representation of equivalence classes of variables. This is an extensive
rewrite, but it brings a number of benefits:
* planner no longer fails in the presence of "incomplete" operator families
that don't offer operators for every possible combination of datatypes.
* avoid generating and then discarding redundant equality clauses.
* remove bogus assumption that derived equalities always use operators
named "=".
* mergejoins can work with a variety of sort orders (e.g., descending) now,
instead of tying each mergejoinable operator to exactly one sort order.
* better recognition of redundant sort columns.
* can make use of equalities appearing underneath an outer join.
currentMarkData from IndexScanDesc to the opaque structs for the
AMs that need this information (currently gist and hash).
Patch from Heikki Linnakangas, fixes by Neil Conway.
is deleted. A backend about to unlink a file now sends a "revoke fsync"
request to the bgwriter to make it clean out pending fsync requests. There
is still a race condition where the bgwriter may try to fsync after the unlink
has happened, but we can resolve that by rechecking the fsync request queue
to see if a revoke request arrived meanwhile. This eliminates the former
kluge of "just assuming" that an ENOENT failure is okay, and lets us handle
the fact that on Windows it might be EACCES too without introducing any
questionable assumptions. After an idea of mine improved by Magnus.
The HEAD patch doesn't apply cleanly to 8.2, but I'll see about a back-port
later. In the meantime this could do with some testing on Windows; I've been
able to force it through the code path via ENOENT, but that doesn't prove that
it actually fixes the Windows problem ...
The implementation is somewhat ugly logic-wise, but I don't see an
easy way to make it more concise.
When writing this, I noticed that my previous implementation of
width_bucket() doesn't handle NaN correctly:
postgres=# select width_bucket('NaN', 1, 5, 5);
width_bucket
--------------
6
(1 row)
AFAICS SQL:2003 does not define a NaN value, so it doesn't address how
width_bucket() should behave here. The patch changes width_bucket() so
that ereport(ERROR) is raised if NaN is specified for the operand or the
lower or upper bounds to width_bucket(). For float8, NaN is disallowed
for any of the floating-point inputs, and +/- infinity is disallowed
for the histogram bounds (but allowed for the operand).
Update docs and regression tests, bump the catversion.
it was checking a pg_constraint OID instead of pg_class OID, resulting in
"relation with OID nnnnn does not exist" failures for anyone who wasn't
owner of the table being examined. Per bug #2848 from Laurence Rowe.
Note: for existing 8.2 installations a simple version update won't fix this;
the easiest fix is to CREATE OR REPLACE this view with the corrected
definition.
accessing it, like DROP DATABASE. This allows the regression tests to pass
with autovacuum enabled, which open the gates for finally enabling autovacuum
by default.
standard convention the 21st century runs from 2001-2100, not 2000-2099,
so make it work like that. Per bug #2885 from Akio Iwaasa.
Backpatch to 8.2, but no further, since this is really a definitional
change; users of older branches are probably more interested in stability.
hold true for operators in a btree operator family. This is mostly to
clarify my own thinking about what the planner can assume for optimization
purposes. (blowing dust off an old abstract-algebra textbook...)
(or other types of pg_class entry): the function pgstat_vacuum_tabstat,
invoked during VACUUM startup, had runtime proportional to the number of
stats table entries times the number of pg_class rows; in other words
O(N^2) if the stats collector's information is reasonably complete.
Replace list searching with a hash table to bring it back to O(N)
behavior. Per report from kim at myemma.com.
Back-patch as far as 8.1; 8.0 and before use different coding here.
which comparison operators to use for plan nodes involving tuple comparison
(Agg, Group, Unique, SetOp). Formerly the executor looked up the default
equality operator for the datatype, which was really pretty shaky, since it's
possible that the data being fed to the node is sorted according to some
nondefault operator class that could have an incompatible idea of equality.
The planner knows what it has sorted by and therefore can provide the right
equality operator to use. Also, this change moves a couple of catalog lookups
out of the executor and into the planner, which should help startup time for
pre-planned queries by some small amount. Modify the planner to remove some
other cavalier assumptions about always being able to use the default
operators. Also add "nulls first/last" info to the Plan node for a mergejoin
--- neither the executor nor the planner can cope yet, but at least the API is
in place.
management. The paper clearly describes many of the ideas embodied in
our current hashing code, but as far as I could find out there is not
a direct code heritage. (Mike Olsen recalls discussion of this paper
at Postgres meetings but believes it "informed the Postgres implementation
probably just at the design level". Margo herself says she wasn't
involved with Postgres' hash code.) Credit where credit is due 'n all
that, even if fifteen years after the fact.
per-column options for btree indexes. The planner's support for this is still
pretty rudimentary; it does not yet know how to plan mergejoins with
nondefault ordering options. The documentation is pretty rudimentary, too.
I'll work on improving that stuff later.
Note incompatible change from prior behavior: ORDER BY ... USING will now be
rejected if the operator is not a less-than or greater-than member of some
btree opclass. This prevents less-than-sane behavior if an operator that
doesn't actually define a proper sort ordering is selected.
when collapsing of JOIN trees is stopped by join_collapse_limit. For instance
a list of 11 LEFT JOINs with limit 8 now produces something like
((1 2 3 4 5 6 7 8) 9 10 11 12)
instead of
(((1 2 3 4 5 6 7 8) (9)) 10 11 12)
The latter structure is really only required for a FULL JOIN.
Noted while studying an example from Shane Ambler.
hash joins with the estimated-larger relation on the inside. There are
several cases where doing that makes perfect sense, and in cases where it
doesn't, the regular cost computation really ought to be able to figure that
out. Make some marginal tweaks in said computation to try to get results
approximating reality a bit better. Per an example from Shane Ambler.
Also, fix an oversight in the original patch to add seq_page_cost: the costs
of spilling a hash join to disk should be scaled by seq_page_cost.
sets the items, and serializes the value back (rather than adding an
arbitrary number of XML preambles as before).
The libxml memory management via palloc had to be disabled because it
crashes when libxml tries to access memory that was helpfully freed
earlier by PostgreSQL. This needs further thought.
an optarg). Add some comments noting that code in three different files has
to be kept in sync. Fix erroneous description of -S switch (it sets work_mem
not silent_mode), and do some light copy-editing elsewhere in postgres-ref.
form '^(foo)$'. Before, these could never be optimized into indexscans.
The recent changes to make psql and pg_dump generate such patterns (for \d
commands and -t and related switches, respectively) therefore represented
a big performance hit for people with large pg_class catalogs, as seen in
recent gripe from Erik Jones. While at it, be more paranoid about
case-sensitivity checking in multibyte encodings, and fix some other
corner cases in which a regex might be interpreted too liberally.
having md.c return a success/failure boolean to smgr.c, which was just going
to elog anyway, let md.c issue the elog messages itself. This allows better
error reporting, particularly in cases such as "short read" or "short write"
which Peter was complaining of. Also, remove the kluge of allowing mdread()
to return zeroes from a read-beyond-EOF: this is now an error condition
except when InRecovery or zero_damaged_pages = true. (Hash indexes used to
require that behavior, but no more.) Also, enforce that mdwrite() is to be
used for rewriting existing blocks while mdextend() is to be used for
extending the relation EOF. This restriction lets us get rid of the old
ad-hoc defense against creating huge files by an accidental reference to
a bogus block number: we'll only create new segments in mdextend() not
mdwrite() or mdread(). (Again, when InRecovery we allow it anyway, since
we need to allow updates of blocks that were later truncated away.)
Also, clean up the original makeshift patch for bug #2737: move the
responsibility for padding relation segments to full length into md.c.
The purpose is to allow autovacuum-esq conditional vacuuming and
clustering using SQL to discover the required stats.
No documentation updates required. Catalog version updated.
Glen Parker
valid result from a computation if one of the input values was infinity.
The previous code assumed an operation that returned infinity was an
overflow.
Handle underflow/overflow consistently, and add checks for aggregate
overflow.
Consistently prevent Inf/Nan from being cast to integer data types.
Fix INT_MIN % -1 to prevent overflow.
Update regression results for new error text.
Per report from Roman Kononov.
pg_opclass during LookupOpclassInfo(), I'd turned pg_opclass_oid_index
into a critical system index. However the problem could only manifest
during a backend's first attempt to load opclass data, and then only
if it had successfully loaded pg_internal.init and subsequently received
a relcache flush; which made it impossible to reproduce in sequential
tests and darn hard even in parallel tests. Memo to self: when
exercising cache flush scenarios, must disable LookupOpclassInfo's
internal cache too.
or contradictory keys even in cross-data-type scenarios. This is another
benefit of the opfamily rewrite: we can find the needed comparison
operators now.
of increasing size, instead of one at a time. This reduces the memory
management overhead when num_temp_buffers is large: in the previous coding
we would actually waste 50% of the space used for temp buffers, because aset.c
would round the individual requests up to 16K. Problem noted while studying
a performance issue reported by Steven Flatt.
Back-patch as far as 8.1 --- older versions used few enough local buffers
that the issue isn't significant for them.
has a small maxBlockSize: the maximum request size that we will treat as a
"chunk" needs to be limited to fit in maxBlockSize. Otherwise we will round
up the request size to the next power of 2, wasting space, which is a bit
pointless if we aren't going to make the blocks big enough to fit additional
stuff in them. The example motivating this is local buffer management, which
makes repeated allocations of 8K (one BLCKSZ buffer) in TopMemoryContext,
which has maxBlockSize = 8K because for the most part allocations there are
small. This leads to each local buffer actually eating 16K of space, which
adds up when there are thousands of them. I intend to change localbuf.c to
aggregate its requests, which will prevent this particular misbehavior, but
it seems likely that similar scenarios could arise elsewhere, so fixing the
core problem seems wise as well.
involving HashAggregate over SubqueryScan (this is the known case, there
may well be more). The bug is only latent in releases before 8.2 since they
didn't try to access tupletable slots' descriptors during ExecDropTupleTable.
The least bogus fix seems to be to make subqueries share the parent query's
memory context, so that tupdescs they create will have the same lifespan as
those of the parent query. There are comments in the code envisioning going
even further by not having a separate child EState at all, but that will
require rethinking executor access to range tables, which I don't want to
tackle right now. Per bug report from Jean-Pierre Pelletier.
ps_TupFromTlist in plan nodes that make use of it. This was being done
correctly in join nodes and Result nodes but not in any relation-scan nodes.
Bug would lead to bogus results if a set-returning function appeared in the
targetlist of a subquery that could be rescanned after partial execution,
for example a subquery within EXISTS(). Bug has been around forever :-(
... surprising it wasn't reported before.
were marked canSetTag. While it's certainly correct to return the result
of the last one that is marked canSetTag, it's less clear what to do when
none of them are. Since plpgsql will complain if zero is returned, the
8.2.0 behavior isn't good. I've fixed it to restore the prior behavior of
returning the physically last query's result code when there are no
canSetTag queries.
Use a TRY block instead of (inadequate) ad-hoc coding to ensure that
libxml is cleaned up after a failure. Report the intended SQLCODE
instead of defaulting to XX000. Avoid risking use of a dangling
pointer by keeping the persistent error buffer in TopMemoryContext.
Be less trusting that error messages don't contain %.
This patch doesn't do anything about changing the way the messages
are put together --- this is just about mechanism.
bletcherous and unsafe manipulation of global encoding setting.
Clean up libxml reporting mechanism a bit (it still looks like a
dangling-pointer crash waiting to happen, though, not to mention
being far less than sane from a localization standpoint).
the XmlExpr code in various lists, use a representation that has some hope
of reverse-listing correctly (though it's still a de-escaping function
shy of correctness), generally try to make it look more like Postgres
coding conventions.
cases. Operator classes now exist within "operator families". While most
families are equivalent to a single class, related classes can be grouped
into one family to represent the fact that they are semantically compatible.
Cross-type operators are now naturally adjunct parts of a family, without
having to wedge them into a particular opclass as we had done originally.
This commit restructures the catalogs and cleans up enough of the fallout so
that everything still works at least as well as before, but most of the work
needed to actually improve the planner's behavior will come later. Also,
there are not yet CREATE/DROP/ALTER OPERATOR FAMILY commands; the only way
to create a new family right now is to allow CREATE OPERATOR CLASS to make
one by default. I owe some more documentation work, too. But that can all
be done in smaller pieces once this infrastructure is in place.
operator strategy numbers, ie, GiST and GIN. This is almost cosmetic
enough to not need a catversion bump, but since the opr_sanity regression
test has to change in sync with the catalog entry, I figured I'd better
do one.
are all in new-in-8.2 logic associated with indexability of ScalarArrayOpExpr
(IN-clauses) or amortization of indexscan costs across repeated indexscans
on the inside of a nestloop. In particular:
Fix some logic errors in the estimation for multiple scans induced by a
ScalarArrayOpExpr indexqual.
Include a small cost component in bitmap index scans to reflect the costs of
manipulating the bitmap itself; this is mainly to prevent a bitmap scan from
appearing to have the same cost as a plain indexscan for fetching a single
tuple.
Also add a per-index-scan-startup CPU cost component; while prior releases
were clearly too pessimistic about the cost of repeated indexscans, the
original 8.2 coding allowed the cost of an indexscan to effectively go to zero
if repeated often enough, which is overly optimistic.
Pay some attention to index correlation when estimating costs for a nestloop
inner indexscan: this is significant when the plan fetches multiple heap
tuples per iteration, since high correlation means those tuples are probably
on the same or adjacent heap pages.
joinclause doesn't use any outer-side vars) requires a "bushy" plan to be
created. The normal heuristic to avoid joins with no joinclause has to be
overridden in that case. Problem is new in 8.2; before that we forced the
outer join order anyway. Per example from Teodor.
representing externally-supplied values, since the APIs that carry such
values only specify type not typmod. However, for PARAM_SUBLINK Params
it is handy to carry the typmod of the sublink's output column. This
is a much cleaner solution for the recently reported 'could not find
pathkey item to sort' and 'failed to find unique expression in subplan
tlist' bugs than my original 8.2-compatible patch. Besides, someday we
might want to support typmods for external parameters ...
in normal operation, and we can avoid rewriting pg_control at every log
segment switch if we don't insist that these values be valid. Reducing
the number of pg_control updates is a good idea for both performance and
reliability. It does make pg_resetxlog's life a bit harder, but that seems
a good tradeoff; and anyway the change to pg_resetxlog amounts to automating
something people formerly needed to do by hand, namely look at the existing
pg_xlog files to make sure the new WAL start point was past them.
In passing, change the wording of xlog.c's "database system was interrupted"
messages: describe the pg_control timestamp as "last known up at" rather than
implying it is the exact time of service interruption. With this change the
timestamp will generally be the time of the last checkpoint, which could be
many minutes before the failure; and we've already seen indications that
people tend to misinterpret the old wording.
initdb forced due to change in pg_control layout. Simon Riggs and Tom Lane
release it in a subtransaction abort, but this neglects possibility that
someone outside SPI already did. Fix is for spi.c to forget about a tuptable
as soon as it's handed it back to the caller.
Per bug #2817 from Michael Andreen.
rearrangeable outer joins and the WHERE clause is non-strict and mentions
only nullable-side relations. New bug in 8.2, caused by new logic to allow
rearranging outer joins. Per bug #2807 from Ross Cohen; thanks to Jeff
Davis for producing a usable test case.
a sublink's test expression have the correct vartypmod, rather than defaulting
to -1. There's at least one place where this is important because we're
expecting these Vars to be exactly equal() to those appearing in the subplan
itself. This is a pretty klugy solution --- it would likely be cleaner to
change Param nodes to include a typmod field --- but we can't do that in the
already-released 8.2 branch.
Per bug report from Hubert Fongarnand.
identify long-running transactions. Since we already need to record
the transaction-start time (e.g. for now()), we don't need any
additional system calls to report this information.
Catversion bumped, initdb required.
Tom Lane
Neil Conway
Peter Eisentraut
Bruce Momjian
Alvaro Herrera
D'Arcy J.M. Cain