1
/*-------------------------------------------------------------------------
4
* Definitions for planner's internal data structures.
7
* Portions Copyright (c) 1996-2005, PostgreSQL Global Development Group
8
* Portions Copyright (c) 1994, Regents of the University of California
10
* $PostgreSQL: pgsql/src/include/nodes/relation.h,v 1.102 2004-12-31 22:03:34 pgsql Exp $
12
*-------------------------------------------------------------------------
17
#include "access/sdir.h"
18
#include "nodes/bitmapset.h"
19
#include "nodes/parsenodes.h"
20
#include "storage/block.h"
25
* Set of relation identifiers (indexes into the rangetable).
28
typedef Bitmapset *Relids;
31
* When looking for a "cheapest path", this enum specifies whether we want
32
* cheapest startup cost or cheapest total cost.
34
typedef enum CostSelector
36
STARTUP_COST, TOTAL_COST
40
* The cost estimate produced by cost_qual_eval() includes both a one-time
41
* (startup) cost, and a per-tuple cost.
43
typedef struct QualCost
45
Cost startup; /* one-time cost */
46
Cost per_tuple; /* per-evaluation cost */
51
* Per-relation information for planning/optimization
53
* For planning purposes, a "base rel" is either a plain relation (a table)
54
* or the output of a sub-SELECT or function that appears in the range table.
55
* In either case it is uniquely identified by an RT index. A "joinrel"
56
* is the joining of two or more base rels. A joinrel is identified by
57
* the set of RT indexes for its component baserels. We create RelOptInfo
58
* nodes for each baserel and joinrel, and store them in the Query's
59
* base_rel_list and join_rel_list respectively.
61
* Note that there is only one joinrel for any given set of component
62
* baserels, no matter what order we assemble them in; so an unordered
63
* set is the right datatype to identify it with.
65
* We also have "other rels", which are like base rels in that they refer to
66
* single RT indexes; but they are not part of the join tree, and are stored
67
* in other_rel_list not base_rel_list.
69
* Currently the only kind of otherrels are those made for child relations
70
* of an inheritance scan (SELECT FROM foo*). The parent table's RTE and
71
* corresponding baserel represent the whole result of the inheritance scan.
72
* The planner creates separate RTEs and associated RelOptInfos for each child
73
* table (including the parent table, in its capacity as a member of the
74
* inheritance set). These RelOptInfos are physically identical to baserels,
75
* but are otherrels because they are not in the main join tree. These added
76
* RTEs and otherrels are used to plan the scans of the individual tables in
77
* the inheritance set; then the parent baserel is given an Append plan
78
* comprising the best plans for the individual child tables.
80
* At one time we also made otherrels to represent join RTEs, for use in
81
* handling join alias Vars. Currently this is not needed because all join
82
* alias Vars are expanded to non-aliased form during preprocess_expression.
84
* Parts of this data structure are specific to various scan and join
85
* mechanisms. It didn't seem worth creating new node types for them.
87
* relids - Set of base-relation identifiers; it is a base relation
88
* if there is just one, a join relation if more than one
89
* rows - estimated number of tuples in the relation after restriction
90
* clauses have been applied (ie, output rows of a plan for it)
91
* width - avg. number of bytes per tuple in the relation after the
92
* appropriate projections have been done (ie, output width)
93
* reltargetlist - List of Var nodes for the attributes we need to
94
* output from this relation (in no particular order)
95
* NOTE: in a child relation, may contain RowExprs
96
* pathlist - List of Path nodes, one for each potentially useful
97
* method of generating the relation
98
* cheapest_startup_path - the pathlist member with lowest startup cost
99
* (regardless of its ordering)
100
* cheapest_total_path - the pathlist member with lowest total cost
101
* (regardless of its ordering)
102
* cheapest_unique_path - for caching cheapest path to produce unique
103
* (no duplicates) output from relation
105
* If the relation is a base relation it will have these fields set:
107
* relid - RTE index (this is redundant with the relids field, but
108
* is provided for convenience of access)
109
* rtekind - distinguishes plain relation, subquery, or function RTE
110
* min_attr, max_attr - range of valid AttrNumbers for rel
111
* attr_needed - array of bitmapsets indicating the highest joinrel
112
* in which each attribute is needed; if bit 0 is set then
113
* the attribute is needed as part of final targetlist
114
* attr_widths - cache space for per-attribute width estimates;
115
* zero means not computed yet
116
* indexlist - list of IndexOptInfo nodes for relation's indexes
117
* (always NIL if it's not a table)
118
* pages - number of disk pages in relation (zero if not a table)
119
* tuples - number of tuples in relation (not considering restrictions)
120
* subplan - plan for subquery (NULL if it's not a subquery)
122
* Note: for a subquery, tuples and subplan are not set immediately
123
* upon creation of the RelOptInfo object; they are filled in when
124
* set_base_rel_pathlist processes the object.
126
* For otherrels that are inheritance children, these fields are filled
127
* in just as for a baserel.
129
* The presence of the remaining fields depends on the restrictions
130
* and joins that the relation participates in:
132
* baserestrictinfo - List of RestrictInfo nodes, containing info about
133
* each qualification clause in which this relation
134
* participates (only used for base rels)
135
* baserestrictcost - Estimated cost of evaluating the baserestrictinfo
136
* clauses at a single tuple (only used for base rels)
137
* outerjoinset - For a base rel: if the rel appears within the nullable
138
* side of an outer join, the set of all relids
139
* participating in the highest such outer join; else NULL.
141
* joininfo - List of JoinInfo nodes, containing info about each join
142
* clause in which this relation participates
143
* index_outer_relids - only used for base rels; set of outer relids
144
* that participate in indexable joinclauses for this rel
145
* index_inner_paths - only used for base rels; list of InnerIndexscanInfo
146
* nodes showing best indexpaths for various subsets of
147
* index_outer_relids.
149
* Note: Keeping a restrictinfo list in the RelOptInfo is useful only for
150
* base rels, because for a join rel the set of clauses that are treated as
151
* restrict clauses varies depending on which sub-relations we choose to join.
152
* (For example, in a 3-base-rel join, a clause relating rels 1 and 2 must be
153
* treated as a restrictclause if we join {1} and {2 3} to make {1 2 3}; but
154
* if we join {1 2} and {3} then that clause will be a restrictclause in {1 2}
155
* and should not be processed again at the level of {1 2 3}.) Therefore,
156
* the restrictinfo list in the join case appears in individual JoinPaths
157
* (field joinrestrictinfo), not in the parent relation. But it's OK for
158
* the RelOptInfo to store the joininfo lists, because those are the same
159
* for a given rel no matter how we form it.
161
* We store baserestrictcost in the RelOptInfo (for base relations) because
162
* we know we will need it at least once (to price the sequential scan)
163
* and may need it multiple times to price index scans.
165
* outerjoinset is used to ensure correct placement of WHERE clauses that
166
* apply to outer-joined relations; we must not apply such WHERE clauses
167
* until after the outer join is performed.
170
typedef enum RelOptKind
174
RELOPT_OTHER_CHILD_REL
177
typedef struct RelOptInfo
181
RelOptKind reloptkind;
183
/* all relations included in this RelOptInfo */
184
Relids relids; /* set of base relids (rangetable indexes) */
186
/* size estimates generated by planner */
187
double rows; /* estimated number of result tuples */
188
int width; /* estimated avg width of result tuples */
190
/* materialization information */
191
List *reltargetlist; /* needed Vars */
192
List *pathlist; /* Path structures */
193
struct Path *cheapest_startup_path;
194
struct Path *cheapest_total_path;
195
struct Path *cheapest_unique_path;
197
/* information about a base rel (not set for join rels!) */
199
RTEKind rtekind; /* RELATION, SUBQUERY, or FUNCTION */
200
AttrNumber min_attr; /* smallest attrno of rel (often <0) */
201
AttrNumber max_attr; /* largest attrno of rel */
202
Relids *attr_needed; /* array indexed [min_attr .. max_attr] */
203
int32 *attr_widths; /* array indexed [min_attr .. max_attr] */
207
struct Plan *subplan; /* if subquery */
209
/* used by various scans and joins: */
210
List *baserestrictinfo; /* RestrictInfo structures (if
212
QualCost baserestrictcost; /* cost of evaluating the above */
213
Relids outerjoinset; /* set of base relids */
214
List *joininfo; /* JoinInfo structures */
216
/* cached info about inner indexscan paths for relation: */
217
Relids index_outer_relids; /* other relids in indexable join
219
List *index_inner_paths; /* InnerIndexscanInfo nodes */
222
* Inner indexscans are not in the main pathlist because they are not
223
* usable except in specific join contexts. We use the
224
* index_inner_paths list just to avoid recomputing the best inner
225
* indexscan repeatedly for similar outer relations. See comments for
226
* InnerIndexscanInfo.
232
* Per-index information for planning/optimization
234
* Prior to Postgres 7.0, RelOptInfo was used to describe both relations
235
* and indexes, but that created confusion without actually doing anything
236
* useful. So now we have a separate IndexOptInfo struct for indexes.
238
* classlist[], indexkeys[], and ordering[] have ncolumns entries.
239
* Zeroes in the indexkeys[] array indicate index columns that are
240
* expressions; there is one element in indexprs for each such column.
242
* Note: for historical reasons, the classlist and ordering arrays have
243
* an extra entry that is always zero. Some code scans until it sees a
244
* zero entry, rather than looking at ncolumns.
246
* The indexprs and indpred expressions have been run through
247
* prepqual.c and eval_const_expressions() for ease of matching to
248
* WHERE clauses. indpred is in implicit-AND form.
251
typedef struct IndexOptInfo
255
Oid indexoid; /* OID of the index relation */
257
/* statistics from pg_class */
258
BlockNumber pages; /* number of disk pages in index */
259
double tuples; /* number of index tuples in index */
261
/* index descriptor information */
262
int ncolumns; /* number of columns in index */
263
Oid *classlist; /* OIDs of operator classes for columns */
264
int *indexkeys; /* column numbers of index's keys, or 0 */
265
Oid *ordering; /* OIDs of sort operators for each column */
266
Oid relam; /* OID of the access method (in pg_am) */
268
RegProcedure amcostestimate; /* OID of the access method's cost fcn */
270
List *indexprs; /* expressions for non-simple index
272
List *indpred; /* predicate if a partial index, else NIL */
274
bool predOK; /* true if predicate matches query */
275
bool unique; /* true if a unique index */
277
/* cached info about inner indexscan paths for index */
278
Relids outer_relids; /* other relids in usable join clauses */
279
List *inner_paths; /* List of InnerIndexscanInfo nodes */
286
* The sort ordering of a path is represented by a list of sublists of
287
* PathKeyItem nodes. An empty list implies no known ordering. Otherwise
288
* the first sublist represents the primary sort key, the second the
289
* first secondary sort key, etc. Each sublist contains one or more
290
* PathKeyItem nodes, each of which can be taken as the attribute that
291
* appears at that sort position. (See the top of optimizer/path/pathkeys.c
292
* for more information.)
295
typedef struct PathKeyItem
299
Node *key; /* the item that is ordered */
300
Oid sortop; /* the ordering operator ('<' op) */
303
* key typically points to a Var node, ie a relation attribute, but it
304
* can also point to an arbitrary expression representing the value
305
* indexed by an index expression.
310
* Type "Path" is used as-is for sequential-scan paths. For other
311
* path types it is the first component of a larger struct.
313
* Note: "pathtype" is the NodeTag of the Plan node we could build from this
314
* Path. It is partially redundant with the Path's NodeTag, but allows us
315
* to use the same Path type for multiple Plan types where there is no need
316
* to distinguish the Plan type during path processing.
323
NodeTag pathtype; /* tag identifying scan/join method */
325
RelOptInfo *parent; /* the relation this path can build */
327
/* estimated execution costs for path (see costsize.c for more info) */
328
Cost startup_cost; /* cost expended before fetching any
330
Cost total_cost; /* total cost (assuming all tuples
333
List *pathkeys; /* sort ordering of path's output */
334
/* pathkeys is a List of Lists of PathKeyItem nodes; see above */
338
* IndexPath represents an index scan. Although an indexscan can only read
339
* a single relation, it can scan it more than once, potentially using a
340
* different index during each scan. The result is the union (OR) of all the
341
* tuples matched during any scan. (The executor is smart enough not to return
342
* the same tuple more than once, even if it is matched in multiple scans.)
344
* 'indexinfo' is a list of IndexOptInfo nodes, one per scan to be performed.
346
* 'indexclauses' is a list of index qualifications, also one per scan.
347
* Each entry in 'indexclauses' is a sublist of qualification clauses to be
348
* used for that scan, with implicit AND semantics across the sublist items.
349
* NOTE that the semantics of the top-level list in 'indexclauses' is OR
350
* combination, while the sublists are implicitly AND combinations!
352
* 'indexquals' has the same structure as 'indexclauses', but it contains
353
* the actual indexqual conditions that can be used with the index(es).
354
* In simple cases this is identical to 'indexclauses', but when special
355
* indexable operators appear in 'indexclauses', they are replaced by the
356
* derived indexscannable conditions in 'indexquals'.
358
* Both 'indexclauses' and 'indexquals' are lists of sublists of RestrictInfo
359
* nodes. (Before 8.0, we kept bare operator expressions in these lists, but
360
* storing RestrictInfos is more efficient since selectivities can be cached.)
362
* 'isjoininner' is TRUE if the path is a nestloop inner scan (that is,
363
* some of the index conditions are join rather than restriction clauses).
365
* 'indexscandir' is one of:
366
* ForwardScanDirection: forward scan of an ordered index
367
* BackwardScanDirection: backward scan of an ordered index
368
* NoMovementScanDirection: scan of an unordered index, or don't care
369
* (The executor doesn't care whether it gets ForwardScanDirection or
370
* NoMovementScanDirection for an indexscan, but the planner wants to
371
* distinguish ordered from unordered indexes for building pathkeys.)
373
* 'rows' is the estimated result tuple count for the indexscan. This
374
* is the same as path.parent->rows for a simple indexscan, but it is
375
* different for a nestloop inner scan, because the additional indexquals
376
* coming from join clauses make the scan more selective than the parent
377
* rel's restrict clauses alone would do.
380
typedef struct IndexPath
387
ScanDirection indexscandir;
388
double rows; /* estimated number of result tuples */
392
* TidPath represents a scan by TID
394
* tideval is an implicitly OR'ed list of quals of the form CTID = something.
395
* Note they are bare quals, not RestrictInfos.
397
typedef struct TidPath
400
List *tideval; /* qual(s) involving CTID = something */
404
* AppendPath represents an Append plan, ie, successive execution of
405
* several member plans. Currently it is only used to handle expansion
406
* of inheritance trees.
408
typedef struct AppendPath
411
List *subpaths; /* list of component Paths */
415
* ResultPath represents use of a Result plan node, either to compute a
416
* variable-free targetlist or to gate execution of a subplan with a
417
* one-time (variable-free) qual condition. Note that in the former case
418
* path.parent will be NULL; in the latter case it is copied from the subpath.
420
* Note that constantqual is a list of bare clauses, not RestrictInfos.
422
typedef struct ResultPath
430
* MaterialPath represents use of a Material plan node, i.e., caching of
431
* the output of its subpath. This is used when the subpath is expensive
432
* and needs to be scanned repeatedly, or when we need mark/restore ability
433
* and the subpath doesn't have it.
435
typedef struct MaterialPath
442
* UniquePath represents elimination of distinct rows from the output of
445
* This is unlike the other Path nodes in that it can actually generate
446
* different plans: either hash-based or sort-based implementation, or a
447
* no-op if the input path can be proven distinct already. The decision
448
* is sufficiently localized that it's not worth having separate Path node
449
* types. (Note: in the no-op case, we could eliminate the UniquePath node
450
* entirely and just return the subpath; but it's convenient to have a
451
* UniquePath in the path tree to signal upper-level routines that the input
452
* is known distinct.)
456
UNIQUE_PATH_NOOP, /* input is known unique already */
457
UNIQUE_PATH_HASH, /* use hashing */
458
UNIQUE_PATH_SORT /* use sorting */
461
typedef struct UniquePath
465
UniquePathMethod umethod;
466
double rows; /* estimated number of result tuples */
470
* All join-type paths share these fields.
473
typedef struct JoinPath
479
Path *outerjoinpath; /* path for the outer side of the join */
480
Path *innerjoinpath; /* path for the inner side of the join */
482
List *joinrestrictinfo; /* RestrictInfos to apply to join */
485
* See the notes for RelOptInfo to understand why joinrestrictinfo is
486
* needed in JoinPath, and can't be merged into the parent RelOptInfo.
491
* A nested-loop path needs no special fields.
494
typedef JoinPath NestPath;
497
* A mergejoin path has these fields.
499
* path_mergeclauses lists the clauses (in the form of RestrictInfos)
500
* that will be used in the merge.
502
* Note that the mergeclauses are a subset of the parent relation's
503
* restriction-clause list. Any join clauses that are not mergejoinable
504
* appear only in the parent's restrict list, and must be checked by a
505
* qpqual at execution time.
507
* outersortkeys (resp. innersortkeys) is NIL if the outer path
508
* (resp. inner path) is already ordered appropriately for the
509
* mergejoin. If it is not NIL then it is a PathKeys list describing
510
* the ordering that must be created by an explicit sort step.
513
typedef struct MergePath
516
List *path_mergeclauses; /* join clauses to be used for
518
List *outersortkeys; /* keys for explicit sort, if any */
519
List *innersortkeys; /* keys for explicit sort, if any */
523
* A hashjoin path has these fields.
525
* The remarks above for mergeclauses apply for hashclauses as well.
527
* Hashjoin does not care what order its inputs appear in, so we have
528
* no need for sortkeys.
531
typedef struct HashPath
534
List *path_hashclauses; /* join clauses used for hashing */
538
* Restriction clause info.
540
* We create one of these for each AND sub-clause of a restriction condition
541
* (WHERE or JOIN/ON clause). Since the restriction clauses are logically
542
* ANDed, we can use any one of them or any subset of them to filter out
543
* tuples, without having to evaluate the rest. The RestrictInfo node itself
544
* stores data used by the optimizer while choosing the best query plan.
546
* If a restriction clause references a single base relation, it will appear
547
* in the baserestrictinfo list of the RelOptInfo for that base rel.
549
* If a restriction clause references more than one base rel, it will
550
* appear in the JoinInfo lists of every RelOptInfo that describes a strict
551
* subset of the base rels mentioned in the clause. The JoinInfo lists are
552
* used to drive join tree building by selecting plausible join candidates.
553
* The clause cannot actually be applied until we have built a join rel
554
* containing all the base rels it references, however.
556
* When we construct a join rel that includes all the base rels referenced
557
* in a multi-relation restriction clause, we place that clause into the
558
* joinrestrictinfo lists of paths for the join rel, if neither left nor
559
* right sub-path includes all base rels referenced in the clause. The clause
560
* will be applied at that join level, and will not propagate any further up
561
* the join tree. (Note: the "predicate migration" code was once intended to
562
* push restriction clauses up and down the plan tree based on evaluation
563
* costs, but it's dead code and is unlikely to be resurrected in the
564
* foreseeable future.)
566
* Note that in the presence of more than two rels, a multi-rel restriction
567
* might reach different heights in the join tree depending on the join
568
* sequence we use. So, these clauses cannot be associated directly with
569
* the join RelOptInfo, but must be kept track of on a per-join-path basis.
571
* When dealing with outer joins we have to be very careful about pushing qual
572
* clauses up and down the tree. An outer join's own JOIN/ON conditions must
573
* be evaluated exactly at that join node, and any quals appearing in WHERE or
574
* in a JOIN above the outer join cannot be pushed down below the outer join.
575
* Otherwise the outer join will produce wrong results because it will see the
576
* wrong sets of input rows. All quals are stored as RestrictInfo nodes
577
* during planning, but there's a flag to indicate whether a qual has been
578
* pushed down to a lower level than its original syntactic placement in the
579
* join tree would suggest. If an outer join prevents us from pushing a qual
580
* down to its "natural" semantic level (the level associated with just the
581
* base rels used in the qual) then the qual will appear in JoinInfo lists
582
* that reference more than just the base rels it actually uses. By
583
* pretending that the qual references all the rels appearing in the outer
584
* join, we prevent it from being evaluated below the outer join's joinrel.
585
* When we do form the outer join's joinrel, we still need to distinguish
586
* those quals that are actually in that join's JOIN/ON condition from those
587
* that appeared higher in the tree and were pushed down to the join rel
588
* because they used no other rels. That's what the is_pushed_down flag is
589
* for; it tells us that a qual came from a point above the join of the
590
* specific set of base rels that it uses (or that the JoinInfo structures
591
* claim it uses). A clause that originally came from WHERE will *always*
592
* have its is_pushed_down flag set; a clause that came from an INNER JOIN
593
* condition, but doesn't use all the rels being joined, will also have
594
* is_pushed_down set because it will get attached to some lower joinrel.
596
* We also store a valid_everywhere flag, which says that the clause is not
597
* affected by any lower-level outer join, and therefore any conditions it
598
* asserts can be presumed true throughout the plan tree.
600
* In general, the referenced clause might be arbitrarily complex. The
601
* kinds of clauses we can handle as indexscan quals, mergejoin clauses,
602
* or hashjoin clauses are fairly limited --- the code for each kind of
603
* path is responsible for identifying the restrict clauses it can use
604
* and ignoring the rest. Clauses not implemented by an indexscan,
605
* mergejoin, or hashjoin will be placed in the plan qual or joinqual field
606
* of the finished Plan node, where they will be enforced by general-purpose
607
* qual-expression-evaluation code. (But we are still entitled to count
608
* their selectivity when estimating the result tuple count, if we
609
* can guess what it is...)
611
* When the referenced clause is an OR clause, we generate a modified copy
612
* in which additional RestrictInfo nodes are inserted below the top-level
613
* OR/AND structure. This is a convenience for OR indexscan processing:
614
* indexquals taken from either the top level or an OR subclause will have
615
* associated RestrictInfo nodes.
618
typedef struct RestrictInfo
622
Expr *clause; /* the represented clause of WHERE or JOIN */
624
bool is_pushed_down; /* TRUE if clause was pushed down in level */
626
bool valid_everywhere; /* TRUE if valid on every level */
629
* This flag is set true if the clause looks potentially useful as a
630
* merge or hash join clause, that is if it is a binary opclause with
631
* nonoverlapping sets of relids referenced in the left and right
632
* sides. (Whether the operator is actually merge or hash joinable
633
* isn't checked, however.)
637
/* The set of relids (varnos) referenced in the clause: */
638
Relids clause_relids;
640
/* These fields are set for any binary opclause: */
641
Relids left_relids; /* relids in left side of clause */
642
Relids right_relids; /* relids in right side of clause */
644
/* This field is NULL unless clause is an OR clause: */
645
Expr *orclause; /* modified clause with RestrictInfos */
647
/* cache space for cost and selectivity */
648
QualCost eval_cost; /* eval cost of clause; -1 if not yet set */
649
Selectivity this_selec; /* selectivity; -1 if not yet set */
651
/* valid if clause is mergejoinable, else InvalidOid: */
652
Oid mergejoinoperator; /* copy of clause operator */
653
Oid left_sortop; /* leftside sortop needed for mergejoin */
654
Oid right_sortop; /* rightside sortop needed for mergejoin */
656
/* cache space for mergeclause processing; NIL if not yet set */
657
List *left_pathkey; /* canonical pathkey for left side */
658
List *right_pathkey; /* canonical pathkey for right side */
660
/* cache space for mergeclause processing; -1 if not yet set */
661
Selectivity left_mergescansel; /* fraction of left side to scan */
662
Selectivity right_mergescansel; /* fraction of right side to scan */
664
/* valid if clause is hashjoinable, else InvalidOid: */
665
Oid hashjoinoperator; /* copy of clause operator */
667
/* cache space for hashclause processing; -1 if not yet set */
668
Selectivity left_bucketsize; /* avg bucketsize of left side */
669
Selectivity right_bucketsize; /* avg bucketsize of right side */
675
* We make a list of these for each RelOptInfo, containing info about
676
* all the join clauses this RelOptInfo participates in. (For this
677
* purpose, a "join clause" is a WHERE clause that mentions both vars
678
* belonging to this relation and vars belonging to relations not yet
679
* joined to it.) We group these clauses according to the set of
680
* other base relations (unjoined relations) mentioned in them.
681
* There is one JoinInfo for each distinct set of unjoined_relids,
682
* and its jinfo_restrictinfo lists the clause(s) that use that set
683
* of other relations.
686
typedef struct JoinInfo
689
Relids unjoined_relids; /* some rels not yet part of my RelOptInfo */
690
List *jinfo_restrictinfo; /* relevant RestrictInfos */
694
* Inner indexscan info.
696
* An inner indexscan is one that uses one or more joinclauses as index
697
* conditions (perhaps in addition to plain restriction clauses). So it
698
* can only be used as the inner path of a nestloop join where the outer
699
* relation includes all other relids appearing in those joinclauses.
700
* The set of usable joinclauses, and thus the best inner indexscan,
701
* thus varies depending on which outer relation we consider; so we have
702
* to recompute the best such path for every join. To avoid lots of
703
* redundant computation, we cache the results of such searches. For
704
* each index we compute the set of possible otherrelids (all relids
705
* appearing in joinquals that could become indexquals for this index).
706
* Two outer relations whose relids have the same intersection with this
707
* set will have the same set of available joinclauses and thus the same
708
* best inner indexscan for that index. Similarly, for each base relation,
709
* we form the union of the per-index otherrelids sets. Two outer relations
710
* with the same intersection with that set will have the same best overall
711
* inner indexscan for the base relation. We use lists of InnerIndexscanInfo
712
* nodes to cache the results of these searches at both the index and
715
* The search key also includes a bool showing whether the join being
716
* considered is an outer join. Since we constrain the join order for
717
* outer joins, I believe that this bool can only have one possible value
718
* for any particular base relation; but store it anyway to avoid confusion.
721
typedef struct InnerIndexscanInfo
724
/* The lookup key: */
725
Relids other_relids; /* a set of relevant other relids */
726
bool isouterjoin; /* true if join is outer */
727
/* Best path for this lookup key: */
728
Path *best_innerpath; /* best inner indexscan, or NULL if none */
729
} InnerIndexscanInfo;
734
* When we convert top-level IN quals into join operations, we must restrict
735
* the order of joining and use special join methods at some join points.
736
* We record information about each such IN clause in an InClauseInfo struct.
737
* These structs are kept in the Query node's in_info_list.
740
typedef struct InClauseInfo
743
Relids lefthand; /* base relids in lefthand expressions */
744
Relids righthand; /* base relids coming from the subselect */
745
List *sub_targetlist; /* targetlist of original RHS subquery */
748
* Note: sub_targetlist is just a list of Vars or expressions; it does
749
* not contain TargetEntry nodes.
753
#endif /* RELATION_H */
added
removed
src/include/nodes/relation.h
