OSDN Git Service

pg-rex/syncrep.git
27 years ago1. BTREE_VERSION_1: using bti_itup->t_tid as unique identifier for a given
Vadim B. Mikheev [Wed, 16 Apr 1997 01:48:29 +0000 (01:48 +0000)]
1. BTREE_VERSION_1: using bti_itup->t_tid as unique identifier for a given
index tuple (logical position within A LEVEL). bti_oid & bti_dummy
taken off from BTItemData.
2. Fix for multi-column indices (nbtsearch.c):
   _bt_binsrch() - for searches on internal pages having keysize <
number of attrs we point at the last item < the scankey, not at the
first item = the scankey;
   _bt_moveright() - if keysize < number of attrs we compare scankey with
_last_ item on current page to decide should we move right or
not.

27 years agoBTREE_VERSION_1: using bti_itup->t_tid as unique identifier for a given
Vadim B. Mikheev [Wed, 16 Apr 1997 01:21:59 +0000 (01:21 +0000)]
BTREE_VERSION_1: using bti_itup->t_tid as unique identifier for a given
index tuple (logical position within A LEVEL). bti_oid & bti_dummy
taken off from BTItemData.

27 years agoA couple of cleanups from Scott Harrison <Scott_Harrison@next.com>
Marc G. Fournier [Tue, 15 Apr 1997 19:08:13 +0000 (19:08 +0000)]
A couple of cleanups from Scott Harrison <Scott_Harrison@next.com>

27 years agoOops, don't forget to remove the CASSERT=true line
Marc G. Fournier [Tue, 15 Apr 1997 18:36:45 +0000 (18:36 +0000)]
Oops, don't forget to remove the CASSERT=true line

27 years agoHave CASSERT enabled/disabled via configure, and passed through config.h
Marc G. Fournier [Tue, 15 Apr 1997 18:35:50 +0000 (18:35 +0000)]
Have CASSERT enabled/disabled via configure, and passed through config.h
instead of as a -D

27 years agoPatch for Solaris 2.5 from: adrian@waltham.harvard.net
Marc G. Fournier [Tue, 15 Apr 1997 18:18:45 +0000 (18:18 +0000)]
Patch for Solaris 2.5 from: adrian@waltham.harvard.net

27 years agoVarious minor HP related patches from:
Marc G. Fournier [Tue, 15 Apr 1997 17:55:37 +0000 (17:55 +0000)]
Various minor HP related patches from:

Morten Kjeldgaard <mok@monster.kemi.aau.dk>

27 years agoA small fix, where default: condition in case had not 'break;'...not required,
Marc G. Fournier [Tue, 15 Apr 1997 17:46:52 +0000 (17:46 +0000)]
A small fix, where default: condition in case had not 'break;'...not required,
but, IMHO, cleaner

27 years agoFrom: Thomas Lockhart <Thomas.G.Lockhart@jpl.nasa.gov>
Marc G. Fournier [Tue, 15 Apr 1997 17:41:44 +0000 (17:41 +0000)]
From: Thomas Lockhart <Thomas.G.Lockhart@jpl.nasa.gov>
Subject: [HACKERS] Money integration patches

Here are patches to integrate the money data type. I have included
some math and aggregate functions and have made the locale support optional
by #ifdef USE_LOCALE bracketing of functions.

Modules affected are:
builtins.h.patch
cash.c.patch
cash.h.patch
main.c.patch
pg_aggregate.h.patch
pg_operator.h.patch
pg_proc.h.patch
pg_type.h.patch

I changed the data type to be pass-by-reference rather than by-value
to pave the way for a larger internal representation (64-bit ints?).
Also, I changed the tabbing of cash.c and cash.h to match most of
the other Postgres source code files (4 space indent, 8 spaces == 1 tab).

The locale stuff should be tested under another convention (Russian?)
but I don't know what the correct results should be so perhaps someone
else can give them a try. Will update docs and regression tests in
the next few days.

27 years ago2. The file /usr/local/pgsql/src/backend/lipq/pgcomprim.c has two
Marc G. Fournier [Tue, 15 Apr 1997 17:25:47 +0000 (17:25 +0000)]
2. The file /usr/local/pgsql/src/backend/lipq/pgcomprim.c has two
invalid macro definitions, the compiler complains about:

"pqcomprim.c", line 48.9: 1506-275 (S) Unexpected text ';' ignored.
"pqcomprim.c", line 61.9: 1506-275 (S) Unexpected text ';' ignored.

The ';' terminating the macro definition ntoh_s(n) on line 27 and
ntoh_l(n) on line 28 should be removed.

Pointed out by: Olaf Mittelstaedt <MSTAEDT@va-sigi.va.fh-ulm.de>

27 years agoFix up the -Wno-error problem for non-gcc compilers
Marc G. Fournier [Tue, 15 Apr 1997 17:24:26 +0000 (17:24 +0000)]
Fix up the -Wno-error problem for non-gcc compilers

Pointed out by many many ppl

27 years agoAdd ccsym to tools to determine OS/compiler specific symbols
Marc G. Fournier [Sun, 13 Apr 1997 17:09:45 +0000 (17:09 +0000)]
Add ccsym to tools to determine OS/compiler specific symbols

Change BSD44_derived to __FreeBSD__ in numutils.c (need to know what
NetBSD is referred to as...someone?)

27 years agoFixes for [] escaping pointed out by Adrian
Marc G. Fournier [Sat, 12 Apr 1997 13:56:44 +0000 (13:56 +0000)]
Fixes for [] escaping pointed out by Adrian

27 years agoscripts required by configure to bypass its inabilty to use []'s inside of
Marc G. Fournier [Sat, 12 Apr 1997 13:22:04 +0000 (13:22 +0000)]
scripts required by configure to bypass its inabilty to use []'s inside of
configure ([] have special means to m4 *sigh*)

27 years agoMake sure that Makefile.${PORTNAME} is link'd and that teh port/*.h
Marc G. Fournier [Sat, 12 Apr 1997 10:35:49 +0000 (10:35 +0000)]
Make sure that Makefile.${PORTNAME} is link'd and that teh port/*.h
file that has moved to include is also link'd properly

27 years agoRemove port specific Makefile 'targets' (ie. shared libraries) from
Marc G. Fournier [Sat, 12 Apr 1997 10:33:10 +0000 (10:33 +0000)]
Remove port specific Makefile 'targets' (ie. shared libraries) from
Makefile.global and move them to seperate 'include' makefiles

Over time, should become even more port specific:

ie. Makefile.BSD44_derived should be broken down into netbsd/freebsd
specific ports

27 years agoMore more of the include files under the include directory
Marc G. Fournier [Sat, 12 Apr 1997 10:21:55 +0000 (10:21 +0000)]
More more of the include files under the include directory

27 years agoremove the port related .h files from the top level directory.
Marc G. Fournier [Sat, 12 Apr 1997 10:20:20 +0000 (10:20 +0000)]
remove the port related .h files from the top level directory.

Moved to include/port

27 years agoFix detection of install vs installbsd
Marc G. Fournier [Sat, 12 Apr 1997 09:39:56 +0000 (09:39 +0000)]
Fix detection of install vs installbsd

Submitted by: adrian@waltham.harvard.net

27 years agoMake sure postgres.h is included before we start checking #ifdef's for other
Marc G. Fournier [Sat, 12 Apr 1997 09:37:31 +0000 (09:37 +0000)]
Make sure postgres.h is included before we start checking #ifdef's for other
header files.

Pointed out by: Edmund Mergl <E.Mergl@bawue.de>

27 years agoGet rid of the .sql files out of the Makefile
Marc G. Fournier [Sat, 12 Apr 1997 09:34:31 +0000 (09:34 +0000)]
Get rid of the .sql files out of the Makefile

Pointed out by: afc@teri.superlink.net

27 years agoMajor cleanup of Install instructions
Marc G. Fournier [Sat, 12 Apr 1997 09:33:25 +0000 (09:33 +0000)]
Major cleanup of Install instructions

Provided by: adrian@waltham.harvard.net

27 years agomodifications to pg_dump towards supporting dumping of ACLs (doesn't work yet!)
Marc G. Fournier [Sat, 12 Apr 1997 09:24:23 +0000 (09:24 +0000)]
modifications to pg_dump towards supporting dumping of ACLs (doesn't work yet!)

modification to c.h so that bool isn't typedef'd under __cplusplus

27 years agoChange BACKEND to CFLAGS
Marc G. Fournier [Fri, 11 Apr 1997 18:55:10 +0000 (18:55 +0000)]
Change BACKEND to CFLAGS

27 years agoFinish removing the TEST_MAIN stuff, which was mean for standalone
Marc G. Fournier [Thu, 10 Apr 1997 20:51:13 +0000 (20:51 +0000)]
Finish removing the TEST_MAIN stuff, which was mean for standalone
testing

27 years agoFix #include "cash.h" to be #include <utils/cash.h>
Marc G. Fournier [Thu, 10 Apr 1997 20:42:35 +0000 (20:42 +0000)]
Fix #include "cash.h" to be #include <utils/cash.h>

Remove the TEST_MAIN stuff at the top...

27 years agoUpdate man page, listing new \z value
Marc G. Fournier [Thu, 10 Apr 1997 11:58:59 +0000 (11:58 +0000)]
Update man page, listing new \z value

27 years agoAdd a \z command to psql that lists off grant/revoke permissions
Marc G. Fournier [Thu, 10 Apr 1997 11:54:29 +0000 (11:54 +0000)]
Add a \z command to psql that lists off grant/revoke permissions

- if someone can pick a better \? for this, plesae let me know...all
  the good ones seem taken :(

27 years agoCopy indexkeys in _copyIndexPath.
Vadim B. Mikheev [Thu, 10 Apr 1997 07:59:09 +0000 (07:59 +0000)]
Copy indexkeys in _copyIndexPath.

27 years agoMuch improved configure that integrates the build script right into it
Marc G. Fournier [Wed, 9 Apr 1997 08:55:32 +0000 (08:55 +0000)]
Much improved configure that integrates the build script right into it

Submitted by: adrian@waltham.harvard.net

27 years agoremove the old regression test files. have copies saved in my directory here,
Marc G. Fournier [Wed, 9 Apr 1997 08:48:14 +0000 (08:48 +0000)]
remove the old regression test files.  have copies saved in my directory here,
but it gets rid of the temptation to modify the old source files :)

27 years agoAdd in D'Arcy's cash code
Marc G. Fournier [Wed, 9 Apr 1997 08:36:21 +0000 (08:36 +0000)]
Add in D'Arcy's cash code

pg_proc.h still needs modifying, but this gets it in there so that we can
get around any compiler bugs.  Will try and get the pg_proc.h entries done
up later tonight...

27 years agoFrom: "Martin S. Utesch" <utesch@aut.tu-freiberg.de>
Marc G. Fournier [Wed, 9 Apr 1997 08:31:29 +0000 (08:31 +0000)]
From: "Martin S. Utesch" <utesch@aut.tu-freiberg.de>
Subject: Re: [HACKERS] GEQO and views (rules)

Oke, this was caused by a classic bug :-/
I thougth, root->base_relation_list_ could be
represented as relid string 1-2-3-4- etc.

Instead, in case of views, the count of relids doesn't start with "1" but
maybe 4-5-6- etc . :-(

GEQO patch follows ... views are now all right.

27 years agoGNUmakefile.in - remove backend/utils/Gen_fmgrtab.sh on distclean
Marc G. Fournier [Wed, 9 Apr 1997 08:29:35 +0000 (08:29 +0000)]
GNUmakefile.in - remove backend/utils/Gen_fmgrtab.sh on distclean
varlena.c - part of Thomas' most recent patch

27 years agoNow we have #define _CPU_INDEX_PAGE_WEIGHT_ 0.033 (/* CPU-index-to-page cost
Vadim B. Mikheev [Wed, 9 Apr 1997 02:24:19 +0000 (02:24 +0000)]
Now we have #define _CPU_INDEX_PAGE_WEIGHT_ 0.033 (/* CPU-index-to-page cost
weighting factor */) in addition to
#define _CPU_PAGE_WEIGHT_  0.065 (/* CPU-heap-to-page cost weighting factor
*/).

27 years agogetattnvals(): if attnvals in pg_attribute is 0 then use
Vadim B. Mikheev [Wed, 9 Apr 1997 02:20:32 +0000 (02:20 +0000)]
getattnvals(): if attnvals in pg_attribute is 0 then use
ATTNVALS_SCALE/reltuples (instead of reltuples).

27 years ago1. Enable to have different _CPU_PAGE_WEIGHT_ for heap and index.
Vadim B. Mikheev [Wed, 9 Apr 1997 02:13:41 +0000 (02:13 +0000)]
1. Enable to have different _CPU_PAGE_WEIGHT_ for heap and index.
2. PageWeights are variables now.
3. Fixed using ceil((double)selec*indextuples) as estimation
for expected heap pages: ceil((double)selec*relpages) now.

27 years agoFix (hack) IndexSelectivity():
Vadim B. Mikheev [Wed, 9 Apr 1997 01:52:04 +0000 (01:52 +0000)]
Fix (hack) IndexSelectivity():
use sum(npages)/((nkeys == 1) ? 1 : nkeys + 1) as expected index page
estimation for multi-key quals - instead of sum(npages).
In old code npages for x > 10 and x < 20 is twice as for x > 10 - cool ?

27 years agoSlight mods to explain (or try to) the new regression tests
Marc G. Fournier [Tue, 8 Apr 1997 19:34:58 +0000 (19:34 +0000)]
Slight mods to explain (or try to) the new regression tests

27 years agoFix btabstimecmp ().
Vadim B. Mikheev [Mon, 7 Apr 1997 06:45:41 +0000 (06:45 +0000)]
Fix btabstimecmp ().

27 years agoGet rid of queries.source...its all in the sql directory.
Marc G. Fournier [Sun, 6 Apr 1997 08:53:34 +0000 (08:53 +0000)]
Get rid of queries.source...its all in the sql directory.

regress.sh modified to get rid of queries.sql tests, as they are
performed vis sql/*.sql

27 years agoMore splits and cleanups...
Marc G. Fournier [Sun, 6 Apr 1997 08:29:57 +0000 (08:29 +0000)]
More splits and cleanups...

Its starting to actually take shape and look as expected...

27 years agoMore splits and cleanups...
Marc G. Fournier [Sun, 6 Apr 1997 06:07:13 +0000 (06:07 +0000)]
More splits and cleanups...

27 years agoAdd in support so that build will at least guess which template file
Marc G. Fournier [Sun, 6 Apr 1997 04:43:28 +0000 (04:43 +0000)]
Add in support so that build will at least guess which template file
should be used :)

27 years agoremove create.{source,sql} as they are now down in the 'sql' directory
Marc G. Fournier [Sat, 5 Apr 1997 21:33:26 +0000 (21:33 +0000)]
remove create.{source,sql} as they are now down in the 'sql' directory
partially split

27 years agoChange Postgres95 to PostgreSQL
Marc G. Fournier [Sat, 5 Apr 1997 21:28:46 +0000 (21:28 +0000)]
Change Postgres95 to PostgreSQL

The whole file needs updating, but will work on that after finishing
with the splits

27 years agoMore splits of the regression tests in order to make them more
Marc G. Fournier [Sat, 5 Apr 1997 21:26:00 +0000 (21:26 +0000)]
More splits of the regression tests in order to make them more
user-friendly (and more useful)

27 years agoPurge out tests/expected that are now in sql/expected subdirectories from
Marc G. Fournier [Sat, 5 Apr 1997 11:58:40 +0000 (11:58 +0000)]
Purge out tests/expected that are now in sql/expected subdirectories from
'master' file

Commit mods to regress.sh so that split out tests are run...look forward
to finding out how to do a proper redirect to continue visual cleanup :)

27 years agoAgain, add more tests
Marc G. Fournier [Sat, 5 Apr 1997 11:26:55 +0000 (11:26 +0000)]
Again, add more tests

27 years agoAdd more expected.out results
Marc G. Fournier [Sat, 5 Apr 1997 11:24:54 +0000 (11:24 +0000)]
Add more expected.out results

27 years agoThere are the broken out 'sql' queries from queries.source
Marc G. Fournier [Sat, 5 Apr 1997 11:08:30 +0000 (11:08 +0000)]
There are the broken out 'sql' queries from queries.source

tests allows us to have a 'for...done' loop inside of regress.sh for
both doing the tests, and determining fail/ok results

27 years agoThese are the broken down 'expected.output' files created so far
Marc G. Fournier [Sat, 5 Apr 1997 11:06:04 +0000 (11:06 +0000)]
These are the broken down 'expected.output' files created so far

27 years agoPrint 'Group' as name of Group plan.
Vadim B. Mikheev [Sat, 5 Apr 1997 06:42:32 +0000 (06:42 +0000)]
Print 'Group' as name of Group plan.

27 years agoChanges for GROUP BY func_results:
Vadim B. Mikheev [Sat, 5 Apr 1997 06:39:58 +0000 (06:39 +0000)]
Changes for GROUP BY func_results:
AddGroupAttrToTlist() is not called from anywhere now.

27 years agoChanges for GROUP BY func_results.
Vadim B. Mikheev [Sat, 5 Apr 1997 06:37:37 +0000 (06:37 +0000)]
Changes for GROUP BY func_results.

27 years agoNow we can GROUP BY func_results.
Vadim B. Mikheev [Sat, 5 Apr 1997 06:29:03 +0000 (06:29 +0000)]
Now we can GROUP BY func_results.

27 years agoFix for 'SET var_name TO var_value': var_name already defined.
Vadim B. Mikheev [Sat, 5 Apr 1997 06:25:59 +0000 (06:25 +0000)]
Fix for 'SET var_name TO var_value': var_name already defined.

27 years agoPut resdom into GroupClause (GROUP BY func_results)
Vadim B. Mikheev [Sat, 5 Apr 1997 06:19:22 +0000 (06:19 +0000)]
Put resdom into GroupClause (GROUP BY func_results)

27 years agoCheck for attributeList is NULL in ConstructTupleDescriptor ().
Vadim B. Mikheev [Sat, 5 Apr 1997 03:36:21 +0000 (03:36 +0000)]
Check for attributeList is NULL in ConstructTupleDescriptor ().

Submitted by Raymond Toy.

27 years agoFrom: Thomas Lockhart <Thomas.G.Lockhart@jpl.nasa.gov>
Marc G. Fournier [Sat, 5 Apr 1997 02:51:41 +0000 (02:51 +0000)]
From: Thomas Lockhart <Thomas.G.Lockhart@jpl.nasa.gov>
Subject: [HACKERS] Fix for European dates

This apparently fixes the European date reading problem reported
by several (European) bleeding edge adopters. I tried a few test
cases and it doesn't break the non-EuroDate cases in my test suite.

27 years agoMove YACC and YFLAGS into the template files
Marc G. Fournier [Fri, 4 Apr 1997 11:23:15 +0000 (11:23 +0000)]
Move YACC and YFLAGS into the template files

Clean up the .sample files...comment out all sample entries except for
the localhost one

27 years agoThis commit represents a clean compile with the new templates under
Marc G. Fournier [Fri, 4 Apr 1997 10:43:16 +0000 (10:43 +0000)]
This commit represents a clean compile with the new templates under
FreeBSD

The Makefile(s) have all been cleaned up such that there is a single
LDFLAGS vs LD_ADD or LDADD or LDFLAGS or LDFLAGS_BE.  The Makefile(s)
should be alot more straightforward then they were before...and
consistent

27 years agoGot AROPT setting backwards...correct it
Marc G. Fournier [Fri, 4 Apr 1997 09:02:34 +0000 (09:02 +0000)]
Got AROPT setting backwards...correct it

27 years agoAdd distclean to target listing
Marc G. Fournier [Fri, 4 Apr 1997 09:01:19 +0000 (09:01 +0000)]
Add distclean to target listing

27 years agoAgain, needs float.h
Marc G. Fournier [Fri, 4 Apr 1997 08:55:29 +0000 (08:55 +0000)]
Again, needs float.h

27 years agoneeds float.h for DBL_MIN under FreeBSD
Marc G. Fournier [Fri, 4 Apr 1997 08:53:08 +0000 (08:53 +0000)]
needs float.h for DBL_MIN under FreeBSD

27 years agoRemove some files that were inadvertantly created
Marc G. Fournier [Fri, 4 Apr 1997 08:16:37 +0000 (08:16 +0000)]
Remove some files that were inadvertantly created
Clean up format of linux-elf

27 years agoOne helluva mess.
Marc G. Fournier [Fri, 4 Apr 1997 07:59:48 +0000 (07:59 +0000)]
One helluva mess.

Further extended Makefile.global/build/configure so that we can
have a 'template' file for each OS (and each version of OS, as in BSDi)
which is used as much as possible to generate Makefile.global

Any future ports should look at using the template file as a basis,
before moving over to Makefile.global.

This will most probably break alot of the ports, atho I've tried to
be very neat about it...

27 years agoInstall os.h when we install the other headers...
Marc G. Fournier [Fri, 4 Apr 1997 02:53:14 +0000 (02:53 +0000)]
Install os.h when we install the other headers...

Pointed out by: System Administrator <sysadmin@sba.miami.edu>

27 years agoMore modifications to make building more interactive:
Marc G. Fournier [Thu, 3 Apr 1997 22:16:34 +0000 (22:16 +0000)]
More modifications to make building more interactive:

Allow installer to change DEF_PGPORT
Allow installer to disable HBA

27 years agoRemove code associated with !ACLGROUP_PATCH, and appropriate #ifdef's
Marc G. Fournier [Thu, 3 Apr 1997 21:31:57 +0000 (21:31 +0000)]
Remove code associated with !ACLGROUP_PATCH, and appropriate #ifdef's

27 years agoVarious improvements to reduce questions :)
Marc G. Fournier [Thu, 3 Apr 1997 21:26:36 +0000 (21:26 +0000)]
Various improvements to reduce questions :)

Remove USE_LOCALE from Makefile.global.in
Add USE_LOCALE to build/configure/config.h

Add check for BUILDRUN in configure to make sure that build is run before
configure

27 years agoFrom: "D'Arcy J.M. Cain" <darcy@druid.net>
Marc G. Fournier [Thu, 3 Apr 1997 19:58:11 +0000 (19:58 +0000)]
From: "D'Arcy J.M. Cain" <darcy@druid.net>
Subject: [HACKERS] timestamp.c changes

I sent in changes previously and they were rejected because they didn't
follow ANSI spec.  Here is the input part of the changes again.  Even
though it allows more flexibility for inputting different formats, it
is also backwards compatible with the standard version.  I have also
not changed the output format so it will still output the ANSI forms.
Is this acceptable to everyone?

27 years agoFrom: Thomas Lockhart <Thomas.G.Lockhart@jpl.nasa.gov>
Marc G. Fournier [Thu, 3 Apr 1997 19:56:47 +0000 (19:56 +0000)]
From: Thomas Lockhart <Thomas.G.Lockhart@jpl.nasa.gov>
Subject: [HACKERS] Aggregate function patches

Here are the aggregate function patches I originally sent in last December.
They fix sum() and avg() behavior for ints and floats when NULL values are
involved.

I was waiting to resubmit these until I had a chance to write a v6.0->v6.1
database upgrade script to ensure that existing v6.0 databases which have
not been reloaded for v6.1 do no break with the new aggregate behavior.
These scripts are included below. It's OK with me if someone wants to do
something different with the upgrade strategy, but something like this
was discussed a few weeks ago.

Also, there were a couple of small items which cropped up in doing a clean
install of 970403 (actually 970402 + 970403 changes since the full 970403
tar file appears to be damaged or at least suspect). They are the first
two patches below and can be omitted if desired (although I think they
aren't dangerous :).

27 years agoremoved as already installed as part of system headers on NetBSD/FreeBSD
Marc G. Fournier [Wed, 2 Apr 1997 18:49:24 +0000 (18:49 +0000)]
removed as already installed as part of system headers on NetBSD/FreeBSD

27 years agoFrom: Thomas Lockhart <Thomas.G.Lockhart@jpl.nasa.gov>
Marc G. Fournier [Wed, 2 Apr 1997 18:36:24 +0000 (18:36 +0000)]
From: Thomas Lockhart <Thomas.G.Lockhart@jpl.nasa.gov>
Subject: [HACKERS] More date time functions

Here are some additional patches mostly related to the date and time
data types. It includes some type conversion routines to move between
the different date types and some other date manipulation routines such
as date_part(units,datetime).

I noticed Edmund Mergl et al's neat trick for getting function overloading
for builtin functions, so started to use that for the date and time stuff.
Later, if someone figures out how to get function overloading directly
for internal C code, then we can move to that technique.

These patches include documentation updates (don't faint!) for the built-in
man page. Doesn't yet include mention of timestamp, since I don't know
much about it and since it may change a bit to become a _real_ ANSI timestamp
which would include parser support for the declaration syntax (what do you
think, Dan?).

The patches were developed on the 970330 release, but have been rebuilt
off of the 970402 release. The first patch below is to get libpq to compile,
on my Linux box, but is not related to the rest of the patches and you can
choose not to apply that one at this time. Thanks in advance, scrappy!

27 years agoFrom: "Martin J. Laubach" <mjl@CSlab.tuwien.ac.at>
Marc G. Fournier [Wed, 2 Apr 1997 18:26:25 +0000 (18:26 +0000)]
From: "Martin J. Laubach" <mjl@CSlab.tuwien.ac.at>
Subject: [HACKERS] libpq variable set patch

  Just a small change, so the automatic variable setting on
connection startup actually works...

27 years agoFrom: "Martin J. Laubach" <mjl@CSlab.tuwien.ac.at>
Marc G. Fournier [Wed, 2 Apr 1997 18:24:52 +0000 (18:24 +0000)]
From: "Martin J. Laubach" <mjl@CSlab.tuwien.ac.at>
Subject: [HACKERS] Patch: SET var TO 'val'

  Here is a patch that adds a "SET variable TO 'somevalue'" capability
to the parser, and then calls the SetPGVariable() function (which does
just issue a elog(NOTICE) to see whether it works).

  That's the framework for adding timezone/date format/language/...
stuff.

27 years agoFrom: Anton de Wet <adw@obsidian.co.za>
Marc G. Fournier [Wed, 2 Apr 1997 18:16:49 +0000 (18:16 +0000)]
From: Anton de Wet <adw@obsidian.co.za>
Subject: [HACKERS] Small patch to pgtclCmds.c

Hi I have made the following small change to the extensions I made to
pgtclCmds.c quite a while ago.

At the moment there is a -assignbyidx option to pg_result assigning the
returned tuples to an array by using the 1st field of the select statement
as the key to the array.
eg "select name,age from vitalstatistics" will result in an array with

myarray(peter) = 32
myarray(paul)  = 45

Often I need to have a pseudo-multi dimentional
array eg. "select name,age from vitalstatistics where occupation='plummer'

I would like to be able to generate an array
newarray(peter,overpaid) = 32

So to add a arbitrary string to the key value I have extended

  pg_result $res -assignbyidx $arrayname

to have an optional argument

  pg_result $res -assignbyidx $arrayname $appendstr

So that that string is appended to the key value.

27 years agoFrom: Oleg Bartunov <oleg@sai.msu.su>
Marc G. Fournier [Wed, 2 Apr 1997 18:13:47 +0000 (18:13 +0000)]
From: Oleg Bartunov <oleg@sai.msu.su>
Subject: [HACKERS] locale patches !

Hi there,

here are little patches to get Postgres 6.1 works with locale stuff.
This is a patch against 970402.tar.gz, there are no problem to apply them
by hand to 6.0 release. Collate stuff tested about 1-2 months in real
working database but I'm sure there must be no problem. US hackers
could vote against locale implementation ( locale for sure will affect to
speed of postgres ), so I introduce variable USE_LOCALE which
controls locale stuff. Non-US users now could use ~* operator
for searching and <order by> for strings with nation alphabet.
Please, don't forget, as I did first time, to set environment variable
LC_CTYPE and LC_COLLATE because backend get locale information from them.
I start postmaster from a little script, assuming that shell is Bash shell
it looks like:

#!/bin/sh

export LC_CTYPE=koi8-r
export LC_COLLATE=koi8-r
postmaster -B 1024 -S -D/usr/local/pgsql/data/ -o '-Fe'

27 years agoNew keyword: SEQUENCE.
Vadim B. Mikheev [Wed, 2 Apr 1997 04:49:13 +0000 (04:49 +0000)]
New keyword: SEQUENCE.

27 years agoManuals for SEQUENCEs.
Vadim B. Mikheev [Wed, 2 Apr 1997 04:20:00 +0000 (04:20 +0000)]
Manuals for SEQUENCEs.

27 years ago\h create/drop sequence.
Vadim B. Mikheev [Wed, 2 Apr 1997 04:18:26 +0000 (04:18 +0000)]
\h create/drop sequence.

27 years agoDumping sequence relations as 'CREATE SEQUENCE ...'.
Vadim B. Mikheev [Wed, 2 Apr 1997 04:17:27 +0000 (04:17 +0000)]
Dumping sequence relations as 'CREATE SEQUENCE ...'.

27 years agoCall DefineSequence () for T_CreateSeqStmt node.
Vadim B. Mikheev [Wed, 2 Apr 1997 04:06:32 +0000 (04:06 +0000)]
Call DefineSequence () for T_CreateSeqStmt node.

27 years agoCan't INSERT/UPDATE/DELETE sequence relation.
Vadim B. Mikheev [Wed, 2 Apr 1997 04:04:11 +0000 (04:04 +0000)]
Can't INSERT/UPDATE/DELETE sequence relation.

27 years agoCREATE/DROP SEQUENCE ...
Vadim B. Mikheev [Wed, 2 Apr 1997 04:01:03 +0000 (04:01 +0000)]
CREATE/DROP SEQUENCE ...
Check nextval/currval permission in analyze.c.

27 years agoCan't COPY TO sequence relation.
Vadim B. Mikheev [Wed, 2 Apr 1997 03:57:06 +0000 (03:57 +0000)]
Can't COPY TO sequence relation.
Can't inherits from ...

27 years agoSequence numbers generators code.
Vadim B. Mikheev [Wed, 2 Apr 1997 03:51:23 +0000 (03:51 +0000)]
Sequence numbers generators code.

27 years agovc_getrels(p, VacRelP) returns NIL for special relations (indices,
Vadim B. Mikheev [Wed, 2 Apr 1997 03:48:01 +0000 (03:48 +0000)]
vc_getrels(p, VacRelP) returns NIL for special relations (indices,
sequences,...) and vc_delhilowstats(NULL->vrl_relid) ...

27 years agoHack for heap_creat to enable relkind to be 'S' for sequences.
Vadim B. Mikheev [Wed, 2 Apr 1997 03:41:16 +0000 (03:41 +0000)]
Hack for heap_creat to enable relkind to be 'S' for sequences.

27 years agoCloseSequences () at xact commit/abort.
Vadim B. Mikheev [Wed, 2 Apr 1997 03:38:02 +0000 (03:38 +0000)]
CloseSequences () at xact commit/abort.

27 years agoNew node T_CreateSeqStmt.
Vadim B. Mikheev [Wed, 2 Apr 1997 03:34:46 +0000 (03:34 +0000)]
New node T_CreateSeqStmt.

27 years agoNew relkind ('S') for sequence relations.
Vadim B. Mikheev [Wed, 2 Apr 1997 03:29:37 +0000 (03:29 +0000)]
New relkind ('S') for sequence relations.
New funcs (nextval & currval) in pg_proc.h

27 years agoPrototypes for sequence.c
Vadim B. Mikheev [Wed, 2 Apr 1997 03:23:38 +0000 (03:23 +0000)]
Prototypes for sequence.c

27 years agoUse $(CC), not gcc, to compile
Marc G. Fournier [Wed, 2 Apr 1997 00:34:23 +0000 (00:34 +0000)]
Use $(CC), not gcc, to compile

Pointed out by: igor@cs.cs.miami.edu

27 years agoMisc port related issues
Marc G. Fournier [Tue, 1 Apr 1997 09:27:11 +0000 (09:27 +0000)]
Misc port related issues

27 years agoSlight Linux related bug pointed out by Gabriel Akos <gabriel@rocker.sch.bme.hu>
Marc G. Fournier [Fri, 28 Mar 1997 09:43:53 +0000 (09:43 +0000)]
Slight Linux related bug pointed out by Gabriel Akos <gabriel@rocker.sch.bme.hu>

27 years agoFrom: Thomas Lockhart <Thomas.G.Lockhart@jpl.nasa.gov>
Marc G. Fournier [Fri, 28 Mar 1997 07:18:06 +0000 (07:18 +0000)]
From: Thomas Lockhart <Thomas.G.Lockhart@jpl.nasa.gov>
Subject: [HACKERS] Small date patches (resubmitted)

Here a some small patches for the date/time code. They set the default
output format for the datetime type to the traditional Postgres
style, and fix a date debugging declaration. I submitted these
a couple of days ago, but they might have gotten lost...

NOTE: the second patch to dt.c is what I believe D'Arcy submitted as well,
      that I claimed was taken out...sorry D'Arcy, my fault :(

27 years agoFrom: Thomas Lockhart <Thomas.G.Lockhart@jpl.nasa.gov>
Marc G. Fournier [Fri, 28 Mar 1997 07:13:21 +0000 (07:13 +0000)]
From: Thomas Lockhart <Thomas.G.Lockhart@jpl.nasa.gov>
Subject: Re: [HACKERS] abstime "now" broken

Yes, I broke 'now' :( with an attempt at a bug fix involving
servers running in the UTC/GMT timezone. These patches fix
the problem, and have been tested in GMT (+00 hours),
PST (-08), and NZT (+12) timezones which exercized the code for
various cases including across day boundaries.  btw, this code
fixes the same type of problem for 'today', 'yesterday', 'tomorrow',
for DATETIME, ABSTIME, DATE and TIME types.

The bugfix itself is quite small, but I have accumulated other
changes in the datetime data type and include them here also.
One set of changes involves printing ISO-formatted dates and
is in response to the helpful information from Kurt Lidl regarding
ANSI SQL dates. I'll send another e-mail sometime soon discussing
more issues he has raised...