1 /*-------------------------------------------------------------------------
4 * definition of the system "statistic" relation (pg_statistic)
5 * along with the relation's initial contents.
8 * Portions Copyright (c) 1996-2017, PostgreSQL Global Development Group
9 * Portions Copyright (c) 1994, Regents of the University of California
11 * src/include/catalog/pg_statistic.h
14 * the genbki.pl script reads this file and generates .bki
15 * information from the DATA() statements.
17 *-------------------------------------------------------------------------
19 #ifndef PG_STATISTIC_H
20 #define PG_STATISTIC_H
22 #include "catalog/genbki.h"
25 * pg_statistic definition. cpp turns this into
26 * typedef struct FormData_pg_statistic
29 #define StatisticRelationId 2619
31 CATALOG(pg_statistic,2619) BKI_WITHOUT_OIDS
33 /* These fields form the unique key for the entry: */
34 Oid starelid; /* relation containing attribute */
35 int16 staattnum; /* attribute (column) stats are for */
36 bool stainherit; /* true if inheritance children are included */
38 /* the fraction of the column's entries that are NULL: */
42 * stawidth is the average width in bytes of non-null entries. For
43 * fixed-width datatypes this is of course the same as the typlen, but for
44 * var-width types it is more useful. Note that this is the average width
45 * of the data as actually stored, post-TOASTing (eg, for a
46 * moved-out-of-line value, only the size of the pointer object is
47 * counted). This is the appropriate definition for the primary use of
48 * the statistic, which is to estimate sizes of in-memory hash tables of
54 * stadistinct indicates the (approximate) number of distinct non-null
55 * data values in the column. The interpretation is:
56 * 0 unknown or not computed
57 * > 0 actual number of distinct values
58 * < 0 negative of multiplier for number of rows
59 * The special negative case allows us to cope with columns that are
60 * unique (stadistinct = -1) or nearly so (for example, a column in which
61 * non-null values appear about twice on the average could be represented
62 * by stadistinct = -0.5 if there are no nulls, or -0.4 if 20% of the
63 * column is nulls). Because the number-of-rows statistic in pg_class may
64 * be updated more frequently than pg_statistic is, it's important to be
65 * able to describe such situations as a multiple of the number of rows,
66 * rather than a fixed number of distinct values. But in other cases a
67 * fixed number is correct (eg, a boolean column).
73 * To allow keeping statistics on different kinds of datatypes,
74 * we do not hard-wire any particular meaning for the remaining
75 * statistical fields. Instead, we provide several "slots" in which
76 * statistical data can be placed. Each slot includes:
77 * kind integer code identifying kind of data (see below)
78 * op OID of associated operator, if needed
79 * numbers float4 array (for statistical values)
80 * values anyarray (for representations of data values)
81 * The ID and operator fields are never NULL; they are zeroes in an
82 * unused slot. The numbers and values fields are NULL in an unused
83 * slot, and might also be NULL in a used slot if the slot kind has
84 * no need for one or the other.
100 #ifdef CATALOG_VARLEN /* variable-length fields start here */
101 float4 stanumbers1[1];
102 float4 stanumbers2[1];
103 float4 stanumbers3[1];
104 float4 stanumbers4[1];
105 float4 stanumbers5[1];
108 * Values in these arrays are values of the column's data type, or of some
109 * related type such as an array element type. We presently have to cheat
110 * quite a bit to allow polymorphic arrays of this kind, but perhaps
111 * someday it'll be a less bogus facility.
119 } FormData_pg_statistic;
121 #define STATISTIC_NUM_SLOTS 5
125 * Form_pg_statistic corresponds to a pointer to a tuple with
126 * the format of pg_statistic relation.
129 typedef FormData_pg_statistic *Form_pg_statistic;
132 * compiler constants for pg_statistic
135 #define Natts_pg_statistic 26
136 #define Anum_pg_statistic_starelid 1
137 #define Anum_pg_statistic_staattnum 2
138 #define Anum_pg_statistic_stainherit 3
139 #define Anum_pg_statistic_stanullfrac 4
140 #define Anum_pg_statistic_stawidth 5
141 #define Anum_pg_statistic_stadistinct 6
142 #define Anum_pg_statistic_stakind1 7
143 #define Anum_pg_statistic_stakind2 8
144 #define Anum_pg_statistic_stakind3 9
145 #define Anum_pg_statistic_stakind4 10
146 #define Anum_pg_statistic_stakind5 11
147 #define Anum_pg_statistic_staop1 12
148 #define Anum_pg_statistic_staop2 13
149 #define Anum_pg_statistic_staop3 14
150 #define Anum_pg_statistic_staop4 15
151 #define Anum_pg_statistic_staop5 16
152 #define Anum_pg_statistic_stanumbers1 17
153 #define Anum_pg_statistic_stanumbers2 18
154 #define Anum_pg_statistic_stanumbers3 19
155 #define Anum_pg_statistic_stanumbers4 20
156 #define Anum_pg_statistic_stanumbers5 21
157 #define Anum_pg_statistic_stavalues1 22
158 #define Anum_pg_statistic_stavalues2 23
159 #define Anum_pg_statistic_stavalues3 24
160 #define Anum_pg_statistic_stavalues4 25
161 #define Anum_pg_statistic_stavalues5 26
164 * Currently, five statistical slot "kinds" are defined by core PostgreSQL,
165 * as documented below. Additional "kinds" will probably appear in
166 * future to help cope with non-scalar datatypes. Also, custom data types
167 * can define their own "kind" codes by mutual agreement between a custom
168 * typanalyze routine and the selectivity estimation functions of the type's
171 * Code reading the pg_statistic relation should not assume that a particular
172 * data "kind" will appear in any particular slot. Instead, search the
173 * stakind fields to see if the desired data is available. (The standard
174 * function get_attstatsslot() may be used for this.)
178 * The present allocation of "kind" codes is:
180 * 1-99: reserved for assignment by the core PostgreSQL project
181 * (values in this range will be documented in this file)
182 * 100-199: reserved for assignment by the PostGIS project
183 * (values to be documented in PostGIS documentation)
184 * 200-299: reserved for assignment by the ESRI ST_Geometry project
185 * (values to be documented in ESRI ST_Geometry documentation)
186 * 300-9999: reserved for future public assignments
188 * For private use you may choose a "kind" code at random in the range
189 * 10000-30000. However, for code that is to be widely disseminated it is
190 * better to obtain a publicly defined "kind" code by request from the
191 * PostgreSQL Global Development Group.
195 * In a "most common values" slot, staop is the OID of the "=" operator
196 * used to decide whether values are the same or not. stavalues contains
197 * the K most common non-null values appearing in the column, and stanumbers
198 * contains their frequencies (fractions of total row count). The values
199 * shall be ordered in decreasing frequency. Note that since the arrays are
200 * variable-size, K may be chosen by the statistics collector. Values should
201 * not appear in MCV unless they have been observed to occur more than once;
202 * a unique column will have no MCV slot.
204 #define STATISTIC_KIND_MCV 1
207 * A "histogram" slot describes the distribution of scalar data. staop is
208 * the OID of the "<" operator that describes the sort ordering. (In theory,
209 * more than one histogram could appear, if a datatype has more than one
210 * useful sort operator.) stavalues contains M (>=2) non-null values that
211 * divide the non-null column data values into M-1 bins of approximately equal
212 * population. The first stavalues item is the MIN and the last is the MAX.
213 * stanumbers is not used and should be NULL. IMPORTANT POINT: if an MCV
214 * slot is also provided, then the histogram describes the data distribution
215 * *after removing the values listed in MCV* (thus, it's a "compressed
216 * histogram" in the technical parlance). This allows a more accurate
217 * representation of the distribution of a column with some very-common
218 * values. In a column with only a few distinct values, it's possible that
219 * the MCV list describes the entire data population; in this case the
220 * histogram reduces to empty and should be omitted.
222 #define STATISTIC_KIND_HISTOGRAM 2
225 * A "correlation" slot describes the correlation between the physical order
226 * of table tuples and the ordering of data values of this column, as seen
227 * by the "<" operator identified by staop. (As with the histogram, more
228 * than one entry could theoretically appear.) stavalues is not used and
229 * should be NULL. stanumbers contains a single entry, the correlation
230 * coefficient between the sequence of data values and the sequence of
231 * their actual tuple positions. The coefficient ranges from +1 to -1.
233 #define STATISTIC_KIND_CORRELATION 3
236 * A "most common elements" slot is similar to a "most common values" slot,
237 * except that it stores the most common non-null *elements* of the column
238 * values. This is useful when the column datatype is an array or some other
239 * type with identifiable elements (for instance, tsvector). staop contains
240 * the equality operator appropriate to the element type. stavalues contains
241 * the most common element values, and stanumbers their frequencies. Unlike
242 * MCV slots, frequencies are measured as the fraction of non-null rows the
243 * element value appears in, not the frequency of all rows. Also unlike
244 * MCV slots, the values are sorted into the element type's default order
245 * (to support binary search for a particular value). Since this puts the
246 * minimum and maximum frequencies at unpredictable spots in stanumbers,
247 * there are two extra members of stanumbers, holding copies of the minimum
248 * and maximum frequencies. Optionally, there can be a third extra member,
249 * which holds the frequency of null elements (expressed in the same terms:
250 * the fraction of non-null rows that contain at least one null element). If
251 * this member is omitted, the column is presumed to contain no null elements.
253 * Note: in current usage for tsvector columns, the stavalues elements are of
254 * type text, even though their representation within tsvector is not
257 #define STATISTIC_KIND_MCELEM 4
260 * A "distinct elements count histogram" slot describes the distribution of
261 * the number of distinct element values present in each row of an array-type
262 * column. Only non-null rows are considered, and only non-null elements.
263 * staop contains the equality operator appropriate to the element type.
264 * stavalues is not used and should be NULL. The last member of stanumbers is
265 * the average count of distinct element values over all non-null rows. The
266 * preceding M (>=2) members form a histogram that divides the population of
267 * distinct-elements counts into M-1 bins of approximately equal population.
268 * The first of these is the minimum observed count, and the last the maximum.
270 #define STATISTIC_KIND_DECHIST 5
273 * A "length histogram" slot describes the distribution of range lengths in
274 * rows of a range-type column. stanumbers contains a single entry, the
275 * fraction of empty ranges. stavalues is a histogram of non-empty lengths, in
276 * a format similar to STATISTIC_KIND_HISTOGRAM: it contains M (>=2) range
277 * values that divide the column data values into M-1 bins of approximately
278 * equal population. The lengths are stored as float8s, as measured by the
279 * range type's subdiff function. Only non-null rows are considered.
281 #define STATISTIC_KIND_RANGE_LENGTH_HISTOGRAM 6
284 * A "bounds histogram" slot is similar to STATISTIC_KIND_HISTOGRAM, but for
285 * a range-type column. stavalues contains M (>=2) range values that divide
286 * the column data values into M-1 bins of approximately equal population.
287 * Unlike a regular scalar histogram, this is actually two histograms combined
288 * into a single array, with the lower bounds of each value forming a
289 * histogram of lower bounds, and the upper bounds a histogram of upper
290 * bounds. Only non-NULL, non-empty ranges are included.
292 #define STATISTIC_KIND_BOUNDS_HISTOGRAM 7
294 #endif /* PG_STATISTIC_H */