php – 将Google搜索查询转换为PostgreSQL“tsquery”
作者：互联网
如何将Google搜索查询转换为可以提供PostgreSQL的to_tsquery()的内容？
如果那里没有现有的库,我应该如何用PHP这样的语言解析Google搜索查询呢？
例如,我想采取以下Google-ish搜索查询：
("used cars" OR "new cars") -ford -mistubishi
并将其转换为to_tsquery() – 友好字符串：
('used cars' | 'new cars') & !ford & !mistubishi
我可以用正则表达式来捏造这个,但这是我能做的最好的.是否有一些强有力的词汇分析方法来解决这个问题？我希望能够支持扩展搜索运算符(如Google的site：和intitle :),它们将应用于不同的数据库字段,因此需要与tsquery字符串分开.
更新：我意识到,对于特殊运算符,这将成为Google到SQL WHERE子句的转换,而不是Google到tsquery的转换.但是WHERE子句可能包含一个或多个tsqueries.
例如,Google风格的查询：
((color:blue OR "4x4") OR style:coupe) -color:red used
应该产生这样的SQL WHERE子句：
WHERE to_tsvector(description) MATCH to_tsquery('used')
AND color <> 'red'
AND ( (color = 'blue' OR to_tsvector(description) MATCH to_tsquery('4x4') )
OR style = 'coupe'
);
我不确定正则表达式是否可以实现上述目标？
解决方法:
老实说,我认为正则表达式是这样的.同样,这是一个有趣的练习.下面的代码是非常原型的 – 事实上,你会发现我甚至没有实现lexer本身 – 我只是伪造了输出.我想继续,但我今天没有更多的业余时间.
此外,在支持其他类型的搜索操作符等方面,肯定还有很多工作要做.
基本上,我们的想法是将某种类型的查询缩小,然后解析为通用格式(在本例中为QueryExpression实例),然后将其作为另一种类型的查询呈现回来.
<?php
ini_set( "display_errors", "on" );
error_reporting( E_ALL );
interface ILexer
{
public function execute( $str );
public function getTokens();
}
interface IParser
{
public function __construct( iLexer $lexer );
public function parse( $input );
public function addToken( $token );
}
class GoogleQueryLexer implements ILexer
{
private $tokenStack = array();
public function execute( $str )
{
$chars = str_split( $str );
foreach ( $chars as $char )
{
// add to self::$tokenStack per your rules
}
//'("used cars" OR "new cars") -ford -mistubishi'
$this->tokenStack = array(
'('
, 'used cars'
, 'or new cars'
, ')'
, '-ford'
, '-mitsubishi'
);
}
public function getTokens()
{
return $this->tokenStack;
}
}
class GoogleQueryParser implements IParser
{
protected $lexer;
public function __construct( iLexer $lexer )
{
$this->lexer = $lexer;
}
public function addToken( $token )
{
$this->tokenStack[] = $token;
}
public function parse( $input )
{
$this->lexer->execute( $input );
$tokens = $this->lexer->getTokens();
$expression = new QueryExpression();
foreach ( $tokens as $token )
{
$expression = $this->processToken( $token, $expression );
}
return $expression;
}
protected function processToken( $token, QueryExpression $expression )
{
switch ( $token )
{
case '(':
return $expression->initiateSubExpression();
break;
case ')':
return $expression->getParentExpression();
break;
default:
$modifier = $token[0];
$phrase = substr( $token, 1 );
switch ( $modifier )
{
case '-':
$expression->addExclusionPhrase( $phrase );
break;
case '+':
$expression->addPhrase( $phrase );
break;
default:
$operator = trim( substr( $token, 0, strpos( $token, ' ' ) ) );
$phrase = trim( substr( $token, strpos( $token, ' ' ) ) );
switch ( strtolower( $operator ) )
{
case 'and':
$expression->addAndPhrase( $phrase );
break;
case 'or':
$expression->addOrPhrase( $phrase );
break;
default:
$expression->addPhrase( $token );
}
}
}
return $expression;
}
}
class QueryExpression
{
protected $phrases = array();
protected $subExpressions = array();
protected $parent;
public function __construct( $parent=null )
{
$this->parent = $parent;
}
public function initiateSubExpression()
{
$expression = new self( $this );
$this->subExpressions[] = $expression;
return $expression;
}
public function getPhrases()
{
return $this->phrases;
}
public function getSubExpressions()
{
return $this->subExpressions;
}
public function getParentExpression()
{
return $this->parent;
}
protected function addQueryPhrase( QueryPhrase $phrase )
{
$this->phrases[] = $phrase;
}
public function addPhrase( $input )
{
$this->addQueryPhrase( new QueryPhrase( $input ) );
}
public function addOrPhrase( $input )
{
$this->addQueryPhrase( new QueryPhrase( $input, QueryPhrase::MODE_OR ) );
}
public function addAndPhrase( $input )
{
$this->addQueryPhrase( new QueryPhrase( $input, QueryPhrase::MODE_AND ) );
}
public function addExclusionPhrase( $input )
{
$this->addQueryPhrase( new QueryPhrase( $input, QueryPhrase::MODE_EXCLUDE ) );
}
}
class QueryPhrase
{
const MODE_DEFAULT = 1;
const MODE_OR = 2;
const MODE_AND = 3;
const MODE_EXCLUDE = 4;
protected $phrase;
protected $mode;
public function __construct( $input, $mode=self::MODE_DEFAULT )
{
$this->phrase = $input;
$this->mode = $mode;
}
public function getMode()
{
return $this->mode;
}
public function __toString()
{
return $this->phrase;
}
}
class TsqueryBuilder
{
protected $expression;
protected $query;
public function __construct( QueryExpression $expression )
{
$this->query = trim( $this->processExpression( $expression ), ' &|' );
}
public function getResult()
{
return $this->query;
}
protected function processExpression( QueryExpression $expression )
{
$query = '';
$phrases = $expression->getPhrases();
$subExpressions = $expression->getSubExpressions();
foreach ( $phrases as $phrase )
{
$format = "'%s' ";
switch ( $phrase->getMode() )
{
case QueryPhrase::MODE_AND :
$format = "& '%s' ";
break;
case QueryPhrase::MODE_OR :
$format = "| '%s' ";
break;
case QueryPhrase::MODE_EXCLUDE :
$format = "& !'%s' ";
break;
}
$query .= sprintf( $format, str_replace( "'", "\\'", $phrase ) );
}
foreach ( $subExpressions as $subExpression )
{
$query .= "& (" . $this->processExpression( $subExpression ) . ")";
}
return $query;
}
}
$parser = new GoogleQueryParser( new GoogleQueryLexer() );
$queryBuilder = new TsqueryBuilder( $parser->parse( '("used cars" OR "new cars") -ford -mistubishi' ) );
echo $queryBuilder->getResult();
标签：php,parsing,postgresql,full-text-search,google-search 来源： https://codeday.me/bug/20190701/1345303.html