repo_zf1/documentation/manual/en/tutorials/lucene-intro.xml

<?xml version="1.0" encoding="UTF-8"?>
<!-- Reviewed: no -->
<sect1 id="learning.lucene.intro">
    <title>Zend_Search_Lucene Introduction</title>

    <para>
        The <classname>Zend_Search_Lucene</classname> component is intended to provide a
        ready-for-use full-text search solution. It doesn't require any <acronym>PHP</acronym>
        extensions<footnote><para>Though some <acronym>UTF-8</acronym> processing functionality
                requires the <emphasis>mbstring</emphasis> extension to be turned
                on</para></footnote> or additional software to be installed, and can be used
        immediately after Zend Framework installation.
    </para>

    <para>
        <classname>Zend_Search_Lucene</classname> is a pure <acronym>PHP</acronym> port of the
        popular open source full-text search engine known as Apache Lucene. See <ulink
            url="http://lucene.apache.org">http://lucene.apache.org/</ulink> for the details.
    </para>

    <para>
        Information must be indexed to be available for searching.
        <classname>Zend_Search_Lucene</classname> and Java Lucene use a document concept known as an
        "atomic indexing item."
    </para>

    <para>
        Each document is a set of fields: &lt;name, value&gt; pairs where name and value are
        <acronym>UTF-8</acronym> strings<footnote><para>Binary strings are also allowed to be used
                as field values</para></footnote>. Any subset of the document fields may be marked
        as "indexed" to include field data in the text indexing process.
    </para>

    <para>
        Field values may or may not be tokenized while indexing. If a field is not tokenized, then
        the field value is stored as one term; otherwise, the current analyzer is used for
        tokenization.
    </para>

    <para>
        Several analyzers are provided within the <classname>Zend_Search_Lucene</classname> package.
        The default analyzer works with <acronym>ASCII</acronym> text (since the
        <acronym>UTF-8</acronym> analyzer needs the <emphasis>mbstring</emphasis> extension to be
        turned on). It is case insensitive, and it skips numbers. Use other analyzers or create your
        own analyzer if you need to change this behavior.
    </para>

    <note>
        <title>Using analyzers during indexing and searching</title>

        <para>
            Important note! Search queries are also tokenized using the "current analyzer", so the
            same analyzer must be set as the default during both the indexing and searching process.
            This will guarantee that source and searched text will be transformed into terms in the
            same way.
        </para>
    </note>

    <para>
        Field values are optionally stored within an index. This allows the original field data to
        be retrieved from the index while searching. This is the only way to associate search
        results with the original data (internal document IDs may be changed after index
        optimization or auto-optimization).
    </para>

    <para>
        The thing that should be remembered is that a Lucene index is not a database. It doesn't
        provide index backup mechanisms except backup of the file system directory. It doesn't
        provide transactional mechanisms though concurrent index update as well as concurrent update
        and read are supported. It doesn't compare with databases in data retrieving speed.
    </para>

    <para>
        So it's good idea:
    </para>

    <itemizedlist>
        <listitem>
            <para>
                <emphasis>Not</emphasis> to use Lucene index as a storage since it may dramatically
                decrease search hit retrieving performance. Store only unique document identifiers
                (doc paths, <acronym>URL</acronym>s, database unique IDs) and associated data within
                an index. E.g. title, annotation, category, language info, avatar. (Note: a field
                may be included in indexing, but not stored, or stored, but not indexed).
            </para>
        </listitem>

        <listitem>
            <para>
                To write functionality that can rebuild an index completely if it's corrupted for
                any reason.
            </para>
        </listitem>
    </itemizedlist>

    <para>
        Individual documents in the index may have completely different sets of fields. The same
        fields in different documents don't need to have the same attributes. E.g. a field may be
        indexed for one document and skipped from indexing for another. The same applies for
        storing, tokenizing, or treating field value as a binary string.
    </para>
</sect1>