17 years agoAdjust pg_trgm expected output for new *.sql file.
Bruce Momjian [Sun, 11 Nov 2007 03:57:06 +0000 (03:57 +0000)]
Adjust pg_trgm expected output for new *.sql file.

17 years agoMake /contrib install/uninstall script consistent:
Bruce Momjian [Sun, 11 Nov 2007 03:25:35 +0000 (03:25 +0000)]
Make /contrib install/uninstall script consistent:

remove transactions
use create or replace function
make formatting consistent
set search patch on first line

Add documentation on modifying *.sql to set the search patch, and
mention that major upgrades should still run the installation scripts.

Some of these issues were spotted by Tom today.

17 years agoRemove references to READMEs from /contrib Makefiles.
Bruce Momjian [Sat, 10 Nov 2007 23:59:52 +0000 (23:59 +0000)]
Remove references to READMEs from /contrib Makefiles.

17 years agoMove most /contrib README files into SGML. Some still need conversion
Bruce Momjian [Sat, 10 Nov 2007 23:30:46 +0000 (23:30 +0000)]
Move most /contrib README files into SGML.  Some still need conversion
or will never be converted.

17 years agoAdd pg_ctl -t/timeout parameter to control amount of time to wait for
Bruce Momjian [Sat, 10 Nov 2007 21:48:51 +0000 (21:48 +0000)]
Add pg_ctl -t/timeout parameter to control amount of time to wait for
start/shutdown.

17 years agoAdd an example of a SQL function with output parameters returning
Tom Lane [Sat, 10 Nov 2007 20:14:36 +0000 (20:14 +0000)]
Add an example of a SQL function with output parameters returning
multiple rows.  I had thought this case was covered, but there was
no example in the obvious section to look in.

17 years agoAdd missing closing / in xsd:restriction, and remove some unnecessary
Tom Lane [Sat, 10 Nov 2007 19:29:54 +0000 (19:29 +0000)]
Add missing closing / in xsd:restriction, and remove some unnecessary
spaces for consistency.  Per bug #3734 from Ben Leslie; fix by
Euler Taveira de Oliveira.

17 years agoHigh availability wording fix.
Bruce Momjian [Sat, 10 Nov 2007 19:19:36 +0000 (19:19 +0000)]
High availability wording fix.

17 years agoUpdate high availability documentation with comments from Markus Schiltknecht.
Bruce Momjian [Sat, 10 Nov 2007 19:14:02 +0000 (19:14 +0000)]
Update high availability documentation with comments from Markus Schiltknecht.

17 years agoxmlGetUTF8Char()'s second argument is both input and output. Fix
Tom Lane [Sat, 10 Nov 2007 18:51:20 +0000 (18:51 +0000)]
xmlGetUTF8Char()'s second argument is both input and output.  Fix
uninitialized value, and avoid invoking the function nine separate
times in the pg_xmlIsNameChar macro.  Should resolve buildfarm failures.
Per report from Ben Leslie.

17 years agoUse strip -x on OS/X-darwin because non-"-x" causes link problems:
Bruce Momjian [Sat, 10 Nov 2007 16:15:23 +0000 (16:15 +0000)]
Use strip -x on OS/X-darwin because non-"-x" causes link problems:

  http://archives.postgresql.org/pgsql-hackers/2007-10/msg01470.php

17 years agoHave text search thesaurus files use "?" for stop words.
Bruce Momjian [Sat, 10 Nov 2007 15:39:34 +0000 (15:39 +0000)]
Have text search thesaurus files use "?" for stop words.

Throw an error for actual stop words, rather than a warning.  This fixes
problems with cache reloading causing warning messages.

Re-enable stop words in regression tests;  was disabled by Tom.

Document "?" as API change.

17 years agoReduce error level of ROLLBACK outside a transaction from WARNING to
Bruce Momjian [Sat, 10 Nov 2007 14:36:44 +0000 (14:36 +0000)]
Reduce error level of ROLLBACK outside a transaction from WARNING to
NOTICE.

17 years agoFix markup problem with recent pg_ctl change.
Bruce Momjian [Sat, 10 Nov 2007 14:07:18 +0000 (14:07 +0000)]
Fix markup problem with recent pg_ctl change.

17 years agoHave crosstab variants treat NULL rowid as a category in its own right,
Joe Conway [Sat, 10 Nov 2007 05:00:41 +0000 (05:00 +0000)]
Have crosstab variants treat NULL rowid as a category in its own right,
per suggestion from Tom Lane. This fixes crash-bug reported by Stefan
Schwarzer.

17 years agoProperly order pg_ctl -c option alphabetically in docs.
Bruce Momjian [Sat, 10 Nov 2007 04:52:17 +0000 (04:52 +0000)]
Properly order pg_ctl -c option alphabetically in docs.

17 years agoRecognize RETURN QUERY via a textual test, so that QUERY doesn't need to be
Tom Lane [Fri, 9 Nov 2007 23:58:32 +0000 (23:58 +0000)]
Recognize RETURN QUERY via a textual test, so that QUERY doesn't need to be
a plpgsql keyword.  This avoids springing a new reserved word on plpgsql
programmers.
For consistency, handle RETURN NEXT the same way.

17 years agoRemove tabs from SGML files.
Bruce Momjian [Fri, 9 Nov 2007 23:52:08 +0000 (23:52 +0000)]
Remove tabs from SGML files.

17 years agoAdd an expected-results file to get regression to pass cleanly in sv_SE
Tom Lane [Fri, 9 Nov 2007 22:52:08 +0000 (22:52 +0000)]
Add an expected-results file to get regression to pass cleanly in sv_SE
locale, which sorts a bit differently.  Per recent gripe from Magnus.

17 years agoClean up ts_locale.h/.c. Fix broken and not-consistent-across-platforms
Tom Lane [Fri, 9 Nov 2007 22:37:35 +0000 (22:37 +0000)]
Clean up ts_locale.h/.c.  Fix broken and not-consistent-across-platforms
behavior of wchar2char/char2wchar; this should resolve bug #3730.  Avoid
excess computations of pg_mblen in t_isalpha and friends.  Const-ify
APIs where possible.

17 years agoPoint to our own UUID page instead of to Wikipedia.
Alvaro Herrera [Fri, 9 Nov 2007 22:22:46 +0000 (22:22 +0000)]
Point to our own UUID page instead of to Wikipedia.

17 years agoSecond pass at improving LIKE/regex estimation in non-C locales. It turns
Tom Lane [Fri, 9 Nov 2007 20:10:02 +0000 (20:10 +0000)]
Second pass at improving LIKE/regex estimation in non-C locales.  It turns
out that it's actually quite likely that a string that is an extension of
the given prefix will sort as larger than the "greater" string our previous
code created.  To provide some defense against that, do the comparisons
against a modified string instead of just the bare prefix.  We tack on
"Z", "z", "y", or "9", whichever is seen as largest in the current locale.
Testing suggests that this is sufficient at least for cases involving
ASCII data.

17 years agoAdd parameter krb_realm used by GSSAPI, SSPI and Kerberos
Magnus Hagander [Fri, 9 Nov 2007 17:31:07 +0000 (17:31 +0000)]
Add parameter krb_realm used by GSSAPI, SSPI and Kerberos
to validate the realm of the connecting user. By default
it's empty meaning no verification, which is the way
Kerberos authentication has traditionally worked in
PostgreSQL.

17 years agoMove pl/proxy into Multi-server parallel query execution documentation
Bruce Momjian [Fri, 9 Nov 2007 16:36:04 +0000 (16:36 +0000)]
Move pl/proxy into Multi-server parallel query execution documentation
section.

17 years agoAllow XML processing instructions starting with "xml" while prohibiting
Peter Eisentraut [Fri, 9 Nov 2007 15:52:51 +0000 (15:52 +0000)]
Allow XML processing instructions starting with "xml" while prohibiting
those being exactly "xml".  Bug #3735 from Ben Leslie

17 years agoIn tsearch code, remove !(A && B) via restructuring, for clarity
Bruce Momjian [Fri, 9 Nov 2007 01:32:22 +0000 (01:32 +0000)]
In tsearch code, remove !(A && B) via restructuring, for clarity

17 years agoIf an index depends on no columns of its table, give it a dependency on the
Tom Lane [Thu, 8 Nov 2007 23:22:54 +0000 (23:22 +0000)]
If an index depends on no columns of its table, give it a dependency on the
whole table instead, to ensure that it goes away when the table is dropped.
Per bug #3723 from Sam Mason.

Backpatch as far as 7.4; AFAICT 7.3 does not have the issue, because it doesn't
have general-purpose expression indexes and so there must be at least one
column referenced by an index.

17 years agoWording improvement for pl/proxy mention.
Bruce Momjian [Thu, 8 Nov 2007 22:14:18 +0000 (22:14 +0000)]
Wording improvement for pl/proxy mention.

17 years agoAdd mention of pl/proxy toolset to docs.
Bruce Momjian [Thu, 8 Nov 2007 22:08:18 +0000 (22:08 +0000)]
Add mention of pl/proxy toolset to docs.

17 years agoFix EquivalenceClass code to handle volatile sort expressions in a more
Tom Lane [Thu, 8 Nov 2007 21:49:48 +0000 (21:49 +0000)]
Fix EquivalenceClass code to handle volatile sort expressions in a more
predictable manner; in particular that if you say ORDER BY output-column-ref,
it will in fact sort by that specific column even if there are multiple
syntactic matches.  An example is
SELECT random() AS a, random() AS b FROM ... ORDER BY b, a;
While the use-case for this might be a bit debatable, it worked as expected
in earlier releases, so we should preserve the behavior for 8.3.  Per my
recent proposal.

While at it, fix convert_subquery_pathkeys() to handle RelabelType stripping
in both directions; it needs this for the same reasons make_sort_from_pathkeys
does.

17 years agoLast week's patch for make_sort_from_pathkeys wasn't good enough: it has
Tom Lane [Thu, 8 Nov 2007 19:25:37 +0000 (19:25 +0000)]
Last week's patch for make_sort_from_pathkeys wasn't good enough: it has
to be able to discard top-level RelabelType nodes on *both* sides of the
equivalence-class-to-target-list comparison, since make_pathkey_from_sortinfo
might either add or remove a RelabelType.  Also fix the latter to do the
removal case cleanly.  Per example from Peter.

17 years agoProperly indent SGML file.
Bruce Momjian [Thu, 8 Nov 2007 19:18:23 +0000 (19:18 +0000)]
Properly indent SGML file.

17 years agoAdd "High Availability, Load Balancing, and Replication Feature Matrix"
Bruce Momjian [Thu, 8 Nov 2007 19:16:30 +0000 (19:16 +0000)]
Add "High Availability, Load Balancing, and Replication Feature Matrix"
table to docs.

17 years agoMove NFS section into "Creating Cluster" documentation section because
Bruce Momjian [Thu, 8 Nov 2007 15:21:03 +0000 (15:21 +0000)]
Move NFS section into "Creating Cluster" documentation section because
it seems more logical there.

17 years agoAfter conferencing again with Bruce, put in more accurate XML error message.
Peter Eisentraut [Thu, 8 Nov 2007 15:16:45 +0000 (15:16 +0000)]
After conferencing again with Bruce, put in more accurate XML error message.

17 years agoAdjust test results for message changes. Darn.
Peter Eisentraut [Thu, 8 Nov 2007 14:56:37 +0000 (14:56 +0000)]
Adjust test results for message changes. Darn.

17 years agoWording changes
Peter Eisentraut [Thu, 8 Nov 2007 14:47:51 +0000 (14:47 +0000)]
Wording changes

17 years agoAdd sentence-ending periods.
Peter Eisentraut [Thu, 8 Nov 2007 14:47:41 +0000 (14:47 +0000)]
Add sentence-ending periods.

17 years agoMention text search instead of tsearch2
Peter Eisentraut [Thu, 8 Nov 2007 13:13:51 +0000 (13:13 +0000)]
Mention text search instead of tsearch2

17 years agoSmall fixes
Peter Eisentraut [Thu, 8 Nov 2007 13:13:28 +0000 (13:13 +0000)]
Small fixes

17 years agoImprove error message
Peter Eisentraut [Thu, 8 Nov 2007 13:12:56 +0000 (13:12 +0000)]
Improve error message

17 years agoAdd missing newlines to error messages, and normalize wording a bit.
Peter Eisentraut [Thu, 8 Nov 2007 10:37:54 +0000 (10:37 +0000)]
Add missing newlines to error messages, and normalize wording a bit.

17 years agoImprove the performance of LIKE/regex estimation in non-C locales, by making
Tom Lane [Wed, 7 Nov 2007 22:37:24 +0000 (22:37 +0000)]
Improve the performance of LIKE/regex estimation in non-C locales, by making
make_greater_string() try harder to generate a string that's actually greater
than its input string.  Before we just assumed that making a string that was
memcmp-greater was enough, but it is easy to generate examples where this is
not so when the locale is not C.  Instead, loop until the relevant comparison
function agrees that the generated string is greater than the input.

Unfortunately this is probably not enough to guarantee that the generated
string is greater than all extensions of the input, so we cannot relax the
restriction to C locale for the LIKE/regex index optimization.  But it should
at least improve the odds of getting a useful selectivity estimate in
prefix_selectivity().  Per example from Guillaume Smet.

Backpatch to 8.1, mainly because that's what the complainant is using...

17 years agoFix patternsel() and callers to do the right thing for NOT LIKE and the other
Tom Lane [Wed, 7 Nov 2007 21:00:37 +0000 (21:00 +0000)]
Fix patternsel() and callers to do the right thing for NOT LIKE and the other
negated-match operators.  patternsel had been using the supplied operator as
though it were a positive-match operator, and thus obtaining a wrong result,
which was even more wrong after the caller subtracted it from 1.  Seems
cleanest to give patternsel an explicit "negate" argument so that it knows
what's going on.  Also install the same factorization scheme for pattern
join selectivity estimators; even though they are just stubs at the
moment, this may keep someone from making the same type of mistake when
they get filled out.  Per report from Greg Mullane.

Backpatch to 8.2 --- previous releases do not show the problem because
patternsel() doesn't actually use the operator directly.

17 years ago<optional> cannot be used here, because psql doesn't handle it.
Peter Eisentraut [Wed, 7 Nov 2007 14:07:53 +0000 (14:07 +0000)]
<optional> cannot be used here, because psql doesn't handle it.

17 years agoConsistent capitalization
Peter Eisentraut [Wed, 7 Nov 2007 14:07:21 +0000 (14:07 +0000)]
Consistent capitalization

17 years agoImprove wording
Peter Eisentraut [Wed, 7 Nov 2007 13:23:20 +0000 (13:23 +0000)]
Improve wording

17 years agoSpecify putting two spaces after sentence-ending period, which is current
Peter Eisentraut [Wed, 7 Nov 2007 13:12:21 +0000 (13:12 +0000)]
Specify putting two spaces after sentence-ending period, which is current
practice also.

17 years agoError message cleanup
Peter Eisentraut [Wed, 7 Nov 2007 13:11:05 +0000 (13:11 +0000)]
Error message cleanup

17 years agoUse "alternative" instead of "alternate" where it is clearer.
Peter Eisentraut [Wed, 7 Nov 2007 12:24:24 +0000 (12:24 +0000)]
Use "alternative" instead of "alternate" where it is clearer.

17 years agoFixed two parser bugs.
Michael Meskes [Tue, 6 Nov 2007 08:32:08 +0000 (08:32 +0000)]
Fixed two parser bugs.