Neil Conway [Mon, 30 May 2005 07:20:59 +0000 (07:20 +0000)]
When enqueueing after-row triggers for updates of a table with a foreign
key, compare the new and old row versions. If the foreign key column has
not changed, we needn't enqueue the trigger, since the update cannot
violate the foreign key. This optimization was previously applied in the
RI trigger function, but it is more efficient to avoid firing the trigger
altogether. Per recent discussion on pgsql-hackers.
Also add a regression test for some unintuitive foreign key behavior, and
refactor some code that deals with the OIDs of the various RI trigger
functions.
Neil Conway [Mon, 30 May 2005 06:52:38 +0000 (06:52 +0000)]
Create separate ON INSERT and ON UPDATE triggers on tables with foreign
keys, rather than a single trigger for both events. This should not change
functionality, but it is more consistent: previously, there were trigger
functions for both "check_insert" and "check_update", but the former was
used for both events.
Bump catalog version number (not strictly necessary, but best to be
cautious).
Tom Lane [Mon, 30 May 2005 01:20:50 +0000 (01:20 +0000)]
Change the UNKNOWN type to have an internal representation matching
cstring, rather than text, so as to eliminate useless conversions
inside the parser. Per recent discussion.
Tom Lane [Mon, 30 May 2005 01:04:44 +0000 (01:04 +0000)]
Skip eval_const_expressions when the query is such that the expression
would be evaluated only once anyway (ie, it's just a SELECT with no
FROM or an INSERT ... VALUES). The planner can't do it any faster than
the executor, so no point in an extra copying of the expression tree.
Tom Lane [Sun, 29 May 2005 23:38:05 +0000 (23:38 +0000)]
Avoid unnecessary fetch from pg_shadow in the normal case in
pg_class_aclmask(). We only need to do this when we have to check
pg_shadow.usecatupd, and that's not relevant unless the target table
is a system catalog. So we can usually avoid one syscache lookup.
Tom Lane [Sun, 29 May 2005 22:45:02 +0000 (22:45 +0000)]
Improve LockAcquire API per my recent proposal. All error conditions
are now reported via elog, eliminating the need to test the result code
at most call sites. Make it possible for the caller to distinguish a
freshly acquired lock from one already held in the current transaction.
Use that capability to avoid redundant AcceptInvalidationMessages() calls
in LockRelation().
Tom Lane [Sun, 29 May 2005 20:38:06 +0000 (20:38 +0000)]
Make superuser.c maintain a simple one-entry cache holding the superuser
status of the most recently queried userid. Since the common pattern is
many successive queries about the same user (ie, the current user) this
can save a lot of syscache probes.
Tom Lane [Sun, 29 May 2005 18:24:14 +0000 (18:24 +0000)]
Remove typeidIsValid() checks in can_coerce_type(). These checks
were pretty expensive and I believe the case they were put in to
defend against can no longer arise, now that we have dependency checks
to prevent deletion of a type entry that is still referenced. Certainly
the example given in the CVS log entry can't happen anymore.
Since this was the only use of typeidIsValid(), remove the routine too.
Tom Lane [Sun, 29 May 2005 17:10:23 +0000 (17:10 +0000)]
expandRTE and get_rte_attribute_type mistakenly always imputed typmod -1
to columns of an RTE that was a function returning RECORD with a column
definition list. Apparently no one has tried to use non-default typmod
with a function returning RECORD before.
Tom Lane [Sun, 29 May 2005 04:23:07 +0000 (04:23 +0000)]
Modify hash_search() API to prevent future occurrences of the error
spotted by Qingqing Zhou. The HASH_ENTER action now automatically
fails with elog(ERROR) on out-of-memory --- which incidentally lets
us eliminate duplicate error checks in quite a bunch of places. If
you really need the old return-NULL-on-out-of-memory behavior, you
can ask for HASH_ENTER_NULL. But there is now an Assert in that path
checking that you aren't hoping to get that behavior in a palloc-based
hash table.
Along the way, remove the old HASH_FIND_SAVE/HASH_REMOVE_SAVED actions,
which were not being used anywhere anymore, and were surely too ugly
and unsafe to want to see revived again.
Tom Lane [Sat, 28 May 2005 17:21:32 +0000 (17:21 +0000)]
Bgwriter should PANIC if it runs out of memory for pending-fsyncs
hash table. This is a pretty unlikely scenario, since the table
should be tiny, but we can't guarantee continued correct operation
if it does occur. Spotted by Qingqing Zhou.
Tom Lane [Sat, 28 May 2005 05:10:47 +0000 (05:10 +0000)]
get_expr_result_type has to be prepared to pull type information
from a RECORD Const node, because that's what it may be faced with
after constant-folding of a function returning RECORD. Per example
from Michael Fuhr.
Bruce Momjian [Sat, 28 May 2005 04:12:13 +0000 (04:12 +0000)]
Remove:
<
< * Add XML output to pg_dump and COPY
<
< We already allow XML to be stored in the database, and XPath queries
< can be used on that data using /contrib/xml2. It also supports XSLT
< transformations.
Tom Lane [Fri, 27 May 2005 23:31:21 +0000 (23:31 +0000)]
Arrange to cache fmgr lookup information for an index's access method
routines in the index's relcache entry, instead of doing a fresh fmgr_info
on every index access. We were already doing this for the index's opclass
support functions; not sure why we didn't think to do it for the AM
functions too. This supersedes the former method of caching (only)
amgettuple in indexscan scan descriptors; it's an improvement because the
function lookup can be amortized across multiple statements instead of
being repeated for each statement. Even though lookup for builtin
functions is pretty cheap, this seems to drop a percent or two off some
simple benchmarks.
Bruce Momjian [Fri, 27 May 2005 22:07:26 +0000 (22:07 +0000)]
Add:
> * Consider sorting hash buckets so entries can be found using a binary
> search, rather than a linear scan
> * In hash indexes, consider storing the hash value with or instead
> of the key itself
Bruce Momjian [Fri, 27 May 2005 22:01:18 +0000 (22:01 +0000)]
Add:
> * Add the features of packages
> o Make private objects accessable only to objects in the same schema
> o Allow current_schema.objname to access current schema objects
> o Add session variables
> o Allow nested schemas
Bruce Momjian [Fri, 27 May 2005 21:31:23 +0000 (21:31 +0000)]
Display only 9 subsecond digits instead of 10 for time values, for
consistency and to prevent rounding for days < 30. Also round off all
trailing zeros, rather than leaving an even number of digits.
Bruce Momjian [Thu, 26 May 2005 15:26:00 +0000 (15:26 +0000)]
Back out:
Display only 9 not 10 digits of precision for timestamp values when
using non-integer timestamps. This prevents the display of rounding
errors for common values like days < 32.
Bruce Momjian [Thu, 26 May 2005 03:48:25 +0000 (03:48 +0000)]
Display only 9 not 10 digits of precision for timestamp values when
using non-integer timestamps. This prevents the display of rounding
errors for common values like days < 32.
Neil Conway [Thu, 26 May 2005 02:04:14 +0000 (02:04 +0000)]
Adjust datetime parsing to be more robust. We now pass the length of the
working buffer into ParseDateTime() and reject too-long input there,
rather than checking the length of the input string before calling
ParseDateTime(). The old method was bogus because ParseDateTime() can use
a variable amount of working space, depending on the content of the
input string (e.g. how many fields need to be NUL terminated). This fixes
a minor stack overrun -- I don't _think_ it's exploitable, although I
won't claim to be an expert.
Along the way, fix a bug reported by Mark Dilger: the working buffer
allocated by interval_in() was too short, which resulted in rejecting
some perfectly valid interval input values. I added a regression test for
this fix.
Tom Lane [Thu, 26 May 2005 01:24:29 +0000 (01:24 +0000)]
Tweak the backend scanner (and psqlscan.l, which must track the backend
scanner anyway) to avoid having any backup states. According to the
flex manual, this should speed things up, and indeed the backend scanner
is about a third faster according to some quick profiling checks.
I haven't tried to measure the speed change in psql, but it probably
is similar.
Bruce Momjian [Wed, 25 May 2005 22:59:33 +0000 (22:59 +0000)]
At the head of wchareq, length of (multibyte) character is compared by
using pg_mblen. Therefore, pg_mblen is executed many times, and it
becomes a bottleneck.
This patch makes a short cut, and reduces execution frequency of
pg_mblen by comparing the first byte first.
Bruce Momjian [Wed, 25 May 2005 22:12:05 +0000 (22:12 +0000)]
Quick patch to adress a recent concern on the mailing list
about adding an errant "TO" when we already have a TO. Since
TO cannot be a valid column name (we must quote it), we can
simply ignore the tab-completion if the previous word
was a "TO".
Tom Lane [Tue, 24 May 2005 18:02:31 +0000 (18:02 +0000)]
Previous fix for "x FULL JOIN y ON true" failed to handle the case
where there was also a WHERE-clause restriction that applied to the
join. The check on restrictlist == NIL is really unnecessary anyway,
because select_mergejoin_clauses already checked for and complained
about any unmergejoinable join clauses. So just take it out.
Tom Lane [Tue, 24 May 2005 16:45:23 +0000 (16:45 +0000)]
Add -I$(srcdir) to CPPFLAGS to make psqlscan.c compile in vpath builds.
Not sure why this hasn't been reported before; perhaps it is not needed
with newer gcc versions, but it definitely fails here.
Tatsuo Ishii [Tue, 24 May 2005 15:45:34 +0000 (15:45 +0000)]
Inserting 5 characters into char(10) does not produce 5 padding spaces
if they are two-byte multibyte characters. Same thing can be happen
if octet_length(multibyte_chars) == n where n is char(n).
Long standing bug since 7.3 days. Per report and fix from Yoshiyuki Asaba.
Tom Lane [Mon, 23 May 2005 03:01:14 +0000 (03:01 +0000)]
Avoid redundant relation lock grabs during planning, and make sure
that we acquire a lock on relations added to the query due to inheritance.
Formerly, no such lock was held throughout planning, which meant that
a schema change could occur to invalidate the plan before it's even
been completed.
Neil Conway [Mon, 23 May 2005 01:50:01 +0000 (01:50 +0000)]
Remove some verbiage describing how min() and max() are slow when applied
to the entire table: as of current sources, they are no longer slow
provided there is an index on the column.
Tom Lane [Sun, 22 May 2005 22:30:20 +0000 (22:30 +0000)]
Teach the planner to remove SubqueryScan nodes from the plan if they
aren't doing anything useful (ie, neither selection nor projection).
Also, extend to SubqueryScan the hacks already in place to avoid
unnecessary ExecProject calls when the result would just be the same
tuple the subquery already delivered. This saves some overhead in
UNION and other set operations, as well as avoiding overhead for
unflatten-able subqueries. Per example from Sokolov Yura.
Neil Conway [Sat, 21 May 2005 12:08:06 +0000 (12:08 +0000)]
Cleanup of GiST extensions in contrib/: now that we always invoke GiST
methods in a short-lived memory context, there is no need for GiST methods
to do their own manual (and error-prone) memory management.
Bruce Momjian [Fri, 20 May 2005 19:18:15 +0000 (19:18 +0000)]
Remove 2-phase description, because it isn't accurate anymore:
<
< This will involve adding a way to respond to commit failure by either
< taking the server into offline/readonly mode or notifying the
< administrator
Neil Conway [Fri, 20 May 2005 01:52:25 +0000 (01:52 +0000)]
Add some links to the CREATE FUNCTION reference page when describing
function definition for particular PLs. Original patch from David
Fetter, editorializing by Neil Conway.
Neil Conway [Fri, 20 May 2005 01:29:56 +0000 (01:29 +0000)]
Implement md5(bytea), update regression tests and documentation. Patch
from Abhijit Menon-Sen, minor editorialization from Neil Conway. Also,
improve md5(text) to allocate a constant-sized buffer on the stack
rather than via palloc.
Tom Lane [Thu, 19 May 2005 23:30:18 +0000 (23:30 +0000)]
Factor out lock cleanup code that is needed in several places in lock.c.
Also, remove the rather useless return value of LockReleaseAll. Change
response to detection of corruption in the shared lock tables to PANIC,
since that is the only way of cleaning up fully.
Originally an idea of Heikki Linnakangas, variously hacked on by
Alvaro Herrera and Tom Lane.
Tom Lane [Thu, 19 May 2005 21:35:48 +0000 (21:35 +0000)]
Split the shared-memory array of PGPROC pointers out of the sinval
communication structure, and make it its own module with its own lock.
This should reduce contention at least a little, and it definitely makes
the code seem cleaner. Per my recent proposal.
Neil Conway [Wed, 18 May 2005 05:01:10 +0000 (05:01 +0000)]
Upon closer inspection, Greg's psql tab completion patch leaks memory.
Fix the leak, and add a comment to note that the return value of
previous_word must be free'd.
Neil Conway [Wed, 18 May 2005 04:47:40 +0000 (04:47 +0000)]
psql tab completion improvements, from Greg Sabino Mullane:
* Made DELETE into "DELETE FROM"
* Moved ANALZYE to the end of the list to ease EXPLAIN / VACUUM
conflicts
* Removed the ANALYZE xx semicolon completion: we don't do that anywhere
else
* Add DECLARE support
* Add parens for DROP AGGREGATE
* Add "CASCADE | RESTRICT" for DROP xx
* Make EXPLAIN <tab> a lot smarter
* GROUP "BY" and ORDER "BY"
* "ISOLATION" becomes "ISOLATION LEVEL"
* Fix error in which REVOKE xx ON yy was receiving "TO", now gets "FROM"
* Add GRANT/REVOKE xx ON yy TO/FROM choices: usernames, GROUP, PUBLIC
* PREPARE xx <tab> AS "SELECT | INSERT | UPDATE | DELETE"
* Add = at end of UPDATE xx SET yy
* Beef up VACUUM stuff
Tom Lane [Tue, 17 May 2005 18:26:23 +0000 (18:26 +0000)]
Add a --dbname option to the pg_regress script, and use pl_regression
for testing PLs and contrib_regression for testing contrib, instead of
overwriting the core system's regression database as formerly done.
Andrew Dunstan
Neil Conway [Tue, 17 May 2005 03:34:18 +0000 (03:34 +0000)]
Cleanup GiST header files. Since GiST extensions are often written as
external projects, we should be careful about what parts of the GiST
API are considered implementation details, and which are part of the
public API. Therefore, I've moved internal-only declarations into
gist_private.h -- future backward-incompatible changes to gist.h should
be made with care, to avoid needlessly breaking external GiST extensions.
Also did some related header cleanup: remove some unnecessary #includes
from gist.h, and remove some unused definitions: isAttByVal(), _gistdump(),
and GISTNStrategies.
Neil Conway [Tue, 17 May 2005 00:59:30 +0000 (00:59 +0000)]
GiST improvements:
- make sure we always invoke user-supplied GiST methods in a short-lived
memory context. This means the backend isn't exposed to any memory leaks
that be in those methods (in fact, it is probably a net loss for most
GiST methods to bother manually freeing memory now). This also means
we can do away with a lot of ugly manual memory management in the
GiST code itself.
- keep the current page of a GiST index scan pinned, rather than doing a
ReadBuffer() for each tuple produced by the scan. Since ReadBuffer() is
expensive, this is a perf. win
- implement dead tuple killing for GiST indexes (which is easy to do, now
that we keep a pin on the current scan page). Now all the builtin indexes
implement dead tuple killing.
Tom Lane [Tue, 17 May 2005 00:43:47 +0000 (00:43 +0000)]
Modify tidbitmap.c to avoid creating a hash table until there is more
than one heap page represented in the bitmap. This is a bit ugly but
it cuts overhead fairly effectively in simple join cases. Per example
from Sergey Koposov.
Tom Lane [Sun, 15 May 2005 21:19:55 +0000 (21:19 +0000)]
Fix latent bug in ExecSeqRestrPos: it leaves the plan node's result slot
in an inconsistent state. (This is only latent because in reality
ExecSeqRestrPos is dead code at the moment ... but someday maybe it won't
be.) Add some comments about what the API for plan node mark/restore
actually is, because it's not immediately obvious.
Tom Lane [Sat, 14 May 2005 20:29:13 +0000 (20:29 +0000)]
Minor speed hacks in AllocSetReset: avoid clearing the freelist headers
when the blocks list is empty (there can surely be no freelist items if
the context contains no memory), and use MemSetAligned not MemSet to
clear the headers (we assume alignof(pointer) >= alignof(int32)).
Per discussion with Atsushi Ogawa. He proposes some further hacking
that I'm not yet sold on, but these two changes are unconditional wins
since there is no case in which they make things slower.