In commit e2c2c2e8b1 I made use of nested
list structures to show which clauses went with which index columns, but
on reflection that's a data structure that only an old-line Lisp hacker
could love. Worse, it adds unnecessary complication to the many places
that don't much care which clauses go with which index columns. Revert
to the previous arrangement of flat lists of clauses, and instead add a
parallel integer list of column numbers. The places that care about the
pairing can chase both lists with forboth(), while the places that don't
care just examine one list the same as before.
The only real downside to this is that there are now two more lists that
need to be passed to amcostestimate functions in case they care about
column matching (which btcostestimate does, so not passing the info is not
an option). Rather than deal with 11-argument amcostestimate functions,
pass just the IndexPath and expect the functions to extract fields from it.
That gets us down to 7 arguments which is better than 11, and it seems
more future-proof against likely additions to the information we keep
about an index path.
It's potentially useful for an index to repeat the same indexable column
or expression in multiple index columns, if the columns have different
opclasses. (If they share opclasses too, the duplicate column is pretty
useless, but nonetheless we've allowed such cases since 9.0.) However,
the planner failed to cope with this, because createplan.c was relying on
simple equal() matching to figure out which index column each index qual
is intended for. We do have that information available upstream in
indxpath.c, though, so the fix is to not flatten the multi-level indexquals
list when putting it into an IndexPath. Then we can rely on the sublist
structure to identify target index columns in createplan.c. There's a
similar issue for index ORDER BYs (the KNNGIST feature), so introduce a
multi-level-list representation for that too. This adds a bit more
representational overhead, but we might more or less buy that back by not
having to search for matching index columns anymore in createplan.c;
likewise btcostestimate saves some cycles.
Per bug #6351 from Christian Rudolph. Likely symptoms include the "btree
index keys must be ordered by attribute" failure shown there, as well as
"operator MMMM is not a member of opfamily NNNN".
Although this is a pre-existing problem that can be demonstrated in 9.0 and
9.1, I'm not going to back-patch it, because the API changes in the planner
seem likely to break things such as index plugins. The corner cases where
this matters seem too narrow to justify possibly breaking things in a minor
release.
While logically correct, these two Asserts could fail depending on the
vagaries of floating-point arithmetic. In particular, on machines with
floating-point registers wider than standard "double" values, it was
possible for the compiler to compare a rounded-to-double value already
stored in memory with an unrounded long double value still in a register.
Given the preceding checks, these assertions aren't adding much, so let's
just get rid of them rather than try to find a compiler-proof fix.
Per report from Pavel Stehule.
Given the lack of previous complaints, and the fact that only developers
would be likely to trip over it, I'm only going to change this in HEAD,
even though the code has been like this for a long time.
Add a column pg_class.relallvisible to remember the number of pages that
were all-visible according to the visibility map as of the last VACUUM
(or ANALYZE, or some other operations that update pg_class.relpages).
Use relallvisible/relpages, instead of an arbitrary constant, to estimate
how many heap page fetches can be avoided during an index-only scan.
This is pretty primitive and will no doubt see refinements once we've
acquired more field experience with the index-only scan mechanism, but
it's way better than using a constant.
Note: I had to adjust an underspecified query in the window.sql regression
test, because it was changing answers when the plan changed to use an
index-only scan. Some of the adjacent tests perhaps should be adjusted
as well, but I didn't do that here.
When a btree index contains all columns required by the query, and the
visibility map shows that all tuples on a target heap page are
visible-to-all, we don't need to fetch that heap page. This patch depends
on the previous patches that made the visibility map reliable.
There's a fair amount left to do here, notably trying to figure out a less
chintzy way of estimating the cost of an index-only scan, but the core
functionality seems ready to commit.
Robert Haas and Ibrar Ahmed, with some previous work by Heikki Linnakangas.
Formerly, set_subquery_pathlist and other creators of plans for subqueries
saved only the rangetable and rowMarks lists from the lower-level
PlannerInfo. But there's no reason not to remember the whole PlannerInfo,
and indeed this turns out to simplify matters in a number of places.
The immediate reason for doing this was so that the subroot will still be
accessible when we're trying to extract column statistics out of an
already-planned subquery. But now that I've done it, it seems like a good
code-beautification effort in its own right.
I also chose to get rid of the transient subrtable and subrowmark fields in
SubqueryScan nodes, in favor of having setrefs.c look up the subquery's
RelOptInfo. That required changing all the APIs in setrefs.c to pass
PlannerInfo not PlannerGlobal, which was a large but quite mechanical
transformation.
One side-effect not foreseen at the beginning is that this finally broke
inheritance_planner's assumption that replanning the same subquery RTE N
times would necessarily give interchangeable results each time. That
assumption was always pretty risky, but now we really have to make a
separate RTE for each instance so that there's a place to carry the
separate subroots.
set_append_rel_pathlist supposed that, while computing per-column width
estimates for the appendrel, it could ignore child rels for which the
translated reltargetlist entry wasn't a Var. This gave rise to completely
silly estimates in some common cases, such as constant outputs from some or
all of the arms of a UNION ALL. Instead, fall back on get_typavgwidth to
estimate from the value's datatype; which might be a poor estimate but at
least it's not completely wacko.
That problem was exposed by an Assert in set_subquery_size_estimates, which
unfortunately was still overoptimistic even with that fix, since we don't
compute attr_widths estimates for appendrels that are entirely excluded by
constraints. So remove the Assert; we'll just fall back on get_typavgwidth
in such cases.
Also, since set_subquery_size_estimates calls set_baserel_size_estimates
which calls set_rel_width, there's no need for set_subquery_size_estimates
to call get_typavgwidth; set_rel_width will handle it for us if we just
leave the estimate set to zero. Remove the unnecessary code.
Per report from Erik Rijkers and subsequent investigation.
A PlaceHolderVar's expression might contain another, lower-level
PlaceHolderVar. If the outer PlaceHolderVar is used, the inner one
certainly will be also, and so we have to make sure that both of them get
into the placeholder_list with correct ph_may_need values during the
initial pre-scan of the query (before deconstruct_jointree starts).
We did this correctly for PlaceHolderVars appearing in the query quals,
but overlooked the issue for those appearing in the top-level targetlist;
with the result that nested placeholders referenced only in the targetlist
did not work correctly, as illustrated in bug #6154.
While at it, add some error checking to find_placeholder_info to ensure
that we don't try to create new placeholders after it's too late to do so;
they have to all be created before deconstruct_jointree starts.
Back-patch to 8.4 where the PlaceHolderVar mechanism was introduced.
The previous coding failed to account properly for the costs of evaluating
the input expressions of aggregates and window functions, as seen in a
recent gripe from Claudio Freire. (I said at the time that it wasn't
counting these costs at all; but on closer inspection, it was effectively
charging these costs once per output tuple. That is completely wrong for
aggregates, and not exactly right for window functions either.)
There was also a hard-wired assumption that aggregates and window functions
had procost 1.0, which is now fixed to respect the actual cataloged costs.
The costing of WindowAgg is still pretty bogus, since it doesn't try to
estimate the effects of spilling data to disk, but that seems like a
separate issue.
Although rowcount estimates really ought not be NaN, a bug elsewhere
could perhaps result in that, and that would cause Assert failure in
cost_mergejoin, which I believe to be the explanation for bug #5977 from
Anton Kuznetsov. Seems like a good idea to expend a couple more cycles
to prevent that, even though the real bug is elsewhere. Not back-patching,
though, because we don't encourage running production systems with
Asserts on.
All expression nodes now have an explicit output-collation field, unless
they are known to only return a noncollatable data type (such as boolean
or record). Also, nodes that can invoke collation-aware functions store
a separate field that is the collation value to pass to the function.
This avoids confusion that arises when a function has collatable inputs
and noncollatable output type, or vice versa.
Also, replace the parser's on-the-fly collation assignment method with
a post-pass over the completed expression tree. This allows us to use
a more complex (and hopefully more nearly spec-compliant) assignment
rule without paying for it in extra storage in every expression node.
Fix assorted bugs in the planner's handling of collations by making
collation one of the defining properties of an EquivalenceClass and
by converting CollateExprs into discardable RelabelType nodes during
expression preprocessing.
While this will give wrong answers when estimating selectivity for a
comparison operator that's using a non-default collation, the estimation
error probably won't be large; and anyway the former approach created
estimation errors of its own by trying to use a histogram that might have
been computed with some other collation. So we'll adopt this simplified
approach for now and perhaps improve it sometime in the future.
This patch incorporates changes from Andres Freund to make sure that
selfuncs.c passes a valid collation OID to any datatype-specific function
it calls, in case that function wants collation information. Said OID will
now always be DEFAULT_COLLATION_OID, but at least we won't get errors.
This commit provides the core code and documentation needed. A contrib
module test case will follow shortly.
Shigeru Hanada, Jan Urbanski, Heikki Linnakangas
This adds collation support for columns and domains, a COLLATE clause
to override it per expression, and B-tree index support.
Peter Eisentraut
reviewed by Pavel Stehule, Itagaki Takahiro, Robert Haas, Noah Misch
This is a heavily revised version of builtin_knngist_core-0.9. The
ordering operators are no longer mixed in with actual quals, which would
have confused not only humans but significant parts of the planner.
Instead, ordering operators are carried separately throughout planning and
execution.
Since the API for ambeginscan and amrescan functions had to be changed
anyway, this commit takes the opportunity to rationalize that a bit.
RelationGetIndexScan no longer forces a premature index_rescan call;
instead, callers of index_beginscan must call index_rescan too. Aside from
making the AM-side initialization logic a bit less peculiar, this has the
advantage that we do not make a useless extra am_rescan call when there are
runtime key values. AMs formerly could not assume that the key values
passed to amrescan were actually valid; now they can.
Teodor Sigaev and Tom Lane
As per the ancient comment for set_rel_width, it really wasn't much good
for relations that aren't plain tables: it would never find any stats and
would always fall back on datatype-based estimates, which are often pretty
silly. Fix that by copying up width estimates from the subquery planning
process.
At some point we might want to do this for CTEs too, but that would be a
significantly more invasive patch because the sub-PlannerInfo is no longer
accessible by the time it's needed. I refrained from doing anything about
that, partly for fear of breaking the unmerged CTE-related patches.
In passing, also generate less bogus width estimates for whole-row Vars.
Per a gripe from Jon Nelson.
It was reporting that these were fully indexed (hence cheap), when of
course they're the exact opposite of that. I'm not certain if the case
would arise in practice, since a clauseless semijoin is hard to produce
in SQL, but if it did happen we'd make some dumb decisions.
This patch eliminates the former need to sort the output of an Append scan
when an ordered scan of an inheritance tree is wanted. This should be
particularly useful for fast-start cases such as queries with LIMIT.
Original patch by Greg Stark, with further hacking by Hans-Jurgen Schonig,
Robert Haas, and Tom Lane.
The logic for determining whether to materialize has been significantly
overhauled for 9.0. In case there should be any doubt about whether
materialization is a win in any particular case, this should provide a
convenient way of seeing what happens without it; but even with enable_material
turned off, we still materialize in cases where it is required for
correctness.
Thanks to Tom Lane for the review.
tuple, instead of the former cpu_tuple_cost. It is sane to charge less than
cpu_tuple_cost because Materialize never does any qual-checking or projection,
so it's got less overhead than most plan node types. In particular, we want
to have the same charge here as is charged for readout in cost_sort. That
avoids the problem recently exhibited by Teodor wherein the planner prefers
a useless sort over a materialize step in a context where a lot of rescanning
will happen. The rescan costs should be just about the same for both node
types, so make their estimates the same.
Not back-patching because all of the current logic for rescan cost estimates
is new in 9.0. The old handling of rescans is sufficiently not-sane that
changing this in that structure is a bit pointless, and might indeed cause
regressions.
This patch only supports seq_page_cost and random_page_cost as parameters,
but it provides the infrastructure to scalably support many more.
In particular, we may want to add support for effective_io_concurrency,
but I'm leaving that as future work for now.
Thanks to Tom Lane for design help and Alvaro Herrera for the review.
mergejoin to shield it from doing mark/restore and refetches. Put an explicit
flag in MergePath so we can centralize the logic that knows about this,
and add costing logic that considers using Materialize even when it's not
forced by the previously-existing considerations. This is in response to
a discussion back in August that suggested that materializing an inner
indexscan can be helpful when the refetch percentage is high enough.
an explicit model of rescan costs being different from first-time costs.
The costing of Material nodes in particular now has some visible relationship
to the actual runtime behavior, where before it was essentially fantasy.
This also fixes up a couple of places where different materialized plan types
were treated differently for no very good reason (probably just oversights).
A couple of the regression tests are affected, because the planner now chooses
to put the other relation on the inside of a nestloop-with-materialize.
So far as I can see both changes are sane, and the planner is now more
consistently following the expectation that it should prefer to materialize
the smaller of two relations.
Per a recent discussion with Robert Haas.
RelOptInfo targetlist. It used to be that the only possibility other than
a Var was a RowExpr representing a whole-row child Var, but as of 8.4's
expanded ability to flatten appendrel members, we can get arbitrary expressions
in there. Use the expression's type info and get_typavgwidth() to produce
an at-least-marginally-sane result. Note that get_typavgwidth()'s fallback
estimate (32 bytes) is the same as what was here before, so there will be
no behavioral change for RowExprs. Noted while looking at recent gripe
about constant quals pushed down to FunctionScan appendrel members ...
not only were we failing to recognize the constant qual, we were getting
the width estimate wrong :-(
joins a bit better, ie, understand the differing cost functions for matched
and unmatched outer tuples. There is more that could be done in cost_hashjoin
but this already helps a great deal. Per discussions with Robert Haas.
"physical tlist" optimization on the outer relation (ie, force a projection
step to occur in its scan). This avoids storing useless column values when
the outer relation's tuples are written to temporary batch files.
Modified version of a patch by Michael Henderson and Ramon Lawrence.
distribution, by creating a special fast path for the (first few) most common
values of the outer relation. Tuples having hashvalues matching the MCVs
are effectively forced to be in the first batch, so that we never write
them out to the batch temp files.
Bryce Cutt and Ramon Lawrence, with some editorialization by me.
keys when considering a semi or anti join. This requires estimating the
selectivity of the merge qual as though it were a regular inner join condition.
To allow caching both that and the real outer-join-aware selectivity, split
RestrictInfo.this_selec into two fields.
This fixes one of the problems reported by Kevin Grittner.
though it is an inner rather than outer join type. This essentially means
that we don't bother to separate "pushed down" qual conditions from actual
join quals at a semijoin plan node; which is okay because the restrictions of
SQL syntax make it impossible to have a pushed-down qual that references the
inner side of a semijoin. This allows noticeably better optimization of
IN/EXISTS cases than we had before, since the equivalence-class machinery can
now use those quals. Also fix a couple of other mistakes that had essentially
disabled the ability to unique-ify the inner relation and then join it to just
a subset of the left-hand relations. An example case using the regression
database is
select * from tenk1 a, tenk1 b
where (a.unique1,b.unique2) in (select unique1,unique2 from tenk1 c);
which is planned reasonably well by 8.3 and earlier but had been forcing a
cartesian join of a/b in CVS HEAD.
that represent some expression that we desire to compute below the top level
of the plan, and then let that value "bubble up" as though it were a plain
Var (ie, a column value).
The immediate application is to allow sub-selects to be flattened even when
they are below an outer join and have non-nullable output expressions.
Formerly we couldn't flatten because such an expression wouldn't properly
go to NULL when evaluated above the outer join. Now, we wrap it in a
PlaceHolderVar and arrange for the actual evaluation to occur below the outer
join. When the resulting Var bubbles up through the join, it will be set to
NULL if necessary, yielding the correct results. This fixes a planner
limitation that's existed since 7.1.
In future we might want to use this mechanism to re-introduce some form of
Hellerstein's "expensive functions" optimization, ie place the evaluation of
an expensive function at the most suitable point in the plan tree.
set_rel_width(). The code had been catering for the possibility of different
varnos in the relation targetlist, but this is impossible for a base relation
(and if it were possible, putting all the widths in the same RelOptInfo would
be wrong anyway).
There are some unimplemented aspects: recursive queries must use UNION ALL
(should allow UNION too), and we don't have SEARCH or CYCLE clauses.
These might or might not get done for 8.4, but even without them it's a
pretty useful feature.
There are also a couple of small loose ends and definitional quibbles,
which I'll send a memo about to pgsql-hackers shortly. But let's land
the patch now so we can get on with other development.
Yoshiyuki Asaba, with lots of help from Tatsuo Ishii and Tom Lane
inserting a materialize node above an inner-side sort node, when the sort is
expected to spill to disk. (The materialize protects the sort from having
to support mark/restore, allowing it to do its final merge pass on-the-fly.)
We neglected to teach cost_mergejoin about that hack, so it was failing to
include the materialize's costs in the estimated cost of the mergejoin.
The materialize's costs are generally going to be pretty negligible in
comparison to the sort's, so this is only a small error and probably not
worth back-patching; but it's still wrong.
In the similar case where a materialize is inserted to protect an inner-side
node that can't do mark/restore at all, it's still true that the materialize
should not spill to disk, and so we should cost it cheaply rather than
expensively.
Noted while thinking about a question from Tom Raney.
into nodes/nodeFuncs, so as to reduce wanton cross-subsystem #includes inside
the backend. There's probably more that should be done along this line,
but this is a start anyway.
subqueries into the same thing you'd have gotten from IN (except always with
unknownEqFalse = true, so as to get the proper semantics for an EXISTS).
I believe this fixes the last case within CVS HEAD in which an EXISTS could
give worse performance than an equivalent IN subquery.
The tricky part of this is that if the upper query probes the EXISTS for only
a few rows, the hashing implementation can actually be worse than the default,
and therefore we need to make a cost-based decision about which way to use.
But at the time when the planner generates plans for subqueries, it doesn't
really know how many times the subquery will be executed. The least invasive
solution seems to be to generate both plans and postpone the choice until
execution. Therefore, in a query that has been optimized this way, EXPLAIN
will show two subplans for the EXISTS, of which only one will actually get
executed.
There is a lot more that could be done based on this infrastructure: in
particular it's interesting to consider switching to the hash plan if we start
out using the non-hashed plan but find a lot more upper rows going by than we
expected. I have therefore left some minor inefficiencies in place, such as
initializing both subplans even though we will currently only use one.
and anti joins. To do this, pass the SpecialJoinInfo struct for the current
join as an additional optional argument to operator join selectivity
estimation functions. This allows the estimator to tell not only what kind
of join is being formed, but which variable is on which side of the join;
a requirement long recognized but not dealt with till now. This also leaves
the door open for future improvements in the estimators, such as accounting
for the null-insertion effects of lower outer joins. I didn't do anything
about that in the current patch but the information is in principle deducible
from what's passed.
The patch also clarifies the definition of join selectivity for semi/anti
joins: it's the fraction of the left input that has (at least one) match
in the right input. This allows getting rid of some very fuzzy thinking
that I had committed in the original 7.4-era IN-optimization patch.
There's probably room to estimate this better than the present patch does,
but at least we know what to estimate.
Since I had to touch CREATE OPERATOR anyway to allow a variant signature
for join estimator functions, I took the opportunity to add a couple of
additional checks that were missing, per my recent message to -hackers:
* Check that estimator functions return float8;
* Require execute permission at the time of CREATE OPERATOR on the
operator's function as well as the estimator functions;
* Require ownership of any pre-existing operator that's modified by
the command.
I also moved the lookup of the functions out of OperatorCreate() and
into operatorcmds.c, since that seemed more consistent with most of
the other catalog object creation processes, eg CREATE TYPE.
Tom Lane
Bruce Momjian
Peter Eisentraut
Magnus Hagander
Robert Haas