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.
the old JOIN_IN code, but antijoins are new functionality.) Teach the planner
to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti
joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL
filters are recognized as being anti joins. Unify the InClauseInfo and
OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change,
it becomes possible to associate a SpecialJoinInfo with every join attempt,
which permits some cleanup of join selectivity estimation. That needs to be
taken much further than this patch does, but the next step is to change the
API for oprjoin selectivity functions, which seems like material for a
separate patch. So for the moment the output size estimates for semi and
especially anti joins are quite bogus.
knowledge up through any joins it participates in. We were doing that already
in some special cases but not in the general case. Also, defend against zero
row estimates for the input relations in cost_mergejoin --- this fix may have
eliminated the only scenario in which that can happen, but be safe. Per
report from Alex Solovey.
the two join variables at both ends: not only trailing rows that need not be
scanned because there cannot be a match on the other side, but initial rows
that will be scanned without possibly having a match. This allows a more
realistic estimate of startup cost to be made, per recent pgsql-performance
discussion. In passing, fix a couple of bugs that had crept into
mergejoinscansel: it was not quite up to speed for the task of estimating
descending-order scans, which is a new requirement in 8.3.
then-delete on the current cursor row. The basic fix is that nodeTidscan.c
has to apply heap_get_latest_tid() to the current-scan-TID obtained from the
cursor query; this ensures we get the latest row version to work with.
However, since that only works if the query plan is a TID scan, we also have
to hack the planner to make sure only that type of plan will be selected.
(Formerly, the planner might decide to apply a seqscan if the table is very
small. This change is probably a Good Thing anyway, since it's hard to see
how a seqscan could really win.) That means the execQual.c code to support
CurrentOfExpr as a regular expression type is dead code, so replace it with
just an elog(). Also, add regression tests covering these cases. Note
that the added tests expose the fact that re-fetching an updated row
misbehaves if the cursor used FOR UPDATE. That's an independent bug that
should be fixed later. Per report from Dharmendra Goyal.
(because they are uncorrelated with the immediate parent query). We were
charging the full run cost to the parent node, disregarding the fact that
only one row need be fetched for EXISTS. While this would only be a
cosmetic issue in most cases, it might possibly affect planning outcomes
if the parent query were itself a subquery to some upper query.
Per recent discussion with Steve Crawford.
Along the way, allow FOR UPDATE in non-WITH-HOLD cursors; there may once
have been a reason to disallow that, but it seems to work now, and it's
really rather necessary if you want to select a row via a cursor and then
update it in a concurrent-safe fashion.
Original patch by Arul Shaji, rather heavily editorialized by Tom Lane.
from the other string-category types; this eliminates a lot of surprising
interpretations that the parser could formerly make when there was no directly
applicable operator.
Create a general mechanism that supports casts to and from the standard string
types (text,varchar,bpchar) for *every* datatype, by invoking the datatype's
I/O functions. These new casts are assignment-only in the to-string direction,
explicit-only in the other, and therefore should create no surprising behavior.
Remove a bunch of thereby-obsoleted datatype-specific casting functions.
The "general mechanism" is a new expression node type CoerceViaIO that can
actually convert between *any* two datatypes if their external text
representations are compatible. This is more general than needed for the
immediate feature, but might be useful in plpgsql or other places in future.
This commit does nothing about the issue that applying the concatenation
operator || to non-text types will now fail, often with strange error messages
due to misinterpreting the operator as array concatenation. Since it often
(not always) worked before, we should either make it succeed or at least give
a more user-friendly error; but details are still under debate.
Peter Eisentraut and Tom Lane
is using mark/restore but not rewind or backward-scan capability. Insert a
materialize plan node between a mergejoin and its inner child if the inner
child is a sort that is expected to spill to disk. The materialize shields
the sort from the need to do mark/restore and thereby allows it to perform
its final merge pass on-the-fly; while the materialize itself is normally
cheap since it won't spill to disk unless the number of tuples with equal
key values exceeds work_mem.
Greg Stark, with some kibitzing from Tom Lane.
need be returned. We keep a heap of the current best N tuples and sift-up
new tuples into it as we scan the input. For M input tuples this means
only about M*log(N) comparisons instead of M*log(M), not to mention a lot
less workspace when N is small --- avoiding spill-to-disk for large M
is actually the most attractive thing about it. Patch includes planner
and executor support for invoking this facility in ORDER BY ... LIMIT
queries. Greg Stark, with some editorialization by moi.
are mostly excluded by constraints: do the CE test a bit earlier to save
some adjust_appendrel_attrs() work on excluded children, and arrange to
use array indexing rather than rt_fetch() to fetch RTEs in the main body
of the planner. The latter is something I'd wanted to do for awhile anyway,
but seeing list_nth_cell() as 35% of the runtime gets one's attention.
seen by code inspecting the expression. The best way to do this seems
to be to drop the original representation as a function invocation, and
instead make a special expression node type that represents applying
the element-type coercion function to each array element. In this way
the element function is exposed and will be checked for volatility.
Per report from Guillaume Smet.
useless substructure for its RangeTblEntry nodes. (I chose to keep using the
same struct node type and just zero out the link fields for unneeded info,
rather than making a separate ExecRangeTblEntry type --- it seemed too
fragile to have two different rangetable representations.)
Along the way, put subplans into a list in the toplevel PlannedStmt node,
and have SubPlan nodes refer to them by list index instead of direct pointers.
Vadim wanted to do that years ago, but I never understood what he was on about
until now. It makes things a *whole* lot more robust, because we can stop
worrying about duplicate processing of subplans during expression tree
traversals. That's been a constant source of bugs, and it's finally gone.
There are some consequent simplifications yet to be made, like not using
a separate EState for subplans in the executor, but I'll tackle that later.
which I had removed in the first cut of the EquivalenceClass rewrite to
simplify that patch a little. But it's still important --- in a four-way
join problem mergejoinscansel() was eating about 40% of the planning time
according to gprof. Also, improve the EquivalenceClass code to re-use
join RestrictInfos rather than generating fresh ones for each join
considered. This saves some memory space but more importantly improves
the effectiveness of caching planning info in RestrictInfos.
columns procost and prorows, to allow simple user adjustment of the estimated
cost of a function call, as well as control of the estimated number of rows
returned by a set-returning function. We might eventually wish to extend this
to allow function-specific estimation routines, but there seems to be
consensus that we should try a simple constant estimate first. In particular
this provides a relatively simple way to control the order in which different
WHERE clauses are applied in a plan node, which is a Good Thing in view of the
fact that the recent EquivalenceClass planner rewrite made that much less
predictable than before.
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.
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.
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.
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.
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.
Bruce Momjian
Robert Haas
Tom Lane