Barry Lind [Fri, 7 Mar 2003 18:39:46 +0000 (18:39 +0000)]
Cleanup and reorganization.
- Added a private api layer (org.postgresql.core.Base*)
- Cleaned up public api (org.postgresql.PG*)
- Added consistent headers and copywrite info
- Removed deprecated Serialize functionality
- Cleaned up imports
- Moved some files to more appropriate locations
Tom Lane [Thu, 6 Mar 2003 22:54:49 +0000 (22:54 +0000)]
Tweak dependency code to suppress NOTICEs generated by new method for
cleaning out temp namespaces. We don't really want the server log to be
cluttered with 'Drop cascades to table foo' every time someone uses a
temp table...
Tom Lane [Thu, 6 Mar 2003 03:16:55 +0000 (03:16 +0000)]
Use poll(2) in preference to select(2), if available. This solves
problems in applications that may have a large number of files open,
such that libpq's socket number exceeds the range supported by fd_set.
From Chris Brown.
Tom Lane [Thu, 6 Mar 2003 00:04:27 +0000 (00:04 +0000)]
Add code to dump contents of free space map into $PGDATA/global/pg_fsm.cache
at database shutdown, and then load it again at database startup. This
preserves our hard-won knowledge of free space across restarts (given
an orderly shutdown, that is).
Tom Lane [Wed, 5 Mar 2003 20:01:04 +0000 (20:01 +0000)]
Turns out new IN implementation has got some problems in an UPDATE or
DELETE with inherited target table. Fix it; add a regression test.
Also, correct ancient misspelling of 'inherited'.
Tom Lane [Wed, 5 Mar 2003 18:38:14 +0000 (18:38 +0000)]
Repair bug reported by Laurent Perez: bad plan generated when UPDATE or
DELETE of an inheritance tree references another inherited relation.
This bug has been latent since 7.1; I'm still not quite sure why 7.1 and
7.2 don't manifest it (at least, they don't crash on a simple test case).
Tom Lane [Tue, 4 Mar 2003 21:51:22 +0000 (21:51 +0000)]
Reimplement free-space-map management as per recent discussions.
Adjustable threshold is gone in favor of keeping track of total requested
page storage and doling out proportional fractions to each relation
(with a minimum amount per relation, and some quantization of the results
to avoid thrashing with small changes in page counts). Provide special-
case code for indexes so as not to waste space storing useless page
free space counts. Restructure internal data storage to be a flat array
instead of list-of-chunks; this may cost a little more work in data
copying when reorganizing, but allows binary search to be used during
lookup_fsm_page_entry().
Tom Lane [Sun, 2 Mar 2003 23:46:34 +0000 (23:46 +0000)]
Be smart about outer-join qualifications that mention only one side of
the join, per recent discussion on pgsql-sql. Not clear that this will
come up often in real queries, but it's not any more expensive to do it
right, so we may as well do it right.
Tom Lane [Sun, 2 Mar 2003 20:45:47 +0000 (20:45 +0000)]
Repair memory leak introduced by recent change to make SPI return a
tupdesc even with zero tuples returned: some plpgsql routines assumed
they didn't need to do SPI_freetuptable() after retrieving no tuples.
Tom Lane [Thu, 27 Feb 2003 21:36:58 +0000 (21:36 +0000)]
Change EXTRACT(EPOCH FROM timestamp) so that a timestamp without time zone
is assumed to be in local time, not GMT. This improves consistency with
other operations, which all assume local timezone when it matters. Per
bug #897.
The quote function should return a string suitable for dropping into a
query string. This fixes a bug where bool types sometimes returned with
a string that could not be dropped into a query.
Tom Lane [Tue, 25 Feb 2003 23:47:43 +0000 (23:47 +0000)]
Remove REWRITE_INVOKE_MAX in favor of making an accurate check for
recursion in RewriteQuery(); also, detect recursion in fireRIRrules(),
so as to catch self-referential views per example from Ryan VanderBijl.
Minor code restructuring to make it easier to catch recursive case.
Tom Lane [Mon, 24 Feb 2003 00:57:17 +0000 (00:57 +0000)]
During VACUUM FULL, truncate off any deletable pages that are at the
end of a btree index. This isn't super-effective, since we won't move
nondeletable pages, but it's better than nothing. Also, improve stats
displayed during VACUUM VERBOSE.
Tom Lane [Sun, 23 Feb 2003 22:43:09 +0000 (22:43 +0000)]
Adjust btbulkdelete logic so that only one WAL record is issued while
deleting multiple index entries on a single index page. This makes for
a very substantial reduction in the amount of WAL traffic during a
large delete operation.
Tom Lane [Sun, 23 Feb 2003 04:48:19 +0000 (04:48 +0000)]
If a shutdown request comes in while we're still starting up, don't
service it until after we execute SetThisStartUpID(). Else shutdown
process will write the wrong SUI into the shutdown checkpoint, which
seems likely to be trouble --- although I've not quite figured out
how significant it really is.
Tom Lane [Sat, 22 Feb 2003 05:57:45 +0000 (05:57 +0000)]
Simplify timezone-handling code per proposal to pghackers: get rid of
setting timezone-related variables during transaction start. They were
not used anyway in platforms that HAVE_TM_ZONE or HAVE_INT_TIMEZONE,
which it appears is *all* the platforms we are currently supporting.
For platforms that have neither, we now only support UTC or numeric-
offset-from-UTC timezones.
Tom Lane [Sat, 22 Feb 2003 00:45:05 +0000 (00:45 +0000)]
More infrastructure for btree compaction project. Tree-traversal code
now knows what to do upon hitting a dead page (in theory anyway, it's
untested...). Add a post-VACUUM-cleanup entry point for index AMs, to
provide a place for dead-page scavenging to happen.
Also, fix oversight that broke btpo_prev links in temporary indexes.
initdb forced due to additions in pg_am.
Tom Lane [Fri, 21 Feb 2003 00:06:22 +0000 (00:06 +0000)]
Make btree index structure adjustments and WAL logging changes needed to
support btree compaction, as per proposal of a few days ago. btree index
pages no longer store parent links, instead they have a level indicator
(counting up from zero for leaf pages). The FixBTree recovery logic is
removed, and replaced by code that detects missing parent-level insertions
during WAL replay. Also, generate appropriate WAL entries when updating
btree metapage and when building a btree index from scratch. I believe
btree indexes are now completely WAL-legal for the first time.
initdb forced due to index and WAL changes.
Tom Lane [Thu, 20 Feb 2003 05:24:55 +0000 (05:24 +0000)]
Fix timestamptz_in so that parsing of 'now'::timestamptz gives right
answer when SET TIMEZONE has been done since the start of the current
transaction. Per bug report from Robert Haas.
I plan some futher cleanup in HEAD, but this is a low-risk patch for
the immediate issue in 7.3.
Bruce Momjian [Wed, 19 Feb 2003 04:04:04 +0000 (04:04 +0000)]
Here's the patch I promised over on HACKERS - it alters the
implementation
of '\e' history tracking for systems that have a readline compatability
library without replace_history_entry. I fall back to pushing the query
onto the history stack after the \e, rather than replacing it.
The patch adds one more place to look for readline headers, and a test
for replace_history_entry. I've only included the patch for configure.in
Bruce Momjian [Wed, 19 Feb 2003 04:02:54 +0000 (04:02 +0000)]
- Modifies LOCKTAG to include a 'classId'. Relation receive a classId of
RelOid_pg_class, and transaction locks XactLockTableId. RelId is renamed
to objId.
- LockObject() and UnlockObject() functions created, and their use
sprinkled throughout the code to do descent locking for domains and
types. They accept lock modes AccessShare and AccessExclusive, as we
only really need a 'read' and 'write' lock at the moment. Most locking
cases are held until the end of the transaction.
This fixes the cases Tom mentioned earlier in regards to locking with
Domains. If the patch is good, I'll work on cleaning up issues with
other database objects that have this problem (most of them).
Bruce Momjian [Wed, 19 Feb 2003 03:54:39 +0000 (03:54 +0000)]
Here's some changes I made last night to psql's common.c (as found in
7.3.2). It removes some code duplication and #ifdeffing, and some
unstructured ugliness such as tacky breaks and an unneeded continue.
Breaks up a large function into smaller functions and reduces required
nesting levels, and kills a variable or two.
Bruce Momjian [Wed, 19 Feb 2003 03:48:11 +0000 (03:48 +0000)]
The following patches eliminate the overflows in the j2date() and date2j()
functions which limited the maximum date for a timestamp to AD 1465001.
The new limit is AD 5874897.
The files affected are:
doc/src/sgml/datatype.sgml:
Documentation change due to patch. Included is a notice about
the reduced range when using an eight-byte integer for timestamps.
src/backend/utils/adt/datetime.c:
Replacement functions for j2date() and date2j() functions.
src/include/utils/datetime.h:
Corrected a bug with the limit on the earliest possible date,
Nov 23,-4713 has a Julian day count of -1. The earliest possible
date should be Nov 24, -4713 with a day count of 0.
src/test/regress/expected/horology-no-DST-before-1970.out:
src/test/regress/expected/horology-solaris-1947.out:
src/test/regress/expected/horology.out:
Copies of expected output for regression testing.
Note: Only horology.out has been physically tested. I do not have access
to a Solaris box and I don't know how to provoke the "pre-1970" test.
src/test/regress/sql/horology.sql:
Added some test cases to check extended range.
Bruce Momjian [Wed, 19 Feb 2003 03:46:00 +0000 (03:46 +0000)]
We just released new version of contrib/btree_gist
(7.3 and current CVS) with support of int8, float4, float8
in addition to int4. Thanks Janko Richter for contribution.
Bruce Momjian [Wed, 19 Feb 2003 03:12:22 +0000 (03:12 +0000)]
This patch adds a note to the documentation describing why the
performance of min() and max() is slow when applied to the entire table,
and suggesting the simple workaround most experienced Pg users
eventually learn about (SELECT xyz ... ORDER BY xyz LIMIT 1).
Bruce Momjian [Wed, 19 Feb 2003 01:36:32 +0000 (01:36 +0000)]
Mark as done, fix grammar:
< * Prevent index uniqueness checks when UPDATE does not modifying column
> * Prevent index uniqueness checks when UPDATE does not modify the column 235c235
< o Make PL/PgSQL %TYPE schema-aware
> o -Make PL/PgSQL %TYPE schema-aware
Tom Lane [Tue, 18 Feb 2003 02:53:29 +0000 (02:53 +0000)]
Async_NotifyHandler must save and restore ImmediateInterruptOK. Fixes
known problem with failure to respond to 'pg_ctl stop -m fast', and
probable problems if SIGINT or SIGTERM arrives while processing a
SIGUSR2 interrupt that arrived while waiting for a new client query.
Bruce Momjian [Mon, 17 Feb 2003 18:48:29 +0000 (18:48 +0000)]
Update marks:
< o Allow CLUSTER to cluster all tables (Alvaro Herrera)
> o -Allow CLUSTER to cluster all tables (Alvaro Herrera) 243c243
< * Allow pg_dump to dump a specific schema (Neil Conway)
> * -Allow pg_dump to dump a specific schema (Neil Conway) 398c398
< * Make IN/NOT IN have similar performance to EXISTS/NOT EXISTS (Tom)
> * -Make IN/NOT IN have similar performance to EXISTS/NOT EXISTS (Tom)
Tom Lane [Sun, 16 Feb 2003 06:06:32 +0000 (06:06 +0000)]
Back off previous patch to skip projection step in scan plan nodes,
in the case where the node immediately above the scan is a Hash, Sort,
or Material node. In these cases it's better to do the projection
so that we don't store unneeded columns in the hash/sort/materialize
table. Per discussion a few days ago with Anagh Lal.
Tom Lane [Sun, 16 Feb 2003 02:30:39 +0000 (02:30 +0000)]
COALESCE() and NULLIF() are now first-class expressions, not macros
that turn into CASE expressions. They evaluate their arguments at most
once. Patch by Kris Jurka, review and (very light) editorializing by me.
Bruce Momjian [Sun, 16 Feb 2003 00:29:49 +0000 (00:29 +0000)]
Update wording:
< * Disallow DROP COLUMN on a column that is part of a multi-column index
> * Require DROP COLUMN CASCADE for a column that is part of a multi-column index
Tom Lane [Sat, 15 Feb 2003 20:12:41 +0000 (20:12 +0000)]
Teach planner how to propagate pathkeys from sub-SELECTs in FROM up to
the outer query. (The implementation is a bit klugy, but it would take
nontrivial restructuring to make it nicer, which this is probably not
worth.) This avoids unnecessary sort steps in examples like
SELECT foo,count(*) FROM (SELECT ... ORDER BY foo,bar) sub GROUP BY foo
which means there is now a reasonable technique for controlling the
order of inputs to custom aggregates, even in the grouping case.
projects / postgresql / log