Baoding-FenShuaJiang
/
Green_GoodERP18_FSJ_Standard
Watch 1
Star 0
Wiki 捐赠
gooderp18绿色标准版
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Tree: 0f75e31513
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '0f75e31513'
${ noResults }
Green_GoodERP18_FSJ_Standard/PostgreSQL/doc/postgresql/html/textsearch-controls.html

543 lines
34KB
Raw Blame History 

  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>12.3. Controlling Text Search</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 V1.79.1" /><link rel="prev" href="textsearch-tables.html" title="12.2. Tables and Indexes" /><link rel="next" href="textsearch-features.html" title="12.4. Additional Features" /></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">12.3. Controlling Text Search</th></tr><tr><td width="10%" align="left"><a accesskey="p" href="textsearch-tables.html" title="12.2. Tables and Indexes">Prev</a> </td><td width="10%" align="left"><a accesskey="u" href="textsearch.html" title="Chapter 12. Full Text Search">Up</a></td><th width="60%" align="center">Chapter 12. Full Text Search</th><td width="10%" align="right"><a accesskey="h" href="index.html" title="PostgreSQL 12.4 Documentation">Home</a></td><td width="10%" align="right"> <a accesskey="n" href="textsearch-features.html" title="12.4. Additional Features">Next</a></td></tr></table><hr></hr></div><div class="sect1" id="TEXTSEARCH-CONTROLS"><div class="titlepage"><div><div><h2 class="title" style="clear: both">12.3. Controlling Text Search</h2></div></div></div><div class="toc"><dl class="toc"><dt><span class="sect2"><a href="textsearch-controls.html#TEXTSEARCH-PARSING-DOCUMENTS">12.3.1. Parsing Documents</a></span></dt><dt><span class="sect2"><a href="textsearch-controls.html#TEXTSEARCH-PARSING-QUERIES">12.3.2. Parsing Queries</a></span></dt><dt><span class="sect2"><a href="textsearch-controls.html#TEXTSEARCH-RANKING">12.3.3. Ranking Search Results</a></span></dt><dt><span class="sect2"><a href="textsearch-controls.html#TEXTSEARCH-HEADLINE">12.3.4. Highlighting Results</a></span></dt></dl></div><p>

  3.    To implement full text searching there must be a function to create a

  4.    <code class="type">tsvector</code> from a document and a <code class="type">tsquery</code> from a

  5.    user query. Also, we need to return results in a useful order, so we need

  6.    a function that compares documents with respect to their relevance to

  7.    the query. It's also important to be able to display the results nicely.

  8.    <span class="productname">PostgreSQL</span> provides support for all of these

  9.    functions.

  10.   </p><div class="sect2" id="TEXTSEARCH-PARSING-DOCUMENTS"><div class="titlepage"><div><div><h3 class="title">12.3.1. Parsing Documents</h3></div></div></div><p>

  11.     <span class="productname">PostgreSQL</span> provides the

  12.     function <code class="function">to_tsvector</code> for converting a document to

  13.     the <code class="type">tsvector</code> data type.

  14.    </p><a id="id-1.5.11.6.3.3" class="indexterm"></a><pre class="synopsis">

  15. to_tsvector([<span class="optional"> <em class="replaceable"><code>config</code></em> <code class="type">regconfig</code>, </span>] <em class="replaceable"><code>document</code></em> <code class="type">text</code>) returns <code class="type">tsvector</code>

  16. </pre><p>

  17.     <code class="function">to_tsvector</code> parses a textual document into tokens,

  18.     reduces the tokens to lexemes, and returns a <code class="type">tsvector</code> which

  19.     lists the lexemes together with their positions in the document.

  20.     The document is processed according to the specified or default

  21.     text search configuration.

  22.     Here is a simple example:

  23. 

  24. </p><pre class="screen">

  25. SELECT to_tsvector('english', 'a fat  cat sat on a mat - it ate a fat rats');

  26.                   to_tsvector

  27. -----------------------------------------------------

  28.  'ate':9 'cat':3 'fat':2,11 'mat':7 'rat':12 'sat':4

  29. </pre><p>

  30.    </p><p>

  31.     In the example above we see that the resulting <code class="type">tsvector</code> does not

  32.     contain the words <code class="literal">a</code>, <code class="literal">on</code>, or

  33.     <code class="literal">it</code>, the word <code class="literal">rats</code> became

  34.     <code class="literal">rat</code>, and the punctuation sign <code class="literal">-</code> was

  35.     ignored.

  36.    </p><p>

  37.     The <code class="function">to_tsvector</code> function internally calls a parser

  38.     which breaks the document text into tokens and assigns a type to

  39.     each token.  For each token, a list of

  40.     dictionaries (<a class="xref" href="textsearch-dictionaries.html" title="12.6. Dictionaries">Section 12.6</a>) is consulted,

  41.     where the list can vary depending on the token type.  The first dictionary

  42.     that <em class="firstterm">recognizes</em> the token emits one or more normalized

  43.     <em class="firstterm">lexemes</em> to represent the token.  For example,

  44.     <code class="literal">rats</code> became <code class="literal">rat</code> because one of the

  45.     dictionaries recognized that the word <code class="literal">rats</code> is a plural

  46.     form of <code class="literal">rat</code>.  Some words are recognized as

  47.     <em class="firstterm">stop words</em> (<a class="xref" href="textsearch-dictionaries.html#TEXTSEARCH-STOPWORDS" title="12.6.1. Stop Words">Section 12.6.1</a>), which

  48.     causes them to be ignored since they occur too frequently to be useful in

  49.     searching.  In our example these are

  50.     <code class="literal">a</code>, <code class="literal">on</code>, and <code class="literal">it</code>.

  51.     If no dictionary in the list recognizes the token then it is also ignored.

  52.     In this example that happened to the punctuation sign <code class="literal">-</code>

  53.     because there are in fact no dictionaries assigned for its token type

  54.     (<code class="literal">Space symbols</code>), meaning space tokens will never be

  55.     indexed. The choices of parser, dictionaries and which types of tokens to

  56.     index are determined by the selected text search configuration (<a class="xref" href="textsearch-configuration.html" title="12.7. Configuration Example">Section 12.7</a>).  It is possible to have

  57.     many different configurations in the same database, and predefined

  58.     configurations are available for various languages. In our example

  59.     we used the default configuration <code class="literal">english</code> for the

  60.     English language.

  61.    </p><p>

  62.     The function <code class="function">setweight</code> can be used to label the

  63.     entries of a <code class="type">tsvector</code> with a given <em class="firstterm">weight</em>,

  64.     where a weight is one of the letters <code class="literal">A</code>, <code class="literal">B</code>,

  65.     <code class="literal">C</code>, or <code class="literal">D</code>.

  66.     This is typically used to mark entries coming from

  67.     different parts of a document, such as title versus body.  Later, this

  68.     information can be used for ranking of search results.

  69.    </p><p>

  70.     Because <code class="function">to_tsvector</code>(<code class="literal">NULL</code>) will

  71.     return <code class="literal">NULL</code>, it is recommended to use

  72.     <code class="function">coalesce</code> whenever a field might be null.

  73.     Here is the recommended method for creating

  74.     a <code class="type">tsvector</code> from a structured document:

  75. 

  76. </p><pre class="programlisting">

  77. UPDATE tt SET ti =

  78.     setweight(to_tsvector(coalesce(title,'')), 'A')    ||

  79.     setweight(to_tsvector(coalesce(keyword,'')), 'B')  ||

  80.     setweight(to_tsvector(coalesce(abstract,'')), 'C') ||

  81.     setweight(to_tsvector(coalesce(body,'')), 'D');

  82. </pre><p>

  83. 

  84.     Here we have used <code class="function">setweight</code> to label the source

  85.     of each lexeme in the finished <code class="type">tsvector</code>, and then merged

  86.     the labeled <code class="type">tsvector</code> values using the <code class="type">tsvector</code>

  87.     concatenation operator <code class="literal">||</code>.  (<a class="xref" href="textsearch-features.html#TEXTSEARCH-MANIPULATE-TSVECTOR" title="12.4.1. Manipulating Documents">Section 12.4.1</a> gives details about these

  88.     operations.)

  89.    </p></div><div class="sect2" id="TEXTSEARCH-PARSING-QUERIES"><div class="titlepage"><div><div><h3 class="title">12.3.2. Parsing Queries</h3></div></div></div><p>

  90.     <span class="productname">PostgreSQL</span> provides the

  91.     functions <code class="function">to_tsquery</code>,

  92.     <code class="function">plainto_tsquery</code>,

  93.     <code class="function">phraseto_tsquery</code> and

  94.     <code class="function">websearch_to_tsquery</code>

  95.     for converting a query to the <code class="type">tsquery</code> data type.

  96.     <code class="function">to_tsquery</code> offers access to more features

  97.     than either <code class="function">plainto_tsquery</code> or

  98.     <code class="function">phraseto_tsquery</code>, but it is less forgiving about its

  99.     input. <code class="function">websearch_to_tsquery</code> is a simplified version

  100.     of <code class="function">to_tsquery</code> with an alternative syntax, similar

  101.     to the one used by web search engines.

  102.    </p><a id="id-1.5.11.6.4.3" class="indexterm"></a><pre class="synopsis">

  103. to_tsquery([<span class="optional"> <em class="replaceable"><code>config</code></em> <code class="type">regconfig</code>, </span>] <em class="replaceable"><code>querytext</code></em> <code class="type">text</code>) returns <code class="type">tsquery</code>

  104. </pre><p>

  105.     <code class="function">to_tsquery</code> creates a <code class="type">tsquery</code> value from

  106.     <em class="replaceable"><code>querytext</code></em>, which must consist of single tokens

  107.     separated by the <code class="type">tsquery</code> operators <code class="literal">&amp;</code> (AND),

  108.     <code class="literal">|</code> (OR), <code class="literal">!</code> (NOT), and

  109.     <code class="literal">&lt;-&gt;</code> (FOLLOWED BY), possibly grouped

  110.     using parentheses.  In other words, the input to

  111.     <code class="function">to_tsquery</code> must already follow the general rules for

  112.     <code class="type">tsquery</code> input, as described in <a class="xref" href="datatype-textsearch.html#DATATYPE-TSQUERY" title="8.11.2. tsquery">Section 8.11.2</a>.  The difference is that while basic

  113.     <code class="type">tsquery</code> input takes the tokens at face value,

  114.     <code class="function">to_tsquery</code> normalizes each token into a lexeme using

  115.     the specified or default configuration, and discards any tokens that are

  116.     stop words according to the configuration.  For example:

  117. 

  118. </p><pre class="screen">

  119. SELECT to_tsquery('english', 'The &amp; Fat &amp; Rats');

  120.   to_tsquery   

  121. ---------------

  122.  'fat' &amp; 'rat'

  123. </pre><p>

  124. 

  125.     As in basic <code class="type">tsquery</code> input, weight(s) can be attached to each

  126.     lexeme to restrict it to match only <code class="type">tsvector</code> lexemes of those

  127.     weight(s).  For example:

  128. 

  129. </p><pre class="screen">

  130. SELECT to_tsquery('english', 'Fat | Rats:AB');

  131.     to_tsquery    

  132. ------------------

  133.  'fat' | 'rat':AB

  134. </pre><p>

  135. 

  136.     Also, <code class="literal">*</code> can be attached to a lexeme to specify prefix matching:

  137. 

  138. </p><pre class="screen">

  139. SELECT to_tsquery('supern:*A &amp; star:A*B');

  140.         to_tsquery        

  141. --------------------------

  142.  'supern':*A &amp; 'star':*AB

  143. </pre><p>

  144. 

  145.     Such a lexeme will match any word in a <code class="type">tsvector</code> that begins

  146.     with the given string.

  147.    </p><p>

  148.     <code class="function">to_tsquery</code> can also accept single-quoted

  149.     phrases.  This is primarily useful when the configuration includes a

  150.     thesaurus dictionary that may trigger on such phrases.

  151.     In the example below, a thesaurus contains the rule <code class="literal">supernovae

  152.     stars : sn</code>:

  153. 

  154. </p><pre class="screen">

  155. SELECT to_tsquery('''supernovae stars'' &amp; !crab');

  156.   to_tsquery

  157. ---------------

  158.  'sn' &amp; !'crab'

  159. </pre><p>

  160. 

  161.     Without quotes, <code class="function">to_tsquery</code> will generate a syntax

  162.     error for tokens that are not separated by an AND, OR, or FOLLOWED BY

  163.     operator.

  164.    </p><a id="id-1.5.11.6.4.7" class="indexterm"></a><pre class="synopsis">

  165. plainto_tsquery([<span class="optional"> <em class="replaceable"><code>config</code></em> <code class="type">regconfig</code>, </span>] <em class="replaceable"><code>querytext</code></em> <code class="type">text</code>) returns <code class="type">tsquery</code>

  166. </pre><p>

  167.     <code class="function">plainto_tsquery</code> transforms the unformatted text

  168.     <em class="replaceable"><code>querytext</code></em> to a <code class="type">tsquery</code> value.

  169.     The text is parsed and normalized much as for <code class="function">to_tsvector</code>,

  170.     then the <code class="literal">&amp;</code> (AND) <code class="type">tsquery</code> operator is

  171.     inserted between surviving words.

  172.    </p><p>

  173.     Example:

  174. 

  175. </p><pre class="screen">

  176. SELECT plainto_tsquery('english', 'The Fat Rats');

  177.  plainto_tsquery 

  178. -----------------

  179.  'fat' &amp; 'rat'

  180. </pre><p>

  181. 

  182.     Note that <code class="function">plainto_tsquery</code> will not

  183.     recognize <code class="type">tsquery</code> operators, weight labels,

  184.     or prefix-match labels in its input:

  185. 

  186. </p><pre class="screen">

  187. SELECT plainto_tsquery('english', 'The Fat &amp; Rats:C');

  188.    plainto_tsquery   

  189. ---------------------

  190.  'fat' &amp; 'rat' &amp; 'c'

  191. </pre><p>

  192. 

  193.     Here, all the input punctuation was discarded as being space symbols.

  194.    </p><a id="id-1.5.11.6.4.11" class="indexterm"></a><pre class="synopsis">

  195. phraseto_tsquery([<span class="optional"> <em class="replaceable"><code>config</code></em> <code class="type">regconfig</code>, </span>] <em class="replaceable"><code>querytext</code></em> <code class="type">text</code>) returns <code class="type">tsquery</code>

  196. </pre><p>

  197.     <code class="function">phraseto_tsquery</code> behaves much like

  198.     <code class="function">plainto_tsquery</code>, except that it inserts

  199.     the <code class="literal">&lt;-&gt;</code> (FOLLOWED BY) operator between

  200.     surviving words instead of the <code class="literal">&amp;</code> (AND) operator.

  201.     Also, stop words are not simply discarded, but are accounted for by

  202.     inserting <code class="literal">&lt;<em class="replaceable"><code>N</code></em>&gt;</code> operators rather

  203.     than <code class="literal">&lt;-&gt;</code> operators.  This function is useful

  204.     when searching for exact lexeme sequences, since the FOLLOWED BY

  205.     operators check lexeme order not just the presence of all the lexemes.

  206.    </p><p>

  207.     Example:

  208. 

  209. </p><pre class="screen">

  210. SELECT phraseto_tsquery('english', 'The Fat Rats');

  211.  phraseto_tsquery

  212. ------------------

  213.  'fat' &lt;-&gt; 'rat'

  214. </pre><p>

  215. 

  216.     Like <code class="function">plainto_tsquery</code>, the

  217.     <code class="function">phraseto_tsquery</code> function will not

  218.     recognize <code class="type">tsquery</code> operators, weight labels,

  219.     or prefix-match labels in its input:

  220. 

  221. </p><pre class="screen">

  222. SELECT phraseto_tsquery('english', 'The Fat &amp; Rats:C');

  223.       phraseto_tsquery

  224. -----------------------------

  225.  'fat' &lt;-&gt; 'rat' &lt;-&gt; 'c'

  226. </pre><p>

  227.    </p><pre class="synopsis">

  228. websearch_to_tsquery([<span class="optional"> <em class="replaceable"><code>config</code></em> <code class="type">regconfig</code>, </span>] <em class="replaceable"><code>querytext</code></em> <code class="type">text</code>) returns <code class="type">tsquery</code>

  229. </pre><p>

  230.     <code class="function">websearch_to_tsquery</code> creates a <code class="type">tsquery</code>

  231.     value from <em class="replaceable"><code>querytext</code></em> using an alternative

  232.     syntax in which simple unformatted text is a valid query.

  233.     Unlike <code class="function">plainto_tsquery</code>

  234.     and <code class="function">phraseto_tsquery</code>, it also recognizes certain

  235.     operators. Moreover, this function should never raise syntax errors,

  236.     which makes it possible to use raw user-supplied input for search.

  237.     The following syntax is supported:

  238.     </p><div class="itemizedlist"><ul class="itemizedlist compact" style="list-style-type: bullet; "><li class="listitem" style="list-style-type: disc"><p>

  239.         <code class="literal">unquoted text</code>: text not inside quote marks will be

  240.         converted to terms separated by <code class="literal">&amp;</code> operators, as

  241.         if processed by

  242.         <code class="function">plainto_tsquery</code>.

  243.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  244.         <code class="literal">"quoted text"</code>: text inside quote marks will be

  245.         converted to terms separated by <code class="literal">&lt;-&gt;</code>

  246.         operators, as if processed by <code class="function">phraseto_tsquery</code>.

  247.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  248.        <code class="literal">OR</code>: logical or will be converted to

  249.        the <code class="literal">|</code> operator.

  250.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  251.        <code class="literal">-</code>: the logical not operator, converted to the

  252.        the <code class="literal">!</code> operator.

  253.       </p></li></ul></div><p>

  254.    </p><p>

  255.     Examples:

  256. </p><pre class="screen">

  257. SELECT websearch_to_tsquery('english', 'The fat rats');

  258.  websearch_to_tsquery

  259. ----------------------

  260.  'fat' &amp; 'rat'

  261. (1 row)

  262. 

  263. SELECT websearch_to_tsquery('english', '"supernovae stars" -crab');

  264.        websearch_to_tsquery

  265. ----------------------------------

  266.  'supernova' &lt;-&gt; 'star' &amp; !'crab'

  267. (1 row)

  268. 

  269. SELECT websearch_to_tsquery('english', '"sad cat" or "fat rat"');

  270.        websearch_to_tsquery

  271. -----------------------------------

  272.  'sad' &lt;-&gt; 'cat' | 'fat' &lt;-&gt; 'rat'

  273. (1 row)

  274. 

  275. SELECT websearch_to_tsquery('english', 'signal -"segmentation fault"');

  276.          websearch_to_tsquery

  277. ---------------------------------------

  278.  'signal' &amp; !( 'segment' &lt;-&gt; 'fault' )

  279. (1 row)

  280. 

  281. SELECT websearch_to_tsquery('english', '""" )( dummy \\ query &lt;-&gt;');

  282.  websearch_to_tsquery

  283. ----------------------

  284.  'dummi' &amp; 'queri'

  285. (1 row)

  286. </pre><p>

  287.     </p></div><div class="sect2" id="TEXTSEARCH-RANKING"><div class="titlepage"><div><div><h3 class="title">12.3.3. Ranking Search Results</h3></div></div></div><p>

  288.     Ranking attempts to measure how relevant documents are to a particular

  289.     query, so that when there are many matches the most relevant ones can be

  290.     shown first.  <span class="productname">PostgreSQL</span> provides two

  291.     predefined ranking functions, which take into account lexical, proximity,

  292.     and structural information; that is, they consider how often the query

  293.     terms appear in the document, how close together the terms are in the

  294.     document, and how important is the part of the document where they occur.

  295.     However, the concept of relevancy is vague and very application-specific.

  296.     Different applications might require additional information for ranking,

  297.     e.g., document modification time.  The built-in ranking functions are only

  298.     examples.  You can write your own ranking functions and/or combine their

  299.     results with additional factors to fit your specific needs.

  300.    </p><p>

  301.     The two ranking functions currently available are:

  302. 

  303.     </p><div class="variablelist"><dl class="variablelist"><dt><span class="term">

  304.        <a id="id-1.5.11.6.5.3.1.1.1.1" class="indexterm"></a>

  305. 

  306.        <code class="literal">ts_rank([<span class="optional"> <em class="replaceable"><code>weights</code></em> <code class="type">float4[]</code>, </span>] <em class="replaceable"><code>vector</code></em> <code class="type">tsvector</code>, <em class="replaceable"><code>query</code></em> <code class="type">tsquery</code> [<span class="optional">, <em class="replaceable"><code>normalization</code></em> <code class="type">integer</code> </span>]) returns <code class="type">float4</code></code>

  307.       </span></dt><dd><p>

  308.         Ranks vectors based on the frequency of their matching lexemes.

  309.        </p></dd><dt><span class="term">

  310.       <a id="id-1.5.11.6.5.3.1.2.1.1" class="indexterm"></a>

  311. 

  312.        <code class="literal">ts_rank_cd([<span class="optional"> <em class="replaceable"><code>weights</code></em> <code class="type">float4[]</code>, </span>] <em class="replaceable"><code>vector</code></em> <code class="type">tsvector</code>, <em class="replaceable"><code>query</code></em> <code class="type">tsquery</code> [<span class="optional">, <em class="replaceable"><code>normalization</code></em> <code class="type">integer</code> </span>]) returns <code class="type">float4</code></code>

  313.       </span></dt><dd><p>

  314.         This function computes the <em class="firstterm">cover density</em>

  315.         ranking for the given document vector and query, as described in

  316.         Clarke, Cormack, and Tudhope's "Relevance Ranking for One to Three

  317.         Term Queries" in the journal "Information Processing and Management",

  318.         1999.  Cover density is similar to <code class="function">ts_rank</code> ranking

  319.         except that the proximity of matching lexemes to each other is

  320.         taken into consideration.

  321.        </p><p>

  322.         This function requires lexeme positional information to perform

  323.         its calculation.  Therefore, it ignores any <span class="quote">“<span class="quote">stripped</span>”</span>

  324.         lexemes in the <code class="type">tsvector</code>.  If there are no unstripped

  325.         lexemes in the input, the result will be zero.  (See <a class="xref" href="textsearch-features.html#TEXTSEARCH-MANIPULATE-TSVECTOR" title="12.4.1. Manipulating Documents">Section 12.4.1</a> for more information

  326.         about the <code class="function">strip</code> function and positional information

  327.         in <code class="type">tsvector</code>s.)

  328.        </p></dd></dl></div><p>

  329. 

  330.    </p><p>

  331.     For both these functions,

  332.     the optional <em class="replaceable"><code>weights</code></em>

  333.     argument offers the ability to weigh word instances more or less

  334.     heavily depending on how they are labeled.  The weight arrays specify

  335.     how heavily to weigh each category of word, in the order:

  336. 

  337. </p><pre class="synopsis">

  338. {D-weight, C-weight, B-weight, A-weight}

  339. </pre><p>

  340. 

  341.     If no <em class="replaceable"><code>weights</code></em> are provided,

  342.     then these defaults are used:

  343. 

  344. </p><pre class="programlisting">

  345. {0.1, 0.2, 0.4, 1.0}

  346. </pre><p>

  347. 

  348.     Typically weights are used to mark words from special areas of the

  349.     document, like the title or an initial abstract, so they can be

  350.     treated with more or less importance than words in the document body.

  351.    </p><p>

  352.     Since a longer document has a greater chance of containing a query term

  353.     it is reasonable to take into account document size, e.g., a hundred-word

  354.     document with five instances of a search word is probably more relevant

  355.     than a thousand-word document with five instances.  Both ranking functions

  356.     take an integer <em class="replaceable"><code>normalization</code></em> option that

  357.     specifies whether and how a document's length should impact its rank.

  358.     The integer option controls several behaviors, so it is a bit mask:

  359.     you can specify one or more behaviors using

  360.     <code class="literal">|</code> (for example, <code class="literal">2|4</code>).

  361. 

  362.     </p><div class="itemizedlist"><ul class="itemizedlist compact" style="list-style-type: bullet; "><li class="listitem" style="list-style-type: disc"><p>

  363.        0 (the default) ignores the document length

  364.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  365.        1 divides the rank by 1 + the logarithm of the document length

  366.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  367.        2 divides the rank by the document length

  368.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  369.        4 divides the rank by the mean harmonic distance between extents

  370.        (this is implemented only by <code class="function">ts_rank_cd</code>)

  371.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  372.        8 divides the rank by the number of unique words in document

  373.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  374.        16 divides the rank by 1 + the logarithm of the number

  375.        of unique words in document

  376.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  377.        32 divides the rank by itself + 1

  378.       </p></li></ul></div><p>

  379. 

  380.     If more than one flag bit is specified, the transformations are

  381.     applied in the order listed.

  382.    </p><p>

  383.     It is important to note that the ranking functions do not use any global

  384.     information, so it is impossible to produce a fair normalization to 1% or

  385.     100% as sometimes desired.  Normalization option 32

  386.     (<code class="literal">rank/(rank+1)</code>) can be applied to scale all ranks

  387.     into the range zero to one, but of course this is just a cosmetic change;

  388.     it will not affect the ordering of the search results.

  389.    </p><p>

  390.     Here is an example that selects only the ten highest-ranked matches:

  391. 

  392. </p><pre class="screen">

  393. SELECT title, ts_rank_cd(textsearch, query) AS rank

  394. FROM apod, to_tsquery('neutrino|(dark &amp; matter)') query

  395. WHERE query @@ textsearch

  396. ORDER BY rank DESC

  397. LIMIT 10;

  398.                      title                     |   rank

  399. -----------------------------------------------+----------

  400.  Neutrinos in the Sun                          |      3.1

  401.  The Sudbury Neutrino Detector                 |      2.4

  402.  A MACHO View of Galactic Dark Matter          |  2.01317

  403.  Hot Gas and Dark Matter                       |  1.91171

  404.  The Virgo Cluster: Hot Plasma and Dark Matter |  1.90953

  405.  Rafting for Solar Neutrinos                   |      1.9

  406.  NGC 4650A: Strange Galaxy and Dark Matter     |  1.85774

  407.  Hot Gas and Dark Matter                       |   1.6123

  408.  Ice Fishing for Cosmic Neutrinos              |      1.6

  409.  Weak Lensing Distorts the Universe            | 0.818218

  410. </pre><p>

  411. 

  412.     This is the same example using normalized ranking:

  413. 

  414. </p><pre class="screen">

  415. SELECT title, ts_rank_cd(textsearch, query, 32 /* rank/(rank+1) */ ) AS rank

  416. FROM apod, to_tsquery('neutrino|(dark &amp; matter)') query

  417. WHERE  query @@ textsearch

  418. ORDER BY rank DESC

  419. LIMIT 10;

  420.                      title                     |        rank

  421. -----------------------------------------------+-------------------

  422.  Neutrinos in the Sun                          | 0.756097569485493

  423.  The Sudbury Neutrino Detector                 | 0.705882361190954

  424.  A MACHO View of Galactic Dark Matter          | 0.668123210574724

  425.  Hot Gas and Dark Matter                       |  0.65655958650282

  426.  The Virgo Cluster: Hot Plasma and Dark Matter | 0.656301290640973

  427.  Rafting for Solar Neutrinos                   | 0.655172410958162

  428.  NGC 4650A: Strange Galaxy and Dark Matter     | 0.650072921219637

  429.  Hot Gas and Dark Matter                       | 0.617195790024749

  430.  Ice Fishing for Cosmic Neutrinos              | 0.615384618911517

  431.  Weak Lensing Distorts the Universe            | 0.450010798361481

  432. </pre><p>

  433.    </p><p>

  434.     Ranking can be expensive since it requires consulting the

  435.     <code class="type">tsvector</code> of each matching document, which can be I/O bound and

  436.     therefore slow. Unfortunately, it is almost impossible to avoid since

  437.     practical queries often result in large numbers of matches.

  438.    </p></div><div class="sect2" id="TEXTSEARCH-HEADLINE"><div class="titlepage"><div><div><h3 class="title">12.3.4. Highlighting Results</h3></div></div></div><p>

  439.     To present search results it is ideal to show a part of each document and

  440.     how it is related to the query. Usually, search engines show fragments of

  441.     the document with marked search terms.  <span class="productname">PostgreSQL</span>

  442.     provides a function <code class="function">ts_headline</code> that

  443.     implements this functionality.

  444.    </p><a id="id-1.5.11.6.6.3" class="indexterm"></a><pre class="synopsis">

  445. ts_headline([<span class="optional"> <em class="replaceable"><code>config</code></em> <code class="type">regconfig</code>, </span>] <em class="replaceable"><code>document</code></em> <code class="type">text</code>, <em class="replaceable"><code>query</code></em> <code class="type">tsquery</code> [<span class="optional">, <em class="replaceable"><code>options</code></em> <code class="type">text</code> </span>]) returns <code class="type">text</code>

  446. </pre><p>

  447.     <code class="function">ts_headline</code> accepts a document along

  448.     with a query, and returns an excerpt from

  449.     the document in which terms from the query are highlighted.  The

  450.     configuration to be used to parse the document can be specified by

  451.     <em class="replaceable"><code>config</code></em>; if <em class="replaceable"><code>config</code></em>

  452.     is omitted, the

  453.     <code class="varname">default_text_search_config</code> configuration is used.

  454.    </p><p>

  455.     If an <em class="replaceable"><code>options</code></em> string is specified it must

  456.     consist of a comma-separated list of one or more

  457.     <em class="replaceable"><code>option</code></em><code class="literal">=</code><em class="replaceable"><code>value</code></em> pairs.

  458.     The available options are:

  459. 

  460.     </p><div class="itemizedlist"><ul class="itemizedlist compact" style="list-style-type: bullet; "><li class="listitem" style="list-style-type: disc"><p>

  461.        <code class="literal">MaxWords</code>, <code class="literal">MinWords</code> (integers):

  462.        these numbers determine the longest and shortest headlines to output.

  463.        The default values are 35 and 15.

  464.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  465.        <code class="literal">ShortWord</code> (integer): words of this length or less

  466.        will be dropped at the start and end of a headline, unless they are

  467.        query terms.  The default value of three eliminates common English

  468.        articles.

  469.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  470.        <code class="literal">HighlightAll</code> (boolean): if

  471.        <code class="literal">true</code> the whole document will be used as the

  472.        headline, ignoring the preceding three parameters.  The default

  473.        is <code class="literal">false</code>.

  474.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  475.        <code class="literal">MaxFragments</code> (integer): maximum number of text

  476.        fragments to display.  The default value of zero selects a

  477.        non-fragment-based headline generation method.  A value greater

  478.        than zero selects fragment-based headline generation (see below).

  479.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  480.        <code class="literal">StartSel</code>, <code class="literal">StopSel</code> (strings):

  481.        the strings with which to delimit query words appearing in the

  482.        document, to distinguish them from other excerpted words.  The

  483.        default values are <span class="quote">“<span class="quote"><code class="literal">&lt;b&gt;</code></span>”</span> and

  484.        <span class="quote">“<span class="quote"><code class="literal">&lt;/b&gt;</code></span>”</span>, which can be suitable

  485.        for HTML output.

  486.       </p></li><li class="listitem" style="list-style-type: disc"><p>

  487.        <code class="literal">FragmentDelimiter</code> (string): When more than one

  488.        fragment is displayed, the fragments will be separated by this string.

  489.        The default is <span class="quote">“<span class="quote"><code class="literal"> ... </code></span>”</span>.

  490.       </p></li></ul></div><p>

  491. 

  492.     These option names are recognized case-insensitively.

  493.     You must double-quote string values if they contain spaces or commas.

  494.    </p><p>

  495.     In non-fragment-based headline

  496.     generation, <code class="function">ts_headline</code> locates matches for the

  497.     given <em class="replaceable"><code>query</code></em> and chooses a

  498.     single one to display, preferring matches that have more query words

  499.     within the allowed headline length.

  500.     In fragment-based headline generation, <code class="function">ts_headline</code>

  501.     locates the query matches and splits each match

  502.     into <span class="quote">“<span class="quote">fragments</span>”</span> of no more than <code class="literal">MaxWords</code>

  503.     words each, preferring fragments with more query words, and when

  504.     possible <span class="quote">“<span class="quote">stretching</span>”</span> fragments to include surrounding

  505.     words.  The fragment-based mode is thus more useful when the query

  506.     matches span large sections of the document, or when it's desirable to

  507.     display multiple matches.

  508.     In either mode, if no query matches can be identified, then a single

  509.     fragment of the first <code class="literal">MinWords</code> words in the document

  510.     will be displayed.

  511.    </p><p>

  512.     For example:

  513. 

  514. </p><pre class="screen">

  515. SELECT ts_headline('english',

  516.   'The most common type of search

  517. is to find all documents containing given query terms

  518. and return them in order of their similarity to the

  519. query.',

  520.   to_tsquery('english', 'query &amp; similarity'));

  521.                         ts_headline

  522. ------------------------------------------------------------

  523.  containing given &lt;b&gt;query&lt;/b&gt; terms                       +

  524.  and return them in order of their &lt;b&gt;similarity&lt;/b&gt; to the+

  525.  &lt;b&gt;query&lt;/b&gt;.

  526. 

  527. SELECT ts_headline('english',

  528.   'Search terms may occur

  529. many times in a document,

  530. requiring ranking of the search matches to decide which

  531. occurrences to display in the result.',

  532.   to_tsquery('english', 'search &amp; term'),

  533.   'MaxFragments=10, MaxWords=7, MinWords=3, StartSel=&lt;&lt;, StopSel=&gt;&gt;');

  534.                         ts_headline

  535. ------------------------------------------------------------

  536.  &lt;&lt;Search&gt;&gt; &lt;&lt;terms&gt;&gt; may occur                            +

  537.  many times ... ranking of the &lt;&lt;search&gt;&gt; matches to decide

  538. </pre><p>

  539.    </p><p>

  540.     <code class="function">ts_headline</code> uses the original document, not a

  541.     <code class="type">tsvector</code> summary, so it can be slow and should be used with

  542.     care.

  543.    </p></div></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="textsearch-tables.html">Prev</a> </td><td width="20%" align="center"><a accesskey="u" href="textsearch.html">Up</a></td><td width="40%" align="right"> <a accesskey="n" href="textsearch-features.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">12.2. Tables and Indexes </td><td width="20%" align="center"><a accesskey="h" href="index.html">Home</a></td><td width="40%" align="right" valign="top"> 12.4. Additional Features</td></tr></table></div></body></html>
上海开阖软件有限公司 沪ICP备12045867号-1