<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<!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>66.3. Extensibility</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="gin-builtin-opclasses.html" title="66.2. Built-in Operator Classes" /><link rel="next" href="gin-implementation.html" title="66.4. Implementation" /></head><body><div xmlns="http://www.w3.org/TR/xhtml1/transitional" class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="5" align="center">66.3. Extensibility</th></tr><tr><td width="10%" align="left"><a accesskey="p" href="gin-builtin-opclasses.html" title="66.2. Built-in Operator Classes">Prev</a> </td><td width="10%" align="left"><a accesskey="u" href="gin.html" title="Chapter 66. GIN Indexes">Up</a></td><th width="60%" align="center">Chapter 66. GIN Indexes</th><td width="10%" align="right"><a accesskey="h" href="index.html" title="PostgreSQL 11.5 Documentation">Home</a></td><td width="10%" align="right"> <a accesskey="n" href="gin-implementation.html" title="66.4. Implementation">Next</a></td></tr></table><hr></hr></div><div class="sect1" id="GIN-EXTENSIBILITY"><div class="titlepage"><div><div><h2 class="title" style="clear: both">66.3. Extensibility</h2></div></div></div><p>
The <acronym class="acronym">GIN</acronym> interface has a high level of abstraction,
requiring the access method implementer only to implement the semantics of
the data type being accessed. The <acronym class="acronym">GIN</acronym> layer itself
takes care of concurrency, logging and searching the tree structure.
</p><p>
All it takes to get a <acronym class="acronym">GIN</acronym> access method working is to
implement a few user-defined methods, which define the behavior of
keys in the tree and the relationships between keys, indexed items,
and indexable queries. In short, <acronym class="acronym">GIN</acronym> combines
extensibility with generality, code reuse, and a clean interface.
</p><p>
There are two methods that an operator class for
<acronym class="acronym">GIN</acronym> must provide:
</p><div class="variablelist"><dl class="variablelist"><dt><span class="term"><code class="function">Datum *extractValue(Datum itemValue, int32 *nkeys,
bool **nullFlags)</code></span></dt><dd><p>
Returns a palloc'd array of keys given an item to be indexed. The
number of returned keys must be stored into <code class="literal">*nkeys</code>.
If any of the keys can be null, also palloc an array of
<code class="literal">*nkeys</code> <code class="type">bool</code> fields, store its address at
<code class="literal">*nullFlags</code>, and set these null flags as needed.
<code class="literal">*nullFlags</code> can be left <code class="symbol">NULL</code> (its initial value)
if all keys are non-null.
The return value can be <code class="symbol">NULL</code> if the item contains no keys.
</p></dd><dt><span class="term"><code class="function">Datum *extractQuery(Datum query, int32 *nkeys,
StrategyNumber n, bool **pmatch, Pointer **extra_data,
bool **nullFlags, int32 *searchMode)</code></span></dt><dd><p>
Returns a palloc'd array of keys given a value to be queried; that is,
<code class="literal">query</code> is the value on the right-hand side of an
indexable operator whose left-hand side is the indexed column.
<code class="literal">n</code> is the strategy number of the operator within the
operator class (see <a class="xref" href="xindex.html#XINDEX-STRATEGIES" title="38.15.2. Index Method Strategies">Section 38.15.2</a>).
Often, <code class="function">extractQuery</code> will need
to consult <code class="literal">n</code> to determine the data type of
<code class="literal">query</code> and the method it should use to extract key values.
The number of returned keys must be stored into <code class="literal">*nkeys</code>.
If any of the keys can be null, also palloc an array of
<code class="literal">*nkeys</code> <code class="type">bool</code> fields, store its address at
<code class="literal">*nullFlags</code>, and set these null flags as needed.
<code class="literal">*nullFlags</code> can be left <code class="symbol">NULL</code> (its initial value)
if all keys are non-null.
The return value can be <code class="symbol">NULL</code> if the <code class="literal">query</code> contains no keys.
</p><p>
<code class="literal">searchMode</code> is an output argument that allows
<code class="function">extractQuery</code> to specify details about how the search
will be done.
If <code class="literal">*searchMode</code> is set to
<code class="literal">GIN_SEARCH_MODE_DEFAULT</code> (which is the value it is
initialized to before call), only items that match at least one of
the returned keys are considered candidate matches.
If <code class="literal">*searchMode</code> is set to
<code class="literal">GIN_SEARCH_MODE_INCLUDE_EMPTY</code>, then in addition to items
containing at least one matching key, items that contain no keys at
all are considered candidate matches. (This mode is useful for
implementing is-subset-of operators, for example.)
If <code class="literal">*searchMode</code> is set to <code class="literal">GIN_SEARCH_MODE_ALL</code>,
then all non-null items in the index are considered candidate
matches, whether they match any of the returned keys or not. (This
mode is much slower than the other two choices, since it requires
scanning essentially the entire index, but it may be necessary to
implement corner cases correctly. An operator that needs this mode
in most cases is probably not a good candidate for a GIN operator
class.)
The symbols to use for setting this mode are defined in
<code class="filename">access/gin.h</code>.
</p><p>
<code class="literal">pmatch</code> is an output argument for use when partial match
is supported. To use it, <code class="function">extractQuery</code> must allocate
an array of <code class="literal">*nkeys</code> bools and store its address at
<code class="literal">*pmatch</code>. Each element of the array should be set to true
if the corresponding key requires partial match, false if not.
If <code class="literal">*pmatch</code> is set to <code class="symbol">NULL</code> then GIN assumes partial match
is not required. The variable is initialized to <code class="symbol">NULL</code> before call,
so this argument can simply be ignored by operator classes that do
not support partial match.
</p><p>
<code class="literal">extra_data</code> is an output argument that allows
<code class="function">extractQuery</code> to pass additional data to the
<code class="function">consistent</code> and <code class="function">comparePartial</code> methods.
To use it, <code class="function">extractQuery</code> must allocate
an array of <code class="literal">*nkeys</code> pointers and store its address at
<code class="literal">*extra_data</code>, then store whatever it wants to into the
individual pointers. The variable is initialized to <code class="symbol">NULL</code> before
call, so this argument can simply be ignored by operator classes that
do not require extra data. If <code class="literal">*extra_data</code> is set, the
whole array is passed to the <code class="function">consistent</code> method, and
the appropriate element to the <code class="function">comparePartial</code> method.
</p></dd></dl></div><p>
An operator class must also provide a function to check if an indexed item
matches the query. It comes in two flavors, a boolean <code class="function">consistent</code>
function, and a ternary <code class="function">triConsistent</code> function.
<code class="function">triConsistent</code> covers the functionality of both, so providing
<code class="function">triConsistent</code> alone is sufficient. However, if the boolean
variant is significantly cheaper to calculate, it can be advantageous to
provide both. If only the boolean variant is provided, some optimizations
that depend on refuting index items before fetching all the keys are
disabled.
</p><div class="variablelist"><dl class="variablelist"><dt><span class="term"><code class="function">bool consistent(bool check[], StrategyNumber n, Datum query,
int32 nkeys, Pointer extra_data[], bool *recheck,
Datum queryKeys[], bool nullFlags[])</code></span></dt><dd><p>
Returns true if an indexed item satisfies the query operator with
strategy number <code class="literal">n</code> (or might satisfy it, if the recheck
indication is returned). This function does not have direct access
to the indexed item's value, since <acronym class="acronym">GIN</acronym> does not
store items explicitly. Rather, what is available is knowledge
about which key values extracted from the query appear in a given
indexed item. The <code class="literal">check</code> array has length
<code class="literal">nkeys</code>, which is the same as the number of keys previously
returned by <code class="function">extractQuery</code> for this <code class="literal">query</code> datum.
Each element of the
<code class="literal">check</code> array is true if the indexed item contains the
corresponding query key, i.e., if (check[i] == true) the i-th key of the
<code class="function">extractQuery</code> result array is present in the indexed item.
The original <code class="literal">query</code> datum is
passed in case the <code class="function">consistent</code> method needs to consult it,
and so are the <code class="literal">queryKeys[]</code> and <code class="literal">nullFlags[]</code>
arrays previously returned by <code class="function">extractQuery</code>.
<code class="literal">extra_data</code> is the extra-data array returned by
<code class="function">extractQuery</code>, or <code class="symbol">NULL</code> if none.
</p><p>
When <code class="function">extractQuery</code> returns a null key in
<code class="literal">queryKeys[]</code>, the corresponding <code class="literal">check[]</code> element
is true if the indexed item contains a null key; that is, the
semantics of <code class="literal">check[]</code> are like <code class="literal">IS NOT DISTINCT
FROM</code>. The <code class="function">consistent</code> function can examine the
corresponding <code class="literal">nullFlags[]</code> element if it needs to tell
the difference between a regular value match and a null match.
</p><p>
On success, <code class="literal">*recheck</code> should be set to true if the heap
tuple needs to be rechecked against the query operator, or false if
the index test is exact. That is, a false return value guarantees
that the heap tuple does not match the query; a true return value with
<code class="literal">*recheck</code> set to false guarantees that the heap tuple does
match the query; and a true return value with
<code class="literal">*recheck</code> set to true means that the heap tuple might match
the query, so it needs to be fetched and rechecked by evaluating the
query operator directly against the originally indexed item.
</p></dd><dt><span class="term"><code class="function">GinTernaryValue triConsistent(GinTernaryValue check[], StrategyNumber n, Datum query,
int32 nkeys, Pointer extra_data[],
Datum queryKeys[], bool nullFlags[])</code></span></dt><dd><p>
<code class="function">triConsistent</code> is similar to <code class="function">consistent</code>,
but instead of booleans in the <code class="literal">check</code> vector, there are
three possible values for each
key: <code class="literal">GIN_TRUE</code>, <code class="literal">GIN_FALSE</code> and
<code class="literal">GIN_MAYBE</code>. <code class="literal">GIN_FALSE</code> and <code class="literal">GIN_TRUE</code>
have the same meaning as regular boolean values, while
<code class="literal">GIN_MAYBE</code> means that the presence of that key is not known.
When <code class="literal">GIN_MAYBE</code> values are present, the function should only
return <code class="literal">GIN_TRUE</code> if the item certainly matches whether or
not the index item contains the corresponding query keys. Likewise, the
function must return <code class="literal">GIN_FALSE</code> only if the item certainly
does not match, whether or not it contains the <code class="literal">GIN_MAYBE</code>
keys. If the result depends on the <code class="literal">GIN_MAYBE</code> entries, i.e.,
the match cannot be confirmed or refuted based on the known query keys,
the function must return <code class="literal">GIN_MAYBE</code>.
</p><p>
When there are no <code class="literal">GIN_MAYBE</code> values in the <code class="literal">check</code>
vector, a <code class="literal">GIN_MAYBE</code> return value is the equivalent of
setting the <code class="literal">recheck</code> flag in the
boolean <code class="function">consistent</code> function.
</p></dd></dl></div><p>
</p><p>
In addition, GIN must have a way to sort the key values stored in the index.
The operator class can define the sort ordering by specifying a comparison
method:
</p><div class="variablelist"><dl class="variablelist"><dt><span class="term"><code class="function">int compare(Datum a, Datum b)</code></span></dt><dd><p>
Compares two keys (not indexed items!) and returns an integer less than
zero, zero, or greater than zero, indicating whether the first key is
less than, equal to, or greater than the second. Null keys are never
passed to this function.
</p></dd></dl></div><p>
Alternatively, if the operator class does not provide a <code class="function">compare</code>
method, GIN will look up the default btree operator class for the index
key data type, and use its comparison function. It is recommended to
specify the comparison function in a GIN operator class that is meant for
just one data type, as looking up the btree operator class costs a few
cycles. However, polymorphic GIN operator classes (such
as <code class="literal">array_ops</code>) typically cannot specify a single comparison
function.
</p><p>
Optionally, an operator class for <acronym class="acronym">GIN</acronym> can supply the
following method:
</p><div class="variablelist"><dl class="variablelist"><dt><span class="term"><code class="function">int comparePartial(Datum partial_key, Datum key, StrategyNumber n,
Pointer extra_data)</code></span></dt><dd><p>
Compare a partial-match query key to an index key. Returns an integer
whose sign indicates the result: less than zero means the index key
does not match the query, but the index scan should continue; zero
means that the index key does match the query; greater than zero
indicates that the index scan should stop because no more matches
are possible. The strategy number <code class="literal">n</code> of the operator
that generated the partial match query is provided, in case its
semantics are needed to determine when to end the scan. Also,
<code class="literal">extra_data</code> is the corresponding element of the extra-data
array made by <code class="function">extractQuery</code>, or <code class="symbol">NULL</code> if none.
Null keys are never passed to this function.
</p></dd></dl></div><p>
</p><p>
To support <span class="quote">“<span class="quote">partial match</span>”</span> queries, an operator class must
provide the <code class="function">comparePartial</code> method, and its
<code class="function">extractQuery</code> method must set the <code class="literal">pmatch</code>
parameter when a partial-match query is encountered. See
<a class="xref" href="gin-implementation.html#GIN-PARTIAL-MATCH" title="66.4.2. Partial Match Algorithm">Section 66.4.2</a> for details.
</p><p>
The actual data types of the various <code class="literal">Datum</code> values mentioned
above vary depending on the operator class. The item values passed to
<code class="function">extractValue</code> are always of the operator class's input type, and
all key values must be of the class's <code class="literal">STORAGE</code> type. The type of
the <code class="literal">query</code> argument passed to <code class="function">extractQuery</code>,
<code class="function">consistent</code> and <code class="function">triConsistent</code> is whatever is the
right-hand input type of the class member operator identified by the
strategy number. This need not be the same as the indexed type, so long as
key values of the correct type can be extracted from it. However, it is
recommended that the SQL declarations of these three support functions use
the opclass's indexed data type for the <code class="literal">query</code> argument, even
though the actual type might be something else depending on the operator.
</p></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="gin-builtin-opclasses.html">Prev</a> </td><td width="20%" align="center"><a accesskey="u" href="gin.html">Up</a></td><td width="40%" align="right"> <a accesskey="n" href="gin-implementation.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">66.2. Built-in Operator Classes </td><td width="20%" align="center"><a accesskey="h" href="index.html">Home</a></td><td width="40%" align="right" valign="top"> 66.4. Implementation</td></tr></table></div></body></html>
