1 <?xml version="1.0" encoding="UTF-8" standalone="no"?>
2 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"><html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>63.1. Basic API Structure for Indexes</title><link rel="stylesheet" type="text/css" href="stylesheet.css" /><link rev="made" href="pgsql-docs@lists.postgresql.org" /><meta name="generator" content="DocBook XSL Stylesheets Vsnapshot" /><link rel="prev" href="indexam.html" title="Chapter 63. Index Access Method Interface Definition" /><link rel="next" href="index-functions.html" title="63.2. Index Access Method Functions" /></head><body id="docContent" class="container-fluid col-10"><div class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="5" align="center">63.1. Basic API Structure for Indexes</th></tr><tr><td width="10%" align="left"><a accesskey="p" href="indexam.html" title="Chapter 63. Index Access Method Interface Definition">Prev</a> </td><td width="10%" align="left"><a accesskey="u" href="indexam.html" title="Chapter 63. Index Access Method Interface Definition">Up</a></td><th width="60%" align="center">Chapter 63. Index Access Method Interface Definition</th><td width="10%" align="right"><a accesskey="h" href="index.html" title="PostgreSQL 18.0 Documentation">Home</a></td><td width="10%" align="right"> <a accesskey="n" href="index-functions.html" title="63.2. Index Access Method Functions">Next</a></td></tr></table><hr /></div><div class="sect1" id="INDEX-API"><div class="titlepage"><div><div><h2 class="title" style="clear: both">63.1. Basic API Structure for Indexes <a href="#INDEX-API" class="id_link">#</a></h2></div></div></div><p>
3 Each index access method is described by a row in the
4 <a class="link" href="catalog-pg-am.html" title="52.3. pg_am"><code class="structname">pg_am</code></a>
5 system catalog. The <code class="structname">pg_am</code> entry
6 specifies a name and a <em class="firstterm">handler function</em> for the index
7 access method. These entries can be created and deleted using the
8 <a class="xref" href="sql-create-access-method.html" title="CREATE ACCESS METHOD"><span class="refentrytitle">CREATE ACCESS METHOD</span></a> and
9 <a class="xref" href="sql-drop-access-method.html" title="DROP ACCESS METHOD"><span class="refentrytitle">DROP ACCESS METHOD</span></a> SQL commands.
11 An index access method handler function must be declared to accept a
12 single argument of type <code class="type">internal</code> and to return the
13 pseudo-type <code class="type">index_am_handler</code>. The argument is a dummy value that
14 simply serves to prevent handler functions from being called directly from
15 SQL commands. The result of the function must be a palloc'd struct of
16 type <code class="structname">IndexAmRoutine</code>, which contains everything
17 that the core code needs to know to make use of the index access method.
18 The <code class="structname">IndexAmRoutine</code> struct, also called the access
19 method's <em class="firstterm">API struct</em>, includes fields specifying assorted
20 fixed properties of the access method, such as whether it can support
21 multicolumn indexes. More importantly, it contains pointers to support
22 functions for the access method, which do all of the real work to access
23 indexes. These support functions are plain C functions and are not
24 visible or callable at the SQL level. The support functions are described
25 in <a class="xref" href="index-functions.html" title="63.2. Index Access Method Functions">Section 63.2</a>.
27 The structure <code class="structname">IndexAmRoutine</code> is defined thus:
28 </p><pre class="programlisting">
29 typedef struct IndexAmRoutine
34 * Total number of strategies (operators) by which we can traverse/search
35 * this AM. Zero if AM does not have a fixed set of strategy assignments.
38 /* total number of support functions that this AM uses */
40 /* opclass options support function number or 0 */
42 /* does AM support ORDER BY indexed column's value? */
44 /* does AM support ORDER BY result of an operator on indexed column? */
46 /* does AM support hashing using API consistent with the hash AM? */
48 /* do operators within an opfamily have consistent equality semantics? */
49 bool amconsistentequality;
50 /* do operators within an opfamily have consistent ordering semantics? */
51 bool amconsistentordering;
52 /* does AM support backward scanning? */
54 /* does AM support UNIQUE indexes? */
56 /* does AM support multi-column indexes? */
58 /* does AM require scans to have a constraint on the first index column? */
60 /* does AM handle ScalarArrayOpExpr quals? */
62 /* does AM handle IS NULL/IS NOT NULL quals? */
64 /* can index storage data type differ from column data type? */
66 /* can an index of this type be clustered on? */
68 /* does AM handle predicate locks? */
70 /* does AM support parallel scan? */
72 /* does AM support parallel build? */
73 bool amcanbuildparallel;
74 /* does AM support columns included with clause INCLUDE? */
76 /* does AM use maintenance_work_mem? */
77 bool amusemaintenanceworkmem;
78 /* does AM summarize tuples, with at least all tuples in the block
79 * summarized in one summary */
81 /* OR of parallel vacuum flags */
82 uint8 amparallelvacuumoptions;
83 /* type of data stored in index, or InvalidOid if variable */
86 /* interface functions */
87 ambuild_function ambuild;
88 ambuildempty_function ambuildempty;
89 aminsert_function aminsert;
90 aminsertcleanup_function aminsertcleanup; /* can be NULL */
91 ambulkdelete_function ambulkdelete;
92 amvacuumcleanup_function amvacuumcleanup;
93 amcanreturn_function amcanreturn; /* can be NULL */
94 amcostestimate_function amcostestimate;
95 amgettreeheight_function amgettreeheight; /* can be NULL */
96 amoptions_function amoptions;
97 amproperty_function amproperty; /* can be NULL */
98 ambuildphasename_function ambuildphasename; /* can be NULL */
99 amvalidate_function amvalidate;
100 amadjustmembers_function amadjustmembers; /* can be NULL */
101 ambeginscan_function ambeginscan;
102 amrescan_function amrescan;
103 amgettuple_function amgettuple; /* can be NULL */
104 amgetbitmap_function amgetbitmap; /* can be NULL */
105 amendscan_function amendscan;
106 ammarkpos_function ammarkpos; /* can be NULL */
107 amrestrpos_function amrestrpos; /* can be NULL */
109 /* interface functions to support parallel index scans */
110 amestimateparallelscan_function amestimateparallelscan; /* can be NULL */
111 aminitparallelscan_function aminitparallelscan; /* can be NULL */
112 amparallelrescan_function amparallelrescan; /* can be NULL */
114 /* interface functions to support planning */
115 amtranslate_strategy_function amtranslatestrategy; /* can be NULL */
116 amtranslate_cmptype_function amtranslatecmptype; /* can be NULL */
120 To be useful, an index access method must also have one or more
121 <em class="firstterm">operator families</em> and
122 <em class="firstterm">operator classes</em> defined in
123 <a class="link" href="catalog-pg-opfamily.html" title="52.35. pg_opfamily"><code class="structname">pg_opfamily</code></a>,
124 <a class="link" href="catalog-pg-opclass.html" title="52.33. pg_opclass"><code class="structname">pg_opclass</code></a>,
125 <a class="link" href="catalog-pg-amop.html" title="52.4. pg_amop"><code class="structname">pg_amop</code></a>, and
126 <a class="link" href="catalog-pg-amproc.html" title="52.5. pg_amproc"><code class="structname">pg_amproc</code></a>.
127 These entries allow the planner
128 to determine what kinds of query qualifications can be used with
129 indexes of this access method. Operator families and classes are described
130 in <a class="xref" href="xindex.html" title="36.16. Interfacing Extensions to Indexes">Section 36.16</a>, which is prerequisite material for reading
133 An individual index is defined by a
134 <a class="link" href="catalog-pg-class.html" title="52.11. pg_class"><code class="structname">pg_class</code></a>
135 entry that describes it as a physical relation, plus a
136 <a class="link" href="catalog-pg-index.html" title="52.26. pg_index"><code class="structname">pg_index</code></a>
137 entry that shows the logical content of the index — that is, the set
138 of index columns it has and the semantics of those columns, as captured by
139 the associated operator classes. The index columns (key values) can be
140 either simple columns of the underlying table or expressions over the table
141 rows. The index access method normally has no interest in where the index
142 key values come from (it is always handed precomputed key values) but it
143 will be very interested in the operator class information in
144 <code class="structname">pg_index</code>. Both of these catalog entries can be
145 accessed as part of the <code class="structname">Relation</code> data structure that is
146 passed to all operations on the index.
148 Some of the flag fields of <code class="structname">IndexAmRoutine</code> have nonobvious
149 implications. The requirements of <code class="structfield">amcanunique</code>
150 are discussed in <a class="xref" href="index-unique-checks.html" title="63.5. Index Uniqueness Checks">Section 63.5</a>.
151 The <code class="structfield">amcanmulticol</code> flag asserts that the
152 access method supports multi-key-column indexes, while
153 <code class="structfield">amoptionalkey</code> asserts that it allows scans
154 where no indexable restriction clause is given for the first index column.
155 When <code class="structfield">amcanmulticol</code> is false,
156 <code class="structfield">amoptionalkey</code> essentially says whether the
157 access method supports full-index scans without any restriction clause.
158 Access methods that support multiple index columns <span class="emphasis"><em>must</em></span>
159 support scans that omit restrictions on any or all of the columns after
160 the first; however they are permitted to require some restriction to
161 appear for the first index column, and this is signaled by setting
162 <code class="structfield">amoptionalkey</code> false.
163 One reason that an index <acronym class="acronym">AM</acronym> might set
164 <code class="structfield">amoptionalkey</code> false is if it doesn't index
165 null values. Since most indexable operators are
166 strict and hence cannot return true for null inputs,
167 it is at first sight attractive to not store index entries for null values:
168 they could never be returned by an index scan anyway. However, this
169 argument fails when an index scan has no restriction clause for a given
170 index column. In practice this means that
171 indexes that have <code class="structfield">amoptionalkey</code> true must
172 index nulls, since the planner might decide to use such an index
173 with no scan keys at all. A related restriction is that an index
174 access method that supports multiple index columns <span class="emphasis"><em>must</em></span>
175 support indexing null values in columns after the first, because the planner
176 will assume the index can be used for queries that do not restrict
177 these columns. For example, consider an index on (a,b) and a query with
178 <code class="literal">WHERE a = 4</code>. The system will assume the index can be
179 used to scan for rows with <code class="literal">a = 4</code>, which is wrong if the
180 index omits rows where <code class="literal">b</code> is null.
181 It is, however, OK to omit rows where the first indexed column is null.
182 An index access method that does index nulls may also set
183 <code class="structfield">amsearchnulls</code>, indicating that it supports
184 <code class="literal">IS NULL</code> and <code class="literal">IS NOT NULL</code> clauses as search
187 The <code class="structfield">amcaninclude</code> flag indicates whether the
188 access method supports <span class="quote">“<span class="quote">included</span>”</span> columns, that is it can
189 store (without processing) additional columns beyond the key column(s).
190 The requirements of the preceding paragraph apply only to the key
191 columns. In particular, the combination
192 of <code class="structfield">amcanmulticol</code>=<code class="literal">false</code>
193 and <code class="structfield">amcaninclude</code>=<code class="literal">true</code> is
194 sensible: it means that there can only be one key column, but there can
195 also be included column(s). Also, included columns must be allowed to be
196 null, independently of <code class="structfield">amoptionalkey</code>.
198 The <code class="structfield">amsummarizing</code> flag indicates whether the
199 access method summarizes the indexed tuples, with summarizing granularity
200 of at least per block.
201 Access methods that do not point to individual tuples, but to block ranges
202 (like <acronym class="acronym">BRIN</acronym>), may allow the <acronym class="acronym">HOT</acronym> optimization
203 to continue. This does not apply to attributes referenced in index
204 predicates, an update of such an attribute always disables <acronym class="acronym">HOT</acronym>.
205 </p></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="indexam.html" title="Chapter 63. Index Access Method Interface Definition">Prev</a> </td><td width="20%" align="center"><a accesskey="u" href="indexam.html" title="Chapter 63. Index Access Method Interface Definition">Up</a></td><td width="40%" align="right"> <a accesskey="n" href="index-functions.html" title="63.2. Index Access Method Functions">Next</a></td></tr><tr><td width="40%" align="left" valign="top">Chapter 63. Index Access Method Interface Definition </td><td width="20%" align="center"><a accesskey="h" href="index.html" title="PostgreSQL 18.0 Documentation">Home</a></td><td width="40%" align="right" valign="top"> 63.2. Index Access Method Functions</td></tr></table></div></body></html>