2010-09-20 22:08:53 +02:00
|
|
|
src/backend/optimizer/README
|
2008-03-20 18:55:15 +01:00
|
|
|
|
|
|
|
Optimizer
|
2008-03-21 14:23:29 +01:00
|
|
|
=========
|
1999-02-09 04:51:42 +01:00
|
|
|
|
2000-02-07 05:41:04 +01:00
|
|
|
These directories take the Query structure returned by the parser, and
|
|
|
|
generate a plan used by the executor. The /plan directory generates the
|
|
|
|
actual output plan, the /path code generates all possible ways to join the
|
2000-11-12 01:37:02 +01:00
|
|
|
tables, and /prep handles various preprocessing steps for special cases.
|
|
|
|
/util is utility stuff. /geqo is the separate "genetic optimization" planner
|
|
|
|
--- it does a semi-random search through the join tree space, rather than
|
|
|
|
exhaustively considering all possible join trees. (But each join considered
|
|
|
|
by /geqo is given to /path to create paths for, so we consider all possible
|
2000-09-29 20:21:41 +02:00
|
|
|
implementation paths for each specific join pair even in GEQO mode.)
|
|
|
|
|
|
|
|
|
|
|
|
Paths and Join Pairs
|
|
|
|
--------------------
|
|
|
|
|
|
|
|
During the planning/optimizing process, we build "Path" trees representing
|
|
|
|
the different ways of doing a query. We select the cheapest Path that
|
|
|
|
generates the desired relation and turn it into a Plan to pass to the
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
executor. (There is pretty nearly a one-to-one correspondence between the
|
2000-09-29 20:21:41 +02:00
|
|
|
Path and Plan trees, but Path nodes omit info that won't be needed during
|
|
|
|
planning, and include info needed for planning that won't be needed by the
|
|
|
|
executor.)
|
|
|
|
|
|
|
|
The optimizer builds a RelOptInfo structure for each base relation used in
|
|
|
|
the query. Base rels are either primitive tables, or subquery subselects
|
|
|
|
that are planned via a separate recursive invocation of the planner. A
|
|
|
|
RelOptInfo is also built for each join relation that is considered during
|
|
|
|
planning. A join rel is simply a combination of base rels. There is only
|
|
|
|
one join RelOptInfo for any given set of baserels --- for example, the join
|
|
|
|
{A B C} is represented by the same RelOptInfo no matter whether we build it
|
|
|
|
by joining A and B first and then adding C, or joining B and C first and
|
|
|
|
then adding A, etc. These different means of building the joinrel are
|
|
|
|
represented as Paths. For each RelOptInfo we build a list of Paths that
|
|
|
|
represent plausible ways to implement the scan or join of that relation.
|
|
|
|
Once we've considered all the plausible Paths for a rel, we select the one
|
|
|
|
that is cheapest according to the planner's cost estimates. The final plan
|
|
|
|
is derived from the cheapest Path for the RelOptInfo that includes all the
|
|
|
|
base rels of the query.
|
|
|
|
|
|
|
|
Possible Paths for a primitive table relation include plain old sequential
|
2005-12-20 03:30:36 +01:00
|
|
|
scan, plus index scans for any indexes that exist on the table, plus bitmap
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
index scans using one or more indexes. Specialized RTE types, such as
|
|
|
|
function RTEs, may have only one possible Path.
|
2000-09-29 20:21:41 +02:00
|
|
|
|
|
|
|
Joins always occur using two RelOptInfos. One is outer, the other inner.
|
|
|
|
Outers drive lookups of values in the inner. In a nested loop, lookups of
|
|
|
|
values in the inner occur by scanning the inner path once per outer tuple
|
|
|
|
to find each matching inner row. In a mergejoin, inner and outer rows are
|
|
|
|
ordered, and are accessed in order, so only one scan is required to perform
|
|
|
|
the entire join: both inner and outer paths are scanned in-sync. (There's
|
|
|
|
not a lot of difference between inner and outer in a mergejoin...) In a
|
|
|
|
hashjoin, the inner is scanned first and all its rows are entered in a
|
|
|
|
hashtable, then the outer is scanned and for each row we lookup the join
|
|
|
|
key in the hashtable.
|
|
|
|
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
A Path for a join relation is actually a tree structure, with the topmost
|
|
|
|
Path node representing the last-applied join method. It has left and right
|
|
|
|
subpaths that represent the scan or join methods used for the two input
|
|
|
|
relations.
|
2000-02-07 05:41:04 +01:00
|
|
|
|
|
|
|
|
|
|
|
Join Tree Construction
|
|
|
|
----------------------
|
|
|
|
|
1999-08-16 04:17:58 +02:00
|
|
|
The optimizer generates optimal query plans by doing a more-or-less
|
2000-09-29 20:21:41 +02:00
|
|
|
exhaustive search through the ways of executing the query. The best Path
|
|
|
|
tree is found by a recursive process:
|
1999-08-16 04:17:58 +02:00
|
|
|
|
|
|
|
1) Take each base relation in the query, and make a RelOptInfo structure
|
|
|
|
for it. Find each potentially useful way of accessing the relation,
|
2008-04-09 03:00:46 +02:00
|
|
|
including sequential and index scans, and make Paths representing those
|
|
|
|
ways. All the Paths made for a given relation are placed in its
|
1999-08-16 04:17:58 +02:00
|
|
|
RelOptInfo.pathlist. (Actually, we discard Paths that are obviously
|
|
|
|
inferior alternatives before they ever get into the pathlist --- what
|
|
|
|
ends up in the pathlist is the cheapest way of generating each potentially
|
2012-01-28 01:26:38 +01:00
|
|
|
useful sort ordering and parameterization of the relation.) Also create a
|
|
|
|
RelOptInfo.joininfo list including all the join clauses that involve this
|
|
|
|
relation. For example, the WHERE clause "tab1.col1 = tab2.col1" generates
|
|
|
|
entries in both tab1 and tab2's joininfo lists.
|
1999-08-16 04:17:58 +02:00
|
|
|
|
2001-01-17 07:41:31 +01:00
|
|
|
If we have only a single base relation in the query, we are done.
|
2000-02-07 05:41:04 +01:00
|
|
|
Otherwise we have to figure out how to join the base relations into a
|
|
|
|
single join relation.
|
|
|
|
|
2005-12-20 03:30:36 +01:00
|
|
|
2) Normally, any explicit JOIN clauses are "flattened" so that we just
|
|
|
|
have a list of relations to join. However, FULL OUTER JOIN clauses are
|
|
|
|
never flattened, and other kinds of JOIN might not be either, if the
|
|
|
|
flattening process is stopped by join_collapse_limit or from_collapse_limit
|
|
|
|
restrictions. Therefore, we end up with a planning problem that contains
|
2007-01-20 21:45:41 +01:00
|
|
|
lists of relations to be joined in any order, where any individual item
|
|
|
|
might be a sub-list that has to be joined together before we can consider
|
|
|
|
joining it to its siblings. We process these sub-problems recursively,
|
|
|
|
bottom up. Note that the join list structure constrains the possible join
|
|
|
|
orders, but it doesn't constrain the join implementation method at each
|
|
|
|
join (nestloop, merge, hash), nor does it say which rel is considered outer
|
|
|
|
or inner at each join. We consider all these possibilities in building
|
|
|
|
Paths. We generate a Path for each feasible join method, and select the
|
|
|
|
cheapest Path.
|
|
|
|
|
|
|
|
For each planning problem, therefore, we will have a list of relations
|
|
|
|
that are either base rels or joinrels constructed per sub-join-lists.
|
|
|
|
We can join these rels together in any order the planner sees fit.
|
2000-09-29 20:21:41 +02:00
|
|
|
The standard (non-GEQO) planner does this as follows:
|
|
|
|
|
Restructure code that is responsible for ensuring that clauseless joins are
considered when it is necessary to do so because of a join-order restriction
(that is, an outer-join or IN-subselect construct). The former coding was a
bit ad-hoc and inconsistent, and it missed some cases, as exposed by Mario
Weilguni's recent bug report. His specific problem was that an IN could be
turned into a "clauseless" join due to constant-propagation removing the IN's
joinclause, and if the IN's subselect involved more than one relation and
there was more than one such IN linking to the same upper relation, then the
only valid join orders involve "bushy" plans but we would fail to consider the
specific paths needed to get there. (See the example case added to the join
regression test.) On examining the code I wonder if there weren't some other
problem cases too; in particular it seems that GEQO was defending against a
different set of corner cases than the main planner was. There was also an
efficiency problem, in that when we did realize we needed a clauseless join
because of an IN, we'd consider clauseless joins against every other relation
whether this was sensible or not. It seems a better design is to use the
outer-join and in-clause lists as a backup heuristic, just as the rule of
joining only where there are joinclauses is a heuristic: we'll join two
relations if they have a usable joinclause *or* this might be necessary to
satisfy an outer-join or IN-clause join order restriction. I refactored the
code to have just one place considering this instead of three, and made sure
that it covered all the cases that any of them had been considering.
Backpatch as far as 8.1 (which has only the IN-clause form of the disease).
By rights 8.0 and 7.4 should have the bug too, but they accidentally fail
to fail, because the joininfo structure used in those releases preserves some
memory of there having once been a joinclause between the inner and outer
sides of an IN, and so it leads the code in the right direction anyway.
I'll be conservative and not touch them.
2007-02-16 01:14:01 +01:00
|
|
|
Consider joining each RelOptInfo to each other RelOptInfo for which there
|
|
|
|
is a usable joinclause, and generate a Path for each possible join method
|
|
|
|
for each such pair. (If we have a RelOptInfo with no join clauses, we have
|
|
|
|
no choice but to generate a clauseless Cartesian-product join; so we
|
|
|
|
consider joining that rel to each other available rel. But in the presence
|
|
|
|
of join clauses we will only consider joins that use available join
|
|
|
|
clauses. Note that join-order restrictions induced by outer joins and
|
2008-08-14 20:48:00 +02:00
|
|
|
IN/EXISTS clauses are also checked, to ensure that we find a workable join
|
|
|
|
order in cases where those restrictions force a clauseless join to be done.)
|
2000-09-29 20:21:41 +02:00
|
|
|
|
2007-01-20 21:45:41 +01:00
|
|
|
If we only had two relations in the list, we are done: we just pick
|
2000-09-29 20:21:41 +02:00
|
|
|
the cheapest path for the join RelOptInfo. If we had more than two, we now
|
1999-08-16 04:17:58 +02:00
|
|
|
need to consider ways of joining join RelOptInfos to each other to make
|
2007-01-20 21:45:41 +01:00
|
|
|
join RelOptInfos that represent more than two list items.
|
2000-02-07 05:41:04 +01:00
|
|
|
|
|
|
|
The join tree is constructed using a "dynamic programming" algorithm:
|
|
|
|
in the first pass (already described) we consider ways to create join rels
|
2007-01-20 21:45:41 +01:00
|
|
|
representing exactly two list items. The second pass considers ways
|
|
|
|
to make join rels that represent exactly three list items; the next pass,
|
2000-09-29 20:21:41 +02:00
|
|
|
four items, etc. The last pass considers how to make the final join
|
2007-01-20 21:45:41 +01:00
|
|
|
relation that includes all list items --- obviously there can be only one
|
2000-02-07 05:41:04 +01:00
|
|
|
join rel at this top level, whereas there can be more than one join rel
|
|
|
|
at lower levels. At each level we use joins that follow available join
|
|
|
|
clauses, if possible, just as described for the first level.
|
1999-08-16 04:17:58 +02:00
|
|
|
|
|
|
|
For example:
|
1999-02-09 04:51:42 +01:00
|
|
|
|
1999-02-15 23:19:01 +01:00
|
|
|
SELECT *
|
|
|
|
FROM tab1, tab2, tab3, tab4
|
|
|
|
WHERE tab1.col = tab2.col AND
|
|
|
|
tab2.col = tab3.col AND
|
|
|
|
tab3.col = tab4.col
|
|
|
|
|
|
|
|
Tables 1, 2, 3, and 4 are joined as:
|
|
|
|
{1 2},{2 3},{3 4}
|
|
|
|
{1 2 3},{2 3 4}
|
|
|
|
{1 2 3 4}
|
2000-02-07 05:41:04 +01:00
|
|
|
(other possibilities will be excluded for lack of join clauses)
|
1999-02-15 23:19:01 +01:00
|
|
|
|
|
|
|
SELECT *
|
|
|
|
FROM tab1, tab2, tab3, tab4
|
|
|
|
WHERE tab1.col = tab2.col AND
|
|
|
|
tab1.col = tab3.col AND
|
|
|
|
tab1.col = tab4.col
|
|
|
|
|
|
|
|
Tables 1, 2, 3, and 4 are joined as:
|
|
|
|
{1 2},{1 3},{1 4}
|
2000-02-07 05:41:04 +01:00
|
|
|
{1 2 3},{1 3 4},{1 2 4}
|
1999-02-15 23:19:01 +01:00
|
|
|
{1 2 3 4}
|
|
|
|
|
2000-02-07 05:41:04 +01:00
|
|
|
We consider left-handed plans (the outer rel of an upper join is a joinrel,
|
2007-01-20 21:45:41 +01:00
|
|
|
but the inner is always a single list item); right-handed plans (outer rel
|
2000-09-29 20:21:41 +02:00
|
|
|
is always a single item); and bushy plans (both inner and outer can be
|
|
|
|
joins themselves). For example, when building {1 2 3 4} we consider
|
|
|
|
joining {1 2 3} to {4} (left-handed), {4} to {1 2 3} (right-handed), and
|
|
|
|
{1 2} to {3 4} (bushy), among other choices. Although the jointree
|
|
|
|
scanning code produces these potential join combinations one at a time,
|
|
|
|
all the ways to produce the same set of joined base rels will share the
|
|
|
|
same RelOptInfo, so the paths produced from different join combinations
|
2005-12-20 03:30:36 +01:00
|
|
|
that produce equivalent joinrels will compete in add_path().
|
2000-02-07 05:41:04 +01:00
|
|
|
|
|
|
|
Once we have built the final join rel, we use either the cheapest path
|
|
|
|
for it or the cheapest path with the desired ordering (if that's cheaper
|
|
|
|
than applying a sort to the cheapest other path).
|
|
|
|
|
2005-12-20 03:30:36 +01:00
|
|
|
If the query contains one-sided outer joins (LEFT or RIGHT joins), or
|
2012-01-28 01:26:38 +01:00
|
|
|
IN or EXISTS WHERE clauses that were converted to semijoins or antijoins,
|
|
|
|
then some of the possible join orders may be illegal. These are excluded
|
|
|
|
by having join_is_legal consult a side list of such "special" joins to see
|
|
|
|
whether a proposed join is illegal. (The same consultation allows it to
|
|
|
|
see which join style should be applied for a valid join, ie, JOIN_INNER,
|
|
|
|
JOIN_LEFT, etc.)
|
2005-12-20 03:30:36 +01:00
|
|
|
|
|
|
|
|
2008-03-20 18:55:15 +01:00
|
|
|
Valid OUTER JOIN Optimizations
|
2005-12-20 03:30:36 +01:00
|
|
|
------------------------------
|
|
|
|
|
|
|
|
The planner's treatment of outer join reordering is based on the following
|
|
|
|
identities:
|
|
|
|
|
|
|
|
1. (A leftjoin B on (Pab)) innerjoin C on (Pac)
|
|
|
|
= (A innerjoin C on (Pac)) leftjoin B on (Pab)
|
|
|
|
|
|
|
|
where Pac is a predicate referencing A and C, etc (in this case, clearly
|
|
|
|
Pac cannot reference B, or the transformation is nonsensical).
|
|
|
|
|
|
|
|
2. (A leftjoin B on (Pab)) leftjoin C on (Pac)
|
|
|
|
= (A leftjoin C on (Pac)) leftjoin B on (Pab)
|
|
|
|
|
|
|
|
3. (A leftjoin B on (Pab)) leftjoin C on (Pbc)
|
|
|
|
= A leftjoin (B leftjoin C on (Pbc)) on (Pab)
|
|
|
|
|
|
|
|
Identity 3 only holds if predicate Pbc must fail for all-null B rows
|
|
|
|
(that is, Pbc is strict for at least one column of B). If Pbc is not
|
|
|
|
strict, the first form might produce some rows with nonnull C columns
|
|
|
|
where the second form would make those entries null.
|
|
|
|
|
|
|
|
RIGHT JOIN is equivalent to LEFT JOIN after switching the two input
|
2009-02-27 23:41:38 +01:00
|
|
|
tables, so the same identities work for right joins.
|
2005-12-20 03:30:36 +01:00
|
|
|
|
|
|
|
An example of a case that does *not* work is moving an innerjoin into or
|
|
|
|
out of the nullable side of an outer join:
|
|
|
|
|
|
|
|
A leftjoin (B join C on (Pbc)) on (Pab)
|
|
|
|
!= (A leftjoin B on (Pab)) join C on (Pbc)
|
|
|
|
|
2009-02-27 23:41:38 +01:00
|
|
|
SEMI joins work a little bit differently. A semijoin can be reassociated
|
2009-07-21 04:02:44 +02:00
|
|
|
into or out of the lefthand side of another semijoin, left join, or
|
|
|
|
antijoin, but not into or out of the righthand side. Likewise, an inner
|
|
|
|
join, left join, or antijoin can be reassociated into or out of the
|
|
|
|
lefthand side of a semijoin, but not into or out of the righthand side.
|
2009-02-27 23:41:38 +01:00
|
|
|
|
|
|
|
ANTI joins work approximately like LEFT joins, except that identity 3
|
|
|
|
fails if the join to C is an antijoin (even if Pbc is strict, and in
|
|
|
|
both the cases where the other join is a leftjoin and where it is an
|
|
|
|
antijoin). So we can't reorder antijoins into or out of the RHS of a
|
|
|
|
leftjoin or antijoin, even if the relevant clause is strict.
|
|
|
|
|
|
|
|
The current code does not attempt to re-order FULL JOINs at all.
|
2005-12-20 03:30:36 +01:00
|
|
|
FULL JOIN ordering is enforced by not collapsing FULL JOIN nodes when
|
2009-02-27 23:41:38 +01:00
|
|
|
translating the jointree to "joinlist" representation. Other types of
|
2005-12-20 03:30:36 +01:00
|
|
|
JOIN nodes are normally collapsed so that they participate fully in the
|
|
|
|
join order search. To avoid generating illegal join orders, the planner
|
2009-02-27 23:41:38 +01:00
|
|
|
creates a SpecialJoinInfo node for each non-inner join, and join_is_legal
|
2005-12-20 03:30:36 +01:00
|
|
|
checks this list to decide if a proposed join is legal.
|
|
|
|
|
2008-08-14 20:48:00 +02:00
|
|
|
What we store in SpecialJoinInfo nodes are the minimum sets of Relids
|
2005-12-20 03:30:36 +01:00
|
|
|
required on each side of the join to form the outer join. Note that
|
|
|
|
these are minimums; there's no explicit maximum, since joining other
|
|
|
|
rels to the OJ's syntactic rels may be legal. Per identities 1 and 2,
|
|
|
|
non-FULL joins can be freely associated into the lefthand side of an
|
2009-02-27 23:41:38 +01:00
|
|
|
OJ, but in some cases they can't be associated into the righthand side.
|
2007-10-26 20:10:50 +02:00
|
|
|
So the restriction enforced by join_is_legal is that a proposed join
|
2007-02-13 03:31:03 +01:00
|
|
|
can't join a rel within or partly within an RHS boundary to one outside
|
2015-08-06 21:35:27 +02:00
|
|
|
the boundary, unless the proposed join is a LEFT join that can associate
|
|
|
|
into the SpecialJoinInfo's RHS using identity 3.
|
|
|
|
|
|
|
|
The use of minimum Relid sets has some pitfalls; consider a query like
|
|
|
|
A leftjoin (B leftjoin (C innerjoin D) on (Pbcd)) on Pa
|
|
|
|
where Pa doesn't mention B/C/D at all. In this case a naive computation
|
|
|
|
would give the upper leftjoin's min LHS as {A} and min RHS as {C,D} (since
|
|
|
|
we know that the innerjoin can't associate out of the leftjoin's RHS, and
|
|
|
|
enforce that by including its relids in the leftjoin's min RHS). And the
|
|
|
|
lower leftjoin has min LHS of {B} and min RHS of {C,D}. Given such
|
|
|
|
information, join_is_legal would think it's okay to associate the upper
|
|
|
|
join into the lower join's RHS, transforming the query to
|
|
|
|
B leftjoin (A leftjoin (C innerjoin D) on Pa) on (Pbcd)
|
2015-10-01 16:31:22 +02:00
|
|
|
which yields totally wrong answers. We prevent that by forcing the min RHS
|
2015-08-06 21:35:27 +02:00
|
|
|
for the upper join to include B. This is perhaps overly restrictive, but
|
|
|
|
such cases don't arise often so it's not clear that it's worth developing a
|
|
|
|
more complicated system.
|
2005-12-20 03:30:36 +01:00
|
|
|
|
2000-09-29 20:21:41 +02:00
|
|
|
|
2008-03-20 18:55:15 +01:00
|
|
|
Pulling Up Subqueries
|
2000-09-29 20:21:41 +02:00
|
|
|
---------------------
|
|
|
|
|
|
|
|
As we described above, a subquery appearing in the range table is planned
|
|
|
|
independently and treated as a "black box" during planning of the outer
|
|
|
|
query. This is necessary when the subquery uses features such as
|
|
|
|
aggregates, GROUP, or DISTINCT. But if the subquery is just a simple
|
|
|
|
scan or join, treating the subquery as a black box may produce a poor plan
|
|
|
|
compared to considering it as part of the entire plan search space.
|
|
|
|
Therefore, at the start of the planning process the planner looks for
|
|
|
|
simple subqueries and pulls them up into the main query's jointree.
|
|
|
|
|
|
|
|
Pulling up a subquery may result in FROM-list joins appearing below the top
|
|
|
|
of the join tree. Each FROM-list is planned using the dynamic-programming
|
|
|
|
search method described above.
|
|
|
|
|
|
|
|
If pulling up a subquery produces a FROM-list as a direct child of another
|
2005-12-20 03:30:36 +01:00
|
|
|
FROM-list, then we can merge the two FROM-lists together. Once that's
|
|
|
|
done, the subquery is an absolutely integral part of the outer query and
|
|
|
|
will not constrain the join tree search space at all. However, that could
|
|
|
|
result in unpleasant growth of planning time, since the dynamic-programming
|
|
|
|
search has runtime exponential in the number of FROM-items considered.
|
|
|
|
Therefore, we don't merge FROM-lists if the result would have too many
|
|
|
|
FROM-items in one list.
|
2000-09-12 23:07:18 +02:00
|
|
|
|
1999-02-08 05:29:25 +01:00
|
|
|
|
1999-02-04 04:19:11 +01:00
|
|
|
Optimizer Functions
|
|
|
|
-------------------
|
|
|
|
|
2000-03-21 06:12:12 +01:00
|
|
|
The primary entry point is planner().
|
|
|
|
|
1997-12-17 19:02:33 +01:00
|
|
|
planner()
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
set up for recursive handling of subqueries
|
2000-03-21 06:12:12 +01:00
|
|
|
-subquery_planner()
|
2008-08-14 20:48:00 +02:00
|
|
|
pull up sublinks and subqueries from rangetable, if possible
|
2000-03-21 06:12:12 +01:00
|
|
|
canonicalize qual
|
2003-12-30 22:49:19 +01:00
|
|
|
Attempt to simplify WHERE clause to the most useful form; this includes
|
|
|
|
flattening nested AND/ORs and detecting clauses that are duplicated in
|
|
|
|
different branches of an OR.
|
|
|
|
simplify constant expressions
|
2000-03-21 06:12:12 +01:00
|
|
|
process sublinks
|
|
|
|
convert Vars of outer query levels into Params
|
2000-11-12 01:37:02 +01:00
|
|
|
--grouping_planner()
|
|
|
|
preprocess target list for non-SELECT queries
|
|
|
|
handle UNION/INTERSECT/EXCEPT, GROUP BY, HAVING, aggregates,
|
|
|
|
ORDER BY, DISTINCT, LIMIT
|
2000-03-21 06:12:12 +01:00
|
|
|
--query_planner()
|
2002-11-06 01:00:45 +01:00
|
|
|
make list of base relations used in query
|
|
|
|
split up the qual into restrictions (a=1) and joins (b=c)
|
|
|
|
find qual clauses that enable merge and hash joins
|
1999-02-15 23:19:01 +01:00
|
|
|
----make_one_rel()
|
|
|
|
set_base_rel_pathlist()
|
2007-09-26 20:51:51 +02:00
|
|
|
find seqscan and all index paths for each base relation
|
1999-02-15 23:19:01 +01:00
|
|
|
find selectivity of columns used in joins
|
2007-09-26 20:51:51 +02:00
|
|
|
make_rel_from_joinlist()
|
|
|
|
hand off join subproblems to a plugin, GEQO, or standard_join_search()
|
|
|
|
-----standard_join_search()
|
|
|
|
call join_search_one_level() for each level of join tree needed
|
|
|
|
join_search_one_level():
|
2000-02-07 05:41:04 +01:00
|
|
|
For each joinrel of the prior level, do make_rels_by_clause_joins()
|
|
|
|
if it has join clauses, or make_rels_by_clauseless_joins() if not.
|
|
|
|
Also generate "bushy plan" joins between joinrels of lower levels.
|
2007-09-26 20:51:51 +02:00
|
|
|
Back at standard_join_search(), apply set_cheapest() to extract the
|
2000-02-07 05:41:04 +01:00
|
|
|
cheapest path for each newly constructed joinrel.
|
|
|
|
Loop back if this wasn't the top join level.
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
Back at grouping_planner:
|
|
|
|
do grouping (GROUP BY) and aggregation
|
|
|
|
do window functions
|
|
|
|
make unique (DISTINCT)
|
|
|
|
do sorting (ORDER BY)
|
|
|
|
do limit (LIMIT/OFFSET)
|
|
|
|
Back at planner():
|
|
|
|
convert finished Path tree into a Plan tree
|
|
|
|
do final cleanup after planning
|
1999-02-03 21:15:53 +01:00
|
|
|
|
|
|
|
|
1999-08-16 04:17:58 +02:00
|
|
|
Optimizer Data Structures
|
|
|
|
-------------------------
|
1999-02-04 04:19:11 +01:00
|
|
|
|
2007-02-19 08:03:34 +01:00
|
|
|
PlannerGlobal - global information for a single planner invocation
|
|
|
|
|
|
|
|
PlannerInfo - information for planning a particular Query (we make
|
|
|
|
a separate PlannerInfo node for each sub-Query)
|
2005-06-06 00:32:58 +02:00
|
|
|
|
1999-02-15 23:19:01 +01:00
|
|
|
RelOptInfo - a relation or joined relations
|
1999-02-04 04:19:11 +01:00
|
|
|
|
2003-01-15 20:35:48 +01:00
|
|
|
RestrictInfo - WHERE clauses, like "x = 3" or "y = z"
|
|
|
|
(note the same structure is used for restriction and
|
|
|
|
join clauses)
|
1999-02-04 04:19:11 +01:00
|
|
|
|
1999-02-15 23:19:01 +01:00
|
|
|
Path - every way to generate a RelOptInfo(sequential,index,joins)
|
2010-10-14 22:56:39 +02:00
|
|
|
SeqScan - represents a sequential scan plan
|
|
|
|
IndexPath - index scan
|
2005-04-21 21:18:13 +02:00
|
|
|
BitmapHeapPath - top of a bitmapped index scan
|
2002-11-06 01:00:45 +01:00
|
|
|
TidPath - scan by CTID
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
SubqueryScanPath - scan a subquery-in-FROM
|
2011-02-20 06:17:18 +01:00
|
|
|
ForeignPath - scan a foreign table
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
CustomPath - for custom scan providers
|
2002-11-06 01:00:45 +01:00
|
|
|
AppendPath - append multiple subpaths together
|
2010-10-14 22:56:39 +02:00
|
|
|
MergeAppendPath - merge multiple subpaths, preserving their common sort order
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
ResultPath - a childless Result plan node (used for FROM-less SELECT)
|
2002-11-30 06:21:03 +01:00
|
|
|
MaterialPath - a Material plan node
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
UniquePath - remove duplicate rows (either by hashing or sorting)
|
|
|
|
GatherPath - collect the results of parallel workers
|
|
|
|
ProjectionPath - a Result plan node with child (used for projection)
|
|
|
|
SortPath - a Sort plan node applied to some sub-path
|
|
|
|
GroupPath - a Group plan node applied to some sub-path
|
|
|
|
UpperUniquePath - a Unique plan node applied to some sub-path
|
|
|
|
AggPath - an Agg plan node applied to some sub-path
|
|
|
|
GroupingSetsPath - an Agg plan node used to implement GROUPING SETS
|
|
|
|
MinMaxAggPath - a Result plan node with subplans performing MIN/MAX
|
|
|
|
WindowAggPath - a WindowAgg plan node applied to some sub-path
|
|
|
|
SetOpPath - a SetOp plan node applied to some sub-path
|
|
|
|
RecursiveUnionPath - a RecursiveUnion plan node applied to two sub-paths
|
|
|
|
LockRowsPath - a LockRows plan node applied to some sub-path
|
|
|
|
ModifyTablePath - a ModifyTable plan node applied to some sub-path(s)
|
|
|
|
LimitPath - a Limit plan node applied to some sub-path
|
1999-08-16 04:17:58 +02:00
|
|
|
NestPath - nested-loop joins
|
1999-02-15 23:19:01 +01:00
|
|
|
MergePath - merge joins
|
|
|
|
HashPath - hash joins
|
1999-02-04 02:47:02 +01:00
|
|
|
|
2007-01-20 21:45:41 +01:00
|
|
|
EquivalenceClass - a data structure representing a set of values known equal
|
|
|
|
|
|
|
|
PathKey - a data structure representing the sort ordering of a path
|
1999-08-16 04:17:58 +02:00
|
|
|
|
|
|
|
The optimizer spends a good deal of its time worrying about the ordering
|
|
|
|
of the tuples returned by a path. The reason this is useful is that by
|
|
|
|
knowing the sort ordering of a path, we may be able to use that path as
|
|
|
|
the left or right input of a mergejoin and avoid an explicit sort step.
|
|
|
|
Nestloops and hash joins don't really care what the order of their inputs
|
|
|
|
is, but mergejoin needs suitably ordered inputs. Therefore, all paths
|
|
|
|
generated during the optimization process are marked with their sort order
|
|
|
|
(to the extent that it is known) for possible use by a higher-level merge.
|
|
|
|
|
|
|
|
It is also possible to avoid an explicit sort step to implement a user's
|
2000-02-07 05:41:04 +01:00
|
|
|
ORDER BY clause if the final path has the right ordering already, so the
|
Simplify query_planner's API by having it return the top-level RelOptInfo.
Formerly, query_planner returned one or possibly two Paths for the topmost
join relation, so that grouping_planner didn't see the join RelOptInfo
(at least not directly; it didn't have any hesitation about examining
cheapest_path->parent, though). However, correct selection of the Paths
involved a significant amount of coupling between query_planner and
grouping_planner, a problem which has gotten worse over time. It seems
best to give up on this API choice and instead return the topmost
RelOptInfo explicitly. Then grouping_planner can pull out the Paths it
wants from the rel's path list. In this way we can remove all knowledge
of grouping behaviors from query_planner.
The only real benefit of the old way is that in the case of an empty
FROM clause, we never made any RelOptInfos at all, just a Path. Now
we have to gin up a dummy RelOptInfo to represent the empty FROM clause.
That's not a very big deal though.
While at it, simplify query_planner's API a bit more by having the caller
set up root->tuple_fraction and root->limit_tuples, rather than passing
those values as separate parameters. Since query_planner no longer does
anything with either value, requiring it to fill the PlannerInfo fields
seemed pretty arbitrary.
This patch just rearranges code; it doesn't (intentionally) change any
behaviors. Followup patches will do more interesting things.
2013-08-05 21:00:57 +02:00
|
|
|
sort ordering is of interest even at the top level. grouping_planner() will
|
2000-02-07 05:41:04 +01:00
|
|
|
look for the cheapest path with a sort order matching the desired order,
|
Simplify query_planner's API by having it return the top-level RelOptInfo.
Formerly, query_planner returned one or possibly two Paths for the topmost
join relation, so that grouping_planner didn't see the join RelOptInfo
(at least not directly; it didn't have any hesitation about examining
cheapest_path->parent, though). However, correct selection of the Paths
involved a significant amount of coupling between query_planner and
grouping_planner, a problem which has gotten worse over time. It seems
best to give up on this API choice and instead return the topmost
RelOptInfo explicitly. Then grouping_planner can pull out the Paths it
wants from the rel's path list. In this way we can remove all knowledge
of grouping behaviors from query_planner.
The only real benefit of the old way is that in the case of an empty
FROM clause, we never made any RelOptInfos at all, just a Path. Now
we have to gin up a dummy RelOptInfo to represent the empty FROM clause.
That's not a very big deal though.
While at it, simplify query_planner's API a bit more by having the caller
set up root->tuple_fraction and root->limit_tuples, rather than passing
those values as separate parameters. Since query_planner no longer does
anything with either value, requiring it to fill the PlannerInfo fields
seemed pretty arbitrary.
This patch just rearranges code; it doesn't (intentionally) change any
behaviors. Followup patches will do more interesting things.
2013-08-05 21:00:57 +02:00
|
|
|
then compare its cost to the cost of using the cheapest-overall path and
|
|
|
|
doing an explicit sort on that.
|
1999-08-16 04:17:58 +02:00
|
|
|
|
|
|
|
When we are generating paths for a particular RelOptInfo, we discard a path
|
|
|
|
if it is more expensive than another known path that has the same or better
|
|
|
|
sort order. We will never discard a path that is the only known way to
|
2000-02-07 05:41:04 +01:00
|
|
|
achieve a given sort order (without an explicit sort, that is). In this
|
|
|
|
way, the next level up will have the maximum freedom to build mergejoins
|
|
|
|
without sorting, since it can pick from any of the paths retained for its
|
|
|
|
inputs.
|
1999-08-16 04:17:58 +02:00
|
|
|
|
2000-07-24 05:11:01 +02:00
|
|
|
|
2007-01-20 21:45:41 +01:00
|
|
|
EquivalenceClasses
|
|
|
|
------------------
|
|
|
|
|
|
|
|
During the deconstruct_jointree() scan of the query's qual clauses, we look
|
|
|
|
for mergejoinable equality clauses A = B whose applicability is not delayed
|
|
|
|
by an outer join; these are called "equivalence clauses". When we find
|
|
|
|
one, we create an EquivalenceClass containing the expressions A and B to
|
|
|
|
record this knowledge. If we later find another equivalence clause B = C,
|
|
|
|
we add C to the existing EquivalenceClass for {A B}; this may require
|
|
|
|
merging two existing EquivalenceClasses. At the end of the scan, we have
|
|
|
|
sets of values that are known all transitively equal to each other. We can
|
|
|
|
therefore use a comparison of any pair of the values as a restriction or
|
|
|
|
join clause (when these values are available at the scan or join, of
|
|
|
|
course); furthermore, we need test only one such comparison, not all of
|
|
|
|
them. Therefore, equivalence clauses are removed from the standard qual
|
|
|
|
distribution process. Instead, when preparing a restriction or join clause
|
|
|
|
list, we examine each EquivalenceClass to see if it can contribute a
|
|
|
|
clause, and if so we select an appropriate pair of values to compare. For
|
|
|
|
example, if we are trying to join A's relation to C's, we can generate the
|
|
|
|
clause A = C, even though this appeared nowhere explicitly in the original
|
|
|
|
query. This may allow us to explore join paths that otherwise would have
|
|
|
|
been rejected as requiring Cartesian-product joins.
|
|
|
|
|
|
|
|
Sometimes an EquivalenceClass may contain a pseudo-constant expression
|
|
|
|
(i.e., one not containing Vars or Aggs of the current query level, nor
|
|
|
|
volatile functions). In this case we do not follow the policy of
|
|
|
|
dynamically generating join clauses: instead, we dynamically generate
|
|
|
|
restriction clauses "var = const" wherever one of the variable members of
|
|
|
|
the class can first be computed. For example, if we have A = B and B = 42,
|
|
|
|
we effectively generate the restriction clauses A = 42 and B = 42, and then
|
|
|
|
we need not bother with explicitly testing the join clause A = B when the
|
|
|
|
relations are joined. In effect, all the class members can be tested at
|
|
|
|
relation-scan level and there's never a need for join tests.
|
|
|
|
|
|
|
|
The precise technical interpretation of an EquivalenceClass is that it
|
|
|
|
asserts that at any plan node where more than one of its member values
|
|
|
|
can be computed, output rows in which the values are not all equal may
|
|
|
|
be discarded without affecting the query result. (We require all levels
|
|
|
|
of the plan to enforce EquivalenceClasses, hence a join need not recheck
|
|
|
|
equality of values that were computable by one of its children.) For an
|
|
|
|
ordinary EquivalenceClass that is "valid everywhere", we can further infer
|
|
|
|
that the values are all non-null, because all mergejoinable operators are
|
|
|
|
strict. However, we also allow equivalence clauses that appear below the
|
|
|
|
nullable side of an outer join to form EquivalenceClasses; for these
|
|
|
|
classes, the interpretation is that either all the values are equal, or
|
|
|
|
all (except pseudo-constants) have gone to null. (This requires a
|
|
|
|
limitation that non-constant members be strict, else they might not go
|
|
|
|
to null when the other members do.) Consider for example
|
|
|
|
|
|
|
|
SELECT *
|
|
|
|
FROM a LEFT JOIN
|
|
|
|
(SELECT * FROM b JOIN c ON b.y = c.z WHERE b.y = 10) ss
|
|
|
|
ON a.x = ss.y
|
|
|
|
WHERE a.x = 42;
|
|
|
|
|
|
|
|
We can form the below-outer-join EquivalenceClass {b.y c.z 10} and thereby
|
|
|
|
apply c.z = 10 while scanning c. (The reason we disallow outerjoin-delayed
|
|
|
|
clauses from forming EquivalenceClasses is exactly that we want to be able
|
|
|
|
to push any derived clauses as far down as possible.) But once above the
|
|
|
|
outer join it's no longer necessarily the case that b.y = 10, and thus we
|
|
|
|
cannot use such EquivalenceClasses to conclude that sorting is unnecessary
|
|
|
|
(see discussion of PathKeys below).
|
|
|
|
|
|
|
|
In this example, notice also that a.x = ss.y (really a.x = b.y) is not an
|
|
|
|
equivalence clause because its applicability to b is delayed by the outer
|
|
|
|
join; thus we do not try to insert b.y into the equivalence class {a.x 42}.
|
|
|
|
But since we see that a.x has been equated to 42 above the outer join, we
|
|
|
|
are able to form a below-outer-join class {b.y 42}; this restriction can be
|
|
|
|
added because no b/c row not having b.y = 42 can contribute to the result
|
|
|
|
of the outer join, and so we need not compute such rows. Now this class
|
|
|
|
will get merged with {b.y c.z 10}, leading to the contradiction 10 = 42,
|
|
|
|
which lets the planner deduce that the b/c join need not be computed at all
|
|
|
|
because none of its rows can contribute to the outer join. (This gets
|
|
|
|
implemented as a gating Result filter, since more usually the potential
|
|
|
|
contradiction involves Param values rather than just Consts, and thus has
|
|
|
|
to be checked at runtime.)
|
|
|
|
|
|
|
|
To aid in determining the sort ordering(s) that can work with a mergejoin,
|
|
|
|
we mark each mergejoinable clause with the EquivalenceClasses of its left
|
|
|
|
and right inputs. For an equivalence clause, these are of course the same
|
|
|
|
EquivalenceClass. For a non-equivalence mergejoinable clause (such as an
|
|
|
|
outer-join qualification), we generate two separate EquivalenceClasses for
|
|
|
|
the left and right inputs. This may result in creating single-item
|
|
|
|
equivalence "classes", though of course these are still subject to merging
|
|
|
|
if other equivalence clauses are later found to bear on the same
|
|
|
|
expressions.
|
|
|
|
|
|
|
|
Another way that we may form a single-item EquivalenceClass is in creation
|
|
|
|
of a PathKey to represent a desired sort order (see below). This is a bit
|
|
|
|
different from the above cases because such an EquivalenceClass might
|
|
|
|
contain an aggregate function or volatile expression. (A clause containing
|
|
|
|
a volatile function will never be considered mergejoinable, even if its top
|
|
|
|
operator is mergejoinable, so there is no way for a volatile expression to
|
|
|
|
get into EquivalenceClasses otherwise. Aggregates are disallowed in WHERE
|
|
|
|
altogether, so will never be found in a mergejoinable clause.) This is just
|
|
|
|
a convenience to maintain a uniform PathKey representation: such an
|
2009-09-29 03:20:34 +02:00
|
|
|
EquivalenceClass will never be merged with any other. Note in particular
|
|
|
|
that a single-item EquivalenceClass {a.x} is *not* meant to imply an
|
|
|
|
assertion that a.x = a.x; the practical effect of this is that a.x could
|
|
|
|
be NULL.
|
2007-01-20 21:45:41 +01:00
|
|
|
|
|
|
|
An EquivalenceClass also contains a list of btree opfamily OIDs, which
|
|
|
|
determines what the equalities it represents actually "mean". All the
|
|
|
|
equivalence clauses that contribute to an EquivalenceClass must have
|
|
|
|
equality operators that belong to the same set of opfamilies. (Note: most
|
|
|
|
of the time, a particular equality operator belongs to only one family, but
|
|
|
|
it's possible that it belongs to more than one. We keep track of all the
|
|
|
|
families to ensure that we can make use of an index belonging to any one of
|
|
|
|
the families for mergejoin purposes.)
|
|
|
|
|
Revisit handling of UNION ALL subqueries with non-Var output columns.
In commit 57664ed25e5dea117158a2e663c29e60b3546e1c I tried to fix a bug
reported by Teodor Sigaev by making non-simple-Var output columns distinct
(by wrapping their expressions with dummy PlaceHolderVar nodes). This did
not work too well. Commit b28ffd0fcc583c1811e5295279e7d4366c3cae6c fixed
some ensuing problems with matching to child indexes, but per a recent
report from Claus Stadler, constraint exclusion of UNION ALL subqueries was
still broken, because constant-simplification didn't handle the injected
PlaceHolderVars well either. On reflection, the original patch was quite
misguided: there is no reason to expect that EquivalenceClass child members
will be distinct. So instead of trying to make them so, we should ensure
that we can cope with the situation when they're not.
Accordingly, this patch reverts the code changes in the above-mentioned
commits (though the regression test cases they added stay). Instead, I've
added assorted defenses to make sure that duplicate EC child members don't
cause any problems. Teodor's original problem ("MergeAppend child's
targetlist doesn't match MergeAppend") is addressed more directly by
revising prepare_sort_from_pathkeys to let the parent MergeAppend's sort
list guide creation of each child's sort list.
In passing, get rid of add_sort_column; as far as I can tell, testing for
duplicate sort keys at this stage is dead code. Certainly it doesn't
trigger often enough to be worth expending cycles on in ordinary queries.
And keeping the test would've greatly complicated the new logic in
prepare_sort_from_pathkeys, because comparing pathkey list entries against
a previous output array requires that we not skip any entries in the list.
Back-patch to 9.1, like the previous patches. The only known issue in
this area that wasn't caused by the ill-advised previous patches was the
MergeAppend planning failure, which of course is not relevant before 9.1.
It's possible that we need some of the new defenses against duplicate child
EC entries in older branches, but until there's some clear evidence of that
I'm going to refrain from back-patching further.
2012-03-16 18:11:12 +01:00
|
|
|
An EquivalenceClass can contain "em_is_child" members, which are copies
|
|
|
|
of members that contain appendrel parent relation Vars, transposed to
|
|
|
|
contain the equivalent child-relation variables or expressions. These
|
|
|
|
members are *not* full-fledged members of the EquivalenceClass and do not
|
|
|
|
affect the class's overall properties at all. They are kept only to
|
|
|
|
simplify matching of child-relation expressions to EquivalenceClasses.
|
|
|
|
Most operations on EquivalenceClasses should ignore child members.
|
|
|
|
|
2007-01-20 21:45:41 +01:00
|
|
|
|
2000-07-24 05:11:01 +02:00
|
|
|
PathKeys
|
|
|
|
--------
|
|
|
|
|
|
|
|
The PathKeys data structure represents what is known about the sort order
|
2007-01-20 21:45:41 +01:00
|
|
|
of the tuples generated by a particular Path. A path's pathkeys field is a
|
|
|
|
list of PathKey nodes, where the n'th item represents the n'th sort key of
|
|
|
|
the result. Each PathKey contains these fields:
|
2000-07-24 05:11:01 +02:00
|
|
|
|
2007-01-20 21:45:41 +01:00
|
|
|
* a reference to an EquivalenceClass
|
|
|
|
* a btree opfamily OID (must match one of those in the EC)
|
|
|
|
* a sort direction (ascending or descending)
|
|
|
|
* a nulls-first-or-last flag
|
|
|
|
|
|
|
|
The EquivalenceClass represents the value being sorted on. Since the
|
|
|
|
various members of an EquivalenceClass are known equal according to the
|
|
|
|
opfamily, we can consider a path sorted by any one of them to be sorted by
|
|
|
|
any other too; this is what justifies referencing the whole
|
|
|
|
EquivalenceClass rather than just one member of it.
|
2000-07-24 05:11:01 +02:00
|
|
|
|
|
|
|
In single/base relation RelOptInfo's, the Paths represent various ways
|
|
|
|
of scanning the relation and the resulting ordering of the tuples.
|
|
|
|
Sequential scan Paths have NIL pathkeys, indicating no known ordering.
|
|
|
|
Index scans have Path.pathkeys that represent the chosen index's ordering,
|
2007-01-20 21:45:41 +01:00
|
|
|
if any. A single-key index would create a single-PathKey list, while a
|
|
|
|
multi-column index generates a list with one element per index column.
|
|
|
|
(Actually, since an index can be scanned either forward or backward, there
|
|
|
|
are two possible sort orders and two possible PathKey lists it can
|
|
|
|
generate.)
|
|
|
|
|
2012-01-28 01:26:38 +01:00
|
|
|
Note that a bitmap scan has NIL pathkeys since we can say nothing about
|
|
|
|
the overall order of its result. Also, an indexscan on an unordered type
|
|
|
|
of index generates NIL pathkeys. However, we can always create a pathkey
|
|
|
|
by doing an explicit sort. The pathkeys for a Sort plan's output just
|
|
|
|
represent the sort key fields and the ordering operators used.
|
2000-07-24 05:11:01 +02:00
|
|
|
|
|
|
|
Things get more interesting when we consider joins. Suppose we do a
|
|
|
|
mergejoin between A and B using the mergeclause A.X = B.Y. The output
|
2007-01-20 21:45:41 +01:00
|
|
|
of the mergejoin is sorted by X --- but it is also sorted by Y. Again,
|
|
|
|
this can be represented by a PathKey referencing an EquivalenceClass
|
|
|
|
containing both X and Y.
|
|
|
|
|
|
|
|
With a little further thought, it becomes apparent that nestloop joins
|
|
|
|
can also produce sorted output. For example, if we do a nestloop join
|
|
|
|
between outer relation A and inner relation B, then any pathkeys relevant
|
|
|
|
to A are still valid for the join result: we have not altered the order of
|
|
|
|
the tuples from A. Even more interesting, if there was an equivalence clause
|
|
|
|
A.X=B.Y, and A.X was a pathkey for the outer relation A, then we can assert
|
|
|
|
that B.Y is a pathkey for the join result; X was ordered before and still
|
|
|
|
is, and the joined values of Y are equal to the joined values of X, so Y
|
2000-07-24 05:11:01 +02:00
|
|
|
must now be ordered too. This is true even though we used neither an
|
2007-01-20 21:45:41 +01:00
|
|
|
explicit sort nor a mergejoin on Y. (Note: hash joins cannot be counted
|
|
|
|
on to preserve the order of their outer relation, because the executor
|
|
|
|
might decide to "batch" the join, so we always set pathkeys to NIL for
|
|
|
|
a hashjoin path.) Exception: a RIGHT or FULL join doesn't preserve the
|
|
|
|
ordering of its outer relation, because it might insert nulls at random
|
|
|
|
points in the ordering.
|
|
|
|
|
|
|
|
In general, we can justify using EquivalenceClasses as the basis for
|
|
|
|
pathkeys because, whenever we scan a relation containing multiple
|
|
|
|
EquivalenceClass members or join two relations each containing
|
|
|
|
EquivalenceClass members, we apply restriction or join clauses derived from
|
|
|
|
the EquivalenceClass. This guarantees that any two values listed in the
|
|
|
|
EquivalenceClass are in fact equal in all tuples emitted by the scan or
|
|
|
|
join, and therefore that if the tuples are sorted by one of the values,
|
|
|
|
they can be considered sorted by any other as well. It does not matter
|
|
|
|
whether the test clause is used as a mergeclause, or merely enforced
|
|
|
|
after-the-fact as a qpqual filter.
|
|
|
|
|
|
|
|
Note that there is no particular difficulty in labeling a path's sort
|
|
|
|
order with a PathKey referencing an EquivalenceClass that contains
|
|
|
|
variables not yet joined into the path's output. We can simply ignore
|
|
|
|
such entries as not being relevant (yet). This makes it possible to
|
|
|
|
use the same EquivalenceClasses throughout the join planning process.
|
|
|
|
In fact, by being careful not to generate multiple identical PathKey
|
|
|
|
objects, we can reduce comparison of EquivalenceClasses and PathKeys
|
|
|
|
to simple pointer comparison, which is a huge savings because add_path
|
|
|
|
has to make a large number of PathKey comparisons in deciding whether
|
|
|
|
competing Paths are equivalently sorted.
|
2000-07-24 05:11:01 +02:00
|
|
|
|
|
|
|
Pathkeys are also useful to represent an ordering that we wish to achieve,
|
|
|
|
since they are easily compared to the pathkeys of a potential candidate
|
2008-08-02 23:32:01 +02:00
|
|
|
path. So, SortGroupClause lists are turned into pathkeys lists for use
|
|
|
|
inside the optimizer.
|
2000-07-24 05:11:01 +02:00
|
|
|
|
2000-12-14 23:30:45 +01:00
|
|
|
An additional refinement we can make is to insist that canonical pathkey
|
2007-01-20 21:45:41 +01:00
|
|
|
lists (sort orderings) do not mention the same EquivalenceClass more than
|
|
|
|
once. For example, in all these cases the second sort column is redundant,
|
|
|
|
because it cannot distinguish values that are the same according to the
|
|
|
|
first sort column:
|
|
|
|
SELECT ... ORDER BY x, x
|
|
|
|
SELECT ... ORDER BY x, x DESC
|
|
|
|
SELECT ... WHERE x = y ORDER BY x, y
|
|
|
|
Although a user probably wouldn't write "ORDER BY x,x" directly, such
|
|
|
|
redundancies are more probable once equivalence classes have been
|
|
|
|
considered. Also, the system may generate redundant pathkey lists when
|
|
|
|
computing the sort ordering needed for a mergejoin. By eliminating the
|
|
|
|
redundancy, we save time and improve planning, since the planner will more
|
|
|
|
easily recognize equivalent orderings as being equivalent.
|
|
|
|
|
|
|
|
Another interesting property is that if the underlying EquivalenceClass
|
|
|
|
contains a constant and is not below an outer join, then the pathkey is
|
|
|
|
completely redundant and need not be sorted by at all! Every row must
|
|
|
|
contain the same constant value, so there's no need to sort. (If the EC is
|
|
|
|
below an outer join, we still have to sort, since some of the rows might
|
|
|
|
have gone to null and others not. In this case we must be careful to pick
|
|
|
|
a non-const member to sort by. The assumption that all the non-const
|
|
|
|
members go to null at the same plan level is critical here, else they might
|
|
|
|
not produce the same sort order.) This might seem pointless because users
|
|
|
|
are unlikely to write "... WHERE x = 42 ORDER BY x", but it allows us to
|
|
|
|
recognize when particular index columns are irrelevant to the sort order:
|
|
|
|
if we have "... WHERE x = 42 ORDER BY y", scanning an index on (x,y)
|
|
|
|
produces correctly ordered data without a sort step. We used to have very
|
|
|
|
ugly ad-hoc code to recognize that in limited contexts, but discarding
|
|
|
|
constant ECs from pathkeys makes it happen cleanly and automatically.
|
|
|
|
|
|
|
|
You might object that a below-outer-join EquivalenceClass doesn't always
|
|
|
|
represent the same values at every level of the join tree, and so using
|
|
|
|
it to uniquely identify a sort order is dubious. This is true, but we
|
|
|
|
can avoid dealing with the fact explicitly because we always consider that
|
|
|
|
an outer join destroys any ordering of its nullable inputs. Thus, even
|
|
|
|
if a path was sorted by {a.x} below an outer join, we'll re-sort if that
|
|
|
|
sort ordering was important; and so using the same PathKey for both sort
|
|
|
|
orderings doesn't create any real problem.
|
|
|
|
|
|
|
|
|
2010-10-29 17:52:16 +02:00
|
|
|
Order of processing for EquivalenceClasses and PathKeys
|
|
|
|
-------------------------------------------------------
|
|
|
|
|
|
|
|
As alluded to above, there is a specific sequence of phases in the
|
|
|
|
processing of EquivalenceClasses and PathKeys during planning. During the
|
|
|
|
initial scanning of the query's quals (deconstruct_jointree followed by
|
|
|
|
reconsider_outer_join_clauses), we construct EquivalenceClasses based on
|
|
|
|
mergejoinable clauses found in the quals. At the end of this process,
|
|
|
|
we know all we can know about equivalence of different variables, so
|
|
|
|
subsequently there will be no further merging of EquivalenceClasses.
|
|
|
|
At that point it is possible to consider the EquivalenceClasses as
|
Postpone creation of pathkeys lists to fix bug #8049.
This patch gets rid of the concept of, and infrastructure for,
non-canonical PathKeys; we now only ever create canonical pathkey lists.
The need for non-canonical pathkeys came from the desire to have
grouping_planner initialize query_pathkeys and related pathkey lists before
calling query_planner. However, since query_planner didn't actually *do*
anything with those lists before they'd been made canonical, we can get rid
of the whole mess by just not creating the lists at all until the point
where we formerly canonicalized them.
There are several ways in which we could implement that without making
query_planner itself deal with grouping/sorting features (which are
supposed to be the province of grouping_planner). I chose to add a
callback function to query_planner's API; other alternatives would have
required adding more fields to PlannerInfo, which while not bad in itself
would create an ABI break for planner-related plugins in the 9.2 release
series. This still breaks ABI for anything that calls query_planner
directly, but it seems somewhat unlikely that there are any such plugins.
I had originally conceived of this change as merely a step on the way to
fixing bug #8049 from Teun Hoogendoorn; but it turns out that this fixes
that bug all by itself, as per the added regression test. The reason is
that now get_eclass_for_sort_expr is adding the ORDER BY expression at the
end of EquivalenceClass creation not the start, and so anything that is in
a multi-member EquivalenceClass has already been created with correct
em_nullable_relids. I am suspicious that there are related scenarios in
which we still need to teach get_eclass_for_sort_expr to compute correct
nullable_relids, but am not eager to risk destabilizing either 9.2 or 9.3
to fix bugs that are only hypothetical. So for the moment, do this and
stop here.
Back-patch to 9.2 but not to earlier branches, since they don't exhibit
this bug for lack of join-clause-movement logic that depends on
em_nullable_relids being correct. (We might have to revisit that choice
if any related bugs turn up.) In 9.2, don't change the signature of
make_pathkeys_for_sortclauses nor remove canonicalize_pathkeys, so as
not to risk more plugin breakage than we have to.
2013-04-29 20:49:01 +02:00
|
|
|
"canonical" and build canonical PathKeys that reference them. At this
|
|
|
|
time we construct PathKeys for the query's ORDER BY and related clauses.
|
|
|
|
(Any ordering expressions that do not appear elsewhere will result in
|
|
|
|
the creation of new EquivalenceClasses, but this cannot result in merging
|
|
|
|
existing classes, so canonical-ness is not lost.)
|
2010-10-29 17:52:16 +02:00
|
|
|
|
|
|
|
Because all the EquivalenceClasses are known before we begin path
|
|
|
|
generation, we can use them as a guide to which indexes are of interest:
|
|
|
|
if an index's column is not mentioned in any EquivalenceClass then that
|
|
|
|
index's sort order cannot possibly be helpful for the query. This allows
|
|
|
|
short-circuiting of much of the processing of create_index_paths() for
|
|
|
|
irrelevant indexes.
|
|
|
|
|
|
|
|
There are some cases where planner.c constructs additional
|
|
|
|
EquivalenceClasses and PathKeys after query_planner has completed.
|
|
|
|
In these cases, the extra ECs/PKs are needed to represent sort orders
|
|
|
|
that were not considered during query_planner. Such situations should be
|
|
|
|
minimized since it is impossible for query_planner to return a plan
|
2015-05-20 15:18:11 +02:00
|
|
|
producing such a sort order, meaning an explicit sort will always be needed.
|
2010-10-29 17:52:16 +02:00
|
|
|
Currently this happens only for queries involving multiple window functions
|
|
|
|
with different orderings, for which extra sorts are needed anyway.
|
2000-12-14 23:30:45 +01:00
|
|
|
|
2002-08-26 00:39:37 +02:00
|
|
|
|
2012-01-28 01:26:38 +01:00
|
|
|
Parameterized Paths
|
|
|
|
-------------------
|
|
|
|
|
|
|
|
The naive way to join two relations using a clause like WHERE A.X = B.Y
|
|
|
|
is to generate a nestloop plan like this:
|
|
|
|
|
|
|
|
NestLoop
|
|
|
|
Filter: A.X = B.Y
|
|
|
|
-> Seq Scan on A
|
|
|
|
-> Seq Scan on B
|
|
|
|
|
|
|
|
We can make this better by using a merge or hash join, but it still
|
|
|
|
requires scanning all of both input relations. If A is very small and B is
|
|
|
|
very large, but there is an index on B.Y, it can be enormously better to do
|
|
|
|
something like this:
|
|
|
|
|
|
|
|
NestLoop
|
|
|
|
-> Seq Scan on A
|
|
|
|
-> Index Scan using B_Y_IDX on B
|
|
|
|
Index Condition: B.Y = A.X
|
|
|
|
|
|
|
|
Here, we are expecting that for each row scanned from A, the nestloop
|
|
|
|
plan node will pass down the current value of A.X into the scan of B.
|
|
|
|
That allows the indexscan to treat A.X as a constant for any one
|
|
|
|
invocation, and thereby use it as an index key. This is the only plan type
|
|
|
|
that can avoid fetching all of B, and for small numbers of rows coming from
|
|
|
|
A, that will dominate every other consideration. (As A gets larger, this
|
|
|
|
gets less attractive, and eventually a merge or hash join will win instead.
|
|
|
|
So we have to cost out all the alternatives to decide what to do.)
|
|
|
|
|
|
|
|
It can be useful for the parameter value to be passed down through
|
|
|
|
intermediate layers of joins, for example:
|
|
|
|
|
|
|
|
NestLoop
|
|
|
|
-> Seq Scan on A
|
|
|
|
Hash Join
|
|
|
|
Join Condition: B.Y = C.W
|
|
|
|
-> Seq Scan on B
|
|
|
|
-> Index Scan using C_Z_IDX on C
|
|
|
|
Index Condition: C.Z = A.X
|
|
|
|
|
2015-02-28 18:43:04 +01:00
|
|
|
If all joins are plain inner joins then this is usually unnecessary,
|
|
|
|
because it's possible to reorder the joins so that a parameter is used
|
2012-01-28 01:26:38 +01:00
|
|
|
immediately below the nestloop node that provides it. But in the
|
2015-02-28 18:43:04 +01:00
|
|
|
presence of outer joins, such join reordering may not be possible.
|
|
|
|
|
|
|
|
Also, the bottom-level scan might require parameters from more than one
|
|
|
|
other relation. In principle we could join the other relations first
|
|
|
|
so that all the parameters are supplied from a single nestloop level.
|
|
|
|
But if those other relations have no join clause in common (which is
|
|
|
|
common in star-schema queries for instance), the planner won't consider
|
|
|
|
joining them directly to each other. In such a case we need to be able
|
|
|
|
to create a plan like
|
|
|
|
|
|
|
|
NestLoop
|
|
|
|
-> Seq Scan on SmallTable1 A
|
|
|
|
NestLoop
|
|
|
|
-> Seq Scan on SmallTable2 B
|
|
|
|
NestLoop
|
|
|
|
-> Index Scan using XYIndex on LargeTable C
|
|
|
|
Index Condition: C.X = A.AID and C.Y = B.BID
|
|
|
|
|
|
|
|
so we should be willing to pass down A.AID through a join even though
|
|
|
|
there is no join order constraint forcing the plan to look like this.
|
|
|
|
|
|
|
|
Before version 9.2, Postgres used ad-hoc methods for planning and
|
|
|
|
executing nestloop queries of this kind, and those methods could not
|
|
|
|
handle passing parameters down through multiple join levels.
|
2012-01-28 01:26:38 +01:00
|
|
|
|
|
|
|
To plan such queries, we now use a notion of a "parameterized path",
|
|
|
|
which is a path that makes use of a join clause to a relation that's not
|
2015-02-28 18:43:04 +01:00
|
|
|
scanned by the path. In the example two above, we would construct a
|
2012-01-28 01:26:38 +01:00
|
|
|
path representing the possibility of doing this:
|
|
|
|
|
|
|
|
-> Index Scan using C_Z_IDX on C
|
|
|
|
Index Condition: C.Z = A.X
|
|
|
|
|
|
|
|
This path will be marked as being parameterized by relation A. (Note that
|
|
|
|
this is only one of the possible access paths for C; we'd still have a
|
|
|
|
plain unparameterized seqscan, and perhaps other possibilities.) The
|
|
|
|
parameterization marker does not prevent joining the path to B, so one of
|
|
|
|
the paths generated for the joinrel {B C} will represent
|
|
|
|
|
|
|
|
Hash Join
|
|
|
|
Join Condition: B.Y = C.W
|
|
|
|
-> Seq Scan on B
|
|
|
|
-> Index Scan using C_Z_IDX on C
|
|
|
|
Index Condition: C.Z = A.X
|
|
|
|
|
|
|
|
This path is still marked as being parameterized by A. When we attempt to
|
|
|
|
join {B C} to A to form the complete join tree, such a path can only be
|
|
|
|
used as the inner side of a nestloop join: it will be ignored for other
|
|
|
|
possible join types. So we will form a join path representing the query
|
|
|
|
plan shown above, and it will compete in the usual way with paths built
|
|
|
|
from non-parameterized scans.
|
|
|
|
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
While all ordinary paths for a particular relation generate the same set
|
|
|
|
of rows (since they must all apply the same set of restriction clauses),
|
|
|
|
parameterized paths typically generate fewer rows than less-parameterized
|
|
|
|
paths, since they have additional clauses to work with. This means we
|
|
|
|
must consider the number of rows generated as an additional figure of
|
|
|
|
merit. A path that costs more than another, but generates fewer rows,
|
|
|
|
must be kept since the smaller number of rows might save work at some
|
|
|
|
intermediate join level. (It would not save anything if joined
|
|
|
|
immediately to the source of the parameters.)
|
|
|
|
|
|
|
|
To keep cost estimation rules relatively simple, we make an implementation
|
|
|
|
restriction that all paths for a given relation of the same parameterization
|
|
|
|
(i.e., the same set of outer relations supplying parameters) must have the
|
|
|
|
same rowcount estimate. This is justified by insisting that each such path
|
|
|
|
apply *all* join clauses that are available with the named outer relations.
|
|
|
|
Different paths might, for instance, choose different join clauses to use
|
|
|
|
as index clauses; but they must then apply any other join clauses available
|
|
|
|
from the same outer relations as filter conditions, so that the set of rows
|
|
|
|
returned is held constant. This restriction doesn't degrade the quality of
|
|
|
|
the finished plan: it amounts to saying that we should always push down
|
|
|
|
movable join clauses to the lowest possible evaluation level, which is a
|
|
|
|
good thing anyway. The restriction is useful in particular to support
|
|
|
|
pre-filtering of join paths in add_path_precheck. Without this rule we
|
|
|
|
could never reject a parameterized path in advance of computing its rowcount
|
|
|
|
estimate, which would greatly reduce the value of the pre-filter mechanism.
|
|
|
|
|
2012-01-28 01:26:38 +01:00
|
|
|
To limit planning time, we have to avoid generating an unreasonably large
|
|
|
|
number of parameterized paths. We do this by only generating parameterized
|
|
|
|
relation scan paths for index scans, and then only for indexes for which
|
|
|
|
suitable join clauses are available. There are also heuristics in join
|
|
|
|
planning that try to limit the number of parameterized paths considered.
|
|
|
|
|
|
|
|
In particular, there's been a deliberate policy decision to favor hash
|
|
|
|
joins over merge joins for parameterized join steps (those occurring below
|
|
|
|
a nestloop that provides parameters to the lower join's inputs). While we
|
|
|
|
do not ignore merge joins entirely, joinpath.c does not fully explore the
|
|
|
|
space of potential merge joins with parameterized inputs. Also, add_path
|
|
|
|
treats parameterized paths as having no pathkeys, so that they compete
|
Fix planner's cost estimation for SEMI/ANTI joins with inner indexscans.
When the inner side of a nestloop SEMI or ANTI join is an indexscan that
uses all the join clauses as indexquals, it can be presumed that both
matched and unmatched outer rows will be processed very quickly: for
matched rows, we'll stop after fetching one row from the indexscan, while
for unmatched rows we'll have an indexscan that finds no matching index
entries, which should also be quick. The planner already knew about this,
but it was nonetheless charging for at least one full run of the inner
indexscan, as a consequence of concerns about the behavior of materialized
inner scans --- but those concerns don't apply in the fast case. If the
inner side has low cardinality (many matching rows) this could make an
indexscan plan look far more expensive than it actually is. To fix,
rearrange the work in initial_cost_nestloop/final_cost_nestloop so that we
don't add the inner scan cost until we've inspected the indexquals, and
then we can add either the full-run cost or just the first tuple's cost as
appropriate.
Experimentation with this fix uncovered another problem: add_path and
friends were coded to disregard cheap startup cost when considering
parameterized paths. That's usually okay (and desirable, because it thins
the path herd faster); but in this fast case for SEMI/ANTI joins, it could
result in throwing away the desired plain indexscan path in favor of a
bitmap scan path before we ever get to the join costing logic. In the
many-matching-rows cases of interest here, a bitmap scan will do a lot more
work than required, so this is a problem. To fix, add a per-relation flag
consider_param_startup that works like the existing consider_startup flag,
but applies to parameterized paths, and set it for relations that are the
inside of a SEMI or ANTI join.
To make this patch reasonably safe to back-patch, care has been taken to
avoid changing the planner's behavior except in the very narrow case of
SEMI/ANTI joins with inner indexscans. There are places in
compare_path_costs_fuzzily and add_path_precheck that are not terribly
consistent with the new approach, but changing them will affect planner
decisions at the margins in other cases, so we'll leave that for a
HEAD-only fix.
Back-patch to 9.3; before that, the consider_startup flag didn't exist,
meaning that the second aspect of the patch would be too invasive.
Per a complaint from Peter Holzer and analysis by Tomas Vondra.
2015-06-03 17:58:47 +02:00
|
|
|
only on cost and rowcount; they don't get preference for producing a
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
special sort order. This creates additional bias against merge joins,
|
|
|
|
since we might discard a path that could have been useful for performing
|
|
|
|
a merge without an explicit sort step. Since a parameterized path must
|
|
|
|
ultimately be used on the inside of a nestloop, where its sort order is
|
|
|
|
uninteresting, these choices do not affect any requirement for the final
|
|
|
|
output order of a query --- they only make it harder to use a merge join
|
|
|
|
at a lower level. The savings in planning work justifies that.
|
2012-01-28 01:26:38 +01:00
|
|
|
|
Fix planner's cost estimation for SEMI/ANTI joins with inner indexscans.
When the inner side of a nestloop SEMI or ANTI join is an indexscan that
uses all the join clauses as indexquals, it can be presumed that both
matched and unmatched outer rows will be processed very quickly: for
matched rows, we'll stop after fetching one row from the indexscan, while
for unmatched rows we'll have an indexscan that finds no matching index
entries, which should also be quick. The planner already knew about this,
but it was nonetheless charging for at least one full run of the inner
indexscan, as a consequence of concerns about the behavior of materialized
inner scans --- but those concerns don't apply in the fast case. If the
inner side has low cardinality (many matching rows) this could make an
indexscan plan look far more expensive than it actually is. To fix,
rearrange the work in initial_cost_nestloop/final_cost_nestloop so that we
don't add the inner scan cost until we've inspected the indexquals, and
then we can add either the full-run cost or just the first tuple's cost as
appropriate.
Experimentation with this fix uncovered another problem: add_path and
friends were coded to disregard cheap startup cost when considering
parameterized paths. That's usually okay (and desirable, because it thins
the path herd faster); but in this fast case for SEMI/ANTI joins, it could
result in throwing away the desired plain indexscan path in favor of a
bitmap scan path before we ever get to the join costing logic. In the
many-matching-rows cases of interest here, a bitmap scan will do a lot more
work than required, so this is a problem. To fix, add a per-relation flag
consider_param_startup that works like the existing consider_startup flag,
but applies to parameterized paths, and set it for relations that are the
inside of a SEMI or ANTI join.
To make this patch reasonably safe to back-patch, care has been taken to
avoid changing the planner's behavior except in the very narrow case of
SEMI/ANTI joins with inner indexscans. There are places in
compare_path_costs_fuzzily and add_path_precheck that are not terribly
consistent with the new approach, but changing them will affect planner
decisions at the margins in other cases, so we'll leave that for a
HEAD-only fix.
Back-patch to 9.3; before that, the consider_startup flag didn't exist,
meaning that the second aspect of the patch would be too invasive.
Per a complaint from Peter Holzer and analysis by Tomas Vondra.
2015-06-03 17:58:47 +02:00
|
|
|
Similarly, parameterized paths do not normally get preference in add_path
|
|
|
|
for having cheap startup cost; that's seldom of much value when on the
|
|
|
|
inside of a nestloop, so it seems not worth keeping extra paths solely for
|
|
|
|
that. An exception occurs for parameterized paths for the RHS relation of
|
|
|
|
a SEMI or ANTI join: in those cases, we can stop the inner scan after the
|
|
|
|
first match, so it's primarily startup not total cost that we care about.
|
|
|
|
|
2012-01-28 01:26:38 +01:00
|
|
|
|
Adjust definition of cheapest_total_path to work better with LATERAL.
In the initial cut at LATERAL, I kept the rule that cheapest_total_path
was always unparameterized, which meant it had to be NULL if the relation
has no unparameterized paths. It turns out to work much more nicely if
we always have *some* path nominated as cheapest-total for each relation.
In particular, let's still say it's the cheapest unparameterized path if
there is one; if not, take the cheapest-total-cost path among those of
the minimum available parameterization. (The first rule is actually
a special case of the second.)
This allows reversion of some temporary lobotomizations I'd put in place.
In particular, the planner can now consider hash and merge joins for
joins below a parameter-supplying nestloop, even if there aren't any
unparameterized paths available. This should bring planning of
LATERAL-containing queries to the same level as queries not using that
feature.
Along the way, simplify management of parameterized paths in add_path()
and friends. In the original coding for parameterized paths in 9.2,
I tried to minimize the logic changes in add_path(), so it just treated
parameterization as yet another dimension of comparison for paths.
We later made it ignore pathkeys (sort ordering) of parameterized paths,
on the grounds that ordering isn't a useful property for the path on the
inside of a nestloop, so we might as well get rid of useless parameterized
paths as quickly as possible. But we didn't take that reasoning as far as
we should have. Startup cost isn't a useful property inside a nestloop
either, so add_path() ought to discount startup cost of parameterized paths
as well. Having done that, the secondary sorting I'd implemented (in
add_parameterized_path) is no longer needed --- any parameterized path that
survives add_path() at all is worth considering at higher levels. So this
should be a bit faster as well as simpler.
2012-08-30 04:05:27 +02:00
|
|
|
LATERAL subqueries
|
|
|
|
------------------
|
|
|
|
|
|
|
|
As of 9.3 we support SQL-standard LATERAL references from subqueries in
|
|
|
|
FROM (and also functions in FROM). The planner implements these by
|
|
|
|
generating parameterized paths for any RTE that contains lateral
|
|
|
|
references. In such cases, *all* paths for that relation will be
|
|
|
|
parameterized by at least the set of relations used in its lateral
|
|
|
|
references. (And in turn, join relations including such a subquery might
|
|
|
|
not have any unparameterized paths.) All the other comments made above for
|
|
|
|
parameterized paths still apply, though; in particular, each such path is
|
|
|
|
still expected to enforce any join clauses that can be pushed down to it,
|
|
|
|
so that all paths of the same parameterization have the same rowcount.
|
|
|
|
|
2013-08-18 02:22:37 +02:00
|
|
|
We also allow LATERAL subqueries to be flattened (pulled up into the parent
|
2013-08-19 19:19:25 +02:00
|
|
|
query) by the optimizer, but only when this does not introduce lateral
|
|
|
|
references into JOIN/ON quals that would refer to relations outside the
|
|
|
|
lowest outer join at/above that qual. The semantics of such a qual would
|
|
|
|
be unclear. Note that even with this restriction, pullup of a LATERAL
|
2013-08-18 02:22:37 +02:00
|
|
|
subquery can result in creating PlaceHolderVars that contain lateral
|
|
|
|
references to relations outside their syntactic scope. We still evaluate
|
|
|
|
such PHVs at their syntactic location or lower, but the presence of such a
|
|
|
|
PHV in the quals or targetlist of a plan node requires that node to appear
|
|
|
|
on the inside of a nestloop join relative to the rel(s) supplying the
|
|
|
|
lateral reference. (Perhaps now that that stuff works, we could relax the
|
|
|
|
pullup restriction?)
|
|
|
|
|
Adjust definition of cheapest_total_path to work better with LATERAL.
In the initial cut at LATERAL, I kept the rule that cheapest_total_path
was always unparameterized, which meant it had to be NULL if the relation
has no unparameterized paths. It turns out to work much more nicely if
we always have *some* path nominated as cheapest-total for each relation.
In particular, let's still say it's the cheapest unparameterized path if
there is one; if not, take the cheapest-total-cost path among those of
the minimum available parameterization. (The first rule is actually
a special case of the second.)
This allows reversion of some temporary lobotomizations I'd put in place.
In particular, the planner can now consider hash and merge joins for
joins below a parameter-supplying nestloop, even if there aren't any
unparameterized paths available. This should bring planning of
LATERAL-containing queries to the same level as queries not using that
feature.
Along the way, simplify management of parameterized paths in add_path()
and friends. In the original coding for parameterized paths in 9.2,
I tried to minimize the logic changes in add_path(), so it just treated
parameterization as yet another dimension of comparison for paths.
We later made it ignore pathkeys (sort ordering) of parameterized paths,
on the grounds that ordering isn't a useful property for the path on the
inside of a nestloop, so we might as well get rid of useless parameterized
paths as quickly as possible. But we didn't take that reasoning as far as
we should have. Startup cost isn't a useful property inside a nestloop
either, so add_path() ought to discount startup cost of parameterized paths
as well. Having done that, the secondary sorting I'd implemented (in
add_parameterized_path) is no longer needed --- any parameterized path that
survives add_path() at all is worth considering at higher levels. So this
should be a bit faster as well as simpler.
2012-08-30 04:05:27 +02:00
|
|
|
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
Post scan/join planning
|
|
|
|
-----------------------
|
|
|
|
|
|
|
|
So far we have discussed only scan/join planning, that is, implementation
|
|
|
|
of the FROM and WHERE clauses of a SQL query. But the planner must also
|
|
|
|
determine how to deal with GROUP BY, aggregation, and other higher-level
|
|
|
|
features of queries; and in many cases there are multiple ways to do these
|
|
|
|
steps and thus opportunities for optimization choices. These steps, like
|
|
|
|
scan/join planning, are handled by constructing Paths representing the
|
|
|
|
different ways to do a step, then choosing the cheapest Path.
|
|
|
|
|
|
|
|
Since all Paths require a RelOptInfo as "parent", we create RelOptInfos
|
|
|
|
representing the outputs of these upper-level processing steps. These
|
|
|
|
RelOptInfos are mostly dummy, but their pathlist lists hold all the Paths
|
|
|
|
considered useful for each step. Currently, we may create these types of
|
|
|
|
additional RelOptInfos during upper-level planning:
|
|
|
|
|
|
|
|
UPPERREL_SETOP result of UNION/INTERSECT/EXCEPT, if any
|
|
|
|
UPPERREL_GROUP_AGG result of grouping/aggregation, if any
|
|
|
|
UPPERREL_WINDOW result of window functions, if any
|
|
|
|
UPPERREL_DISTINCT result of "SELECT DISTINCT", if any
|
|
|
|
UPPERREL_ORDERED result of ORDER BY, if any
|
|
|
|
UPPERREL_FINAL result of any remaining top-level actions
|
|
|
|
|
|
|
|
UPPERREL_FINAL is used to represent any final processing steps, currently
|
|
|
|
LockRows (SELECT FOR UPDATE), LIMIT/OFFSET, and ModifyTable. There is no
|
|
|
|
flexibility about the order in which these steps are done, and thus no need
|
|
|
|
to subdivide this stage more finely.
|
|
|
|
|
|
|
|
These "upper relations" are identified by the UPPERREL enum values shown
|
|
|
|
above, plus a relids set, which allows there to be more than one upperrel
|
|
|
|
of the same kind. We use NULL for the relids if there's no need for more
|
|
|
|
than one upperrel of the same kind. Currently, in fact, the relids set
|
|
|
|
is vestigial because it's always NULL, but that's expected to change in
|
|
|
|
future. For example, in planning set operations, we might need the relids
|
|
|
|
to denote which subset of the leaf SELECTs has been combined in a
|
|
|
|
particular group of Paths that are competing with each other.
|
|
|
|
|
|
|
|
The result of subquery_planner() is always returned as a set of Paths
|
|
|
|
stored in the UPPERREL_FINAL rel with NULL relids. The other types of
|
|
|
|
upperrels are created only if needed for the particular query.
|
|
|
|
|
|
|
|
The upper-relation infrastructure is designed so that things will work
|
|
|
|
properly if a particular upper relation is created and Paths are added
|
|
|
|
to it sooner than would normally happen. This allows, for example,
|
|
|
|
for an FDW's GetForeignPaths function to insert a Path representing
|
|
|
|
remote aggregation into the UPPERREL_GROUP_AGG upperrel, if it notices
|
|
|
|
that the query represents an aggregation that could be done entirely on
|
|
|
|
the foreign server. That Path will then compete with Paths representing
|
|
|
|
local aggregation on a regular scan of the foreign table, once the core
|
2016-03-15 01:04:44 +01:00
|
|
|
planner reaches the point of considering aggregation. (In practice,
|
|
|
|
it will usually be more convenient for FDWs to detect such cases in a
|
|
|
|
GetForeignUpperPaths callback; but that still represents injecting a
|
|
|
|
Path before the core code has touched the corresponding upperrel.)
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
|
|
|
|
|
2016-01-20 20:29:22 +01:00
|
|
|
Parallel Query and Partial Paths
|
|
|
|
--------------------------------
|
|
|
|
|
|
|
|
Parallel query involves dividing up the work that needs to be performed
|
|
|
|
either by an entire query or some portion of the query in such a way that
|
|
|
|
some of that work can be done by one or more worker processes, which are
|
|
|
|
called parallel workers. Parallel workers are a subtype of dynamic
|
|
|
|
background workers; see src/backend/access/transam/README.parallel for a
|
|
|
|
fuller description. Academic literature on parallel query suggests that
|
|
|
|
that parallel execution strategies can be divided into essentially two
|
|
|
|
categories: pipelined parallelism, where the execution of the query is
|
|
|
|
divided into multiple stages and each stage is handled by a separate
|
|
|
|
process; and partitioning parallelism, where the data is split between
|
|
|
|
multiple processes and each process handles a subset of it. The
|
|
|
|
literature, however, suggests that gains from pipeline parallelism are
|
|
|
|
often very limited due to the difficulty of avoiding pipeline stalls.
|
|
|
|
Consequently, we do not currently attempt to generate query plans that
|
|
|
|
use this technique.
|
|
|
|
|
2016-03-08 03:48:17 +01:00
|
|
|
Instead, we focus on partitioning parallelism, which does not require
|
2016-01-20 20:29:22 +01:00
|
|
|
that the underlying table be partitioned. It only requires that (1)
|
|
|
|
there is some method of dividing the data from at least one of the base
|
|
|
|
tables involved in the relation across multiple processes, (2) allowing
|
|
|
|
each process to handle its own portion of the data, and then (3)
|
|
|
|
collecting the results. Requirements (2) and (3) is satisfied by the
|
|
|
|
executor node Gather, which launches any number of worker processes and
|
|
|
|
executes its single child plan in all of them (and perhaps in the leader
|
|
|
|
also, if the children aren't generating enough data to keep the leader
|
|
|
|
busy). Requirement (1) is handled by the SeqScan node: when invoked
|
|
|
|
with parallel_aware = true, this node will, in effect, partition the
|
|
|
|
table on a block by block basis, returning a subset of the tuples from
|
|
|
|
the relation in each worker where that SeqScan is executed. A similar
|
|
|
|
scheme could be (and probably should be) implemented for bitmap heap
|
|
|
|
scans.
|
|
|
|
|
|
|
|
Just as we do for non-parallel access methods, we build Paths to
|
|
|
|
represent access strategies that can be used in a parallel plan. These
|
|
|
|
are, in essence, the same strategies that are available in the
|
|
|
|
non-parallel plan, but there is an important difference: a path that
|
|
|
|
will run beneath a Gather node returns only a subset of the query
|
|
|
|
results in each worker, not all of them. To form a path that can
|
|
|
|
actually be executed, the (rather large) cost of the Gather node must be
|
|
|
|
accounted for. For this reason among others, paths intended to run
|
|
|
|
beneath a Gather node - which we call "partial" paths since they return
|
|
|
|
only a subset of the results in each worker - must be kept separate from
|
|
|
|
ordinary paths (see RelOptInfo's partial_pathlist and the function
|
|
|
|
add_partial_path).
|
|
|
|
|
|
|
|
One of the keys to making parallel query effective is to run as much of
|
|
|
|
the query in parallel as possible. Therefore, we expect it to generally
|
|
|
|
be desirable to postpone the Gather stage until as near to the top of the
|
|
|
|
plan as possible. Expanding the range of cases in which more work can be
|
|
|
|
pushed below the Gather (and costly them accurately) is likely to keep us
|
|
|
|
busy for a long time to come.
|