1
/*-------------------------------------------------------------------------
4
* definition of the system "statistic" relation (pg_statistic)
5
* along with the relation's initial contents.
8
* Portions Copyright (c) 1996-2005, PostgreSQL Global Development Group
9
* Portions Copyright (c) 1994, Regents of the University of California
11
* $PostgreSQL: pgsql/src/include/catalog/pg_statistic.h,v 1.28 2004-12-31 22:03:26 pgsql Exp $
14
* the genbki.sh script reads this file and generates .bki
15
* information from the DATA() statements.
17
*-------------------------------------------------------------------------
19
#ifndef PG_STATISTIC_H
20
#define PG_STATISTIC_H
23
* postgres.h contains the system type definitions and the
24
* CATALOG(), BOOTSTRAP and DATA() sugar words so this file
25
* can be read by both genbki.sh and the C compiler.
30
* Keep C compiler happy with anyarray, below. This will need to go elsewhere
31
* if we ever use anyarray for more than pg_statistic.
33
typedef struct varlena anyarray;
36
* pg_statistic definition. cpp turns this into
37
* typedef struct FormData_pg_statistic
40
CATALOG(pg_statistic) BKI_WITHOUT_OIDS
42
/* These fields form the unique key for the entry: */
43
Oid starelid; /* relation containing attribute */
44
int2 staattnum; /* attribute (column) stats are for */
46
/* the fraction of the column's entries that are NULL: */
50
* stawidth is the average width in bytes of non-null entries. For
51
* fixed-width datatypes this is of course the same as the typlen, but
52
* for var-width types it is more useful. Note that this is the
53
* average width of the data as actually stored, post-TOASTing (eg,
54
* for a moved-out-of-line value, only the size of the pointer object
55
* is counted). This is the appropriate definition for the primary
56
* use of the statistic, which is to estimate sizes of in-memory hash
62
* stadistinct indicates the (approximate) number of distinct non-null
63
* data values in the column. The interpretation is:
64
* 0 unknown or not computed
65
* > 0 actual number of distinct values
66
* < 0 negative of multiplier for number of rows
67
* The special negative case allows us to cope with columns that are
68
* unique (stadistinct = -1) or nearly so (for example, a column in
69
* which values appear about twice on the average could be represented
70
* by stadistinct = -0.5). Because the number-of-rows statistic in
71
* pg_class may be updated more frequently than pg_statistic is, it's
72
* important to be able to describe such situations as a multiple of
73
* the number of rows, rather than a fixed number of distinct values.
74
* But in other cases a fixed number is correct (eg, a boolean column).
80
* To allow keeping statistics on different kinds of datatypes,
81
* we do not hard-wire any particular meaning for the remaining
82
* statistical fields. Instead, we provide several "slots" in which
83
* statistical data can be placed. Each slot includes:
84
* kind integer code identifying kind of data
85
* op OID of associated operator, if needed
86
* numbers float4 array (for statistical values)
87
* values anyarray (for representations of data values)
88
* The ID and operator fields are never NULL; they are zeroes in an
89
* unused slot. The numbers and values fields are NULL in an unused
90
* slot, and might also be NULL in a used slot if the slot kind has
91
* no need for one or the other.
106
* THE REST OF THESE ARE VARIABLE LENGTH FIELDS, and may even be
107
* absent (NULL). They cannot be accessed as C struct entries; you
108
* have to use the full field access machinery (heap_getattr) for
109
* them. We declare them here for the catalog machinery.
112
float4 stanumbers1[1];
113
float4 stanumbers2[1];
114
float4 stanumbers3[1];
115
float4 stanumbers4[1];
118
* Values in these arrays are values of the column's data type. We
119
* presently have to cheat quite a bit to allow polymorphic arrays of
120
* this kind, but perhaps someday it'll be a less bogus facility.
126
} FormData_pg_statistic;
128
#define STATISTIC_NUM_SLOTS 4
131
* Form_pg_statistic corresponds to a pointer to a tuple with
132
* the format of pg_statistic relation.
135
typedef FormData_pg_statistic *Form_pg_statistic;
138
* compiler constants for pg_statistic
141
#define Natts_pg_statistic 21
142
#define Anum_pg_statistic_starelid 1
143
#define Anum_pg_statistic_staattnum 2
144
#define Anum_pg_statistic_stanullfrac 3
145
#define Anum_pg_statistic_stawidth 4
146
#define Anum_pg_statistic_stadistinct 5
147
#define Anum_pg_statistic_stakind1 6
148
#define Anum_pg_statistic_stakind2 7
149
#define Anum_pg_statistic_stakind3 8
150
#define Anum_pg_statistic_stakind4 9
151
#define Anum_pg_statistic_staop1 10
152
#define Anum_pg_statistic_staop2 11
153
#define Anum_pg_statistic_staop3 12
154
#define Anum_pg_statistic_staop4 13
155
#define Anum_pg_statistic_stanumbers1 14
156
#define Anum_pg_statistic_stanumbers2 15
157
#define Anum_pg_statistic_stanumbers3 16
158
#define Anum_pg_statistic_stanumbers4 17
159
#define Anum_pg_statistic_stavalues1 18
160
#define Anum_pg_statistic_stavalues2 19
161
#define Anum_pg_statistic_stavalues3 20
162
#define Anum_pg_statistic_stavalues4 21
165
* Currently, three statistical slot "kinds" are defined: most common values,
166
* histogram, and correlation. Additional "kinds" will probably appear in
167
* future to help cope with non-scalar datatypes. Also, custom data types
168
* can define their own "kind" codes by mutual agreement between a custom
169
* typanalyze routine and the selectivity estimation functions of the type's
172
* Code reading the pg_statistic relation should not assume that a particular
173
* data "kind" will appear in any particular slot. Instead, search the
174
* stakind fields to see if the desired data is available. (The standard
175
* function get_attstatsslot() may be used for this.)
179
* The present allocation of "kind" codes is:
181
* 1-99: reserved for assignment by the core PostgreSQL project
182
* (values in this range will be documented in this file)
183
* 100-199: reserved for assignment by the PostGIS project
184
* (values to be documented in PostGIS documentation)
185
* 200-9999: reserved for future public assignments
187
* For private use you may choose a "kind" code at random in the range
188
* 10000-30000. However, for code that is to be widely disseminated it is
189
* better to obtain a publicly defined "kind" code by request from the
190
* PostgreSQL Global Development Group.
194
* In a "most common values" slot, staop is the OID of the "=" operator
195
* used to decide whether values are the same or not. stavalues contains
196
* the K most common non-null values appearing in the column, and stanumbers
197
* contains their frequencies (fractions of total row count). The values
198
* shall be ordered in decreasing frequency. Note that since the arrays are
199
* variable-size, K may be chosen by the statistics collector. Values should
200
* not appear in MCV unless they have been observed to occur more than once;
201
* a unique column will have no MCV slot.
203
#define STATISTIC_KIND_MCV 1
206
* A "histogram" slot describes the distribution of scalar data. staop is
207
* the OID of the "<" operator that describes the sort ordering. (In theory,
208
* more than one histogram could appear, if a datatype has more than one
209
* useful sort operator.) stavalues contains M (>=2) non-null values that
210
* divide the non-null column data values into M-1 bins of approximately equal
211
* population. The first stavalues item is the MIN and the last is the MAX.
212
* stanumbers is not used and should be NULL. IMPORTANT POINT: if an MCV
213
* slot is also provided, then the histogram describes the data distribution
214
* *after removing the values listed in MCV* (thus, it's a "compressed
215
* histogram" in the technical parlance). This allows a more accurate
216
* representation of the distribution of a column with some very-common
217
* values. In a column with only a few distinct values, it's possible that
218
* the MCV list describes the entire data population; in this case the
219
* histogram reduces to empty and should be omitted.
221
#define STATISTIC_KIND_HISTOGRAM 2
224
* A "correlation" slot describes the correlation between the physical order
225
* of table tuples and the ordering of data values of this column, as seen
226
* by the "<" operator identified by staop. (As with the histogram, more
227
* than one entry could theoretically appear.) stavalues is not used and
228
* should be NULL. stanumbers contains a single entry, the correlation
229
* coefficient between the sequence of data values and the sequence of
230
* their actual tuple positions. The coefficient ranges from +1 to -1.
232
#define STATISTIC_KIND_CORRELATION 3
234
#endif /* PG_STATISTIC_H */