PostgreSQL Source Code  git master
pg_statistic.h
Go to the documentation of this file.
1 /*-------------------------------------------------------------------------
2  *
3  * pg_statistic.h
4  * definition of the "statistics" system catalog (pg_statistic)
5  *
6  *
7  * Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
8  * Portions Copyright (c) 1994, Regents of the University of California
9  *
10  * src/include/catalog/pg_statistic.h
11  *
12  * NOTES
13  * The Catalog.pm module reads this file and derives schema
14  * information.
15  *
16  *-------------------------------------------------------------------------
17  */
18 #ifndef PG_STATISTIC_H
19 #define PG_STATISTIC_H
20 
21 #include "catalog/genbki.h"
22 #include "catalog/pg_statistic_d.h"
23 
24 /* ----------------
25  * pg_statistic definition. cpp turns this into
26  * typedef struct FormData_pg_statistic
27  * ----------------
28  */
29 CATALOG(pg_statistic,2619,StatisticRelationId)
30 {
31  /* These fields form the unique key for the entry: */
32  Oid starelid BKI_LOOKUP(pg_class); /* relation containing
33  * attribute */
34  int16 staattnum; /* attribute (column) stats are for */
35  bool stainherit; /* true if inheritance children are included */
36 
37  /* the fraction of the column's entries that are NULL: */
38  float4 stanullfrac;
39 
40  /*
41  * stawidth is the average width in bytes of non-null entries. For
42  * fixed-width datatypes this is of course the same as the typlen, but for
43  * var-width types it is more useful. Note that this is the average width
44  * of the data as actually stored, post-TOASTing (eg, for a
45  * moved-out-of-line value, only the size of the pointer object is
46  * counted). This is the appropriate definition for the primary use of
47  * the statistic, which is to estimate sizes of in-memory hash tables of
48  * tuples.
49  */
50  int32 stawidth;
51 
52  /* ----------------
53  * stadistinct indicates the (approximate) number of distinct non-null
54  * data values in the column. The interpretation is:
55  * 0 unknown or not computed
56  * > 0 actual number of distinct values
57  * < 0 negative of multiplier for number of rows
58  * The special negative case allows us to cope with columns that are
59  * unique (stadistinct = -1) or nearly so (for example, a column in which
60  * non-null values appear about twice on the average could be represented
61  * by stadistinct = -0.5 if there are no nulls, or -0.4 if 20% of the
62  * column is nulls). Because the number-of-rows statistic in pg_class may
63  * be updated more frequently than pg_statistic is, it's important to be
64  * able to describe such situations as a multiple of the number of rows,
65  * rather than a fixed number of distinct values. But in other cases a
66  * fixed number is correct (eg, a boolean column).
67  * ----------------
68  */
69  float4 stadistinct;
70 
71  /* ----------------
72  * To allow keeping statistics on different kinds of datatypes,
73  * we do not hard-wire any particular meaning for the remaining
74  * statistical fields. Instead, we provide several "slots" in which
75  * statistical data can be placed. Each slot includes:
76  * kind integer code identifying kind of data (see below)
77  * op OID of associated operator, if needed
78  * coll OID of relevant collation, or 0 if none
79  * numbers float4 array (for statistical values)
80  * values anyarray (for representations of data values)
81  * The ID, operator, and collation fields are never NULL; they are zeroes
82  * in an unused slot. The numbers and values fields are NULL in an
83  * unused slot, and might also be NULL in a used slot if the slot kind
84  * has no need for one or the other.
85  * ----------------
86  */
87 
88  int16 stakind1;
89  int16 stakind2;
90  int16 stakind3;
91  int16 stakind4;
92  int16 stakind5;
93 
94  Oid staop1 BKI_LOOKUP_OPT(pg_operator);
95  Oid staop2 BKI_LOOKUP_OPT(pg_operator);
96  Oid staop3 BKI_LOOKUP_OPT(pg_operator);
97  Oid staop4 BKI_LOOKUP_OPT(pg_operator);
98  Oid staop5 BKI_LOOKUP_OPT(pg_operator);
99 
100  Oid stacoll1 BKI_LOOKUP_OPT(pg_collation);
101  Oid stacoll2 BKI_LOOKUP_OPT(pg_collation);
102  Oid stacoll3 BKI_LOOKUP_OPT(pg_collation);
103  Oid stacoll4 BKI_LOOKUP_OPT(pg_collation);
104  Oid stacoll5 BKI_LOOKUP_OPT(pg_collation);
105 
106 #ifdef CATALOG_VARLEN /* variable-length fields start here */
107  float4 stanumbers1[1];
108  float4 stanumbers2[1];
109  float4 stanumbers3[1];
110  float4 stanumbers4[1];
111  float4 stanumbers5[1];
112 
113  /*
114  * Values in these arrays are values of the column's data type, or of some
115  * related type such as an array element type. We presently have to cheat
116  * quite a bit to allow polymorphic arrays of this kind, but perhaps
117  * someday it'll be a less bogus facility.
118  */
119  anyarray stavalues1;
120  anyarray stavalues2;
121  anyarray stavalues3;
122  anyarray stavalues4;
123  anyarray stavalues5;
124 #endif
126 
127 #define STATISTIC_NUM_SLOTS 5
128 
129 
130 /* ----------------
131  * Form_pg_statistic corresponds to a pointer to a tuple with
132  * the format of pg_statistic relation.
133  * ----------------
134  */
136 
137 DECLARE_TOAST(pg_statistic, 2840, 2841);
138 
139 DECLARE_UNIQUE_INDEX_PKEY(pg_statistic_relid_att_inh_index, 2696, StatisticRelidAttnumInhIndexId, on pg_statistic using btree(starelid oid_ops, staattnum int2_ops, stainherit bool_ops));
140 
141 DECLARE_FOREIGN_KEY((starelid, staattnum), pg_attribute, (attrelid, attnum));
142 
143 #ifdef EXPOSE_TO_CLIENT_CODE
144 
145 /*
146  * Several statistical slot "kinds" are defined by core PostgreSQL, as
147  * documented below. Also, custom data types can define their own "kind"
148  * codes by mutual agreement between a custom typanalyze routine and the
149  * selectivity estimation functions of the type's operators.
150  *
151  * Code reading the pg_statistic relation should not assume that a particular
152  * data "kind" will appear in any particular slot. Instead, search the
153  * stakind fields to see if the desired data is available. (The standard
154  * function get_attstatsslot() may be used for this.)
155  */
156 
157 /*
158  * The present allocation of "kind" codes is:
159  *
160  * 1-99: reserved for assignment by the core PostgreSQL project
161  * (values in this range will be documented in this file)
162  * 100-199: reserved for assignment by the PostGIS project
163  * (values to be documented in PostGIS documentation)
164  * 200-299: reserved for assignment by the ESRI ST_Geometry project
165  * (values to be documented in ESRI ST_Geometry documentation)
166  * 300-9999: reserved for future public assignments
167  *
168  * For private use you may choose a "kind" code at random in the range
169  * 10000-30000. However, for code that is to be widely disseminated it is
170  * better to obtain a publicly defined "kind" code by request from the
171  * PostgreSQL Global Development Group.
172  */
173 
174 /*
175  * In a "most common values" slot, staop is the OID of the "=" operator
176  * used to decide whether values are the same or not, and stacoll is the
177  * collation used (same as column's collation). stavalues contains
178  * the K most common non-null values appearing in the column, and stanumbers
179  * contains their frequencies (fractions of total row count). The values
180  * shall be ordered in decreasing frequency. Note that since the arrays are
181  * variable-size, K may be chosen by the statistics collector. Values should
182  * not appear in MCV unless they have been observed to occur more than once;
183  * a unique column will have no MCV slot.
184  */
185 #define STATISTIC_KIND_MCV 1
186 
187 /*
188  * A "histogram" slot describes the distribution of scalar data. staop is
189  * the OID of the "<" operator that describes the sort ordering, and stacoll
190  * is the relevant collation. (In theory more than one histogram could appear,
191  * if a datatype has more than one useful sort operator or we care about more
192  * than one collation. Currently the collation will always be that of the
193  * underlying column.) stavalues contains M (>=2) non-null values that
194  * divide the non-null column data values into M-1 bins of approximately equal
195  * population. The first stavalues item is the MIN and the last is the MAX.
196  * stanumbers is not used and should be NULL. IMPORTANT POINT: if an MCV
197  * slot is also provided, then the histogram describes the data distribution
198  * *after removing the values listed in MCV* (thus, it's a "compressed
199  * histogram" in the technical parlance). This allows a more accurate
200  * representation of the distribution of a column with some very-common
201  * values. In a column with only a few distinct values, it's possible that
202  * the MCV list describes the entire data population; in this case the
203  * histogram reduces to empty and should be omitted.
204  */
205 #define STATISTIC_KIND_HISTOGRAM 2
206 
207 /*
208  * A "correlation" slot describes the correlation between the physical order
209  * of table tuples and the ordering of data values of this column, as seen
210  * by the "<" operator identified by staop with the collation identified by
211  * stacoll. (As with the histogram, more than one entry could theoretically
212  * appear.) stavalues is not used and should be NULL. stanumbers contains
213  * a single entry, the correlation coefficient between the sequence of data
214  * values and the sequence of their actual tuple positions. The coefficient
215  * ranges from +1 to -1.
216  */
217 #define STATISTIC_KIND_CORRELATION 3
218 
219 /*
220  * A "most common elements" slot is similar to a "most common values" slot,
221  * except that it stores the most common non-null *elements* of the column
222  * values. This is useful when the column datatype is an array or some other
223  * type with identifiable elements (for instance, tsvector). staop contains
224  * the equality operator appropriate to the element type, and stacoll
225  * contains the collation to use with it. stavalues contains
226  * the most common element values, and stanumbers their frequencies. Unlike
227  * MCV slots, frequencies are measured as the fraction of non-null rows the
228  * element value appears in, not the frequency of all rows. Also unlike
229  * MCV slots, the values are sorted into the element type's default order
230  * (to support binary search for a particular value). Since this puts the
231  * minimum and maximum frequencies at unpredictable spots in stanumbers,
232  * there are two extra members of stanumbers, holding copies of the minimum
233  * and maximum frequencies. Optionally, there can be a third extra member,
234  * which holds the frequency of null elements (expressed in the same terms:
235  * the fraction of non-null rows that contain at least one null element). If
236  * this member is omitted, the column is presumed to contain no null elements.
237  *
238  * Note: in current usage for tsvector columns, the stavalues elements are of
239  * type text, even though their representation within tsvector is not
240  * exactly text.
241  */
242 #define STATISTIC_KIND_MCELEM 4
243 
244 /*
245  * A "distinct elements count histogram" slot describes the distribution of
246  * the number of distinct element values present in each row of an array-type
247  * column. Only non-null rows are considered, and only non-null elements.
248  * staop contains the equality operator appropriate to the element type,
249  * and stacoll contains the collation to use with it.
250  * stavalues is not used and should be NULL. The last member of stanumbers is
251  * the average count of distinct element values over all non-null rows. The
252  * preceding M (>=2) members form a histogram that divides the population of
253  * distinct-elements counts into M-1 bins of approximately equal population.
254  * The first of these is the minimum observed count, and the last the maximum.
255  */
256 #define STATISTIC_KIND_DECHIST 5
257 
258 /*
259  * A "length histogram" slot describes the distribution of range lengths in
260  * rows of a range-type column. stanumbers contains a single entry, the
261  * fraction of empty ranges. stavalues is a histogram of non-empty lengths, in
262  * a format similar to STATISTIC_KIND_HISTOGRAM: it contains M (>=2) range
263  * values that divide the column data values into M-1 bins of approximately
264  * equal population. The lengths are stored as float8s, as measured by the
265  * range type's subdiff function. Only non-null rows are considered.
266  */
267 #define STATISTIC_KIND_RANGE_LENGTH_HISTOGRAM 6
268 
269 /*
270  * A "bounds histogram" slot is similar to STATISTIC_KIND_HISTOGRAM, but for
271  * a range-type column. stavalues contains M (>=2) range values that divide
272  * the column data values into M-1 bins of approximately equal population.
273  * Unlike a regular scalar histogram, this is actually two histograms combined
274  * into a single array, with the lower bounds of each value forming a
275  * histogram of lower bounds, and the upper bounds a histogram of upper
276  * bounds. Only non-NULL, non-empty ranges are included.
277  */
278 #define STATISTIC_KIND_BOUNDS_HISTOGRAM 7
279 
280 #endif /* EXPOSE_TO_CLIENT_CODE */
281 
282 #endif /* PG_STATISTIC_H */
signed short int16
Definition: c.h:477
signed int int32
Definition: c.h:478
float float4
Definition: c.h:613
#define BKI_LOOKUP(catalog)
Definition: genbki.h:46
#define BKI_LOOKUP_OPT(catalog)
Definition: genbki.h:47
int16 attnum
Definition: pg_attribute.h:74
DECLARE_FOREIGN_KEY((starelid, staattnum), pg_attribute,(attrelid, attnum))
CATALOG(pg_statistic, 2619, StatisticRelationId)
Definition: pg_statistic.h:29
FormData_pg_statistic
Definition: pg_statistic.h:125
DECLARE_TOAST(pg_statistic, 2840, 2841)
DECLARE_UNIQUE_INDEX_PKEY(pg_statistic_relid_att_inh_index, 2696, StatisticRelidAttnumInhIndexId, on pg_statistic using btree(starelid oid_ops, staattnum int2_ops, stainherit bool_ops))
FormData_pg_statistic * Form_pg_statistic
Definition: pg_statistic.h:135
unsigned int Oid
Definition: postgres_ext.h:31