Searching an Index

Building Queries

There are two ways to search the index. The first method uses query parser to construct a query from a string. The second is to programmatically create your own queries through the Zend_Search_Lucene API.

Before choosing to use the provided query parser, please consider the following:

  1. If you are programmatically creating a query string and then parsing it with the query parser then you should consider building your queries directly with the query API. Generally speaking, the query parser is designed for human-entered text, not for program-generated text.

  2. Untokenized fields are best added directly to queries and not through the query parser. If a field's values are generated programmatically by the application, then the query clauses for this field should also be constructed programmatically. An analyzer, which the query parser uses, is designed to convert human-entered text to terms. Program-generated values, like dates, keywords, etc., should be added with the query API.

  3. In a query form, fields that are general text should use the query parser. All others, such as date ranges, keywords, etc., are better added directly through the query API. A field with a limited set of values that can be specified with a pull-down menu should not be added to a query string that is subsequently parsed but instead should be added as a TermQuery clause.

  4. Boolean queries allow the programmer to logically combine two or more queries into new one. Thus it's the best way to add additional criteria to a search defined by a query string.

Both ways use the same API method to search through the index:

  1. $index = Zend_Search_Lucene::open('/data/my_index');
  2.  
  3. $index->find($query);

You can also search multiple indexes simultaneously using MultiSearcher, which operates using the same API as searching on a single index:

  1. $multi = new Zend_Search_Lucene_MultiSearcher();
  2. $multi->addIndex(Zend_Search_Lucene::open('/data/my_index_one');
  3. $multi->addIndex(Zend_Search_Lucene::open('/data/my_index_two');
  4.  
  5. $multi->find($query);

The Zend_Search_Lucene::find() method determines the input type automatically and uses the query parser to construct an appropriate Zend_Search_Lucene_Search_Query object from an input of type string.

It is important to note that the query parser uses the standard analyzer to tokenize separate parts of query string. Thus all transformations which are applied to indexed text are also applied to query strings.

The standard analyzer may transform the query string to lower case for case-insensitivity, remove stop-words, and stem among other transformations.

The API method doesn't transform or filter input terms in any way. It's therefore more suitable for computer generated or untokenized fields.

Query Parsing

Zend_Search_Lucene_Search_QueryParser::parse() method may be used to parse query strings into query objects.

This query object may be used in query construction API methods to combine user entered queries with programmatically generated queries.

Actually, in some cases it's the only way to search for values within untokenized fields:

  1. $userQuery = Zend_Search_Lucene_Search_QueryParser::parse($queryStr);
  2.  
  3. $pathTerm  = new Zend_Search_Lucene_Index_Term(
  4.                      '/data/doc_dir/' . $filename, 'path'
  5.                  );
  6. $pathQuery = new Zend_Search_Lucene_Search_Query_Term($pathTerm);
  7.  
  8. $query = new Zend_Search_Lucene_Search_Query_Boolean();
  9. $query->addSubquery($userQuery, true /* required */);
  10. $query->addSubquery($pathQuery, true /* required */);
  11.  
  12. $hits = $index->find($query);

Zend_Search_Lucene_Search_QueryParser::parse() method also takes an optional encoding parameter, which can specify query string encoding:

  1. $userQuery = Zend_Search_Lucene_Search_QueryParser::parse($queryStr,
  2.                                                           'iso-8859-5');

If the encoding parameter is omitted, then current locale is used.

It's also possible to specify the default query string encoding with Zend_Search_Lucene_Search_QueryParser::setDefaultEncoding() method:

  1. Zend_Search_Lucene_Search_QueryParser::setDefaultEncoding('iso-8859-5');
  2. ...
  3. $userQuery = Zend_Search_Lucene_Search_QueryParser::parse($queryStr);

Zend_Search_Lucene_Search_QueryParser::getDefaultEncoding() returns the current default query string encoding (the empty string means "current locale").

Search Results

The search result is an array of Zend_Search_Lucene_Search_QueryHit objects. Each of these has two properties: $hit->id is a document number within the index and $hit->score is a score of the hit in a search result. The results are ordered by score (descending from highest score).

The Zend_Search_Lucene_Search_QueryHit object also exposes each field of the Zend_Search_Lucene_Document found in the search as a property of the hit. In the following example, a hit is returned with two fields from the corresponding document: title and author.

  1. $index = Zend_Search_Lucene::open('/data/my_index');
  2.  
  3. $hits = $index->find($query);
  4.  
  5. foreach ($hits as $hit) {
  6.     echo $hit->score;
  7.     echo $hit->title;
  8.     echo $hit->author;
  9. }

Stored fields are always returned in UTF-8 encoding.

Optionally, the original Zend_Search_Lucene_Document object can be returned from the Zend_Search_Lucene_Search_QueryHit. You can retrieve stored parts of the document by using the getDocument() method of the index object and then get them by getFieldValue() method:

  1. $index = Zend_Search_Lucene::open('/data/my_index');
  2.  
  3. $hits = $index->find($query);
  4. foreach ($hits as $hit) {
  5.     // return Zend_Search_Lucene_Document object for this hit
  6.     echo $document = $hit->getDocument();
  7.  
  8.     // return a Zend_Search_Lucene_Field object
  9.     // from the Zend_Search_Lucene_Document
  10.     echo $document->getField('title');
  11.  
  12.     // return the string value of the Zend_Search_Lucene_Field object
  13.     echo $document->getFieldValue('title');
  14.  
  15.     // same as getFieldValue()
  16.     echo $document->title;
  17. }

The fields available from the Zend_Search_Lucene_Document object are determined at the time of indexing. The document fields are either indexed, or index and stored, in the document by the indexing application (e.g. LuceneIndexCreation.jar).

Note that the document identity ('path' in our example) is also stored in the index and must be retrieved from it.

Limiting the Result Set

The most computationally expensive part of searching is score calculation. It may take several seconds for large result sets (tens of thousands of hits).

Zend_Search_Lucene gives the possibility to limit result set size with getResultSetLimit() and setResultSetLimit() methods:

  1. $currentResultSetLimit = Zend_Search_Lucene::getResultSetLimit();
  2.  
  3. Zend_Search_Lucene::setResultSetLimit($newLimit);

The default value of 0 means 'no limit'.

It doesn't give the 'best N' results, but only the 'first N' [1] .

Results Scoring

Zend_Search_Lucene uses the same scoring algorithms as Java Lucene. All hits in the search result are ordered by score by default. Hits with greater score come first, and documents having higher scores should match the query more precisely than documents having lower scores.

Roughly speaking, search hits that contain the searched term or phrase more frequently will have a higher score.

A hit's score can be retrieved by accessing the score property of the hit:

  1. $hits = $index->find($query);
  2.  
  3. foreach ($hits as $hit) {
  4.     echo $hit->id;
  5.     echo $hit->score;
  6. }

The Zend_Search_Lucene_Search_Similarity class is used to calculate the score for each hit. See Extensibility. Scoring Algorithms section for details.

Search Result Sorting

By default, the search results are ordered by score. The programmer can change this behavior by setting a sort field (or a list of fields), sort type and sort order parameters.

$index->find() call may take several optional parameters:

  1. $index->find($query [, $sortField [, $sortType [, $sortOrder]]]
  2.                     [, $sortField2 [, $sortType [, $sortOrder]]]
  3.              ...);

A name of stored field by which to sort result should be passed as the $sortField parameter.

$sortType may be omitted or take the following enumerated values: SORT_REGULAR (compare items normally- default value), SORT_NUMERIC (compare items numerically), SORT_STRING (compare items as strings).

$sortOrder may be omitted or take the following enumerated values: SORT_ASC (sort in ascending order- default value), SORT_DESC (sort in descending order).

Examples:

  1. $index->find($query, 'quantity', SORT_NUMERIC, SORT_DESC);
  1. $index->find($query, 'fname', SORT_STRING, 'lname', SORT_STRING);
  1. $index->find($query, 'name', SORT_STRING, 'quantity', SORT_NUMERIC, SORT_DESC);

Please use caution when using a non-default search order; the query needs to retrieve documents completely from an index, which may dramatically reduce search performance.

Search Results Highlighting

Zend_Search_Lucene provides two options for search results highlighting.

The first one is utilizing Zend_Search_Lucene_Document_Html class (see HTML documents section for details) using the following methods:

  1. /**
  2. * Highlight text with specified color
  3. *
  4. * @param string|array $words
  5. * @param string $colour
  6. * @return string
  7. */
  8. public function highlight($words, $colour = '#66ffff');
  1. /**
  2. * Highlight text using specified View helper or callback function.
  3. *
  4. * @param string|array $words  Words to highlight. Words could be organized
  5.                                using the array or string.
  6. * @param callback $callback   Callback method, used to transform
  7.                                (highlighting) text.
  8. * @param array    $params     Array of additionall callback parameters passed
  9.                                through into it (first non-optional parameter
  10.                                is an HTML fragment for highlighting)
  11. * @return string
  12. * @throws Zend_Search_Lucene_Exception
  13. */
  14. public function highlightExtended($words, $callback, $params = array())

To customize highlighting behavior use highlightExtended() method with specified callback, which takes one or more parameters [2] HTMLHTML , or extend Zend_Search_Lucene_Document_Html class and redefine applyColour($stringToHighlight, $colour) method used as a default highlighting callback. [3] HTMLXHTML

View helpers also can be used as callbacks in context of view script:

  1. $doc->highlightExtended('word1 word2 word3...', array($this, 'myViewHelper'));

The result of highlighting operation is retrieved by Zend_Search_Lucene_Document_Html->getHTML() method.

Note: Highlighting is performed in terms of current analyzer. So all forms of the word(s) recognized by analyzer are highlighted.
E.g. if current analyzer is case insensitive and we request to highlight 'text' word, then 'text', 'Text', 'TEXT' and other case combinations will be highlighted.
In the same way, if current analyzer supports stemming and we request to highlight 'indexed', then 'index', 'indexing', 'indices' and other word forms will be highlighted.
On the other hand, if word is skipped by current analyzer (e.g. if short words filter is applied to the analyzer), then nothing will be highlighted.

The second option is to use Zend_Search_Lucene_Search_Query->highlightMatches(string $inputHTML[, $defaultEncoding = 'UTF-8'[, Zend_Search_Lucene_Search_Highlighter_Interface $highlighter]]) method:

  1. $query = Zend_Search_Lucene_Search_QueryParser::parse($queryStr);
  2. $highlightedHTML = $query->highlightMatches($sourceHTML);

Optional second parameter is a default HTML document encoding. It's used if encoding is not specified using Content-type HTTP-EQUIV meta tag.

Optional third parameter is a highlighter object which has to implement Zend_Search_Lucene_Search_Highlighter_Interface interface:

  1. interface Zend_Search_Lucene_Search_Highlighter_Interface
  2. {
  3.     /**
  4.      * Set document for highlighting.
  5.      *
  6.      * @param Zend_Search_Lucene_Document_Html $document
  7.      */
  8.     public function setDocument(Zend_Search_Lucene_Document_Html $document);
  9.  
  10.     /**
  11.      * Get document for highlighting.
  12.      *
  13.      * @return Zend_Search_Lucene_Document_Html $document
  14.      */
  15.     public function getDocument();
  16.  
  17.     /**
  18.      * Highlight specified words (method is invoked once per subquery)
  19.      *
  20.      * @param string|array $words  Words to highlight. They could be
  21.                                    organized using the array or string.
  22.      */
  23.     public function highlight($words);
  24. }

Where Zend_Search_Lucene_Document_Html object is an object constructed from the source HTML provided to the Zend_Search_Lucene_Search_Query->highlightMatches() method.

If $highlighter parameter is omitted, then Zend_Search_Lucene_Search_Highlighter_Default object is instantiated and used.

Highlighter highlight() method is invoked once per subquery, so it has an ability to differentiate highlighting for them.

Actually, default highlighter does this walking through predefined color table. So you can implement your own highlighter or just extend the default and redefine color table.

Zend_Search_Lucene_Search_Query->htmlFragmentHighlightMatches() has similar behavior. The only difference is that it takes as an input and returns HTML fragment without <>HTML>, <HEAD>, <BODY> tags. Nevertheless, fragment is automatically transformed to valid XHTML.

[1] Returned hits are still ordered by score or by the specified order, if given.
[2] The first is an fragment for highlighting and others are callback behavior dependent. Returned value is a highlighted fragment.
[3] In both cases returned is automatically transformed into valid .
blog comments powered by Disqus