<!-- doc/src/sgml/pgtrgm.sgml -->
<sect1 id="pgtrgm" xreflabel="pg_trgm">
<title>pg_trgm &mdash;
support for similarity of text using trigram matching</title>
<indexterm zone="pgtrgm">
<primary>pg_trgm</primary>
</indexterm>
<para>
The <filename>pg_trgm</filename> module provides functions and operators
for determining the similarity of
alphanumeric text based on trigram matching, as
well as index operator classes that support fast searching for similar
strings.
</para>
<para>
This module is considered <quote>trusted</quote>, that is, it can be
installed by non-superusers who have <literal>CREATE</literal> privilege
on the current database.
</para>
<sect2 id="pgtrgm-concepts">
<title>Trigram (or Trigraph) Concepts</title>
<para>
A trigram is a group of three consecutive characters taken
from a string. We can measure the similarity of two strings by
counting the number of trigrams they share. This simple idea
turns out to be very effective for measuring the similarity of
words in many natural languages.
</para>
<note>
<para>
<filename>pg_trgm</filename> ignores non-word characters
(non-alphanumerics) when extracting trigrams from a string.
Each word is considered to have two spaces
prefixed and one space suffixed when determining the set
of trigrams contained in the string.
For example, the set of trigrams in the string
<quote><literal>cat</literal></quote> is
<quote><literal> c</literal></quote>,
<quote><literal> ca</literal></quote>,
<quote><literal>cat</literal></quote>, and
<quote><literal>at </literal></quote>.
The set of trigrams in the string
<quote><literal>foo|bar</literal></quote> is
<quote><literal> f</literal></quote>,
<quote><literal> fo</literal></quote>,
<quote><literal>foo</literal></quote>,
<quote><literal>oo </literal></quote>,
<quote><literal> b</literal></quote>,
<quote><literal> ba</literal></quote>,
<quote><literal>bar</literal></quote>, and
<quote><literal>ar </literal></quote>.
</para>
</note>
</sect2>
<sect2 id="pgtrgm-funcs-ops">
<title>Functions and Operators</title>
<para>
The functions provided by the <filename>pg_trgm</filename> module
are shown in <xref linkend="pgtrgm-func-table"/>, the operators
in <xref linkend="pgtrgm-op-table"/>.
</para>
<table id="pgtrgm-func-table">
<title><filename>pg_trgm</filename> Functions</title>
<tgroup cols="1">
<thead>
<row>
<entry role="func_table_entry"><para role="func_signature">
Function
</para>
<para>
Description
</para></entry>
</row>
</thead>
<tbody>
<row>
<entry role="func_table_entry"><para role="func_signature">
<indexterm><primary>similarity</primary></indexterm>
<function>similarity</function> ( <type>text</type>, <type>text</type> )
<returnvalue>real</returnvalue>
</para>
<para>
Returns a number that indicates how similar the two arguments are.
The range of the result is zero (indicating that the two strings are
completely dissimilar) to one (indicating that the two strings are
identical).
</para></entry>
</row>
<row>
<entry role="func_table_entry"><para role="func_signature">
<indexterm><primary>show_trgm</primary></indexterm>
<function>show_trgm</function> ( <type>text</type> )
<returnvalue>text[]</returnvalue>
</para>
<para>
Returns an array of all the trigrams in the given string.
(In practice this is seldom useful except for debugging.)
</para></entry>
</row>
<row>
<entry role="func_table_entry"><para role="func_signature">
<indexterm><primary>word_similarity</primary></indexterm>
<function>word_similarity</function> ( <type>text</type>, <type>text</type> )
<returnvalue>real</returnvalue>
</para>
<para>
Returns a number that indicates the greatest similarity between
the set of trigrams in the first string and any continuous extent
of an ordered set of trigrams in the second string. For details, see
the explanation below.
</para></entry>
</row>
<row>
<entry role="func_table_entry"><para role="func_signature">
<indexterm><primary>strict_word_similarity</primary></indexterm>
<function>strict_word_similarity</function> ( <type>text</type>, <type>text</type> )
<returnvalue>real</returnvalue>
</para>
<para>
Same as <function>word_similarity</function>, but forces
extent boundaries to match word boundaries. Since we don't have
cross-word trigrams, this function actually returns greatest similarity
between first string and any continuous extent of words of the second
string.
</para></entry>
</row>
<row>
<entry role="func_table_entry"><para role="func_signature">
<indexterm><primary>show_limit</primary></indexterm>
<function>show_limit</function> ()
<returnvalue>real</returnvalue>
</para>
<para>
Returns the current similarity threshold used by the <literal>%</literal>
operator. This sets the minimum similarity between
two words for them to be considered similar enough to
be misspellings of each other, for example.
(<emphasis>Deprecated</emphasis>; instead use <command>SHOW</command>
<varname>pg_trgm.similarity_threshold</varname>.)
</para></entry>
</row>
<row>
<entry role="func_table_entry"><para role="func_signature">
<indexterm><primary>set_limit</primary></indexterm>
<function>set_limit</function> ( <type>real</type> )
<returnvalue>real</returnvalue>
</para>
<para>
Sets the current similarity threshold that is used by the <literal>%</literal>
operator. The threshold must be between 0 and 1 (default is 0.3).
Returns the same value passed in.
(<emphasis>Deprecated</emphasis>; instead use <command>SET</command>
<varname>pg_trgm.similarity_threshold</varname>.)
</para></entry>
</row>
</tbody>
</tgroup>
</table>
<para>
Consider the following example:
<programlisting>
# SELECT word_similarity('word', 'two words');
word_similarity
-----------------
0.8
(1 row)
</programlisting>
In the first string, the set of trigrams is
<literal>{" w"," wo","wor","ord","rd "}</literal>.
In the second string, the ordered set of trigrams is
<literal>{" t"," tw","two","wo "," w"," wo","wor","ord","rds","ds "}</literal>.
The most similar extent of an ordered set of trigrams in the second string
is <literal>{" w"," wo","wor","ord"}</literal>, and the similarity is
<literal>0.8</literal>.
</para>
<para>
This function returns a value that can be approximately understood as the
greatest similarity between the first string and any substring of the second
string. However, this function does not add padding to the boundaries of
the extent. Thus, the number of additional characters present in the
second string is not considered, except for the mismatched word boundaries.
</para>
<para>
At the same time, <function>strict_word_similarity</function>
selects an extent of words in the second string. In the example above,
<function>strict_word_similarity</function> would select the
extent of a single word <literal>'words'</literal>, whose set of trigrams is
<literal>{" w"," wo","wor","ord","rds","ds "}</literal>.
<programlisting>
# SELECT strict_word_similarity('word', 'two words'), similarity('word', 'words');
strict_word_similarity | similarity
------------------------+------------
0.571429 | 0.571429
(1 row)
</programlisting>
</para>
<para>
Thus, the <function>strict_word_similarity</function> function
is useful for finding the similarity to whole words, while
<function>word_similarity</function> is more suitable for
finding the similarity for parts of words.
</para>
<table id="pgtrgm-op-table">
<title><filename>pg_trgm</filename> Operators</title>
<tgroup cols="1">
<thead>
<row>
<entry role="func_table_entry"><para role="func_signature">
Operator
</para>
<para>
Description
</para></entry>
</row>
</thead>
<tbody>
<row>
<entry role="func_table_entry"><para role="func_signature">
<type>text</type> <literal>%</literal> <type>text</type>
<returnvalue>boolean</returnvalue>
</para>
<para>
Returns <literal>true</literal> if its arguments have a similarity
that is greater than the current similarity threshold set by
<varname>pg_trgm.similarity_threshold</varname>.
</para></entry>
</row>
<row>
<entry role="func_table_entry"><para role="func_signature">
<type>text</type> <literal>&lt;%</literal> <type>text</type>
<returnvalue>boolean</returnvalue>
</para>
<para>
Returns <literal>true</literal> if the similarity between the trigram
set in the first argument and a continuous extent of an ordered trigram
set in the second argument is greater than the current word similarity
threshold set by <varname>pg_trgm.word_similarity_threshold</varname>
parameter.
</para></entry>
</row>
<row>
<entry role="func_table_entry"><para role="func_signature">
<type>text</type> <literal>%&gt;</literal> <type>text</type>
<returnvalue>boolean</returnvalue>
</para>
<para>
Commutator of the <literal>&lt;%</literal> operator.
</para></entry>
</row>
<row>
<entry role="func_table_entry"><para role="func_signature">
<type>text</type> <literal>&lt;&lt;%</literal> <type>text</type>
<returnvalue>boolean</returnvalue>
</para>
<para>
Returns <literal>true</literal> if its second argument has a continuous
extent of an ordered trigram set that matches word boundaries,
and its similarity to the trigram set of the first argument is greater
than the current strict word similarity threshold set by the
<varname>pg_trgm.strict_word_similarity_threshold</varname> parameter.
</para></entry>
</row>
<row>
<entry role="func_table_entry"><para role="func_signature">
<type>text</type> <literal>%&gt;&gt;</literal> <type>text</type>
<returnvalue>boolean</returnvalue>
</para>
<para>
Commutator of the <literal>&lt;&lt;%</literal> operator.
</para></entry>
</row>
<row>
<entry role="func_table_entry"><para role="func_signature">
<type>text</type> <literal>&lt;-&gt;</literal> <type>text</type>
<returnvalue>real</returnvalue>
</para>
<para>
Returns the <quote>distance</quote> between the arguments, that is
one minus the <function>similarity()</function> value.
</para></entry>
</row>
<row>
<entry role="func_table_entry"><para role="func_signature">
<type>text</type> <literal>&lt;&lt;-&gt;</literal> <type>text</type>
<returnvalue>real</returnvalue>
</para>
<para>
Returns the <quote>distance</quote> between the arguments, that is
one minus the <function>word_similarity()</function> value.
</para></entry>
</row>
<row>
<entry role="func_table_entry"><para role="func_signature">
<type>text</type> <literal>&lt;-&gt;&gt;</literal> <type>text</type>
<returnvalue>real</returnvalue>
</para>
<para>
Commutator of the <literal>&lt;&lt;-&gt;</literal> operator.
</para></entry>
</row>
<row>
<entry role="func_table_entry"><para role="func_signature">
<type>text</type> <literal>&lt;&lt;&lt;-&gt;</literal> <type>text</type>
<returnvalue>real</returnvalue>
</para>
<para>
Returns the <quote>distance</quote> between the arguments, that is
one minus the <function>strict_word_similarity()</function> value.
</para></entry>
</row>
<row>
<entry role="func_table_entry"><para role="func_signature">
<type>text</type> <literal>&lt;-&gt;&gt;&gt;</literal> <type>text</type>
<returnvalue>real</returnvalue>
</para>
<para>
Commutator of the <literal>&lt;&lt;&lt;-&gt;</literal> operator.
</para></entry>
</row>
</tbody>
</tgroup>
</table>
</sect2>
<sect2 id="pgtrgm-guc">
<title>GUC Parameters</title>
<variablelist>
<varlistentry id="guc-pgtrgm-similarity-threshold" xreflabel="pg_trgm.similarity_threshold">
<term>
<varname>pg_trgm.similarity_threshold</varname> (<type>real</type>)
<indexterm>
<primary><varname>pg_trgm.similarity_threshold</varname> configuration parameter</primary>
</indexterm>
</term>
<listitem>
<para>
Sets the current similarity threshold that is used by the <literal>%</literal>
operator. The threshold must be between 0 and 1 (default is 0.3).
</para>
</listitem>
</varlistentry>
<varlistentry id="guc-pgtrgm-word-similarity-threshold" xreflabel="pg_trgm.word_similarity_threshold">
<term>
<varname>pg_trgm.word_similarity_threshold</varname> (<type>real</type>)
<indexterm>
<primary><varname>pg_trgm.word_similarity_threshold</varname> configuration parameter</primary>
</indexterm>
</term>
<listitem>
<para>
Sets the current word similarity threshold that is used by the
<literal>&lt;%</literal> and <literal>%&gt;</literal> operators. The threshold
must be between 0 and 1 (default is 0.6).
</para>
</listitem>
</varlistentry>
<varlistentry id="guc-pgtrgm-strict-word-similarity-threshold" xreflabel="pg_trgm.strict_word_similarity_threshold">
<term>
<varname>pg_trgm.strict_word_similarity_threshold</varname> (<type>real</type>)
<indexterm>
<primary><varname>pg_trgm.strict_word_similarity_threshold</varname> configuration parameter</primary>
</indexterm>
</term>
<listitem>
<para>
Sets the current strict word similarity threshold that is used by the
<literal>&lt;&lt;%</literal> and <literal>%&gt;&gt;</literal> operators. The threshold
must be between 0 and 1 (default is 0.5).
</para>
</listitem>
</varlistentry>
</variablelist>
</sect2>
<sect2 id="pgtrgm-index">
<title>Index Support</title>
<para>
The <filename>pg_trgm</filename> module provides GiST and GIN index
operator classes that allow you to create an index over a text column for
the purpose of very fast similarity searches. These index types support
the above-described similarity operators, and additionally support
trigram-based index searches for <literal>LIKE</literal>, <literal>ILIKE</literal>,
<literal>~</literal>, <literal>~*</literal> and <literal>=</literal> queries.
The similarity comparisons are case-insensitive in a default build of
<filename>pg_trgm</filename>.
Inequality operators are not supported.
Note that those indexes may not be as efficient as regular B-tree indexes
for equality operator.
</para>
<para>
Example:
<programlisting>
CREATE TABLE test_trgm (t text);
CREATE INDEX trgm_idx ON test_trgm USING GIST (t gist_trgm_ops);
</programlisting>
or
<programlisting>
CREATE INDEX trgm_idx ON test_trgm USING GIN (t gin_trgm_ops);
</programlisting>
</para>
<para>
<literal>gist_trgm_ops</literal> GiST opclass approximates a set of
trigrams as a bitmap signature. Its optional integer parameter
<literal>siglen</literal> determines the
signature length in bytes. The default length is 12 bytes.
Valid values of signature length are between 1 and 2024 bytes. Longer
signatures lead to a more precise search (scanning a smaller fraction of the index and
fewer heap pages), at the cost of a larger index.
</para>
<para>
Example of creating such an index with a signature length of 32 bytes:
</para>
<programlisting>
CREATE INDEX trgm_idx ON test_trgm USING GIST (t gist_trgm_ops(siglen=32));
</programlisting>
<para>
At this point, you will have an index on the <structfield>t</structfield> column that
you can use for similarity searching. A typical query is
<programlisting>
SELECT t, similarity(t, '<replaceable>word</replaceable>') AS sml
FROM test_trgm
WHERE t % '<replaceable>word</replaceable>'
ORDER BY sml DESC, t;
</programlisting>
This will return all values in the text column that are sufficiently
similar to <replaceable>word</replaceable>, sorted from best match to worst. The
index will be used to make this a fast operation even over very large data
sets.
</para>
<para>
A variant of the above query is
<programlisting>
SELECT t, t &lt;-&gt; '<replaceable>word</replaceable>' AS dist
FROM test_trgm
ORDER BY dist LIMIT 10;
</programlisting>
This can be implemented quite efficiently by GiST indexes, but not
by GIN indexes. It will usually beat the first formulation when only
a small number of the closest matches is wanted.
</para>
<para>
Also you can use an index on the <structfield>t</structfield> column for word
similarity or strict word similarity. Typical queries are:
<programlisting>
SELECT t, word_similarity('<replaceable>word</replaceable>', t) AS sml
FROM test_trgm
WHERE '<replaceable>word</replaceable>' &lt;% t
ORDER BY sml DESC, t;
</programlisting>
and
<programlisting>
SELECT t, strict_word_similarity('<replaceable>word</replaceable>', t) AS sml
FROM test_trgm
WHERE '<replaceable>word</replaceable>' &lt;&lt;% t
ORDER BY sml DESC, t;
</programlisting>
This will return all values in the text column for which there is a
continuous extent in the corresponding ordered trigram set that is
sufficiently similar to the trigram set of <replaceable>word</replaceable>,
sorted from best match to worst. The index will be used to make this
a fast operation even over very large data sets.
</para>
<para>
Possible variants of the above queries are:
<programlisting>
SELECT t, '<replaceable>word</replaceable>' &lt;&lt;-&gt; t AS dist
FROM test_trgm
ORDER BY dist LIMIT 10;
</programlisting>
and
<programlisting>
SELECT t, '<replaceable>word</replaceable>' &lt;&lt;&lt;-&gt; t AS dist
FROM test_trgm
ORDER BY dist LIMIT 10;
</programlisting>
This can be implemented quite efficiently by GiST indexes, but not
by GIN indexes.
</para>
<para>
Beginning in <productname>PostgreSQL</productname> 9.1, these index types also support
index searches for <literal>LIKE</literal> and <literal>ILIKE</literal>, for example
<programlisting>
SELECT * FROM test_trgm WHERE t LIKE '%foo%bar';
</programlisting>
The index search works by extracting trigrams from the search string
and then looking these up in the index. The more trigrams in the search
string, the more effective the index search is. Unlike B-tree based
searches, the search string need not be left-anchored.
</para>
<para>
Beginning in <productname>PostgreSQL</productname> 9.3, these index types also support
index searches for regular-expression matches
(<literal>~</literal> and <literal>~*</literal> operators), for example
<programlisting>
SELECT * FROM test_trgm WHERE t ~ '(foo|bar)';
</programlisting>
The index search works by extracting trigrams from the regular expression
and then looking these up in the index. The more trigrams that can be
extracted from the regular expression, the more effective the index search
is. Unlike B-tree based searches, the search string need not be
left-anchored.
</para>
<para>
For both <literal>LIKE</literal> and regular-expression searches, keep in mind
that a pattern with no extractable trigrams will degenerate to a full-index
scan.
</para>
<para>
The choice between GiST and GIN indexing depends on the relative
performance characteristics of GiST and GIN, which are discussed elsewhere.
</para>
</sect2>
<sect2 id="pgtrgm-text-search">
<title>Text Search Integration</title>
<para>
Trigram matching is a very useful tool when used in conjunction
with a full text index. In particular it can help to recognize
misspelled input words that will not be matched directly by the
full text search mechanism.
</para>
<para>
The first step is to generate an auxiliary table containing all
the unique words in the documents:
<programlisting>
CREATE TABLE words AS SELECT word FROM
ts_stat('SELECT to_tsvector(''simple'', bodytext) FROM documents');
</programlisting>
where <structname>documents</structname> is a table that has a text field
<structfield>bodytext</structfield> that we wish to search. The reason for using
the <literal>simple</literal> configuration with the <function>to_tsvector</function>
function, instead of using a language-specific configuration,
is that we want a list of the original (unstemmed) words.
</para>
<para>
Next, create a trigram index on the word column:
<programlisting>
CREATE INDEX words_idx ON words USING GIN (word gin_trgm_ops);
</programlisting>
Now, a <command>SELECT</command> query similar to the previous example can
be used to suggest spellings for misspelled words in user search terms.
A useful extra test is to require that the selected words are also of
similar length to the misspelled word.
</para>
<note>
<para>
Since the <structname>words</structname> table has been generated as a separate,
static table, it will need to be periodically regenerated so that
it remains reasonably up-to-date with the document collection.
Keeping it exactly current is usually unnecessary.
</para>
</note>
</sect2>
<sect2 id="pgtrgm-references">
<title>References</title>
<para>
GiST Development Site
<ulink url="http://www.sai.msu.su/~megera/postgres/gist/"></ulink>
</para>
<para>
Tsearch2 Development Site
<ulink url="http://www.sai.msu.su/~megera/postgres/gist/tsearch/V2/"></ulink>
</para>
</sect2>
<sect2 id="pgtrgm-authors">
<title>Authors</title>
<para>
Oleg Bartunov <email>oleg@sai.msu.su</email>, Moscow, Moscow University, Russia
</para>
<para>
Teodor Sigaev <email>teodor@sigaev.ru</email>, Moscow, Delta-Soft Ltd.,Russia
</para>
<para>
Alexander Korotkov <email>a.korotkov@postgrespro.ru</email>, Moscow, Postgres Professional, Russia
</para>
<para>
Documentation: Christopher Kings-Lynne
</para>
<para>
This module is sponsored by Delta-Soft Ltd., Moscow, Russia.
</para>
</sect2>
</sect1>