Tom Lane [Sun, 19 Sep 1999 17:20:58 +0000 (17:20 +0000)]
Remove incorrect 'Assert(targetList != NULL)'. An
INSERT ... DEFAULT VALUES statement does indeed have a null targetlist,
at least during parse and rewrite stages.
Tom Lane [Sat, 18 Sep 1999 19:10:19 +0000 (19:10 +0000)]
Update regress test expected outputs for small changes in
error message wording, due to most cases of no-such-relation now being
detected in central heap_open code rather than on an ad-hoc basis.
Tom Lane [Sat, 18 Sep 1999 19:08:25 +0000 (19:08 +0000)]
Mega-commit to make heap_open/heap_openr/heap_close take an
additional argument specifying the kind of lock to acquire/release (or
'NoLock' to do no lock processing). Ensure that all relations are locked
with some appropriate lock level before being examined --- this ensures
that relevant shared-inval messages have been processed and should prevent
problems caused by concurrent VACUUM. Fix several bugs having to do with
mismatched increment/decrement of relation ref count and mismatched
heap_open/close (which amounts to the same thing). A bogus ref count on
a relation doesn't matter much *unless* a SI Inval message happens to
arrive at the wrong time, which is probably why we got away with this
sloppiness for so long. Repair missing grab of AccessExclusiveLock in
DROP TABLE, ALTER/RENAME TABLE, etc, as noted by Hiroshi.
Recommend 'make clean all' after pulling this update; I modified the
Relation struct layout slightly.
Will post further discussion to pghackers list shortly.
Hiroshi Inoue wrote:
>
> Hello Vadim,
>
> I have a question about vacuum.
>
> VACUUM has a phase like commit which calls TransactionIdCommit().
> But if elog(ERROR) occured after that,the status of transaction is
> changed from XID_COMMIT to XID_ABORT.
>
> Seems to me this causes inconsistency.
> Shoudn't AbortTransaction() be changed not to call TransacionIdAbort()
> in case of vacuum.
Allow CASE statement to contain *only* untyped result clauses or nulls.
Almost worked before, but forgot one place to check.
Reported by Tatsuo Ishii.
Still does not do the right thing if inserting into a non-string target
column. Should look for a type coersion later, but doesn't.
Tom Lane [Mon, 13 Sep 1999 03:00:19 +0000 (03:00 +0000)]
Patch to cure O(N^2) behavior in libpq when reading a long
message under a kernel that only returns one packet per recv() call. This
didn't use to matter much, but it starts to get annoying with multi-megabyte
EXPLAIN VERBOSE responses...
Tom Lane [Mon, 13 Sep 1999 00:17:25 +0000 (00:17 +0000)]
First cut at doing something reasonable with OR-of-ANDs WHERE
conditions. There are some pretty bogus heuristics in prepqual.c that
try to decide whether to output CNF or DNF format; they need to be replaced,
likely. Right now the code is probably too willing to choose DNF form,
which might hurt performance in some cases that used to work OK.
But at least we have a foundation to build on.
Tom Lane [Sun, 12 Sep 1999 18:08:17 +0000 (18:08 +0000)]
Further improvements in cnfify: reduce amount of self-recursion
in or_normalize, remove detection of duplicate subexpressions (since it's
highly unlikely to be worth the amount of time it takes), and introduce
a dnfify() entry point so that unintelligible backwards logic in UNION
processing can be eliminated. This is just an intermediate step ---
next thing is to look at not forcing the qual into CNF form when it would
be better off in DNF form.
Tom Lane [Sat, 11 Sep 1999 22:28:11 +0000 (22:28 +0000)]
Eliminate query length limitation imposed by pg_client_to_server
and pg_server_to_client. Eliminate copy.c's restriction on the length
of a single attribute.
Tom Lane [Sat, 11 Sep 1999 22:02:51 +0000 (22:02 +0000)]
Update protocol doc to emphasize that backend can return
arbitrarily long strings, and frontends should cope gracefully. Goes
along with long query and long error message changes...
Tom Lane [Sat, 11 Sep 1999 19:06:42 +0000 (19:06 +0000)]
Eliminate elog()'s hardwired limit on length of an error message.
This change seems necessary in conjunction with long queries, and it
cleans up some bogosity in connection with long EXPLAIN texts anyway.
Note that current libpq will accept any length error message (at least
until it runs out of memory); prior versions have a limit of 8K, but
will cleanly discard excess error text, so there shouldn't be any
big compatibility problems with old clients.
Tom Lane [Thu, 9 Sep 1999 16:25:35 +0000 (16:25 +0000)]
Repair incorrect cleanup of heap memory allocation during
transaction abort --- before it only worked if there was exactly one level
of allocation context stacked in the blank portal. Now it does the right
thing for any depth, including zero...
Tom Lane [Thu, 9 Sep 1999 02:36:04 +0000 (02:36 +0000)]
Repair error noticed by Roberto Cornacchia: selectivity code
was rejecting negative attnums as bogus, which of course they are not.
Add code to get_attdisbursion to produce a useful value for OID attribute,
since VACUUM does not store stats for system attributes.
Also, repair bug that's been in eqjoinsel for a long time: it was taking
the max of the two columns' disbursions, whereas it should use the min.
Tom Lane [Tue, 7 Sep 1999 03:47:06 +0000 (03:47 +0000)]
Performance improvements in cnfify(): get rid of exponential
space consumption in pull_args, and avoid doing the full CNF transform on
operands of operator clauses, where it's really not particularly helpful.
This answers the TODO item about large numbers of OR clauses, at least
partially. I was able to do a ten-thousand-OR-clause query with about
20Mb memory consumption ... it took an obscenely long time, but it worked...
Tom Lane [Tue, 7 Sep 1999 00:13:27 +0000 (00:13 +0000)]
Mike Ansley's fixes for long queries. This change just
corrects flex myinput() routine so that it doesn't assume there is only
one bufferload of data. We still have the issue of getting rid of
YY_USES_REJECT so that the scanner can cope with tokens larger than its
initial buffer size.
Tom Lane [Mon, 6 Sep 1999 21:16:20 +0000 (21:16 +0000)]
Mark bpchareq not hashjoinable, since it strips trailing blanks
before comparison; if fields being joined are different widths then hashing
will yield wrong answer. Also, remove hashjoinable mark from all uses of
array_eq, because array structures may have padding bytes between elements
and the pad bytes are of uncertain content. This could be revisited if
array code is cleaned up.
Modify opr_sanity regress test to complain if array_eq operator is marked
hashjoinable.
Tom Lane [Mon, 6 Sep 1999 19:37:38 +0000 (19:37 +0000)]
I finally understood what sinvaladt.c is doing --- and it
offended my aesthestic sensibility that there was so much unreadable code
doing so little. Rewritten code is about half the size, faster, and
(I hope) much more intelligible.
Tom Lane [Mon, 6 Sep 1999 19:33:16 +0000 (19:33 +0000)]
Fix relcache.c so that local relations (those created during
current transaction) are not flushed by shared-cache-inval reset message.
SI reset actually works now, for probably the first time in a long time.
I was able to run initdb and regression tests with a 16-element SI message
array, with a lot of NOTICE: cache state reset messages but no crashes.
Tom Lane [Mon, 6 Sep 1999 18:13:02 +0000 (18:13 +0000)]
RelationCacheInvalidate thought there were 7 nailed-in-cache
system tables, but actually there are only 6 --- see RelationInitialize.
Kinda makes you wonder how long ago this code was last executed...
Tom Lane [Sun, 5 Sep 1999 23:24:53 +0000 (23:24 +0000)]
Per Tatsuo's recommendation, change mdopen so that it won't
automatically create the file, except during bootstrap mode where that
seems to be necessary.
Tom Lane [Sun, 5 Sep 1999 17:43:47 +0000 (17:43 +0000)]
Disallow DROP TABLE/DROP INDEX inside a transaction block.
We can't support these properly, since once the relation's physical files
are unlinked, there's no way to roll back the transaction. I suppose
we could postpone the unlink till transaction commit, but then what of
BEGIN; DROP TABLE foo; CREATE TABLE foo; ?
The code does allow dropping a table/index created in the current
transaction block, however, since the post-abort state would be that
the table doesn't exist anyway.
Tom Lane [Sat, 4 Sep 1999 21:47:23 +0000 (21:47 +0000)]
In RelationNameGetRelation(), replace temp table name by
real name before doing lookup. We only want to index temp tables by their
real names in the relcache, to ensure there's not more than one relcache
entry for them.
Tom Lane [Sat, 4 Sep 1999 18:42:15 +0000 (18:42 +0000)]
Modify RelationFlushRelation so that if the relcache entry
has positive refcount, it is rebuilt from pg_class data. This ensures
that relcache entries will track changes made by other backends. Formerly,
a shared inval report would just be ignored if it happened to arrive while
the relcache entry was in use. Also, fix relcache to reset ref counts
to zero during transaction abort. Finally, change LockRelation() so that
it checks for shared inval reports after obtaining the lock. In this way,
once any kind of lock has been obtained on a rel, we can trust the relcache
entry to be up-to-date.
Tom Lane [Sat, 4 Sep 1999 18:36:45 +0000 (18:36 +0000)]
Modify sinval so that InvalidateSharedInvalid() does not hold
the SInval spinlock while it is calling the passed invalFunction or
resetFunction. This is necessary to avoid deadlock with lmgr change;
InvalidateSharedInvalid can be called recursively now. It should be
a good performance improvement anyway --- holding a spinlock for more
than a very short interval is a no-no.
Fix for perl5 on BSD/OS breaks most other platforms, so back it out.
istm that this would be a job for configure.
Most modern OSes actually use perl5 by default ;)
Tom Lane [Thu, 2 Sep 1999 02:57:50 +0000 (02:57 +0000)]
Repair a bunch of problems in md.c. This builds on Hiroshi's
insight that RelationFlushRelation ought to invoke smgrclose, and that the
way to make that work is to ensure that mdclose doesn't fail if the relation
is already closed (or unlinked, if we are looking at a DROP TABLE). While
I was testing that, I was able to identify several problems that we had
with multiple-segment relations. The system is now able to do initdb and
pass the regression tests with a very small segment size (I had it set to
64Kb per segment for testing). I don't believe that ever worked before.
File descriptor leaks seem to be gone too.
I have partially addressed the concerns we had about mdtruncate(), too.
On a Win32 or NFS filesystem it is not possible to unlink a file that
another backend is holding open, so what md.c now does is to truncate
unwanted files to zero length before trying to unlink them. The other
backends will be forced to close their open files by relation cache
invalidation --- but I think it would take considerable work to make
that happen before vacuum truncates the relation rather than after.
Leaving zero-length files lying around seems a usable compromise.
Fix wording on allowed/forbidden keyword usage.
Thanks to Michael Deck <deckm@cleansoft.com> for the tipoff.
Add more examples for language components.
Tom Lane [Tue, 31 Aug 1999 01:37:37 +0000 (01:37 +0000)]
Update frontend libpq to remove limits on query lengths,
error/notice message lengths, and number of fields per tuple. Add
pqexpbuffer.c/.h, a frontend version of backend's stringinfo module.
This is first step in applying Mike Ansley's long-query patches,
even though he didn't do any of these particular changes...
Tom Lane [Sun, 29 Aug 1999 01:35:11 +0000 (01:35 +0000)]
Correct broken entries for pg_proc OIDs 1364 (time(abstime))
and 1370 (timestamp(datetime)). This does not force an initdb, exactly,
but you won't see the effects of the bug fix until you do one.
BTW, OID 1358 for timespan(time) is still broken:
select timespan('21:11:26'::time);
ERROR: No such function 'time_timespan' with the specified attributes
But I couldn't figure out what it ought to be defined as, so I left it be.
Tom Lane [Sat, 28 Aug 1999 03:59:05 +0000 (03:59 +0000)]
Fix several problems in rule deparsing: didn't handle array
references or CASE expressions, didn't parenthesize complex expressions
properly. Also, always output variable references as fully qualified
names to eliminate ambiguity bug recently reported. (This could be
smarter, but reliability comes first.)
Tom Lane [Thu, 26 Aug 1999 05:09:06 +0000 (05:09 +0000)]
Clean up some mistakes in handling of uplevel Vars in planner.
Most parts of the planner should ignore, or indeed never even see, uplevel
Vars because they will be or have been replaced by Params. There were a
couple of places that got it wrong though, probably my fault from recent
changes...
Tom Lane [Thu, 26 Aug 1999 04:59:15 +0000 (04:59 +0000)]
Clean up some bugs in oper_select_candidate(), notably the
last loop which would return the *first* surviving-to-that-point candidate
regardless of which one actually passed the test. This was producing
such curious results as 'oid % 2' getting translated to 'int2(oid) % 2'.
Tom Lane [Wed, 25 Aug 1999 23:21:43 +0000 (23:21 +0000)]
Revise implementation of SubLinks so that there is a consistent,
documented intepretation of the lefthand and oper fields. Fix a number of
obscure problems while at it --- for example, the old code failed if the parser
decided to insert a type-coercion function just below the operator of a
SubLink.
CAUTION: this will break stored rules that contain subplans. You may
need to initdb.
Tatsuo Ishii [Wed, 25 Aug 1999 12:18:31 +0000 (12:18 +0000)]
Add new vpl_num_allocated_pages member to VPageListData.
It will keep track the number of pages allocated so that
vacuum could allocate twice of the previous allocation.
This will greatly reduce the total memory consumption of
vacuum.
Tom Lane [Tue, 24 Aug 1999 20:11:19 +0000 (20:11 +0000)]
Alter AllocSet routines so that requests larger than
ALLOC_BIGCHUNK_LIMIT are always allocated as separate malloc() blocks,
and are free()d immediately upon pfree(). Also, if such a chunk is enlarged
with repalloc(), translate the operation into a realloc() so as to
minimize memory usage. Of course, these large chunks still get freed
automatically if the alloc set is reset.
I have set ALLOC_BIGCHUNK_LIMIT at 64K for now, but perhaps another
size would be better?
Tom Lane [Tue, 24 Aug 1999 00:09:56 +0000 (00:09 +0000)]
coerce_type() failed to guard against trying to convert a NULL
constant to a different type. Not sure that this could happen in ordinary
parser usage, but it can in some new code I'm working on...
Tom Lane [Mon, 23 Aug 1999 23:48:39 +0000 (23:48 +0000)]
Remove bogus code in oper_exact --- if it didn't find an exact
match then it tried for a self-commutative operator with the reversed input
data types. This is pretty silly; there could never be such an operator,
except maybe in binary-compatible-type scenarios, and we have oper_inexact
for that. Besides which, the oprsanity regress test would complain about
such an operator. Remove nonfunctional code and simplify routine calling
convention accordingly.
Tom Lane [Sun, 22 Aug 1999 20:15:04 +0000 (20:15 +0000)]
Further planner/optimizer cleanups. Move all set_tlist_references
and fix_opids processing to a single recursive pass over the plan tree
executed at the very tail end of planning, rather than haphazardly here
and there at different places. Now that tlist Vars do not get modified
until the very end, it's possible to get rid of the klugy var_equal and
match_varid partial-matching routines, and just use plain equal()
throughout the optimizer. This is a step towards allowing merge and
hash joins to be done on expressions instead of only Vars ...
Tom Lane [Sat, 21 Aug 1999 03:49:17 +0000 (03:49 +0000)]
Major revision of sort-node handling: push knowledge of query
sort order down into planner, instead of handling it only at the very top
level of the planner. This fixes many things. An explicit sort is now
avoided if there is a cheaper alternative (typically an indexscan) not
only for ORDER BY, but also for the internal sort of GROUP BY. It works
even when there is no other reason (such as a WHERE condition) to consider
the indexscan. It works for indexes on functions. It works for indexes
on functions, backwards. It's just so cool...
CAUTION: I have changed the representation of SortClause nodes, therefore
THIS UPDATE BREAKS STORED RULES. You will need to initdb.
Tom Lane [Sat, 21 Aug 1999 03:06:58 +0000 (03:06 +0000)]
Cleanups for int8: guard against null inputs in comparison
operators (and some other places), fix rangechecks in int8 to int4
conversion (same problem we recently figured out in pg_atoi).
Tom Lane [Wed, 18 Aug 1999 04:15:16 +0000 (04:15 +0000)]
Remove extraneous SeqScan node that make_noname was inserting
above a Sort or Materialize node. As far as I can tell, the only place
that actually needed that was set_tlist_references, which was being lazy
about checking to see if it had a noname node to fix or not...
Tom Lane [Tue, 17 Aug 1999 21:21:22 +0000 (21:21 +0000)]
Add script that runs the regression tests with all valid
combinations of query-plan-type backend options. Good for testing
planner/optimizer. Tedious, though.