* Definitions for planner's internal data structures.
*
*
- * Portions Copyright (c) 1996-2009, PostgreSQL Global Development Group
+ * Portions Copyright (c) 1996-2015, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
- * $PostgreSQL: pgsql/src/include/nodes/relation.h,v 1.170 2009/03/05 23:06:45 tgl Exp $
+ * src/include/nodes/relation.h
*
*-------------------------------------------------------------------------
*/
#define RELATION_H
#include "access/sdir.h"
-#include "nodes/bitmapset.h"
+#include "lib/stringinfo.h"
#include "nodes/params.h"
#include "nodes/parsenodes.h"
#include "storage/block.h"
Cost per_tuple; /* per-evaluation cost */
} QualCost;
+/*
+ * Costing aggregate function execution requires these statistics about
+ * the aggregates to be executed by a given Agg node. Note that the costs
+ * include the execution costs of the aggregates' argument expressions as
+ * well as the aggregate functions themselves.
+ */
+typedef struct AggClauseCosts
+{
+ int numAggs; /* total number of aggregate functions */
+ int numOrderedAggs; /* number w/ DISTINCT/ORDER BY/WITHIN GROUP */
+ QualCost transCost; /* total per-input-row execution costs */
+ Cost finalCost; /* total per-aggregated-row costs */
+ Size transitionSpace; /* space for pass-by-ref transition data */
+} AggClauseCosts;
+
/*----------
* PlannerGlobal
ParamListInfo boundParams; /* Param values provided to planner() */
- List *paramlist; /* to keep track of cross-level Params */
-
List *subplans; /* Plans for SubPlan nodes */
- List *subrtables; /* Rangetables for SubPlan nodes */
+ List *subroots; /* PlannerInfos for SubPlan nodes */
Bitmapset *rewindPlanIDs; /* indices of subplans that require REWIND */
List *finalrtable; /* "flat" rangetable for executor */
+ List *finalrowmarks; /* "flat" list of PlanRowMarks */
+
+ List *resultRelations; /* "flat" list of integer RT indexes */
+
List *relationOids; /* OIDs of relations the plan depends on */
List *invalItems; /* other dependencies, as PlanInvalItems */
+ int nParamExec; /* number of PARAM_EXEC Params used */
+
Index lastPHId; /* highest PlaceHolderVar ID assigned */
+ Index lastRowMarkId; /* highest PlanRowMark ID assigned */
+
+ int lastPlanNodeId; /* highest plan node ID assigned */
+
bool transientPlan; /* redo plan when TransactionXmin changes? */
+
+ bool hasRowSecurity; /* row security applied? */
+
+ bool parallelModeOK; /* parallel mode potentially OK? */
+
+ bool parallelModeNeeded; /* parallel mode actually required? */
} PlannerGlobal;
/* macro for fetching the Plan associated with a SubPlan node */
*
* This struct is conventionally called "root" in all the planner routines.
* It holds links to all of the planner's working state, in addition to the
- * original Query. Note that at present the planner extensively modifies
+ * original Query. Note that at present the planner extensively modifies
* the passed-in Query data structure; someday that should stop.
*----------
*/
Index query_level; /* 1 at the outermost Query */
- struct PlannerInfo *parent_root; /* NULL at outermost Query */
+ struct PlannerInfo *parent_root; /* NULL at outermost Query */
+
+ /*
+ * plan_params contains the expressions that this query level needs to
+ * make available to a lower query level that is currently being planned.
+ * outer_params contains the paramIds of PARAM_EXEC Params that outer
+ * query levels will make available to this query level.
+ */
+ List *plan_params; /* list of PlannerParamItems, see below */
+ Bitmapset *outer_params;
/*
* simple_rel_array holds pointers to "base rels" and "other rels" (see
- * comments for RelOptInfo for more info). It is indexed by rangetable
+ * comments for RelOptInfo for more info). It is indexed by rangetable
* index (so entry 0 is always wasted). Entries can be NULL when an RTE
* does not correspond to a base relation, such as a join RTE or an
* unreferenced view RTE; or if the RelOptInfo hasn't been made yet.
*/
RangeTblEntry **simple_rte_array; /* rangetable as an array */
+ /*
+ * all_baserels is a Relids set of all base relids (but not "other"
+ * relids) in the query; that is, the Relids identifier of the final join
+ * we need to form. This is computed in make_one_rel, just before we
+ * start making Paths.
+ */
+ Relids all_baserels;
+
+ /*
+ * nullable_baserels is a Relids set of base relids that are nullable by
+ * some outer join in the jointree; these are rels that are potentially
+ * nullable below the WHERE clause, SELECT targetlist, etc. This is
+ * computed in deconstruct_jointree.
+ */
+ Relids nullable_baserels;
+
/*
* join_rel_list is a list of all join-relation RelOptInfos we have
* considered in this planning run. For small problems we just scan the
* list to do lookups, but when there are many join relations we build a
* hash table for faster lookups. The hash table is present and valid
- * when join_rel_hash is not NULL. Note that we still maintain the list
+ * when join_rel_hash is not NULL. Note that we still maintain the list
* even when using the hash table for lookups; this simplifies life for
* GEQO.
*/
List *join_rel_list; /* list of join-relation RelOptInfos */
struct HTAB *join_rel_hash; /* optional hashtable for join relations */
- List *resultRelations; /* integer list of RT indexes, or NIL */
-
- List *returningLists; /* list of lists of TargetEntry, or NIL */
+ /*
+ * When doing a dynamic-programming-style join search, join_rel_level[k]
+ * is a list of all join-relation RelOptInfos of level k, and
+ * join_cur_level is the current level. New join-relation RelOptInfos are
+ * automatically added to the join_rel_level[join_cur_level] list.
+ * join_rel_level is NULL if not in use.
+ */
+ List **join_rel_level; /* lists of join-relation RelOptInfos */
+ int join_cur_level; /* index of list being extended */
List *init_plans; /* init SubPlans for query */
List *cte_plan_ids; /* per-CTE-item list of subplan IDs */
+ List *multiexpr_params; /* List of Lists of Params for
+ * MULTIEXPR subquery outputs */
+
List *eq_classes; /* list of active EquivalenceClasses */
List *canon_pathkeys; /* list of "canonical" PathKeys */
List *full_join_clauses; /* list of RestrictInfos for
* mergejoinable full join clauses */
- List *join_info_list; /* list of SpecialJoinInfos */
+ List *join_info_list; /* list of SpecialJoinInfos */
+
+ List *lateral_info_list; /* list of LateralJoinInfos */
List *append_rel_list; /* list of AppendRelInfos */
- List *placeholder_list; /* list of PlaceHolderInfos */
+ List *rowMarks; /* list of PlanRowMarks */
+
+ List *placeholder_list; /* list of PlaceHolderInfos */
List *query_pathkeys; /* desired pathkeys for query_planner(), and
- * actual pathkeys afterwards */
+ * actual pathkeys after planning */
- List *group_pathkeys; /* groupClause pathkeys, if any */
+ List *group_pathkeys; /* groupClause pathkeys, if any */
List *window_pathkeys; /* pathkeys of bottom window, if any */
- List *distinct_pathkeys; /* distinctClause pathkeys, if any */
- List *sort_pathkeys; /* sortClause pathkeys, if any */
+ List *distinct_pathkeys; /* distinctClause pathkeys, if any */
+ List *sort_pathkeys; /* sortClause pathkeys, if any */
+
+ List *minmax_aggs; /* List of MinMaxAggInfos */
List *initial_rels; /* RelOptInfos we are now trying to join */
double total_table_pages; /* # of pages in all tables of query */
double tuple_fraction; /* tuple_fraction passed to query_planner */
+ double limit_tuples; /* limit_tuples passed to query_planner */
+ bool hasInheritedTarget; /* true if parse->resultRelation is an
+ * inheritance child rel */
bool hasJoinRTEs; /* true if any RTEs are RTE_JOIN kind */
+ bool hasLateralRTEs; /* true if any RTEs are marked LATERAL */
+ bool hasDeletedRTEs; /* true if any RTE was deleted from jointree */
bool hasHavingQual; /* true if havingQual was non-null */
bool hasPseudoConstantQuals; /* true if any RestrictInfo has
* pseudoconstant = true */
bool hasRecursion; /* true if planning a recursive WITH item */
/* These fields are used only when hasRecursion is true: */
- int wt_param_id; /* PARAM_EXEC ID for the work table */
+ int wt_param_id; /* PARAM_EXEC ID for the work table */
struct Plan *non_recursive_plan; /* plan for non-recursive term */
+
+ /* These fields are workspace for createplan.c */
+ Relids curOuterRels; /* outer rels above current node */
+ List *curOuterParams; /* not-yet-assigned NestLoopParams */
+
+ /* optional private data for join_search_hook, e.g., GEQO */
+ void *join_search_private;
+
+ /* for GroupingFunc fixup in setrefs */
+ AttrNumber *grouping_map;
} PlannerInfo;
*
* We also have "other rels", which are like base rels in that they refer to
* single RT indexes; but they are not part of the join tree, and are given
- * a different RelOptKind to identify them.
+ * a different RelOptKind to identify them. Lastly, there is a RelOptKind
+ * for "dead" relations, which are base rels that we have proven we don't
+ * need to join after all.
*
* Currently the only kind of otherrels are those made for member relations
* of an "append relation", that is an inheritance set or UNION ALL subquery.
* An append relation has a parent RTE that is a base rel, which represents
- * the entire append relation. The member RTEs are otherrels. The parent
+ * the entire append relation. The member RTEs are otherrels. The parent
* is present in the query join tree but the members are not. The member
* RTEs and otherrels are used to plan the scans of the individual tables or
- * subqueries of the append set; then the parent baserel is given an Append
- * plan comprising the best plans for the individual member rels. (See
- * comments for AppendRelInfo for more information.)
+ * subqueries of the append set; then the parent baserel is given Append
+ * and/or MergeAppend paths comprising the best paths for the individual
+ * member rels. (See comments for AppendRelInfo for more information.)
*
* At one time we also made otherrels to represent join RTEs, for use in
* handling join alias Vars. Currently this is not needed because all join
* alias Vars are expanded to non-aliased form during preprocess_expression.
*
* Parts of this data structure are specific to various scan and join
- * mechanisms. It didn't seem worth creating new node types for them.
+ * mechanisms. It didn't seem worth creating new node types for them.
*
* relids - Set of base-relation identifiers; it is a base relation
* if there is just one, a join relation if more than one
* clauses have been applied (ie, output rows of a plan for it)
* width - avg. number of bytes per tuple in the relation after the
* appropriate projections have been done (ie, output width)
+ * consider_startup - true if there is any value in keeping plain paths for
+ * this rel on the basis of having cheap startup cost
+ * consider_param_startup - the same for parameterized paths
* reltargetlist - List of Var and PlaceHolderVar nodes for the values
* we need to output from this relation.
* List is in no particular order, but all rels of an
* appendrel set must use corresponding orders.
- * NOTE: in a child relation, may contain RowExpr or
- * ConvertRowtypeExpr representing a whole-row Var.
+ * NOTE: in an appendrel child relation, may contain
+ * arbitrary expressions pulled up from a subquery!
* pathlist - List of Path nodes, one for each potentially useful
* method of generating the relation
+ * ppilist - ParamPathInfo nodes for parameterized Paths, if any
* cheapest_startup_path - the pathlist member with lowest startup cost
- * (regardless of its ordering)
+ * (regardless of ordering) among the unparameterized paths;
+ * or NULL if there is no unparameterized path
* cheapest_total_path - the pathlist member with lowest total cost
- * (regardless of its ordering)
+ * (regardless of ordering) among the unparameterized paths;
+ * or if there is no unparameterized path, the path with lowest
+ * total cost among the paths with minimum parameterization
* cheapest_unique_path - for caching cheapest path to produce unique
- * (no duplicates) output from relation
+ * (no duplicates) output from relation; NULL if not yet requested
+ * cheapest_parameterized_paths - best paths for their parameterizations;
+ * always includes cheapest_total_path, even if that's unparameterized
*
* If the relation is a base relation it will have these fields set:
*
* the attribute is needed as part of final targetlist
* attr_widths - cache space for per-attribute width estimates;
* zero means not computed yet
+ * lateral_vars - lateral cross-references of rel, if any (list of
+ * Vars and PlaceHolderVars)
+ * lateral_relids - required outer rels for LATERAL, as a Relids set
+ * (for child rels this can be more than lateral_vars)
+ * lateral_referencers - relids of rels that reference this one laterally
* indexlist - list of IndexOptInfo nodes for relation's indexes
* (always NIL if it's not a table)
* pages - number of disk pages in relation (zero if not a table)
* tuples - number of tuples in relation (not considering restrictions)
+ * allvisfrac - fraction of disk pages that are marked all-visible
* subplan - plan for subquery (NULL if it's not a subquery)
- * subrtable - rangetable for subquery (NIL if it's not a subquery)
+ * subroot - PlannerInfo for subquery (NULL if it's not a subquery)
+ * subplan_params - list of PlannerParamItems to be passed to subquery
*
- * Note: for a subquery, tuples and subplan are not set immediately
+ * Note: for a subquery, tuples, subplan, subroot are not set immediately
* upon creation of the RelOptInfo object; they are filled in when
- * set_base_rel_pathlist processes the object.
+ * set_subquery_pathlist processes the object.
*
* For otherrels that are appendrel members, these fields are filled
* in just as for a baserel.
*
+ * If the relation is either a foreign table or a join of foreign tables that
+ * all belong to the same foreign server, these fields will be set:
+ *
+ * serverid - OID of foreign server, if foreign table (else InvalidOid)
+ * fdwroutine - function hooks for FDW, if foreign table (else NULL)
+ * fdw_private - private state for FDW, if foreign table (else NULL)
+ *
* The presence of the remaining fields depends on the restrictions
* and joins that the relation participates in:
*
* note this excludes clauses that might be derivable from
* EquivalenceClasses)
* has_eclass_joins - flag that EquivalenceClass joins are possible
- * index_outer_relids - only used for base rels; set of outer relids
- * that participate in indexable joinclauses for this rel
- * index_inner_paths - only used for base rels; list of InnerIndexscanInfo
- * nodes showing best indexpaths for various subsets of
- * index_outer_relids.
*
* Note: Keeping a restrictinfo list in the RelOptInfo is useful only for
* base rels, because for a join rel the set of clauses that are treated as
{
RELOPT_BASEREL,
RELOPT_JOINREL,
- RELOPT_OTHER_MEMBER_REL
+ RELOPT_OTHER_MEMBER_REL,
+ RELOPT_DEADREL
} RelOptKind;
typedef struct RelOptInfo
double rows; /* estimated number of result tuples */
int width; /* estimated avg width of result tuples */
+ /* per-relation planner control flags */
+ bool consider_startup; /* keep cheap-startup-cost paths? */
+ bool consider_param_startup; /* ditto, for parameterized paths? */
+
/* materialization information */
List *reltargetlist; /* Vars to be output by scan of relation */
List *pathlist; /* Path structures */
+ List *ppilist; /* ParamPathInfos used in pathlist */
struct Path *cheapest_startup_path;
struct Path *cheapest_total_path;
struct Path *cheapest_unique_path;
+ List *cheapest_parameterized_paths;
/* information about a base rel (not set for join rels!) */
Index relid;
+ Oid reltablespace; /* containing tablespace */
RTEKind rtekind; /* RELATION, SUBQUERY, or FUNCTION */
AttrNumber min_attr; /* smallest attrno of rel (often <0) */
AttrNumber max_attr; /* largest attrno of rel */
Relids *attr_needed; /* array indexed [min_attr .. max_attr] */
int32 *attr_widths; /* array indexed [min_attr .. max_attr] */
+ List *lateral_vars; /* LATERAL Vars and PHVs referenced by rel */
+ Relids lateral_relids; /* minimum parameterization of rel */
+ Relids lateral_referencers; /* rels that reference me laterally */
List *indexlist; /* list of IndexOptInfo */
- BlockNumber pages;
+ BlockNumber pages; /* size estimates derived from pg_class */
double tuples;
+ double allvisfrac;
+ /* use "struct Plan" to avoid including plannodes.h here */
struct Plan *subplan; /* if subquery */
- List *subrtable; /* if subquery */
+ PlannerInfo *subroot; /* if subquery */
+ List *subplan_params; /* if subquery */
+
+ /* Information about foreign tables and foreign joins */
+ Oid serverid; /* identifies server for the table or join */
+ /* use "struct FdwRoutine" to avoid including fdwapi.h here */
+ struct FdwRoutine *fdwroutine;
+ void *fdw_private;
/* used by various scans and joins: */
List *baserestrictinfo; /* RestrictInfo structures (if base
List *joininfo; /* RestrictInfo structures for join clauses
* involving this rel */
bool has_eclass_joins; /* T means joininfo is incomplete */
-
- /* cached info about inner indexscan paths for relation: */
- Relids index_outer_relids; /* other relids in indexable join
- * clauses */
- List *index_inner_paths; /* InnerIndexscanInfo nodes */
-
- /*
- * Inner indexscans are not in the main pathlist because they are not
- * usable except in specific join contexts. We use the index_inner_paths
- * list just to avoid recomputing the best inner indexscan repeatedly for
- * similar outer relations. See comments for InnerIndexscanInfo.
- */
} RelOptInfo;
/*
* IndexOptInfo
* Per-index information for planning/optimization
*
- * Prior to Postgres 7.0, RelOptInfo was used to describe both relations
- * and indexes, but that created confusion without actually doing anything
- * useful. So now we have a separate IndexOptInfo struct for indexes.
+ * indexkeys[], indexcollations[], opfamily[], and opcintype[]
+ * each have ncolumns entries.
*
- * opfamily[], indexkeys[], opcintype[], fwdsortop[], revsortop[],
- * and nulls_first[] each have ncolumns entries.
- * Note: for historical reasons, the opfamily array has an extra entry
- * that is always zero. Some code scans until it sees a zero entry,
- * rather than looking at ncolumns.
+ * sortopfamily[], reverse_sort[], and nulls_first[] likewise have
+ * ncolumns entries, if the index is ordered; but if it is unordered,
+ * those pointers are NULL.
*
* Zeroes in the indexkeys[] array indicate index columns that are
* expressions; there is one element in indexprs for each such column.
*
- * For an unordered index, the sortop arrays contains zeroes. Note that
- * fwdsortop[] and nulls_first[] describe the sort ordering of a forward
- * indexscan; we can also consider a backward indexscan, which will
- * generate sort order described by revsortop/!nulls_first.
+ * For an ordered index, reverse_sort[] and nulls_first[] describe the
+ * sort ordering of a forward indexscan; we can also consider a backward
+ * indexscan, which will generate the reverse ordering.
*
* The indexprs and indpred expressions have been run through
* prepqual.c and eval_const_expressions() for ease of matching to
* WHERE clauses. indpred is in implicit-AND form.
+ *
+ * indextlist is a TargetEntry list representing the index columns.
+ * It provides an equivalent base-relation Var for each simple column,
+ * and links to the matching indexprs element for each expression column.
*/
typedef struct IndexOptInfo
{
NodeTag type;
Oid indexoid; /* OID of the index relation */
+ Oid reltablespace; /* tablespace of index (not table) */
RelOptInfo *rel; /* back-link to index's table */
- /* statistics from pg_class */
+ /* index-size statistics (from pg_class and elsewhere) */
BlockNumber pages; /* number of disk pages in index */
double tuples; /* number of index tuples in index */
+ int tree_height; /* index tree height, or -1 if unknown */
/* index descriptor information */
int ncolumns; /* number of columns in index */
- Oid *opfamily; /* OIDs of operator families for columns */
int *indexkeys; /* column numbers of index's keys, or 0 */
+ Oid *indexcollations; /* OIDs of collations of index columns */
+ Oid *opfamily; /* OIDs of operator families for columns */
Oid *opcintype; /* OIDs of opclass declared input data types */
- Oid *fwdsortop; /* OIDs of sort operators for each column */
- Oid *revsortop; /* OIDs of sort operators for backward scan */
+ Oid *sortopfamily; /* OIDs of btree opfamilies, if orderable */
+ bool *reverse_sort; /* is sort order descending? */
bool *nulls_first; /* do NULLs come first in the sort order? */
+ bool *canreturn; /* which index cols can be returned in an
+ * index-only scan? */
Oid relam; /* OID of the access method (in pg_am) */
RegProcedure amcostestimate; /* OID of the access method's cost fcn */
List *indexprs; /* expressions for non-simple index columns */
List *indpred; /* predicate if a partial index, else NIL */
+ List *indextlist; /* targetlist representing index columns */
+
bool predOK; /* true if predicate matches query */
bool unique; /* true if a unique index */
+ bool immediate; /* is uniqueness enforced immediately? */
+ bool hypothetical; /* true if index doesn't really exist */
+ bool amcanorderbyop; /* does AM support order by operator result? */
bool amoptionalkey; /* can query omit key for the first column? */
- bool amsearchnulls; /* can AM search for NULL index entries? */
+ bool amsearcharray; /* can AM handle ScalarArrayOpExpr quals? */
+ bool amsearchnulls; /* can AM search for NULL/NOT NULL entries? */
bool amhasgettuple; /* does AM have amgettuple interface? */
- bool amhasgetbitmap; /* does AM have amgetbitmap interface? */
+ bool amhasgetbitmap; /* does AM have amgetbitmap interface? */
} IndexOptInfo;
* require merging two existing EquivalenceClasses. At the end of the qual
* distribution process, we have sets of values that are known all transitively
* equal to each other, where "equal" is according to the rules of the btree
- * operator family(s) shown in ec_opfamilies. (We restrict an EC to contain
- * only equalities whose operators belong to the same set of opfamilies. This
- * could probably be relaxed, but for now it's not worth the trouble, since
- * nearly all equality operators belong to only one btree opclass anyway.)
+ * operator family(s) shown in ec_opfamilies, as well as the collation shown
+ * by ec_collation. (We restrict an EC to contain only equalities whose
+ * operators belong to the same set of opfamilies. This could probably be
+ * relaxed, but for now it's not worth the trouble, since nearly all equality
+ * operators belong to only one btree opclass anyway. Similarly, we suppose
+ * that all or none of the input datatypes are collatable, so that a single
+ * collation value is sufficient.)
*
* We also use EquivalenceClasses as the base structure for PathKeys, letting
* us represent knowledge about different sort orderings being equivalent.
* Since every PathKey must reference an EquivalenceClass, we will end up
* with single-member EquivalenceClasses whenever a sort key expression has
- * not been equivalenced to anything else. It is also possible that such an
+ * not been equivalenced to anything else. It is also possible that such an
* EquivalenceClass will contain a volatile expression ("ORDER BY random()"),
* which is a case that can't arise otherwise since clauses containing
* volatile functions are never considered mergejoinable. We mark such
* We allow equality clauses appearing below the nullable side of an outer join
* to form EquivalenceClasses, but these have a slightly different meaning:
* the included values might be all NULL rather than all the same non-null
- * values. See src/backend/optimizer/README for more on that point.
+ * values. See src/backend/optimizer/README for more on that point.
*
* NB: if ec_merged isn't NULL, this class has been merged into another, and
* should be ignored in favor of using the pointed-to class.
NodeTag type;
List *ec_opfamilies; /* btree operator family OIDs */
+ Oid ec_collation; /* collation, if datatypes are collatable */
List *ec_members; /* list of EquivalenceMembers */
List *ec_sources; /* list of generating RestrictInfos */
List *ec_derives; /* list of derived RestrictInfos */
- Relids ec_relids; /* all relids appearing in ec_members */
+ Relids ec_relids; /* all relids appearing in ec_members, except
+ * for child members (see below) */
bool ec_has_const; /* any pseudoconstants in ec_members? */
bool ec_has_volatile; /* the (sole) member is a volatile expr */
bool ec_below_outer_join; /* equivalence applies below an OJ */
* EquivalenceMember - one member expression of an EquivalenceClass
*
* em_is_child signifies that this element was built by transposing a member
- * for an inheritance parent relation to represent the corresponding expression
- * on an inheritance child. The element should be ignored for all purposes
- * except constructing inner-indexscan paths for the child relation. (Other
- * types of join are driven from transposed joininfo-list entries.) Note
- * that the EC's ec_relids field does NOT include the child relation.
+ * for an appendrel parent relation to represent the corresponding expression
+ * for an appendrel child. These members are used for determining the
+ * pathkeys of scans on the child relation and for explicitly sorting the
+ * child when necessary to build a MergeAppend path for the whole appendrel
+ * tree. An em_is_child member has no impact on the properties of the EC as a
+ * whole; in particular the EC's ec_relids field does NOT include the child
+ * relation. An em_is_child member should never be marked em_is_const nor
+ * cause ec_has_const or ec_has_volatile to be set, either. Thus, em_is_child
+ * members are not really full-fledged members of the EC, but just reflections
+ * or doppelgangers of real members. Most operations on EquivalenceClasses
+ * should ignore em_is_child members, and those that don't should test
+ * em_relids to make sure they only consider relevant members.
*
* em_datatype is usually the same as exprType(em_expr), but can be
* different when dealing with a binary-compatible opfamily; in particular
- * anyarray_ops would never work without this. Use em_datatype when
+ * anyarray_ops would never work without this. Use em_datatype when
* looking up a specific btree operator to work with this expression.
*/
typedef struct EquivalenceMember
Expr *em_expr; /* the expression represented */
Relids em_relids; /* all relids appearing in em_expr */
+ Relids em_nullable_relids; /* nullable by lower outer joins */
bool em_is_const; /* expression is pseudoconstant? */
bool em_is_child; /* derived version for a child relation? */
Oid em_datatype; /* the "nominal type" used by the opfamily */
* represents the primary sort key, the second the first secondary sort key,
* etc. The value being sorted is represented by linking to an
* EquivalenceClass containing that value and including pk_opfamily among its
- * ec_opfamilies. This is a convenient method because it makes it trivial
- * to detect equivalent and closely-related orderings. (See optimizer/README
- * for more information.)
+ * ec_opfamilies. The EquivalenceClass tells which collation to use, too.
+ * This is a convenient method because it makes it trivial to detect
+ * equivalent and closely-related orderings. (See optimizer/README for more
+ * information.)
*
* Note: pk_strategy is either BTLessStrategyNumber (for ASC) or
- * BTGreaterStrategyNumber (for DESC). We assume that all ordering-capable
+ * BTGreaterStrategyNumber (for DESC). We assume that all ordering-capable
* index types will use btree-compatible strategy numbers.
*/
-
typedef struct PathKey
{
NodeTag type;
bool pk_nulls_first; /* do NULLs come before normal values? */
} PathKey;
+
+/*
+ * ParamPathInfo
+ *
+ * All parameterized paths for a given relation with given required outer rels
+ * link to a single ParamPathInfo, which stores common information such as
+ * the estimated rowcount for this parameterization. We do this partly to
+ * avoid recalculations, but mostly to ensure that the estimated rowcount
+ * is in fact the same for every such path.
+ *
+ * Note: ppi_clauses is only used in ParamPathInfos for base relation paths;
+ * in join cases it's NIL because the set of relevant clauses varies depending
+ * on how the join is formed. The relevant clauses will appear in each
+ * parameterized join path's joinrestrictinfo list, instead.
+ */
+typedef struct ParamPathInfo
+{
+ NodeTag type;
+
+ Relids ppi_req_outer; /* rels supplying parameters used by path */
+ double ppi_rows; /* estimated number of result tuples */
+ List *ppi_clauses; /* join clauses available from outer rels */
+} ParamPathInfo;
+
+
/*
* Type "Path" is used as-is for sequential-scan paths, as well as some other
* simple plan types that we don't need any extra information in the path for.
* For other path types it is the first component of a larger struct.
*
- * Note: "pathtype" is the NodeTag of the Plan node we could build from this
- * Path. It is partially redundant with the Path's NodeTag, but allows us
- * to use the same Path type for multiple Plan types where there is no need
- * to distinguish the Plan type during path processing.
+ * "pathtype" is the NodeTag of the Plan node we could build from this Path.
+ * It is partially redundant with the Path's NodeTag, but allows us to use
+ * the same Path type for multiple Plan types when there is no need to
+ * distinguish the Plan type during path processing.
+ *
+ * "param_info", if not NULL, links to a ParamPathInfo that identifies outer
+ * relation(s) that provide parameter values to each scan of this path.
+ * That means this path can only be joined to those rels by means of nestloop
+ * joins with this path on the inside. Also note that a parameterized path
+ * is responsible for testing all "movable" joinclauses involving this rel
+ * and the specified outer rel(s).
+ *
+ * "rows" is the same as parent->rows in simple paths, but in parameterized
+ * paths and UniquePaths it can be less than parent->rows, reflecting the
+ * fact that we've filtered by extra join conditions or removed duplicates.
+ *
+ * "pathkeys" is a List of PathKey nodes (see above), describing the sort
+ * ordering of the path's output rows.
*/
-
typedef struct Path
{
NodeTag type;
NodeTag pathtype; /* tag identifying scan/join method */
RelOptInfo *parent; /* the relation this path can build */
+ ParamPathInfo *param_info; /* parameterization info, or NULL if none */
- /* estimated execution costs for path (see costsize.c for more info) */
+ /* estimated size/costs for path (see costsize.c for more info) */
+ double rows; /* estimated number of result tuples */
Cost startup_cost; /* cost expended before fetching any tuples */
Cost total_cost; /* total cost (assuming all tuples fetched) */
/* pathkeys is a List of PathKey nodes; see above */
} Path;
+/* Macro for extracting a path's parameterization relids; beware double eval */
+#define PATH_REQ_OUTER(path) \
+ ((path)->param_info ? (path)->param_info->ppi_req_outer : (Relids) NULL)
+
/*----------
* IndexPath represents an index scan over a single index.
*
+ * This struct is used for both regular indexscans and index-only scans;
+ * path.pathtype is T_IndexScan or T_IndexOnlyScan to show which is meant.
+ *
* 'indexinfo' is the index to be scanned.
*
* 'indexclauses' is a list of index qualification clauses, with implicit
* AND semantics across the list. Each clause is a RestrictInfo node from
- * the query's WHERE or JOIN conditions.
+ * the query's WHERE or JOIN conditions. An empty list implies a full
+ * index scan.
*
* 'indexquals' has the same structure as 'indexclauses', but it contains
- * the actual indexqual conditions that can be used with the index.
+ * the actual index qual conditions that can be used with the index.
* In simple cases this is identical to 'indexclauses', but when special
* indexable operators appear in 'indexclauses', they are replaced by the
* derived indexscannable conditions in 'indexquals'.
*
- * 'isjoininner' is TRUE if the path is a nestloop inner scan (that is,
- * some of the index conditions are join rather than restriction clauses).
- * Note that the path costs will be calculated differently from a plain
- * indexscan in this case, and in addition there's a special 'rows' value
- * different from the parent RelOptInfo's (see below).
+ * 'indexqualcols' is an integer list of index column numbers (zero-based)
+ * of the same length as 'indexquals', showing which index column each qual
+ * is meant to be used with. 'indexquals' is required to be ordered by
+ * index column, so 'indexqualcols' must form a nondecreasing sequence.
+ * (The order of multiple quals for the same index column is unspecified.)
+ *
+ * 'indexorderbys', if not NIL, is a list of ORDER BY expressions that have
+ * been found to be usable as ordering operators for an amcanorderbyop index.
+ * The list must match the path's pathkeys, ie, one expression per pathkey
+ * in the same order. These are not RestrictInfos, just bare expressions,
+ * since they generally won't yield booleans. Also, unlike the case for
+ * quals, it's guaranteed that each expression has the index key on the left
+ * side of the operator.
+ *
+ * 'indexorderbycols' is an integer list of index column numbers (zero-based)
+ * of the same length as 'indexorderbys', showing which index column each
+ * ORDER BY expression is meant to be used with. (There is no restriction
+ * on which index column each ORDER BY can be used with.)
*
* 'indexscandir' is one of:
* ForwardScanDirection: forward scan of an ordered index
* 'indextotalcost' and 'indexselectivity' are saved in the IndexPath so that
* we need not recompute them when considering using the same index in a
* bitmap index/heap scan (see BitmapHeapPath). The costs of the IndexPath
- * itself represent the costs of an IndexScan plan type.
- *
- * 'rows' is the estimated result tuple count for the indexscan. This
- * is the same as path.parent->rows for a simple indexscan, but it is
- * different for a nestloop inner scan, because the additional indexquals
- * coming from join clauses make the scan more selective than the parent
- * rel's restrict clauses alone would do.
+ * itself represent the costs of an IndexScan or IndexOnlyScan plan type.
*----------
*/
typedef struct IndexPath
IndexOptInfo *indexinfo;
List *indexclauses;
List *indexquals;
- bool isjoininner;
+ List *indexqualcols;
+ List *indexorderbys;
+ List *indexorderbycols;
ScanDirection indexscandir;
Cost indextotalcost;
Selectivity indexselectivity;
- double rows; /* estimated number of result tuples */
} IndexPath;
/*
*
* The individual indexscans are represented by IndexPath nodes, and any
* logic on top of them is represented by a tree of BitmapAndPath and
- * BitmapOrPath nodes. Notice that we can use the same IndexPath node both
- * to represent a regular IndexScan plan, and as the child of a BitmapHeapPath
- * that represents scanning the same index using a BitmapIndexScan. The
- * startup_cost and total_cost figures of an IndexPath always represent the
- * costs to use it as a regular IndexScan. The costs of a BitmapIndexScan
- * can be computed using the IndexPath's indextotalcost and indexselectivity.
- *
- * BitmapHeapPaths can be nestloop inner indexscans. The isjoininner and
- * rows fields serve the same purpose as for plain IndexPaths.
+ * BitmapOrPath nodes. Notice that we can use the same IndexPath node both
+ * to represent a regular (or index-only) index scan plan, and as the child
+ * of a BitmapHeapPath that represents scanning the same index using a
+ * BitmapIndexScan. The startup_cost and total_cost figures of an IndexPath
+ * always represent the costs to use it as a regular (or index-only)
+ * IndexScan. The costs of a BitmapIndexScan can be computed using the
+ * IndexPath's indextotalcost and indexselectivity.
*/
typedef struct BitmapHeapPath
{
Path path;
Path *bitmapqual; /* IndexPath, BitmapAndPath, BitmapOrPath */
- bool isjoininner; /* T if it's a nestloop inner scan */
- double rows; /* estimated number of result tuples */
} BitmapHeapPath;
/*
List *tidquals; /* qual(s) involving CTID = something */
} TidPath;
+/*
+ * ForeignPath represents a potential scan of a foreign table
+ *
+ * fdw_private stores FDW private data about the scan. While fdw_private is
+ * not actually touched by the core code during normal operations, it's
+ * generally a good idea to use a representation that can be dumped by
+ * nodeToString(), so that you can examine the structure during debugging
+ * with tools like pprint().
+ */
+typedef struct ForeignPath
+{
+ Path path;
+ List *fdw_private;
+} ForeignPath;
+
+/*
+ * CustomPath represents a table scan done by some out-of-core extension.
+ *
+ * We provide a set of hooks here - which the provider must take care to set
+ * up correctly - to allow extensions to supply their own methods of scanning
+ * a relation. For example, a provider might provide GPU acceleration, a
+ * cache-based scan, or some other kind of logic we haven't dreamed up yet.
+ *
+ * CustomPaths can be injected into the planning process for a relation by
+ * set_rel_pathlist_hook functions.
+ *
+ * Core code must avoid assuming that the CustomPath is only as large as
+ * the structure declared here; providers are allowed to make it the first
+ * element in a larger structure. (Since the planner never copies Paths,
+ * this doesn't add any complication.) However, for consistency with the
+ * FDW case, we provide a "custom_private" field in CustomPath; providers
+ * may prefer to use that rather than define another struct type.
+ */
+struct CustomPath;
+
+#define CUSTOMPATH_SUPPORT_BACKWARD_SCAN 0x0001
+#define CUSTOMPATH_SUPPORT_MARK_RESTORE 0x0002
+
+typedef struct CustomPathMethods
+{
+ const char *CustomName;
+
+ /* Convert Path to a Plan */
+ struct Plan *(*PlanCustomPath) (PlannerInfo *root,
+ RelOptInfo *rel,
+ struct CustomPath *best_path,
+ List *tlist,
+ List *clauses,
+ List *custom_plans);
+ /* Optional: print additional fields besides "private" */
+ void (*TextOutCustomPath) (StringInfo str,
+ const struct CustomPath *node);
+} CustomPathMethods;
+
+typedef struct CustomPath
+{
+ Path path;
+ uint32 flags; /* mask of CUSTOMPATH_* flags, see above */
+ List *custom_paths; /* list of child Path nodes, if any */
+ List *custom_private;
+ const CustomPathMethods *methods;
+} CustomPath;
+
/*
* AppendPath represents an Append plan, ie, successive execution of
* several member plans.
#define IS_DUMMY_PATH(p) \
(IsA((p), AppendPath) && ((AppendPath *) (p))->subpaths == NIL)
+/* A relation that's been proven empty will have one path that is dummy */
+#define IS_DUMMY_REL(r) \
+ ((r)->cheapest_total_path != NULL && \
+ IS_DUMMY_PATH((r)->cheapest_total_path))
+
+/*
+ * MergeAppendPath represents a MergeAppend plan, ie, the merging of sorted
+ * results from several member plans to produce similarly-sorted output.
+ */
+typedef struct MergeAppendPath
+{
+ Path path;
+ List *subpaths; /* list of component Paths */
+ double limit_tuples; /* hard limit on output tuples, or -1 */
+} MergeAppendPath;
+
/*
* ResultPath represents use of a Result plan node to compute a variable-free
* targetlist with no underlying tables (a "SELECT expressions" query).
*
* This is unlike the other Path nodes in that it can actually generate
* different plans: either hash-based or sort-based implementation, or a
- * no-op if the input path can be proven distinct already. The decision
+ * no-op if the input path can be proven distinct already. The decision
* is sufficiently localized that it's not worth having separate Path node
* types. (Note: in the no-op case, we could eliminate the UniquePath node
* entirely and just return the subpath; but it's convenient to have a
UniquePathMethod umethod;
List *in_operators; /* equality operators of the IN clause */
List *uniq_exprs; /* expressions to be made unique */
- double rows; /* estimated number of result tuples */
} UniquePath;
+/*
+ * GatherPath runs several copies of a plan in parallel and collects the
+ * results. The parallel leader may also execute the plan, unless the
+ * single_copy flag is set.
+ */
+typedef struct GatherPath
+{
+ Path path;
+ Path *subpath; /* path for each worker */
+ int num_workers; /* number of workers sought to help */
+ bool single_copy; /* path must not be executed >1x */
+} GatherPath;
+
/*
* All join-type paths share these fields.
*/
List *joinrestrictinfo; /* RestrictInfos to apply to join */
/*
- * See the notes for RelOptInfo to understand why joinrestrictinfo is
- * needed in JoinPath, and can't be merged into the parent RelOptInfo.
+ * See the notes for RelOptInfo and ParamPathInfo to understand why
+ * joinrestrictinfo is needed in JoinPath, and can't be merged into the
+ * parent RelOptInfo.
*/
} JoinPath;
/*
* A mergejoin path has these fields.
*
+ * Unlike other path types, a MergePath node doesn't represent just a single
+ * run-time plan node: it can represent up to four. Aside from the MergeJoin
+ * node itself, there can be a Sort node for the outer input, a Sort node
+ * for the inner input, and/or a Material node for the inner input. We could
+ * represent these nodes by separate path nodes, but considering how many
+ * different merge paths are investigated during a complex join problem,
+ * it seems better to avoid unnecessary palloc overhead.
+ *
* path_mergeclauses lists the clauses (in the form of RestrictInfos)
* that will be used in the merge.
*
* outersortkeys (resp. innersortkeys) is NIL if the outer path
* (resp. inner path) is already ordered appropriately for the
* mergejoin. If it is not NIL then it is a PathKeys list describing
- * the ordering that must be created by an explicit sort step.
+ * the ordering that must be created by an explicit Sort node.
+ *
+ * materialize_inner is TRUE if a Material node should be placed atop the
+ * inner input. This may appear with or without an inner Sort step.
*/
typedef struct MergePath
List *path_mergeclauses; /* join clauses to be used for merge */
List *outersortkeys; /* keys for explicit sort, if any */
List *innersortkeys; /* keys for explicit sort, if any */
+ bool materialize_inner; /* add Materialize to inner? */
} MergePath;
/*
{
JoinPath jpath;
List *path_hashclauses; /* join clauses used for hashing */
+ int num_batches; /* number of batches expected */
} HashPath;
/*
* When we construct a join rel that includes all the base rels referenced
* in a multi-relation restriction clause, we place that clause into the
* joinrestrictinfo lists of paths for the join rel, if neither left nor
- * right sub-path includes all base rels referenced in the clause. The clause
+ * right sub-path includes all base rels referenced in the clause. The clause
* will be applied at that join level, and will not propagate any further up
* the join tree. (Note: the "predicate migration" code was once intended to
* push restriction clauses up and down the plan tree based on evaluation
* that appeared elsewhere in the tree and were pushed down to the join rel
* because they used no other rels. That's what the is_pushed_down flag is
* for; it tells us that a qual is not an OUTER JOIN qual for the set of base
- * rels listed in required_relids. A clause that originally came from WHERE
+ * rels listed in required_relids. A clause that originally came from WHERE
* or an INNER JOIN condition will *always* have its is_pushed_down flag set.
* It's possible for an OUTER JOIN clause to be marked is_pushed_down too,
* if we decide that it can be pushed down into the nullable side of the join.
*
* RestrictInfo nodes also contain an outerjoin_delayed flag, which is true
* if the clause's applicability must be delayed due to any outer joins
- * appearing below its own syntactic level (ie, it references any Vars from
- * the nullable side of any lower outer join).
+ * appearing below it (ie, it has to be postponed to some join level higher
+ * than the set of relations it actually references).
*
- * In general, the referenced clause might be arbitrarily complex. The
+ * There is also an outer_relids field, which is NULL except for outer join
+ * clauses; for those, it is the set of relids on the outer side of the
+ * clause's outer join. (These are rels that the clause cannot be applied to
+ * in parameterized scans, since pushing it into the join's outer side would
+ * lead to wrong answers.)
+ *
+ * There is also a nullable_relids field, which is the set of rels the clause
+ * references that can be forced null by some outer join below the clause.
+ *
+ * outerjoin_delayed = true is subtly different from nullable_relids != NULL:
+ * a clause might reference some nullable rels and yet not be
+ * outerjoin_delayed because it also references all the other rels of the
+ * outer join(s). A clause that is not outerjoin_delayed can be enforced
+ * anywhere it is computable.
+ *
+ * In general, the referenced clause might be arbitrarily complex. The
* kinds of clauses we can handle as indexscan quals, mergejoin clauses,
* or hashjoin clauses are limited (e.g., no volatile functions). The code
* for each kind of path is responsible for identifying the restrict clauses
*
* The pseudoconstant flag is set true if the clause contains no Vars of
* the current query level and no volatile functions. Such a clause can be
- * pulled out and used as a one-time qual in a gating Result node. We keep
+ * pulled out and used as a one-time qual in a gating Result node. We keep
* pseudoconstant clauses in the same lists as other RestrictInfos so that
* the regular clause-pushing machinery can assign them to the correct join
* level, but they need to be treated specially for cost and selectivity
*
* When join clauses are generated from EquivalenceClasses, there may be
* several equally valid ways to enforce join equivalence, of which we need
- * apply only one. We mark clauses of this kind by setting parent_ec to
+ * apply only one. We mark clauses of this kind by setting parent_ec to
* point to the generating EquivalenceClass. Multiple clauses with the same
* parent_ec in the same join are redundant.
*/
bool is_pushed_down; /* TRUE if clause was pushed down in level */
- bool outerjoin_delayed; /* TRUE if delayed by lower outer join */
+ bool outerjoin_delayed; /* TRUE if delayed by lower outer join */
bool can_join; /* see comment above */
/* The set of relids required to evaluate the clause: */
Relids required_relids;
+ /* If an outer-join clause, the outer-side relations, else NULL: */
+ Relids outer_relids;
+
+ /* The relids used in the clause that are nullable by lower outer joins: */
+ Relids nullable_relids;
+
/* These fields are set for any binary opclause: */
Relids left_relids; /* relids in left side of clause */
Relids right_relids; /* relids in right side of clause */
/* cache space for cost and selectivity */
QualCost eval_cost; /* eval cost of clause; -1 if not yet set */
Selectivity norm_selec; /* selectivity for "normal" (JOIN_INNER)
- * semantics; -1 if not yet set; >1 means
- * a redundant clause */
- Selectivity outer_selec; /* selectivity for outer join semantics;
- * -1 if not yet set */
+ * semantics; -1 if not yet set; >1 means a
+ * redundant clause */
+ Selectivity outer_selec; /* selectivity for outer join semantics; -1 if
+ * not yet set */
/* valid if clause is mergejoinable, else NIL */
List *mergeopfamilies; /* opfamilies containing clause operator */
{
/* Ordering details (cache lookup key) */
Oid opfamily; /* btree opfamily defining the ordering */
+ Oid collation; /* collation for the ordering */
int strategy; /* sort direction (ASC or DESC) */
bool nulls_first; /* do NULLs come before normal values? */
/* Results */
Selectivity rightendsel; /* last-join fraction for clause right side */
} MergeScanSelCache;
-/*
- * Inner indexscan info.
- *
- * An inner indexscan is one that uses one or more joinclauses as index
- * conditions (perhaps in addition to plain restriction clauses). So it
- * can only be used as the inner path of a nestloop join where the outer
- * relation includes all other relids appearing in those joinclauses.
- * The set of usable joinclauses, and thus the best inner indexscan,
- * thus varies depending on which outer relation we consider; so we have
- * to recompute the best such paths for every join. To avoid lots of
- * redundant computation, we cache the results of such searches. For
- * each relation we compute the set of possible otherrelids (all relids
- * appearing in joinquals that could become indexquals for this table).
- * Two outer relations whose relids have the same intersection with this
- * set will have the same set of available joinclauses and thus the same
- * best inner indexscans for the inner relation. By taking the intersection
- * before scanning the cache, we avoid recomputing when considering
- * join rels that differ only by the inclusion of irrelevant other rels.
- *
- * The search key also includes a bool showing whether the join being
- * considered is an outer join. Since we constrain the join order for
- * outer joins, I believe that this bool can only have one possible value
- * for any particular lookup key; but store it anyway to avoid confusion.
- */
-
-typedef struct InnerIndexscanInfo
-{
- NodeTag type;
- /* The lookup key: */
- Relids other_relids; /* a set of relevant other relids */
- bool isouterjoin; /* true if join is outer */
- /* Best paths for this lookup key (NULL if no available indexscans): */
- Path *cheapest_startup_innerpath; /* cheapest startup cost */
- Path *cheapest_total_innerpath; /* cheapest total cost */
-} InnerIndexscanInfo;
-
/*
* Placeholder node for an expression to be evaluated below the top level
* of a plan tree. This is used during planning to represent the contained
* "Special join" info.
*
* One-sided outer joins constrain the order of joining partially but not
- * completely. We flatten such joins into the planner's top-level list of
+ * completely. We flatten such joins into the planner's top-level list of
* relations to join, but record information about each outer join in a
* SpecialJoinInfo struct. These structs are kept in the PlannerInfo node's
* join_info_list.
* to be evaluated after this join is formed (because it references the RHS).
* Any outer joins that have such a clause and this join in their RHS cannot
* commute with this join, because that would leave noplace to check the
- * pushed-down clause. (We don't track this for FULL JOINs, either.)
+ * pushed-down clause. (We don't track this for FULL JOINs, either.)
*
- * join_quals is an implicit-AND list of the quals syntactically associated
- * with the join (they may or may not end up being applied at the join level).
- * This is just a side list and does not drive actual application of quals.
- * For JOIN_SEMI joins, this is cleared to NIL in create_unique_path() if
- * the join is found not to be suitable for a uniqueify-the-RHS plan.
+ * For a semijoin, we also extract the join operators and their RHS arguments
+ * and set semi_operators, semi_rhs_exprs, semi_can_btree, and semi_can_hash.
+ * This is done in support of possibly unique-ifying the RHS, so we don't
+ * bother unless at least one of semi_can_btree and semi_can_hash can be set
+ * true. (You might expect that this information would be computed during
+ * join planning; but it's helpful to have it available during planning of
+ * parameterized table scans, so we store it in the SpecialJoinInfo structs.)
*
* jointype is never JOIN_RIGHT; a RIGHT JOIN is handled by switching
* the inputs to make it a LEFT JOIN. So the allowed values of jointype
* SpecialJoinInfos with jointype == JOIN_INNER for outer joins, since for
* cost estimation purposes it is sometimes useful to know the join size under
* plain innerjoin semantics. Note that lhs_strict, delay_upper_joins, and
- * join_quals are not set meaningfully within such structs.
+ * of course the semi_xxx fields are not set meaningfully within such structs.
*/
typedef struct SpecialJoinInfo
JoinType jointype; /* always INNER, LEFT, FULL, SEMI, or ANTI */
bool lhs_strict; /* joinclause is strict for some LHS rel */
bool delay_upper_joins; /* can't commute with upper RHS */
- List *join_quals; /* join quals, in implicit-AND list format */
+ /* Remaining fields are set only for JOIN_SEMI jointype: */
+ bool semi_can_btree; /* true if semi_operators are all btree */
+ bool semi_can_hash; /* true if semi_operators are all hash */
+ List *semi_operators; /* OIDs of equality join operators */
+ List *semi_rhs_exprs; /* righthand-side expressions of these ops */
} SpecialJoinInfo;
+/*
+ * "Lateral join" info.
+ *
+ * Lateral references constrain the join order in a way that's somewhat like
+ * outer joins, though different in detail. We construct a LateralJoinInfo
+ * for each lateral cross-reference, placing them in the PlannerInfo node's
+ * lateral_info_list.
+ *
+ * For unflattened LATERAL RTEs, we generate LateralJoinInfo(s) in which
+ * lateral_rhs is the relid of the LATERAL baserel, and lateral_lhs is a set
+ * of relids of baserels it references, all of which must be present on the
+ * LHS to compute a parameter needed by the RHS. Typically, lateral_lhs is
+ * a singleton, but it can include multiple rels if the RHS references a
+ * PlaceHolderVar with a multi-rel ph_eval_at level. We disallow joining to
+ * only part of the LHS in such cases, since that would result in a join tree
+ * with no convenient place to compute the PHV.
+ *
+ * When an appendrel contains lateral references (eg "LATERAL (SELECT x.col1
+ * UNION ALL SELECT y.col2)"), the LateralJoinInfos reference the parent
+ * baserel not the member otherrels, since it is the parent relid that is
+ * considered for joining purposes.
+ *
+ * If any LATERAL RTEs were flattened into the parent query, it is possible
+ * that the query now contains PlaceHolderVars containing lateral references,
+ * representing expressions that need to be evaluated at particular spots in
+ * the jointree but contain lateral references to Vars from elsewhere. These
+ * give rise to LateralJoinInfos in which lateral_rhs is the evaluation point
+ * of a PlaceHolderVar and lateral_lhs is the set of lateral rels it needs.
+ */
+
+typedef struct LateralJoinInfo
+{
+ NodeTag type;
+ Relids lateral_lhs; /* rels needed to compute a lateral value */
+ Relids lateral_rhs; /* rel where lateral value is needed */
+} LateralJoinInfo;
+
/*
* Append-relation info.
*
/*
* For an inheritance appendrel, the parent and child are both regular
* relations, and we store their rowtype OIDs here for use in translating
- * whole-row Vars. For a UNION-ALL appendrel, the parent and child are
+ * whole-row Vars. For a UNION-ALL appendrel, the parent and child are
* both subqueries with no named rowtype, and we store InvalidOid here.
*/
Oid parent_reltype; /* OID of parent's composite type */
* Vars are special-cased, and system columns (attno < 0) need no special
* translation since their attnos are the same for all tables.
*
- * Caution: the Vars have varlevelsup = 0. Be careful to adjust as needed
+ * Caution: the Vars have varlevelsup = 0. Be careful to adjust as needed
* when copying into a subquery.
*/
List *translated_vars; /* Expressions in the child's Vars */
* then allow it to bubble up like a Var until the ph_needed join level.
* ph_needed has the same definition as attr_needed for a regular Var.
*
+ * The PlaceHolderVar's expression might contain LATERAL references to vars
+ * coming from outside its syntactic scope. If so, those rels are *not*
+ * included in ph_eval_at, but they are recorded in ph_lateral.
+ *
+ * Notice that when ph_eval_at is a join rather than a single baserel, the
+ * PlaceHolderInfo may create constraints on join order: the ph_eval_at join
+ * has to be formed below any outer joins that should null the PlaceHolderVar.
+ *
* We create a PlaceHolderInfo only after determining that the PlaceHolderVar
- * is actually referenced in the plan tree.
+ * is actually referenced in the plan tree, so that unreferenced placeholders
+ * don't result in unnecessary constraints on join order.
*/
typedef struct PlaceHolderInfo
Index phid; /* ID for PH (unique within planner run) */
PlaceHolderVar *ph_var; /* copy of PlaceHolderVar tree */
Relids ph_eval_at; /* lowest level we can evaluate value at */
+ Relids ph_lateral; /* relids of contained lateral refs, if any */
Relids ph_needed; /* highest level the value is needed at */
int32 ph_width; /* estimated attribute width */
} PlaceHolderInfo;
/*
- * glob->paramlist keeps track of the PARAM_EXEC slots that we have decided
- * we need for the query. At runtime these slots are used to pass values
- * either down into subqueries (for outer references in subqueries) or up out
- * of subqueries (for the results of a subplan). The n'th entry in the list
- * (n counts from 0) corresponds to Param->paramid = n.
- *
- * Each paramlist item shows the absolute query level it is associated with,
- * where the outermost query is level 1 and nested subqueries have higher
- * numbers. The item the parameter slot represents can be one of three kinds:
- *
- * A Var: the slot represents a variable of that level that must be passed
- * down because subqueries have outer references to it. The varlevelsup
- * value in the Var will always be zero.
+ * For each potentially index-optimizable MIN/MAX aggregate function,
+ * root->minmax_aggs stores a MinMaxAggInfo describing it.
+ */
+typedef struct MinMaxAggInfo
+{
+ NodeTag type;
+
+ Oid aggfnoid; /* pg_proc Oid of the aggregate */
+ Oid aggsortop; /* Oid of its sort operator */
+ Expr *target; /* expression we are aggregating on */
+ PlannerInfo *subroot; /* modified "root" for planning the subquery */
+ Path *path; /* access path for subquery */
+ Cost pathcost; /* estimated cost to fetch first row */
+ Param *param; /* param for subplan's output */
+} MinMaxAggInfo;
+
+/*
+ * At runtime, PARAM_EXEC slots are used to pass values around from one plan
+ * node to another. They can be used to pass values down into subqueries (for
+ * outer references in subqueries), or up out of subqueries (for the results
+ * of a subplan), or from a NestLoop plan node into its inner relation (when
+ * the inner scan is parameterized with values from the outer relation).
+ * The planner is responsible for assigning nonconflicting PARAM_EXEC IDs to
+ * the PARAM_EXEC Params it generates.
+ *
+ * Outer references are managed via root->plan_params, which is a list of
+ * PlannerParamItems. While planning a subquery, each parent query level's
+ * plan_params contains the values required from it by the current subquery.
+ * During create_plan(), we use plan_params to track values that must be
+ * passed from outer to inner sides of NestLoop plan nodes.
+ *
+ * The item a PlannerParamItem represents can be one of three kinds:
+ *
+ * A Var: the slot represents a variable of this level that must be passed
+ * down because subqueries have outer references to it, or must be passed
+ * from a NestLoop node to its inner scan. The varlevelsup value in the Var
+ * will always be zero.
+ *
+ * A PlaceHolderVar: this works much like the Var case, except that the
+ * entry is a PlaceHolderVar node with a contained expression. The PHV
+ * will have phlevelsup = 0, and the contained expression is adjusted
+ * to match in level.
*
* An Aggref (with an expression tree representing its argument): the slot
* represents an aggregate expression that is an outer reference for some
* subquery. The Aggref itself has agglevelsup = 0, and its argument tree
* is adjusted to match in level.
*
- * A Param: the slot holds the result of a subplan (it is a setParam item
- * for that subplan). The absolute level shown for such items corresponds
- * to the parent query of the subplan.
- *
- * Note: we detect duplicate Var parameters and coalesce them into one slot,
- * but we do not do this for Aggref or Param slots.
+ * Note: we detect duplicate Var and PlaceHolderVar parameters and coalesce
+ * them into one slot, but we do not bother to do that for Aggrefs.
+ * The scope of duplicate-elimination only extends across the set of
+ * parameters passed from one query level into a single subquery, or for
+ * nestloop parameters across the set of nestloop parameters used in a single
+ * query level. So there is no possibility of a PARAM_EXEC slot being used
+ * for conflicting purposes.
+ *
+ * In addition, PARAM_EXEC slots are assigned for Params representing outputs
+ * from subplans (values that are setParam items for those subplans). These
+ * IDs need not be tracked via PlannerParamItems, since we do not need any
+ * duplicate-elimination nor later processing of the represented expressions.
+ * Instead, we just record the assignment of the slot number by incrementing
+ * root->glob->nParamExec.
*/
typedef struct PlannerParamItem
{
NodeTag type;
- Node *item; /* the Var, Aggref, or Param */
- Index abslevel; /* its absolute query level */
+ Node *item; /* the Var, PlaceHolderVar, or Aggref */
+ int paramId; /* its assigned PARAM_EXEC slot number */
} PlannerParamItem;
+/*
+ * When making cost estimates for a SEMI or ANTI join, there are some
+ * correction factors that are needed in both nestloop and hash joins
+ * to account for the fact that the executor can stop scanning inner rows
+ * as soon as it finds a match to the current outer row. These numbers
+ * depend only on the selected outer and inner join relations, not on the
+ * particular paths used for them, so it's worthwhile to calculate them
+ * just once per relation pair not once per considered path. This struct
+ * is filled by compute_semi_anti_join_factors and must be passed along
+ * to the join cost estimation functions.
+ *
+ * outer_match_frac is the fraction of the outer tuples that are
+ * expected to have at least one match.
+ * match_count is the average number of matches expected for
+ * outer tuples that have at least one match.
+ */
+typedef struct SemiAntiJoinFactors
+{
+ Selectivity outer_match_frac;
+ Selectivity match_count;
+} SemiAntiJoinFactors;
+
+/*
+ * Struct for extra information passed to subroutines of add_paths_to_joinrel
+ *
+ * restrictlist contains all of the RestrictInfo nodes for restriction
+ * clauses that apply to this join
+ * mergeclause_list is a list of RestrictInfo nodes for available
+ * mergejoin clauses in this join
+ * sjinfo is extra info about special joins for selectivity estimation
+ * semifactors is as shown above (only valid for SEMI or ANTI joins)
+ * param_source_rels are OK targets for parameterization of result paths
+ * extra_lateral_rels are additional parameterization for result paths
+ */
+typedef struct JoinPathExtraData
+{
+ List *restrictlist;
+ List *mergeclause_list;
+ SpecialJoinInfo *sjinfo;
+ SemiAntiJoinFactors semifactors;
+ Relids param_source_rels;
+ Relids extra_lateral_rels;
+} JoinPathExtraData;
+
+/*
+ * For speed reasons, cost estimation for join paths is performed in two
+ * phases: the first phase tries to quickly derive a lower bound for the
+ * join cost, and then we check if that's sufficient to reject the path.
+ * If not, we come back for a more refined cost estimate. The first phase
+ * fills a JoinCostWorkspace struct with its preliminary cost estimates
+ * and possibly additional intermediate values. The second phase takes
+ * these values as inputs to avoid repeating work.
+ *
+ * (Ideally we'd declare this in cost.h, but it's also needed in pathnode.h,
+ * so seems best to put it here.)
+ */
+typedef struct JoinCostWorkspace
+{
+ /* Preliminary cost estimates --- must not be larger than final ones! */
+ Cost startup_cost; /* cost expended before fetching any tuples */
+ Cost total_cost; /* total cost (assuming all tuples fetched) */
+
+ /* Fields below here should be treated as private to costsize.c */
+ Cost run_cost; /* non-startup cost components */
+
+ /* private for cost_nestloop code */
+ Cost inner_run_cost; /* also used by cost_mergejoin code */
+ Cost inner_rescan_run_cost;
+
+ /* private for cost_mergejoin code */
+ double outer_rows;
+ double inner_rows;
+ double outer_skip_rows;
+ double inner_skip_rows;
+
+ /* private for cost_hashjoin code */
+ int numbuckets;
+ int numbatches;
+} JoinCostWorkspace;
+
#endif /* RELATION_H */
projects / postgresql / blobdiff