1 /*-------------------------------------------------------------------------
4 * Selectivity functions for standard operators, and assorted
5 * infrastructure for selectivity and cost estimation.
8 * Portions Copyright (c) 1996-2019, PostgreSQL Global Development Group
9 * Portions Copyright (c) 1994, Regents of the University of California
11 * src/include/utils/selfuncs.h
13 *-------------------------------------------------------------------------
18 #include "access/htup.h"
20 #include "nodes/pathnodes.h"
24 * Note: the default selectivity estimates are not chosen entirely at random.
25 * We want them to be small enough to ensure that indexscans will be used if
26 * available, for typical table densities of ~100 tuples/page. Thus, for
27 * example, 0.01 is not quite small enough, since that makes it appear that
28 * nearly all pages will be hit anyway. Also, since we sometimes estimate
29 * eqsel as 1/num_distinct, we probably want DEFAULT_NUM_DISTINCT to equal
33 /* default selectivity estimate for equalities such as "A = b" */
34 #define DEFAULT_EQ_SEL 0.005
36 /* default selectivity estimate for inequalities such as "A < b" */
37 #define DEFAULT_INEQ_SEL 0.3333333333333333
39 /* default selectivity estimate for range inequalities "A > b AND A < c" */
40 #define DEFAULT_RANGE_INEQ_SEL 0.005
42 /* default selectivity estimate for pattern-match operators such as LIKE */
43 #define DEFAULT_MATCH_SEL 0.005
45 /* default number of distinct values in a table */
46 #define DEFAULT_NUM_DISTINCT 200
48 /* default selectivity estimate for boolean and null test nodes */
49 #define DEFAULT_UNK_SEL 0.005
50 #define DEFAULT_NOT_UNK_SEL (1.0 - DEFAULT_UNK_SEL)
54 * Clamp a computed probability estimate (which may suffer from roundoff or
55 * estimation errors) to valid range. Argument must be a float variable.
57 #define CLAMP_PROBABILITY(p) \
66 /* Return data from examine_variable and friends */
67 typedef struct VariableStatData
69 Node *var; /* the Var or expression tree */
70 RelOptInfo *rel; /* Relation, or NULL if not identifiable */
71 HeapTuple statsTuple; /* pg_statistic tuple, or NULL if none */
72 /* NB: if statsTuple!=NULL, it must be freed when caller is done */
73 void (*freefunc) (HeapTuple tuple); /* how to free statsTuple */
74 Oid vartype; /* exposed type of expression */
75 Oid atttype; /* actual type (after stripping relabel) */
76 int32 atttypmod; /* actual typmod (after stripping relabel) */
77 bool isunique; /* matches unique index or DISTINCT clause */
78 bool acl_ok; /* result of ACL check on table or column */
81 #define ReleaseVariableStats(vardata) \
83 if (HeapTupleIsValid((vardata).statsTuple)) \
84 (vardata).freefunc((vardata).statsTuple); \
89 * genericcostestimate is a general-purpose estimator that can be used for
90 * most index types. In some cases we use genericcostestimate as the base
91 * code and then incorporate additional index-type-specific knowledge in
92 * the type-specific calling function. To avoid code duplication, we make
93 * genericcostestimate return a number of intermediate values as well as
94 * its preliminary estimates of the output cost values. The GenericCosts
95 * struct includes all these values.
97 * Callers should initialize all fields of GenericCosts to zero. In addition,
98 * they can set numIndexTuples to some positive value if they have a better
99 * than default way of estimating the number of leaf index tuples visited.
103 /* These are the values the cost estimator must return to the planner */
104 Cost indexStartupCost; /* index-related startup cost */
105 Cost indexTotalCost; /* total index-related scan cost */
106 Selectivity indexSelectivity; /* selectivity of index */
107 double indexCorrelation; /* order correlation of index */
109 /* Intermediate values we obtain along the way */
110 double numIndexPages; /* number of leaf pages visited */
111 double numIndexTuples; /* number of leaf tuples visited */
112 double spc_random_page_cost; /* relevant random_page_cost value */
113 double num_sa_scans; /* # indexscans from ScalarArrayOpExprs */
116 /* Hooks for plugins to get control when we ask for stats */
117 typedef bool (*get_relation_stats_hook_type) (PlannerInfo *root,
120 VariableStatData *vardata);
121 extern PGDLLIMPORT get_relation_stats_hook_type get_relation_stats_hook;
122 typedef bool (*get_index_stats_hook_type) (PlannerInfo *root,
124 AttrNumber indexattnum,
125 VariableStatData *vardata);
126 extern PGDLLIMPORT get_index_stats_hook_type get_index_stats_hook;
128 /* Functions in selfuncs.c */
130 extern void examine_variable(PlannerInfo *root, Node *node, int varRelid,
131 VariableStatData *vardata);
132 extern bool statistic_proc_security_check(VariableStatData *vardata, Oid func_oid);
133 extern bool get_restriction_variable(PlannerInfo *root, List *args,
135 VariableStatData *vardata, Node **other,
137 extern void get_join_variables(PlannerInfo *root, List *args,
138 SpecialJoinInfo *sjinfo,
139 VariableStatData *vardata1,
140 VariableStatData *vardata2,
141 bool *join_is_reversed);
142 extern double get_variable_numdistinct(VariableStatData *vardata,
144 extern double mcv_selectivity(VariableStatData *vardata, FmgrInfo *opproc,
145 Datum constval, bool varonleft,
147 extern double histogram_selectivity(VariableStatData *vardata, FmgrInfo *opproc,
148 Datum constval, bool varonleft,
149 int min_hist_size, int n_skip,
151 extern double ineq_histogram_selectivity(PlannerInfo *root,
152 VariableStatData *vardata,
153 FmgrInfo *opproc, bool isgt, bool iseq,
154 Datum constval, Oid consttype);
155 extern double var_eq_const(VariableStatData *vardata, Oid oproid,
156 Datum constval, bool constisnull,
157 bool varonleft, bool negate);
158 extern double var_eq_non_const(VariableStatData *vardata, Oid oproid,
160 bool varonleft, bool negate);
162 extern Selectivity boolvarsel(PlannerInfo *root, Node *arg, int varRelid);
163 extern Selectivity booltestsel(PlannerInfo *root, BoolTestType booltesttype,
164 Node *arg, int varRelid,
165 JoinType jointype, SpecialJoinInfo *sjinfo);
166 extern Selectivity nulltestsel(PlannerInfo *root, NullTestType nulltesttype,
167 Node *arg, int varRelid,
168 JoinType jointype, SpecialJoinInfo *sjinfo);
169 extern Selectivity scalararraysel(PlannerInfo *root,
170 ScalarArrayOpExpr *clause,
172 int varRelid, JoinType jointype, SpecialJoinInfo *sjinfo);
173 extern int estimate_array_length(Node *arrayexpr);
174 extern Selectivity rowcomparesel(PlannerInfo *root,
175 RowCompareExpr *clause,
176 int varRelid, JoinType jointype, SpecialJoinInfo *sjinfo);
178 extern void mergejoinscansel(PlannerInfo *root, Node *clause,
179 Oid opfamily, int strategy, bool nulls_first,
180 Selectivity *leftstart, Selectivity *leftend,
181 Selectivity *rightstart, Selectivity *rightend);
183 extern double estimate_num_groups(PlannerInfo *root, List *groupExprs,
184 double input_rows, List **pgset);
186 extern void estimate_hash_bucket_stats(PlannerInfo *root,
187 Node *hashkey, double nbuckets,
188 Selectivity *mcv_freq,
189 Selectivity *bucketsize_frac);
190 extern double estimate_hashagg_tablesize(Path *path,
191 const AggClauseCosts *agg_costs,
194 extern List *get_quals_from_indexclauses(List *indexclauses);
195 extern Cost index_other_operands_eval_cost(PlannerInfo *root,
197 extern List *add_predicate_to_index_quals(IndexOptInfo *index,
199 extern void genericcostestimate(PlannerInfo *root, IndexPath *path,
201 GenericCosts *costs);
203 /* Functions in array_selfuncs.c */
205 extern Selectivity scalararraysel_containment(PlannerInfo *root,
206 Node *leftop, Node *rightop,
207 Oid elemtype, bool isEquality, bool useOr,
210 #endif /* SELFUNCS_H */