repo_zf1/documentation/manual/en/ref/performance-localization.xml

<?xml version="1.0" encoding="UTF-8"?>
<!-- Reviewed: no -->
<sect1 id="performance.localization">
    <title>Internationalization (i18n) and Localization (l10n)</title>

    <para>
        Internationalizing and localizing a site are fantastic ways to expand
        your audience and ensure that all visitors can get to the information
        they need. However, it often comes with a performance penalty. Below
        are some strategies you can employ to reduce the overhead of i18n and
        l10n.
    </para>

    <sect2 id="performance.localization.translationadapter">
        <title>Which translation adapter should I use?</title>

        <para>
            Not all translation adapters are made equal. Some have more
            features than others, and some perform better than others.
            Additionally, you may have business requirements that force you to
            use a particular adapter. However, if you have a choice, which
            adapters are fastest?
        </para>

        <sect3 id="performance.localization.translationadapter.fastest">
            <title>Use non-XML translation adapters for greatest speed</title>

            <para>
                Zend Framework ships with a variety of translation adapters.
                Fully half of them utilize an <acronym>XML</acronym> format, incurring memory and
                performance overhead. Fortunately, there are several adapters
                that utilize other formats that can be parsed much more
                quickly. In order of speed, from fastest to slowest, they are:
            </para>

            <itemizedlist>
                <listitem><para>
                    <emphasis>Array</emphasis>: this is the fastest, as it is,
                    by definition, parsed into a native <acronym>PHP</acronym> format immediately
                    on inclusion.
                </para></listitem>

                <listitem><para>
                    <emphasis><acronym>CSV</acronym></emphasis>: uses
                    <methodname>fgetcsv()</methodname> to parse a <acronym>CSV</acronym> file and
                    transform it into a native <acronym>PHP</acronym> format.
                </para></listitem>

                <listitem><para>
                    <emphasis><acronym>INI</acronym></emphasis>: uses
                    <methodname>parse_ini_file()</methodname> to parse an <acronym>INI</acronym>
                    file and transform it into a native <acronym>PHP</acronym> format. This and the
                    <acronym>CSV</acronym> adapter are roughly equivalent performance-wise.
                </para></listitem>

                <listitem><para>
                    <emphasis>Gettext</emphasis>: The gettext adapter from Zend Framework
                    does <emphasis>not</emphasis> use the gettext
                    extension as it is not thread safe and does not allow
                    specifying more than one locale per server. As a result, it
                    is slower than using the gettext extension directly, but,
                    because the gettext format is binary, it's faster to parse
                    than <acronym>XML</acronym>.
                </para></listitem>
            </itemizedlist>

            <para>
                If high performance is one of your concerns, we suggest
                utilizing one of the above adapters.
            </para>
        </sect3>
    </sect2>

    <sect2 id="performance.localization.cache">
        <title>How can I make translation and localization even faster?</title>

        <para>
            Maybe, for business reasons, you're limited to an <acronym>XML</acronym>-based
            translation adapter. Or perhaps you'd like to speed things up even
            more. Or perhaps you want to make l10n operations faster. How can
            you do this?
        </para>

        <sect3 id="performance.localization.cache.usage">
            <title>Use translation and localization caches</title>

            <para>
                Both <classname>Zend_Translate</classname> and <classname>Zend_Locale</classname>
                implement caching functionality that can greatly affect
                performance. In the case of each, the major bottleneck is
                typically reading the files, not the actual lookups; using a
                cache eliminates the need to read the translation and/or
                localization files.
            </para>

            <para>
                You can read about caching of translation and localization
                strings in the following locations:
            </para>

            <itemizedlist>
                <listitem>
                    <para>
                        <link linkend="zend.translate.adapter.caching"><classname>
                        Zend_Translate</classname> adapter caching</link>
                    </para>
                </listitem>

                <listitem>
                    <para>
                        <link linkend="zend.locale.cache"><classname>Zend_Locale</classname>
                        caching</link>
                    </para>
                </listitem>
            </itemizedlist>
        </sect3>
    </sect2>
</sect1>
<!--
vim:se ts=4 sw=4 et:
-->