rdelaage
/
postgresql
mirror of https://git.postgresql.org/git/postgresql.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Move full text search operators, functions, and data type sections into 

the main documentation, out of its own text search chapter.
This commit is contained in:
Bruce Momjian 2007-08-29 20:37:14 +00:00
parent 8bc225e799
commit bb8f629c7a
3 changed files with 1147 additions and 1136 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

145
doc/src/sgml/datatype.sgml
View File

@ -1,4 +1,4 @@
<!-- $PostgreSQL: pgsql/doc/src/sgml/datatype.sgml,v 1.207 2007/08/21 01:11:11 tgl Exp $ -->
<!-- $PostgreSQL: pgsql/doc/src/sgml/datatype.sgml,v 1.208 2007/08/29 20:37:14 momjian Exp $ -->
<chapter id="datatype">
<title id="datatype-title">Data Types</title>
@ -234,6 +234,18 @@
<entry>date and time, including time zone</entry>
</row>
<row>
<entry><type>tsquery</type></entry>
<entry></entry>
<entry>full text search query</entry>
</row>
<row>
<entry><type>tsvector</type></entry>
<entry></entry>
<entry>full text search document</entry>
</row>
<row>
<entry><type>uuid</type></entry>
<entry></entry>
@ -3264,6 +3276,137 @@ a0eebc999c0b4ef8bb6d6bb9bd380a11
</para>
</sect1>
<sect1 id="datatype-textsearch">
<title>Full Text Search</title>
<variablelist>
<indexterm zone="datatype-textsearch">
<primary>tsvector</primary>
</indexterm>
<varlistentry>
<term><firstterm>tsvector</firstterm></term>
<listitem>
<para>
<type>tsvector</type> is a data type that represents a document and is
optimized for full text searching. In the simplest case,
<type>tsvector</type> is a sorted list of lexemes, so even without indexes
full text searches perform better than standard <literal>~</literal> and
<literal>LIKE</literal> operations:
<programlisting>
SELECT 'a fat cat sat on a mat and ate a fat rat'::tsvector;
tsvector
----------------------------------------------------
'a' 'on' 'and' 'ate' 'cat' 'fat' 'mat' 'rat' 'sat'
</programlisting>
Notice, that <literal>space</literal> is also a lexeme:
<programlisting>
SELECT 'space '' '' is a lexeme'::tsvector;
tsvector
----------------------------------
'a' 'is' ' ' 'space' 'lexeme'
</programlisting>
Each lexeme, optionally, can have positional information which is used for
<varname>proximity ranking</varname>:
<programlisting>
SELECT 'a:1 fat:2 cat:3 sat:4 on:5 a:6 mat:7 and:8 ate:9 a:10 fat:11 rat:12'::tsvector;
tsvector
-------------------------------------------------------------------------------
'a':1,6,10 'on':5 'and':8 'ate':9 'cat':3 'fat':2,11 'mat':7 'rat':12 'sat':4
</programlisting>
Each lexeme position also can be labeled as <literal>A</literal>,
<literal>B</literal>, <literal>C</literal>, <literal>D</literal>,
where <literal>D</literal> is the default. These labels can be used to group
lexemes into different <emphasis>importance</emphasis> or
<emphasis>rankings</emphasis>, for example to reflect document structure.
Actual values can be assigned at search time and used during the calculation
of the document rank. This is very useful for controlling search results.
</para>
<para>
The concatenation operator, e.g. <literal>tsvector || tsvector</literal>,
can "construct" a document from several parts. The order is important if
<type>tsvector</type> contains positional information. Of course,
it is also possible to build a document using different tables:
<programlisting>
SELECT 'fat:1 cat:2'::tsvector || 'fat:1 rat:2'::tsvector;
?column?
---------------------------
'cat':2 'fat':1,3 'rat':4
SELECT 'fat:1 rat:2'::tsvector || 'fat:1 cat:2'::tsvector;
?column?
---------------------------
'cat':4 'fat':1,3 'rat':2
</programlisting>
</para>
</listitem>
</varlistentry>
<indexterm zone="datatype-textsearch">
<primary>tsquery</primary>
</indexterm>
<varlistentry>
<term><firstterm>tsquery</firstterm></term>
<listitem>
<para>
<type>tsquery</type> is a data type for textual queries which supports
the boolean operators <literal>&amp;</literal> (AND), <literal>|</literal> (OR),
and parentheses. A <type>tsquery</type> consists of lexemes
(optionally labeled by letters) with boolean operators in between:
<programlisting>
SELECT 'fat &amp; cat'::tsquery;
tsquery
---------------
'fat' &amp; 'cat'
SELECT 'fat:ab &amp; cat'::tsquery;
tsquery
------------------
'fat':AB &amp; 'cat'
</programlisting>
Labels can be used to restrict the search region, which allows the
development of different search engines using the same full text index.
</para>
<para>
<type>tsqueries</type> can be concatenated using <literal>&amp;&amp;</literal> (AND)
and <literal>||</literal> (OR) operators:
<programlisting>
SELECT 'a &amp; b'::tsquery &amp;&amp; 'c | d'::tsquery;
?column?
---------------------------
'a' &amp; 'b' &amp; ( 'c' | 'd' )
SELECT 'a &amp; b'::tsquery || 'c|d'::tsquery;
?column?
---------------------------
'a' &amp; 'b' | ( 'c' | 'd' )
</programlisting>
</para>
</listitem>
</varlistentry>
</variablelist>
</sect1>
<sect1 id="datatype-xml">
<title><acronym>XML</> Type</title>

916
doc/src/sgml/func.sgml
View File

@ -1,4 +1,4 @@
<!-- $PostgreSQL: pgsql/doc/src/sgml/func.sgml,v 1.388 2007/08/21 01:11:11 tgl Exp $ -->
<!-- $PostgreSQL: pgsql/doc/src/sgml/func.sgml,v 1.389 2007/08/29 20:37:14 momjian Exp $ -->
<chapter id="functions">
<title>Functions and Operators</title>
@ -7551,6 +7551,920 @@ CREATE TYPE rainbow AS ENUM ('red', 'orange', 'yellow', 'green', 'blue', 'purple
</sect1>
<sect1 id="functions-textsearch">
<title>Full Text Search Operators and Functions</title>
<para>
This section outlines all the functions and operators that are available
for full text searching.
</para>
<para>
Full text search vectors and queries both use lexemes, but for different
purposes. A <type>tsvector</type> represents the lexemes (tokens) parsed
out of a document, with an optional position. A <type>tsquery</type>
specifies a boolean condition using lexemes.
</para>
<para>
All of the following functions that accept a configuration argument can
use a textual configuration name to select a configuration. If the option
is omitted the configuration specified by
<varname>default_text_search_config</> is used. For more information on
configuration, see <xref linkend="textsearch-tables-configuration">.
</para>
<sect2 id="functions-textsearch-search-operator">
<title>Search</title>
<para>The operator <literal>@@</> is used to perform full text
searches:
</para>
<variablelist>
<varlistentry>
<indexterm zone="functions-textsearch-search-operator">
<primary>TSVECTOR @@ TSQUERY</primary>
</indexterm>
<term>
<synopsis>
<!-- why allow such combinations? -->
TSVECTOR @@ TSQUERY
TSQUERY @@ TSVECTOR
</synopsis>
</term>
<listitem>
<para>
Returns <literal>true</literal> if <literal>TSQUERY</literal> is contained
in <literal>TSVECTOR</literal>, and <literal>false</literal> if not:
<programlisting>
SELECT 'a fat cat sat on a mat and ate a fat rat'::tsvector @@ 'cat &amp; rat'::tsquery;
?column?
----------
t
SELECT 'a fat cat sat on a mat and ate a fat rat'::tsvector @@ 'fat &amp; cow'::tsquery;
?column?
----------
f
</programlisting>
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-search-operator">
<primary>TEXT @@ TSQUERY</primary>
</indexterm>
<term>
<synopsis>
text @@ tsquery
</synopsis>
</term>
<listitem>
<para>
Returns <literal>true</literal> if <literal>TSQUERY</literal> is contained
in <literal>TEXT</literal>, and <literal>false</literal> if not:
<programlisting>
SELECT 'a fat cat sat on a mat and ate a fat rat'::text @@ 'cat &amp; rat'::tsquery;
?column?
----------
t
SELECT 'a fat cat sat on a mat and ate a fat rat'::text @@ 'cat &amp; cow'::tsquery;
?column?
----------
f
</programlisting>
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-search-operator">
<primary>TEXT @@ TEXT</primary>
</indexterm>
<term>
<synopsis>
<!-- this is very confusing because there is no rule suggesting which is
first. -->
text @@ text
</synopsis>
</term>
<listitem>
<para>
Returns <literal>true</literal> if the right
argument (the query) is contained in the left argument, and
<literal>false</literal> otherwise:
<programlisting>
SELECT 'a fat cat sat on a mat and ate a fat rat' @@ 'cat rat';
?column?
----------
t
SELECT 'a fat cat sat on a mat and ate a fat rat' @@ 'cat cow';
?column?
----------
f
</programlisting>
</para>
</listitem>
</varlistentry>
</variablelist>
<para>
For index support of full text operators consult <xref linkend="textsearch-indexes">.
</para>
</sect2>
<sect2 id="functions-textsearch-tsvector">
<title>tsvector</title>
<variablelist>
<varlistentry>
<indexterm zone="functions-textsearch-tsvector">
<primary>to_tsvector</primary>
</indexterm>
<term>
<synopsis>
to_tsvector(<optional><replaceable class="PARAMETER">config_name</replaceable></optional>, <replaceable class="PARAMETER">document</replaceable> TEXT) returns TSVECTOR
</synopsis>
</term>
<listitem>
<para>
Parses a document into tokens, reduces the tokens to lexemes, and returns a
<type>tsvector</type> which lists the lexemes together with their positions in the document
in lexicographic order.
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsvector">
<primary>strip</primary>
</indexterm>
<term>
<synopsis>
strip(<replaceable class="PARAMETER">vector</replaceable> TSVECTOR) returns TSVECTOR
</synopsis>
</term>
<listitem>
<para>
Returns a vector which lists the same lexemes as the given vector, but
which lacks any information about where in the document each lexeme
appeared. While the returned vector is useless for relevance ranking it
will usually be much smaller.
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsvector">
<primary>setweight</primary>
</indexterm>
<term>
<synopsis>
setweight(<replaceable class="PARAMETER">vector</replaceable> TSVECTOR, <replaceable class="PARAMETER">letter</replaceable>) returns TSVECTOR
</synopsis>
</term>
<listitem>
<para>
This function returns a copy of the input vector in which every location
has been labeled with either the letter <literal>A</literal>,
<literal>B</literal>, or <literal>C</literal>, or the default label
<literal>D</literal> (which is the default for new vectors
and as such is usually not displayed). These labels are retained
when vectors are concatenated, allowing words from different parts of a
document to be weighted differently by ranking functions.
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsvector">
<primary>tsvector concatenation</primary>
</indexterm>
<term>
<synopsis>
<replaceable class="PARAMETER">vector1</replaceable> || <replaceable class="PARAMETER">vector2</replaceable>
tsvector_concat(<replaceable class="PARAMETER">vector1</replaceable> TSVECTOR, <replaceable class="PARAMETER">vector2</replaceable> TSVECTOR) returns TSVECTOR
</synopsis>
</term>
<listitem>
<para>
Returns a vector which combines the lexemes and positional information of
the two vectors given as arguments. Positional weight labels (described
in the previous paragraph) are retained during the concatenation. This
has at least two uses. First, if some sections of your document need to be
parsed with different configurations than others, you can parse them
separately and then concatenate the resulting vectors. Second, you can
weigh words from one section of your document differently than the others
by parsing the sections into separate vectors and assigning each vector
a different position label with the <function>setweight()</function>
function. You can then concatenate them into a single vector and provide
a weights argument to the <function>ts_rank()</function> function that assigns
different weights to positions with different labels.
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsvector">
<primary>length(tsvector)</primary>
</indexterm>
<term>
<synopsis>
length(<replaceable class="PARAMETER">vector</replaceable> TSVECTOR) returns INT4
</synopsis>
</term>
<listitem>
<para>
Returns the number of lexemes stored in the vector.
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsvector">
<primary>text::tsvector</primary>
</indexterm>
<term>
<synopsis>
<replaceable>text</replaceable>::TSVECTOR returns TSVECTOR
</synopsis>
</term>
<listitem>
<para>
Directly casting <type>text</type> to a <type>tsvector</type> allows you
to directly inject lexemes into a vector with whatever positions and
positional weights you choose to specify. The text should be formatted to
match the way a vector is displayed by <literal>SELECT</literal>.
<!-- TODO what a strange definition, I think something like
"input format" or so should be used (and defined somewhere, didn't see
it yet) -->
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsvector">
<primary>trigger</primary>
<secondary>for updating a derived tsvector column</secondary>
</indexterm>
<term>
<synopsis>
tsvector_update_trigger(<replaceable class="PARAMETER">tsvector_column_name</replaceable>, <replaceable class="PARAMETER">config_name</replaceable>, <replaceable class="PARAMETER">text_column_name</replaceable> <optional>, ... </optional>)
tsvector_update_trigger_column(<replaceable class="PARAMETER">tsvector_column_name</replaceable>, <replaceable class="PARAMETER">config_column_name</replaceable>, <replaceable class="PARAMETER">text_column_name</replaceable> <optional>, ... </optional>)
</synopsis>
</term>
<listitem>
<para>
Two built-in trigger functions are available to automatically update a
<type>tsvector</> column from one or more textual columns. An example
of their use is:
<programlisting>
CREATE TABLE tblMessages (
strMessage text,
tsv tsvector
);
CREATE TRIGGER tsvectorupdate BEFORE INSERT OR UPDATE
ON tblMessages FOR EACH ROW EXECUTE PROCEDURE
tsvector_update_trigger(tsv, 'pg_catalog.english', strMessage);
</programlisting>
Having created this trigger, any change in <structfield>strMessage</>
will be automatically reflected into <structfield>tsv</>.
</para>
<para>
Both triggers require you to specify the text search configuration to
be used to perform the conversion. For
<function>tsvector_update_trigger</>, the configuration name is simply
given as the second trigger argument. It must be schema-qualified as
shown above, so that the trigger behavior will not change with changes
in <varname>search_path</>. For
<function>tsvector_update_trigger_column</>, the second trigger argument
is the name of another table column, which must be of type
<type>regconfig</>. This allows a per-row selection of configuration
to be made.
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsvector">
<primary>ts_stat</primary>
</indexterm>
<term>
<synopsis>
ts_stat(<replaceable class="PARAMETER">sqlquery</replaceable> text <optional>, <replaceable class="PARAMETER">weights</replaceable> text </optional>) returns SETOF statinfo
</synopsis>
</term>
<listitem>
<para>
Here <type>statinfo</type> is a type, defined as:
<programlisting>
CREATE TYPE statinfo AS (word text, ndoc integer, nentry integer);
</programlisting>
and <replaceable>sqlquery</replaceable> is a text value containing a SQL query
which returns a single <type>tsvector</type> column. <function>ts_stat</>
executes the query and returns statistics about the resulting
<type>tsvector</type> data, i.e., the number of documents, <literal>ndoc</>,
and the total number of words in the collection, <literal>nentry</>. It is
useful for checking your configuration and to find stop word candidates. For
example, to find the ten most frequent words:
<programlisting>
SELECT * FROM ts_stat('SELECT vector from apod')
ORDER BY ndoc DESC, nentry DESC, word
LIMIT 10;
</programlisting>
Optionally, one can specify <replaceable>weights</replaceable> to obtain
statistics about words with a specific <replaceable>weight</replaceable>:
<programlisting>
SELECT * FROM ts_stat('SELECT vector FROM apod','a')
ORDER BY ndoc DESC, nentry DESC, word
LIMIT 10;
</programlisting>
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsvector">
<primary>Btree operations for tsvector</primary>
</indexterm>
<term>
<synopsis>
TSVECTOR &lt; TSVECTOR
TSVECTOR &lt;= TSVECTOR
TSVECTOR = TSVECTOR
TSVECTOR &gt;= TSVECTOR
TSVECTOR &gt; TSVECTOR
</synopsis>
</term>
<listitem>
<para>
All btree operations are defined for the <type>tsvector</type> type.
<type>tsvector</>s are compared with each other using
<emphasis>lexicographical</emphasis> ordering.
<!-- TODO of the output representation or something else? -->
</para>
</listitem>
</varlistentry>
</variablelist>
</sect2>
<sect2 id="functions-textsearch-tsquery">
<title>tsquery</title>
<variablelist>
<varlistentry>
<indexterm zone="functions-textsearch-tsquery">
<primary>to_tsquery</primary>
</indexterm>
<term>
<synopsis>
to_tsquery(<optional><replaceable class="PARAMETER">config_name</replaceable></optional>, <replaceable class="PARAMETER">querytext</replaceable> text) returns TSQUERY
</synopsis>
</term>
<listitem>
<para>
Accepts <replaceable>querytext</replaceable>, which should consist of single tokens
separated by the boolean operators <literal>&amp;</literal> (and), <literal>|</literal>
(or) and <literal>!</literal> (not), which can be grouped using parentheses.
In other words, <function>to_tsquery</function> expects already parsed text.
Each token is reduced to a lexeme using the specified or current configuration.
A weight class can be assigned to each lexeme entry to restrict the search region
(see <function>setweight</function> for an explanation). For example:
<programlisting>
'fat:a &amp; rats'
</programlisting>
The <function>to_tsquery</function> function can also accept a <literal>text
string</literal>. In this case <replaceable>querytext</replaceable> should
be quoted. This may be useful, for example, to use with a thesaurus
dictionary. In the example below, a thesaurus contains rule <literal>supernovae
stars : sn</literal>:
<programlisting>
SELECT to_tsquery('''supernovae stars'' &amp; !crab');
to_tsquery
---------------
'sn' &amp; !'crab'
</programlisting>
Without quotes <function>to_tsquery</function> will generate a syntax error.
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsquery">
<primary>plainto_tsquery</primary>
</indexterm>
<term>
<synopsis>
plainto_tsquery(<optional><replaceable class="PARAMETER">config_name</replaceable></optional>, <replaceable class="PARAMETER">querytext</replaceable> text) returns TSQUERY
</synopsis>
</term>
<listitem>
<para>
Transforms unformatted text <replaceable>querytext</replaceable> to <type>tsquery</type>.
It is the same as <function>to_tsquery</function> but accepts <literal>text</literal>
without quotes and will call the parser to break it into tokens.
<function>plainto_tsquery</function> assumes the <literal>&amp;</literal> boolean
operator between words and does not recognize weight classes.
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsquery">
<primary>querytree</primary>
</indexterm>
<term>
<synopsis>
querytree(<replaceable class="PARAMETER">query</replaceable> TSQUERY) returns TEXT
</synopsis>
</term>
<listitem>
<para>
This returns the query used for searching an index. It can be used to test
for an empty query. The <command>SELECT</> below returns <literal>NULL</>,
which corresponds to an empty query since GIN indexes do not support queries with negation
<!-- TODO or "negated queries" (depending on what the correct rule is) -->
(a full index scan is inefficient):
<programlisting>
SELECT querytree(to_tsquery('!defined'));
querytree
-----------
</programlisting>
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsquery">
<primary>text::tsquery casting</primary>
</indexterm>
<term>
<synopsis>
<replaceable class="PARAMETER">text</replaceable>::TSQUERY returns TSQUERY
</synopsis>
</term>
<listitem>
<para>
Directly casting <replaceable>text</replaceable> to a <type>tsquery</type>
allows you to directly inject lexemes into a query using whatever positions
and positional weight flags you choose to specify. The text should be
formatted to match the way a vector is displayed by
<literal>SELECT</literal>.
<!-- TODO what a strange definition, I think something like
"input format" or so should be used (and defined somewhere, didn't see
it yet) -->
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsquery">
<primary>numnode</primary>
</indexterm>
<term>
<synopsis>
numnode(<replaceable class="PARAMETER">query</replaceable> TSQUERY) returns INTEGER
</synopsis>
</term>
<listitem>
<para>
This returns the number of nodes in a query tree. This function can be
used to determine if <replaceable>query</replaceable> is meaningful
(returns &gt; 0), or contains only stop words (returns 0):
<programlisting>
SELECT numnode(plainto_tsquery('the any'));
NOTICE: query contains only stopword(s) or does not contain lexeme(s), ignored
numnode
---------
0
SELECT numnode(plainto_tsquery('the table'));
numnode
---------
1
SELECT numnode(plainto_tsquery('long table'));
numnode
---------
3
</programlisting>
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsquery">
<primary>TSQUERY &amp;&amp; TSQUERY</primary>
</indexterm>
<term>
<synopsis>
TSQUERY &amp;&amp; TSQUERY returns TSQUERY
</synopsis>
</term>
<listitem>
<para>
Returns <literal>AND</literal>-ed TSQUERY
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsquery">
<primary>TSQUERY || TSQUERY</primary>
</indexterm>
<term>
<synopsis>
TSQUERY || TSQUERY returns TSQUERY
</synopsis>
</term>
<listitem>
<para>
Returns <literal>OR</literal>-ed TSQUERY
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsquery">
<primary>!! TSQUERY</primary>
</indexterm>
<term>
<synopsis>
!! TSQUERY returns TSQUERY
</synopsis>
</term>
<listitem>
<para>
negation of TSQUERY
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsquery">
<primary>Btree operations for tsquery</primary>
</indexterm>
<term>
<synopsis>
TSQUERY &lt; TSQUERY
TSQUERY &lt;= TSQUERY
TSQUERY = TSQUERY
TSQUERY &gt;= TSQUERY
TSQUERY &gt; TSQUERY
</synopsis>
</term>
<listitem>
<para>
All btree operations are defined for the <type>tsquery</type> type.
tsqueries are compared to each other using <emphasis>lexicographical</emphasis>
ordering.
</para>
</listitem>
</varlistentry>
</variablelist>
<sect3 id="functions-textsearch-queryrewriting">
<title>Query Rewriting</title>
<para>
Query rewriting is a set of functions and operators for the
<type>tsquery</type> data type. It allows control at search
<emphasis>query time</emphasis> without reindexing (the opposite of the
thesaurus). For example, you can expand the search using synonyms
(<literal>new york</>, <literal>big apple</>, <literal>nyc</>,
<literal>gotham</>) or narrow the search to direct the user to some hot
topic.
</para>
<para>
The <function>ts_rewrite()</function> function changes the original query by
replacing part of the query with some other string of type <type>tsquery</type>,
as defined by the rewrite rule. Arguments to <function>ts_rewrite()</function>
can be names of columns of type <type>tsquery</type>.
</para>
<programlisting>
CREATE TABLE aliases (t TSQUERY PRIMARY KEY, s TSQUERY);
INSERT INTO aliases VALUES('a', 'c');
</programlisting>
<variablelist>
<varlistentry>
<indexterm zone="functions-textsearch-tsquery">
<primary>ts_rewrite</primary>
</indexterm>
<term>
<synopsis>
ts_rewrite (<replaceable class="PARAMETER">query</replaceable> TSQUERY, <replaceable class="PARAMETER">target</replaceable> TSQUERY, <replaceable class="PARAMETER">sample</replaceable> TSQUERY) returns TSQUERY
</synopsis>
</term>
<listitem>
<para>
<programlisting>
SELECT ts_rewrite('a &amp; b'::tsquery, 'a'::tsquery, 'c'::tsquery);
ts_rewrite
------------
'b' &amp; 'c'
</programlisting>
</para>
</listitem>
</varlistentry>
<varlistentry>
<term>
<synopsis>
ts_rewrite(ARRAY[<replaceable class="PARAMETER">query</replaceable> TSQUERY, <replaceable class="PARAMETER">target</replaceable> TSQUERY, <replaceable class="PARAMETER">sample</replaceable> TSQUERY]) returns TSQUERY
</synopsis>
</term>
<listitem>
<para>
<programlisting>
SELECT ts_rewrite(ARRAY['a &amp; b'::tsquery, t,s]) FROM aliases;
ts_rewrite
------------
'b' &amp; 'c'
</programlisting>
</para>
</listitem>
</varlistentry>
<varlistentry>
<term>
<synopsis>
ts_rewrite (<replaceable class="PARAMETER">query</> TSQUERY,<literal>'SELECT target ,sample FROM test'</literal>::text) returns TSQUERY
</synopsis>
</term>
<listitem>
<para>
<programlisting>
SELECT ts_rewrite('a &amp; b'::tsquery, 'SELECT t,s FROM aliases');
ts_rewrite
------------
'b' &amp; 'c'
</programlisting>
</para>
</listitem>
</varlistentry>
</variablelist>
<para>
What if there are several instances of rewriting? For example, query
<literal>'a &amp; b'</literal> can be rewritten as
<literal>'b &amp; c'</literal> and <literal>'cc'</literal>.
<programlisting>
SELECT * FROM aliases;
t | s
-----------+------
'a' | 'c'
'x' | 'z'
'a' &amp; 'b' | 'cc'
</programlisting>
This ambiguity can be resolved by specifying a sort order:
<programlisting>
SELECT ts_rewrite('a &amp; b', 'SELECT t, s FROM aliases ORDER BY t DESC');
ts_rewrite
---------
'cc'
SELECT ts_rewrite('a &amp; b', 'SELECT t, s FROM aliases ORDER BY t ASC');
ts_rewrite
--------------
'b' &amp; 'c'
</programlisting>
</para>
<para>
Let's consider a real-life astronomical example. We'll expand query
<literal>supernovae</literal> using table-driven rewriting rules:
<programlisting>
CREATE TABLE aliases (t tsquery primary key, s tsquery);
INSERT INTO aliases VALUES(to_tsquery('supernovae'), to_tsquery('supernovae|sn'));
SELECT ts_rewrite(to_tsquery('supernovae'), 'SELECT * FROM aliases') &amp;&amp; to_tsquery('crab');
?column?
-------------------------------
( 'supernova' | 'sn' ) &amp; 'crab'
</programlisting>
Notice, that we can change the rewriting rule online<!-- TODO maybe use another word for "online"? -->:
<programlisting>
UPDATE aliases SET s=to_tsquery('supernovae|sn &amp; !nebulae') WHERE t=to_tsquery('supernovae');
SELECT ts_rewrite(to_tsquery('supernovae'), 'SELECT * FROM aliases') &amp;&amp; to_tsquery('crab');
?column?
-----------------------------------------------
'supernova' | 'sn' &amp; !'nebula' ) &amp; 'crab'
</programlisting>
</para>
</sect3>
<sect3 id="functions-textsearch-tsquery-ops">
<title>Operators For tsquery</title>
<para>
Rewriting can be slow for many rewriting rules since it checks every rule
for a possible hit. To filter out obvious non-candidate rules there are containment
operators for the <type>tsquery</type> type. In the example below, we select only those
rules which might contain the original query:
<programlisting>
SELECT ts_rewrite(ARRAY['a &amp; b'::tsquery, t,s])
FROM aliases
WHERE 'a &amp; b' @> t;
ts_rewrite
------------
'b' &amp; 'c'
</programlisting>
</para>
<para>
Two operators are defined for <type>tsquery</type>:
</para>
<variablelist>
<varlistentry>
<indexterm zone="functions-textsearch-tsquery">
<primary>TSQUERY @&gt; TSQUERY</primary>
</indexterm>
<term>
<synopsis>
TSQUERY @&gt; TSQUERY
</synopsis>
</term>
<listitem>
<para>
Returns <literal>true</literal> if the right argument might be contained in left argument.
</para>
</listitem>
</varlistentry>
<varlistentry>
<indexterm zone="functions-textsearch-tsquery">
<primary>tsquery &lt;@ tsquery</primary>
</indexterm>
<term>
<synopsis>
TSQUERY &lt;@ TSQUERY
</synopsis>
</term>
<listitem>
<para>
Returns <literal>true</literal> if the left argument might be contained in right argument.
</para>
</listitem>
</varlistentry>
</variablelist>
</sect3>
<sect3 id="functions-textsearch-tsqueryindex">
<title>Index For tsquery</title>
<para>
To speed up operators <literal>&lt;@</> and <literal>@&gt;</literal> for
<type>tsquery</type> one can use a <acronym>GiST</acronym> index with
a <literal>tsquery_ops</literal> opclass:
<programlisting>
CREATE INDEX t_idx ON aliases USING gist (t tsquery_ops);
</programlisting>
</para>
</sect3>
</sect2>
</sect1>
<sect1 id="functions-xml">
<title>XML Functions</title>

1222
doc/src/sgml/textsearch.sgml
View File

File diff suppressed because it is too large Load Diff