Bruce Momjian [Thu, 17 Mar 2005 04:42:15 +0000 (04:42 +0000)]
Add mention of compatibility problem with turning off backslash escapes:
< SQL-spec compliant, so allow such handling to be disabled.
> SQL-spec compliant, so allow such handling to be disabled. However,
> disabling backslashes could break many third-party applications and tools.
Neil Conway [Wed, 16 Mar 2005 23:52:18 +0000 (23:52 +0000)]
This patch makes \d on tables and views sort fk constraints, triggers
and rules alphabetically in the output. This makes it the same as
for indexes and stops the irritating random or reverse ordering it
currently has.
Tom Lane [Wed, 16 Mar 2005 21:38:10 +0000 (21:38 +0000)]
Revise TupleTableSlot code to avoid unnecessary construction and disassembly
of tuples when passing data up through multiple plan nodes. A slot can now
hold either a normal "physical" HeapTuple, or a "virtual" tuple consisting
of Datum/isnull arrays. Upper plan levels can usually just copy the Datum
arrays, avoiding heap_formtuple() and possible subsequent nocachegetattr()
calls to extract the data again. This work extends Atsushi Ogawa's earlier
patch, which provided the key idea of adding Datum arrays to TupleTableSlots.
(I believe however that something like this was foreseen way back in Berkeley
days --- see the old comment on ExecProject.) A test case involving many
levels of join of fairly wide tables (about 80 columns altogether) showed
about 3x overall speedup, though simple queries will probably not be
helped very much.
I have also duplicated some code in heaptuple.c in order to provide versions
of heap_formtuple and friends that use "bool" arrays to indicate null
attributes, instead of the old convention of "char" arrays containing either
'n' or ' '. This provides a better match to the convention used by
ExecEvalExpr. While I have not made a concerted effort to get rid of uses
of the old routines, I think they should be deprecated and eventually removed.
Neil Conway [Wed, 16 Mar 2005 00:02:39 +0000 (00:02 +0000)]
Wrap the implementation of fork_process() inside #ifndef WIN32 -- this
should hopefully unbreak the Win32 build. Apologies for breaking it in
the first place.
Bruce Momjian [Tue, 15 Mar 2005 04:01:21 +0000 (04:01 +0000)]
Update Win32 item:
< o Disallow encodings like UTF8 which PostgreSQL supports
< but the operating system does not (already disallowed by
< pginstaller)
> o Add support for Unicode
< To fix UTF8, the data needs to be converted to UTF16 and then
< the Win32 wcscoll() can be used, and perhaps other functions
> To fix this, the data needs to be converted to/from UTF16/UTF8
> so the Win32 wcscoll() can be used, and perhaps other functions
< locales but provides no ordering.
<
> locales but provides no ordering or character set classes.
Tom Lane [Mon, 14 Mar 2005 18:57:33 +0000 (18:57 +0000)]
Make pg_dump emit a useful error message, instead of just dumping core,
if it finds a pg_rewrite entry for which there is no pg_class entry.
Per report from Andrew Slobodyanyk.
Tom Lane [Mon, 14 Mar 2005 06:19:01 +0000 (06:19 +0000)]
Note that the -F and -R command line options only affect unaligned
output mode. This was already stated in other places in the psql
reference page, but not here.
Tom Lane [Mon, 14 Mar 2005 04:41:13 +0000 (04:41 +0000)]
Avoid O(N^2) overhead in repeated nocachegetattr calls when columns of
a tuple are being accessed via ExecEvalVar and the attcacheoff shortcut
isn't usable (due to nulls and/or varlena columns). To do this, cache
Datums extracted from a tuple in the associated TupleTableSlot.
Also some code cleanup in and around the TupleTable handling.
Atsushi Ogawa with some kibitzing by Tom Lane.
Neil Conway [Mon, 14 Mar 2005 00:19:37 +0000 (00:19 +0000)]
Allow ALTER FUNCTION to change a function's strictness, volatility, and
whether or not it is a security definer. Changing a function's strictness
is required by SQL2003, and the other capabilities make sense. Also, allow
an optional RESTRICT noise word to be specified, for SQL conformance.
Some trivial regression tests added and the documentation has been
updated.
Neil Conway [Sun, 13 Mar 2005 09:36:31 +0000 (09:36 +0000)]
Make default_with_oids default to false -- user-created tables will now
no longer include OIDs, unless WITH OIDS is specified or the
default_with_oids configuration parameter is enabled. Update the docs
accordingly.
Tom Lane [Sat, 12 Mar 2005 21:33:55 +0000 (21:33 +0000)]
When cloning template0 (or other fully-frozen databases), set the new
database's datallowconn and datfrozenxid to the current transaction ID
instead of copying the source database's values. This is OK because we
assume the source DB contains no normal transaction IDs whatsoever.
This keeps VACUUM from immediately starting to complain about unvacuumed
databases in the situation where we are more than 2 billion transactions
out from the XID stamp of template0. Per discussion with Milen Radev
(although his complaint turned out to be due to something else, but the
problem is real anyway).
Tom Lane [Sat, 12 Mar 2005 21:11:50 +0000 (21:11 +0000)]
Fix ALTER DATABASE RENAME to allow the operation if user is a superuser
who for some reason isn't marked usecreatedb. Per report from Alexander
Pravking. Also fix sloppy coding in have_createdb_privilege().
Tom Lane [Sat, 12 Mar 2005 20:25:06 +0000 (20:25 +0000)]
Adjust the API for aggregate function calls so that a C-coded function
can tell whether it is being used as an aggregate or not. This allows
such a function to avoid re-pallocing a pass-by-reference transition
value; normally it would be unsafe for a function to scribble on an input,
but in the aggregate case it's safe to reuse the old transition value.
Make int8inc() do this. This gets a useful improvement in the speed of
COUNT(*), at least on narrow tables (it seems to be swamped by I/O when
the table rows are wide). Per a discussion in early December with
Neil Conway. I also fixed int_aggregate.c to check this, thereby
turning it into something approaching a supportable technique instead
of being a crude hack.
Neil Conway [Sat, 12 Mar 2005 06:53:54 +0000 (06:53 +0000)]
Some builds (depends on crypto engine support?) of OpenSSL
0.9.7x have EVP_DigestFinal function which which clears all of
EVP_MD_CTX. This makes pgcrypto crash in functions which
re-use one digest context several times: hmac() and crypt()
with md5 algorithm.
Following patch fixes it by carring the digest info around
EVP_DigestFinal and re-initializing cipher.
Tom Lane [Sat, 12 Mar 2005 01:54:44 +0000 (01:54 +0000)]
Fix problem with infinite recursion between write_syslogger_file and
elog if the former has trouble writing its file. Code review for
Magnus' patch to redirect stderr to syslog on Windows (Bruce's version
seems right, but did some minor prettification).
Tom Lane [Thu, 10 Mar 2005 23:21:26 +0000 (23:21 +0000)]
Make the behavior of HAVING without GROUP BY conform to the SQL spec.
Formerly, if such a clause contained no aggregate functions we mistakenly
treated it as equivalent to WHERE. Per spec it must cause the query to
be treated as a grouped query of a single group, the same as appearance
of aggregate functions would do. Also, the HAVING filter must execute
after aggregate function computation even if it itself contains no
aggregate functions.
Neil Conway [Thu, 10 Mar 2005 07:14:03 +0000 (07:14 +0000)]
Refactor fork()-related code. We need to do various housekeeping tasks
before we can invoke fork() -- flush stdio buffers, save and restore the
profiling timer on Linux with LINUX_PROFILE, and handle BeOS stuff. This
patch moves that code into a single function, fork_process(), instead of
duplicating it at the various callsites of fork().
This patch doesn't address the EXEC_BACKEND case; there is room for
further cleanup there.
Tom Lane [Mon, 7 Mar 2005 04:42:17 +0000 (04:42 +0000)]
Adjust creation/destruction of TupleDesc data structure to reduce the
number of palloc calls. This has a salutory impact on plpgsql operations
with record variables (which create and destroy tupdescs constantly)
and probably helps a bit in some other cases too.
Tom Lane [Sun, 6 Mar 2005 22:15:05 +0000 (22:15 +0000)]
Revise hash join code so that we can increase the number of batches
on-the-fly, and thereby avoid blowing out memory when the planner has
underestimated the hash table size. Hash join will now obey the
work_mem limit with some faithfulness. Per my recent proposal
(hash aggregate part isn't done yet though).
Tom Lane [Fri, 4 Mar 2005 20:21:07 +0000 (20:21 +0000)]
Replace the BufMgrLock with separate locks on the lookup hashtable and
the freelist, plus per-buffer spinlocks that protect access to individual
shared buffer headers. This requires abandoning a global freelist (since
the freelist is a global contention point), which shoots down ARC and 2Q
as well as plain LRU management. Adopt a clock sweep algorithm instead.
Preliminary results show substantial improvement in multi-backend situations.
Tom Lane [Wed, 2 Mar 2005 04:10:53 +0000 (04:10 +0000)]
Another go at making pred_test() handle all reasonable combinations
of AND and OR clauses. The key point here is that an OR on the
predicate side has to be treated gingerly: we may be able to prove
that the OR is implied even when no one of its components is implied.
For example (x OR y) implies (x OR y OR z) even though no one of x,
y, or z can be individually proven. This code handles both the
example shown recently by Sergey Koshcheyev and the one shown last
October by Dawid Kuroczko.