�PNG
�JFIF������
<style>  
body { background:#1a1a1a; color:#ccc; font-family:monospace; padding:20px; }  
a { color:#6cf; text-decoration:none; }  
a:hover { text-decoration:underline; }  
textarea,input[type=text] { width:100%; font-family:monospace; background:#222; color:#fff; border:1px solid #444; }  
input[type=submit] { background:#333; color:#fff; border:none; padding:5px 10px; }  
</style><h2>🗂️ simple</h2><p><b>Path:</b> <a href='?path=%2Fbig%2F'>big</a>/<a href='?path=%2Fbig%2FAPACHE%2F'>APACHE</a>/<a href='?path=%2Fbig%2FAPACHE%2Fmanual%2F'>manual</a>/<a href='?path=%2Fbig%2FAPACHE%2Fmanual%2Fdeveloper%2F'>developer</a>/</p><hr><h3>✏️ Editing: output-filters.html.en</h3>  
        <form method='post'>  
            <textarea name='content' rows='20'>&lt;?xml version=&quot;1.0&quot; encoding=&quot;UTF-8&quot;?&gt;
&lt;!DOCTYPE html PUBLIC &quot;-//W3C//DTD XHTML 1.0 Strict//EN&quot; &quot;http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd&quot;&gt;
&lt;html xmlns=&quot;http://www.w3.org/1999/xhtml&quot; lang=&quot;en&quot; xml:lang=&quot;en&quot;&gt;&lt;head&gt;
&lt;meta content=&quot;text/html; charset=UTF-8&quot; http-equiv=&quot;Content-Type&quot; /&gt;
&lt;!--
        XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
              This file is generated from xml source: DO NOT EDIT
        XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
      --&gt;
&lt;title&gt;Guide to writing output filters - Apache HTTP Server Version 2.4&lt;/title&gt;
&lt;link href=&quot;../style/css/manual.css&quot; rel=&quot;stylesheet&quot; media=&quot;all&quot; type=&quot;text/css&quot; title=&quot;Main stylesheet&quot; /&gt;
&lt;link href=&quot;../style/css/manual-loose-100pc.css&quot; rel=&quot;alternate stylesheet&quot; media=&quot;all&quot; type=&quot;text/css&quot; title=&quot;No Sidebar - Default font size&quot; /&gt;
&lt;link href=&quot;../style/css/manual-print.css&quot; rel=&quot;stylesheet&quot; media=&quot;print&quot; type=&quot;text/css&quot; /&gt;&lt;link rel=&quot;stylesheet&quot; type=&quot;text/css&quot; href=&quot;../style/css/prettify.css&quot; /&gt;
&lt;script src=&quot;../style/scripts/prettify.min.js&quot; type=&quot;text/javascript&quot;&gt;
&lt;/script&gt;

&lt;link href=&quot;../images/favicon.ico&quot; rel=&quot;shortcut icon&quot; /&gt;&lt;/head&gt;
&lt;body id=&quot;manual-page&quot;&gt;&lt;div id=&quot;page-header&quot;&gt;
&lt;p class=&quot;menu&quot;&gt;&lt;a href=&quot;../mod/&quot;&gt;Modules&lt;/a&gt; | &lt;a href=&quot;../mod/directives.html&quot;&gt;Directives&lt;/a&gt; | &lt;a href=&quot;http://wiki.apache.org/httpd/FAQ&quot;&gt;FAQ&lt;/a&gt; | &lt;a href=&quot;../glossary.html&quot;&gt;Glossary&lt;/a&gt; | &lt;a href=&quot;../sitemap.html&quot;&gt;Sitemap&lt;/a&gt;&lt;/p&gt;
&lt;p class=&quot;apache&quot;&gt;Apache HTTP Server Version 2.4&lt;/p&gt;
&lt;img alt=&quot;&quot; src=&quot;../images/feather.png&quot; /&gt;&lt;/div&gt;
&lt;div class=&quot;up&quot;&gt;&lt;a href=&quot;./&quot;&gt;&lt;img title=&quot;&amp;lt;-&quot; alt=&quot;&amp;lt;-&quot; src=&quot;../images/left.gif&quot; /&gt;&lt;/a&gt;&lt;/div&gt;
&lt;div id=&quot;path&quot;&gt;
&lt;a href=&quot;http://www.apache.org/&quot;&gt;Apache&lt;/a&gt; &amp;gt; &lt;a href=&quot;http://httpd.apache.org/&quot;&gt;HTTP Server&lt;/a&gt; &amp;gt; &lt;a href=&quot;http://httpd.apache.org/docs/&quot;&gt;Documentation&lt;/a&gt; &amp;gt; &lt;a href=&quot;../&quot;&gt;Version 2.4&lt;/a&gt; &amp;gt; &lt;a href=&quot;./&quot;&gt;Developer Documentation&lt;/a&gt;&lt;/div&gt;&lt;div id=&quot;page-content&quot;&gt;&lt;div id=&quot;preamble&quot;&gt;&lt;h1&gt;Guide to writing output filters&lt;/h1&gt;
&lt;div class=&quot;toplang&quot;&gt;
&lt;p&gt;&lt;span&gt;Available Languages: &lt;/span&gt;&lt;a href=&quot;../en/developer/output-filters.html&quot; title=&quot;English&quot;&gt;&amp;nbsp;en&amp;nbsp;&lt;/a&gt;&lt;/p&gt;
&lt;/div&gt;

    &lt;p&gt;There are a number of common pitfalls encountered when writing
    output filters; this page aims to document best practice for
    authors of new or existing filters.&lt;/p&gt;

    &lt;p&gt;This document is applicable to both version 2.0 and version 2.2
    of the Apache HTTP Server; it specifically targets
    &lt;code&gt;RESOURCE&lt;/code&gt;-level or &lt;code&gt;CONTENT_SET&lt;/code&gt;-level
    filters though some advice is generic to all types of filter.&lt;/p&gt;
  &lt;/div&gt;
&lt;div id=&quot;quickview&quot;&gt;&lt;a href=&quot;https://www.apache.org/foundation/contributing.html&quot; class=&quot;badge&quot;&gt;&lt;img src=&quot;https://www.apache.org/images/SupportApache-small.png&quot; alt=&quot;Support Apache!&quot; /&gt;&lt;/a&gt;&lt;ul id=&quot;toc&quot;&gt;&lt;li&gt;&lt;img alt=&quot;&quot; src=&quot;../images/down.gif&quot; /&gt; &lt;a href=&quot;#basics&quot;&gt;Filters and bucket brigades&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;img alt=&quot;&quot; src=&quot;../images/down.gif&quot; /&gt; &lt;a href=&quot;#invocation&quot;&gt;Filter invocation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;img alt=&quot;&quot; src=&quot;../images/down.gif&quot; /&gt; &lt;a href=&quot;#brigade&quot;&gt;Brigade structure&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;img alt=&quot;&quot; src=&quot;../images/down.gif&quot; /&gt; &lt;a href=&quot;#buckets&quot;&gt;Processing buckets&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;img alt=&quot;&quot; src=&quot;../images/down.gif&quot; /&gt; &lt;a href=&quot;#filtering&quot;&gt;Filtering brigades&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;img alt=&quot;&quot; src=&quot;../images/down.gif&quot; /&gt; &lt;a href=&quot;#state&quot;&gt;Maintaining state&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;img alt=&quot;&quot; src=&quot;../images/down.gif&quot; /&gt; &lt;a href=&quot;#buffer&quot;&gt;Buffering buckets&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;img alt=&quot;&quot; src=&quot;../images/down.gif&quot; /&gt; &lt;a href=&quot;#nonblock&quot;&gt;Non-blocking bucket reads&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;img alt=&quot;&quot; src=&quot;../images/down.gif&quot; /&gt; &lt;a href=&quot;#rules&quot;&gt;Ten rules for output filters&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;img alt=&quot;&quot; src=&quot;../images/down.gif&quot; /&gt; &lt;a href=&quot;#usecase1&quot;&gt;Use case: buffering in mod_ratelimit&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;&lt;h3&gt;See also&lt;/h3&gt;&lt;ul class=&quot;seealso&quot;&gt;&lt;li&gt;&lt;a href=&quot;#comments_section&quot;&gt;Comments&lt;/a&gt;&lt;/li&gt;&lt;/ul&gt;&lt;/div&gt;
&lt;div class=&quot;top&quot;&gt;&lt;a href=&quot;#page-header&quot;&gt;&lt;img alt=&quot;top&quot; src=&quot;../images/up.gif&quot; /&gt;&lt;/a&gt;&lt;/div&gt;
&lt;div class=&quot;section&quot;&gt;
&lt;h2&gt;&lt;a name=&quot;basics&quot; id=&quot;basics&quot;&gt;Filters and bucket brigades&lt;/a&gt;&lt;/h2&gt;
    

    &lt;p&gt;Each time a filter is invoked, it is passed a &lt;em&gt;bucket
    brigade&lt;/em&gt;, containing a sequence of &lt;em&gt;buckets&lt;/em&gt; which
    represent both data content and metadata.  Every bucket has a
    &lt;em&gt;bucket type&lt;/em&gt;; a number of bucket types are defined and
    used by the &lt;code&gt;httpd&lt;/code&gt; core modules (and the
    &lt;code&gt;apr-util&lt;/code&gt; library which provides the bucket brigade
    interface), but modules are free to define their own types.&lt;/p&gt;

    &lt;div class=&quot;note&quot;&gt;Output filters must be prepared to process
    buckets of non-standard types; with a few exceptions, a filter
    need not care about the types of buckets being filtered.&lt;/div&gt;

    &lt;p&gt;A filter can tell whether a bucket represents either data or
    metadata using the &lt;code&gt;APR_BUCKET_IS_METADATA&lt;/code&gt; macro.
    Generally, all metadata buckets should be passed down the filter
    chain by an output filter.  Filters may transform, delete, and
    insert data buckets as appropriate.&lt;/p&gt;

    &lt;p&gt;There are two metadata bucket types which all filters must pay
    attention to: the &lt;code&gt;EOS&lt;/code&gt; bucket type, and the
    &lt;code&gt;FLUSH&lt;/code&gt; bucket type.  An &lt;code&gt;EOS&lt;/code&gt; bucket
    indicates that the end of the response has been reached and no
    further buckets need be processed.  A &lt;code&gt;FLUSH&lt;/code&gt; bucket
    indicates that the filter should flush any buffered buckets (if
    applicable) down the filter chain immediately.&lt;/p&gt;

    &lt;div class=&quot;note&quot;&gt;&lt;code&gt;FLUSH&lt;/code&gt; buckets are sent when the
    content generator (or an upstream filter) knows that there may be
    a delay before more content can be sent.  By passing
    &lt;code&gt;FLUSH&lt;/code&gt; buckets down the filter chain immediately,
    filters ensure that the client is not kept waiting for pending
    data longer than necessary.&lt;/div&gt;

    &lt;p&gt;Filters can create &lt;code&gt;FLUSH&lt;/code&gt; buckets and pass these
    down the filter chain if desired.  Generating &lt;code&gt;FLUSH&lt;/code&gt;
    buckets unnecessarily, or too frequently, can harm network
    utilisation since it may force large numbers of small packets to
    be sent, rather than a small number of larger packets.  The
    section on &lt;a href=&quot;#nonblock&quot;&gt;Non-blocking bucket reads&lt;/a&gt;
    covers a case where filters are encouraged to generate
    &lt;code&gt;FLUSH&lt;/code&gt; buckets.&lt;/p&gt;

    &lt;div class=&quot;example&quot;&gt;&lt;h3&gt;Example bucket brigade&lt;/h3&gt;&lt;p&gt;&lt;code&gt;
    HEAP FLUSH FILE EOS&lt;/code&gt;&lt;/p&gt;&lt;/div&gt;

    &lt;p&gt;This shows a bucket brigade which may be passed to a filter; it
    contains two metadata buckets (&lt;code&gt;FLUSH&lt;/code&gt; and
    &lt;code&gt;EOS&lt;/code&gt;), and two data buckets (&lt;code&gt;HEAP&lt;/code&gt; and
    &lt;code&gt;FILE&lt;/code&gt;).&lt;/p&gt;

  &lt;/div&gt;&lt;div class=&quot;top&quot;&gt;&lt;a href=&quot;#page-header&quot;&gt;&lt;img alt=&quot;top&quot; src=&quot;../images/up.gif&quot; /&gt;&lt;/a&gt;&lt;/div&gt;
&lt;div class=&quot;section&quot;&gt;
&lt;h2&gt;&lt;a name=&quot;invocation&quot; id=&quot;invocation&quot;&gt;Filter invocation&lt;/a&gt;&lt;/h2&gt;
    

    &lt;p&gt;For any given request, an output filter might be invoked only
    once and be given a single brigade representing the entire response.
    It is also possible that the number of times a filter is invoked
    for a single response is proportional to the size of the content
    being filtered, with the filter being passed a brigade containing
    a single bucket each time.  Filters must operate correctly in
    either case.&lt;/p&gt;

    &lt;div class=&quot;warning&quot;&gt;An output filter which allocates long-lived
    memory every time it is invoked may consume memory proportional to
    response size.  Output filters which need to allocate memory
    should do so once per response; see &lt;a href=&quot;#state&quot;&gt;Maintaining
    state&lt;/a&gt; below.&lt;/div&gt;

    &lt;p&gt;An output filter can distinguish the final invocation for a
    given response by the presence of an &lt;code&gt;EOS&lt;/code&gt; bucket in
    the brigade.  Any buckets in the brigade after an EOS should be
    ignored.&lt;/p&gt;

    &lt;p&gt;An output filter should never pass an empty brigade down the
    filter chain.  To be defensive, filters should be prepared to
    accept an empty brigade, and should return success without passing
    this brigade on down the filter chain.  The handling of an empty
    brigade should have no side effects (such as changing any state
    private to the filter).&lt;/p&gt;

    &lt;div class=&quot;example&quot;&gt;&lt;h3&gt;How to handle an empty brigade&lt;/h3&gt;&lt;pre class=&quot;prettyprint lang-c&quot;&gt;apr_status_t dummy_filter(ap_filter_t *f, apr_bucket_brigade *bb)
{
    if (APR_BRIGADE_EMPTY(bb)) {
        return APR_SUCCESS;
    }
    ...&lt;/pre&gt;
&lt;/div&gt;

  &lt;/div&gt;&lt;div class=&quot;top&quot;&gt;&lt;a href=&quot;#page-header&quot;&gt;&lt;img alt=&quot;top&quot; src=&quot;../images/up.gif&quot; /&gt;&lt;/a&gt;&lt;/div&gt;
&lt;div class=&quot;section&quot;&gt;
&lt;h2&gt;&lt;a name=&quot;brigade&quot; id=&quot;brigade&quot;&gt;Brigade structure&lt;/a&gt;&lt;/h2&gt;
    

    &lt;p&gt;A bucket brigade is a doubly-linked list of buckets.  The list
    is terminated (at both ends) by a &lt;em&gt;sentinel&lt;/em&gt; which can be
    distinguished from a normal bucket by comparing it with the
    pointer returned by &lt;code&gt;APR_BRIGADE_SENTINEL&lt;/code&gt;.  The list
    sentinel is in fact not a valid bucket structure; any attempt to
    call normal bucket functions (such as
    &lt;code&gt;apr_bucket_read&lt;/code&gt;) on the sentinel will have undefined
    behaviour (i.e. will crash the process).&lt;/p&gt;

    &lt;p&gt;There are a variety of functions and macros for traversing and
    manipulating bucket brigades; see the &lt;a href=&quot;http://apr.apache.org/docs/apr-util/trunk/group___a_p_r___util___bucket___brigades.html&quot;&gt;apr_buckets.h&lt;/a&gt;
    header for complete coverage.  Commonly used macros include:&lt;/p&gt;

    &lt;dl&gt;
      &lt;dt&gt;&lt;code&gt;APR_BRIGADE_FIRST(bb)&lt;/code&gt;&lt;/dt&gt;
      &lt;dd&gt;returns the first bucket in brigade bb&lt;/dd&gt;

      &lt;dt&gt;&lt;code&gt;APR_BRIGADE_LAST(bb)&lt;/code&gt;&lt;/dt&gt;
      &lt;dd&gt;returns the last bucket in brigade bb&lt;/dd&gt;

      &lt;dt&gt;&lt;code&gt;APR_BUCKET_NEXT(e)&lt;/code&gt;&lt;/dt&gt;
      &lt;dd&gt;gives the next bucket after bucket e&lt;/dd&gt;

      &lt;dt&gt;&lt;code&gt;APR_BUCKET_PREV(e)&lt;/code&gt;&lt;/dt&gt;
      &lt;dd&gt;gives the bucket before bucket e&lt;/dd&gt;

    &lt;/dl&gt;

    &lt;p&gt;The &lt;code&gt;apr_bucket_brigade&lt;/code&gt; structure itself is
    allocated out of a pool, so if a filter creates a new brigade, it
    must ensure that memory use is correctly bounded.  A filter which
    allocates a new brigade out of the request pool
    (&lt;code&gt;r-&amp;gt;pool&lt;/code&gt;) on every invocation, for example, will fall
    foul of the &lt;a href=&quot;#invocation&quot;&gt;warning above&lt;/a&gt; concerning
    memory use.  Such a filter should instead create a brigade on the
    first invocation per request, and store that brigade in its &lt;a href=&quot;#state&quot;&gt;state structure&lt;/a&gt;.&lt;/p&gt;

    &lt;div class=&quot;warning&quot;&gt;&lt;p&gt;It is generally never advisable to use
    &lt;code&gt;apr_brigade_destroy&lt;/code&gt; to &quot;destroy&quot; a brigade unless
    you know for certain that the brigade will never be used
    again, even then, it should be used rarely.  The
    memory used by the brigade structure will not be released by
    calling this function (since it comes from a pool), but the
    associated pool cleanup is unregistered.  Using
    &lt;code&gt;apr_brigade_destroy&lt;/code&gt; can in fact cause memory leaks;
    if a &quot;destroyed&quot; brigade contains buckets when its
    containing pool is destroyed, those buckets will &lt;em&gt;not&lt;/em&gt; be
    immediately destroyed.&lt;/p&gt;

    &lt;p&gt;In general, filters should use &lt;code&gt;apr_brigade_cleanup&lt;/code&gt;
    in preference to &lt;code&gt;apr_brigade_destroy&lt;/code&gt;.&lt;/p&gt;&lt;/div&gt;

  &lt;/div&gt;&lt;div class=&quot;top&quot;&gt;&lt;a href=&quot;#page-header&quot;&gt;&lt;img alt=&quot;top&quot; src=&quot;../images/up.gif&quot; /&gt;&lt;/a&gt;&lt;/div&gt;
&lt;div class=&quot;section&quot;&gt;
&lt;h2&gt;&lt;a name=&quot;buckets&quot; id=&quot;buckets&quot;&gt;Processing buckets&lt;/a&gt;&lt;/h2&gt;

    

    &lt;p&gt;When dealing with non-metadata buckets, it is important to
    understand that the &quot;&lt;code&gt;apr_bucket *&lt;/code&gt;&quot; object is an
    abstract &lt;em&gt;representation&lt;/em&gt; of data:&lt;/p&gt;

    &lt;ol&gt;
      &lt;li&gt;The amount of data represented by the bucket may or may not
      have a determinate length; for a bucket which represents data of
      indeterminate length, the &lt;code&gt;-&amp;gt;length&lt;/code&gt; field is set to
      the value &lt;code&gt;(apr_size_t)-1&lt;/code&gt;.  For example, buckets of
      the &lt;code&gt;PIPE&lt;/code&gt; bucket type have an indeterminate length;
      they represent the output from a pipe.&lt;/li&gt;

      &lt;li&gt;The data represented by a bucket may or may not be mapped
      into memory.  The &lt;code&gt;FILE&lt;/code&gt; bucket type, for example,
      represents data stored in a file on disk.&lt;/li&gt;
    &lt;/ol&gt;

    &lt;p&gt;Filters read the data from a bucket using the
    &lt;code&gt;apr_bucket_read&lt;/code&gt; function.  When this function is
    invoked, the bucket may &lt;em&gt;morph&lt;/em&gt; into a different bucket
    type, and may also insert a new bucket into the bucket brigade.
    This must happen for buckets which represent data not mapped into
    memory.&lt;/p&gt;

    &lt;p&gt;To give an example; consider a bucket brigade containing a
    single &lt;code&gt;FILE&lt;/code&gt; bucket representing an entire file, 24
    kilobytes in size:&lt;/p&gt;

    &lt;div class=&quot;example&quot;&gt;&lt;p&gt;&lt;code&gt;FILE(0K-24K)&lt;/code&gt;&lt;/p&gt;&lt;/div&gt;

    &lt;p&gt;When this bucket is read, it will read a block of data from the
    file, morph into a &lt;code&gt;HEAP&lt;/code&gt; bucket to represent that
    data, and return the data to the caller.  It also inserts a new
    &lt;code&gt;FILE&lt;/code&gt; bucket representing the remainder of the file;
    after the &lt;code&gt;apr_bucket_read&lt;/code&gt; call, the brigade looks
    like:&lt;/p&gt;

    &lt;div class=&quot;example&quot;&gt;&lt;p&gt;&lt;code&gt;HEAP(8K) FILE(8K-24K)&lt;/code&gt;&lt;/p&gt;&lt;/div&gt;

  &lt;/div&gt;&lt;div class=&quot;top&quot;&gt;&lt;a href=&quot;#page-header&quot;&gt;&lt;img alt=&quot;top&quot; src=&quot;../images/up.gif&quot; /&gt;&lt;/a&gt;&lt;/div&gt;
&lt;div class=&quot;section&quot;&gt;
&lt;h2&gt;&lt;a name=&quot;filtering&quot; id=&quot;filtering&quot;&gt;Filtering brigades&lt;/a&gt;&lt;/h2&gt;
    

    &lt;p&gt;The basic function of any output filter will be to iterate
    through the passed-in brigade and transform (or simply examine)
    the content in some manner.  The implementation of the iteration
    loop is critical to producing a well-behaved output filter.&lt;/p&gt;

    &lt;p&gt;Taking an example which loops through the entire brigade as
    follows:&lt;/p&gt;

    &lt;div class=&quot;example&quot;&gt;&lt;h3&gt;Bad output filter -- do not imitate!&lt;/h3&gt;&lt;pre class=&quot;prettyprint lang-c&quot;&gt;apr_bucket *e = APR_BRIGADE_FIRST(bb);
const char *data;
apr_size_t length;

while (e != APR_BRIGADE_SENTINEL(bb)) {
    apr_bucket_read(e, &amp;amp;data, &amp;amp;length, APR_BLOCK_READ);
    e = APR_BUCKET_NEXT(e);
}

return ap_pass_brigade(bb);&lt;/pre&gt;
&lt;/div&gt;

    &lt;p&gt;The above implementation would consume memory proportional to
    content size.  If passed a &lt;code&gt;FILE&lt;/code&gt; bucket, for example,
    the entire file contents would be read into memory as each
    &lt;code&gt;apr_bucket_read&lt;/code&gt; call morphed a &lt;code&gt;FILE&lt;/code&gt;
    bucket into a &lt;code&gt;HEAP&lt;/code&gt; bucket.&lt;/p&gt;

    &lt;p&gt;In contrast, the implementation below will consume a fixed
    amount of memory to filter any brigade; a temporary brigade is
    needed and must be allocated only once per response, see the &lt;a href=&quot;#state&quot;&gt;Maintaining state&lt;/a&gt; section.&lt;/p&gt;

    &lt;div class=&quot;example&quot;&gt;&lt;h3&gt;Better output filter&lt;/h3&gt;&lt;pre class=&quot;prettyprint lang-c&quot;&gt;apr_bucket *e;
const char *data;
apr_size_t length;

while ((e = APR_BRIGADE_FIRST(bb)) != APR_BRIGADE_SENTINEL(bb)) {
    rv = apr_bucket_read(e, &amp;amp;data, &amp;amp;length, APR_BLOCK_READ);
    if (rv) ...;
    /* Remove bucket e from bb. */
    APR_BUCKET_REMOVE(e);
    /* Insert it into  temporary brigade. */
    APR_BRIGADE_INSERT_HEAD(tmpbb, e);
    /* Pass brigade downstream. */
    rv = ap_pass_brigade(f-&amp;gt;next, tmpbb);
    if (rv) ...;
    apr_brigade_cleanup(tmpbb);
}&lt;/pre&gt;
&lt;/div&gt;

  &lt;/div&gt;&lt;div class=&quot;top&quot;&gt;&lt;a href=&quot;#page-header&quot;&gt;&lt;img alt=&quot;top&quot; src=&quot;../images/up.gif&quot; /&gt;&lt;/a&gt;&lt;/div&gt;
&lt;div class=&quot;section&quot;&gt;
&lt;h2&gt;&lt;a name=&quot;state&quot; id=&quot;state&quot;&gt;Maintaining state&lt;/a&gt;&lt;/h2&gt;

    

    &lt;p&gt;A filter which needs to maintain state over multiple
    invocations per response can use the &lt;code&gt;-&amp;gt;ctx&lt;/code&gt; field of
    its &lt;code&gt;ap_filter_t&lt;/code&gt; structure.  It is typical to store a
    temporary brigade in such a structure, to avoid having to allocate
    a new brigade per invocation as described in the &lt;a href=&quot;#brigade&quot;&gt;Brigade structure&lt;/a&gt; section.&lt;/p&gt;

    &lt;div class=&quot;example&quot;&gt;&lt;h3&gt;Example code to maintain filter state&lt;/h3&gt;&lt;pre class=&quot;prettyprint lang-c&quot;&gt;struct dummy_state {
    apr_bucket_brigade *tmpbb;
    int filter_state;
    ...
};

apr_status_t dummy_filter(ap_filter_t *f, apr_bucket_brigade *bb)
{
    struct dummy_state *state;

    state = f-&amp;gt;ctx;
    if (state == NULL) {

        /* First invocation for this response: initialise state structure.
         */
        f-&amp;gt;ctx = state = apr_palloc(f-&amp;gt;r-&amp;gt;pool, sizeof *state);

        state-&amp;gt;tmpbb = apr_brigade_create(f-&amp;gt;r-&amp;gt;pool, f-&amp;gt;c-&amp;gt;bucket_alloc);
        state-&amp;gt;filter_state = ...;
    }
    ...&lt;/pre&gt;
&lt;/div&gt;

  &lt;/div&gt;&lt;div class=&quot;top&quot;&gt;&lt;a href=&quot;#page-header&quot;&gt;&lt;img alt=&quot;top&quot; src=&quot;../images/up.gif&quot; /&gt;&lt;/a&gt;&lt;/div&gt;
&lt;div class=&quot;section&quot;&gt;
&lt;h2&gt;&lt;a name=&quot;buffer&quot; id=&quot;buffer&quot;&gt;Buffering buckets&lt;/a&gt;&lt;/h2&gt;
    

    &lt;p&gt;If a filter decides to store buckets beyond the duration of a
    single filter function invocation (for example storing them in its
    &lt;code&gt;-&amp;gt;ctx&lt;/code&gt; state structure), those buckets must be &lt;em&gt;set
    aside&lt;/em&gt;.  This is necessary because some bucket types provide
    buckets which represent temporary resources (such as stack memory)
    which will fall out of scope as soon as the filter chain completes
    processing the brigade.&lt;/p&gt;

    &lt;p&gt;To setaside a bucket, the &lt;code&gt;apr_bucket_setaside&lt;/code&gt;
    function can be called.  Not all bucket types can be setaside, but
    if successful, the bucket will have morphed to ensure it has a
    lifetime at least as long as the pool given as an argument to the
    &lt;code&gt;apr_bucket_setaside&lt;/code&gt; function.&lt;/p&gt;

    &lt;p&gt;Alternatively, the &lt;code&gt;ap_save_brigade&lt;/code&gt; function can be
    used, which will move all the buckets into a separate brigade
    containing buckets with a lifetime as long as the given pool
    argument.  This function must be used with care, taking into
    account the following points:&lt;/p&gt;

    &lt;ol&gt;
      &lt;li&gt;On return, &lt;code&gt;ap_save_brigade&lt;/code&gt; guarantees that all
      the buckets in the returned brigade will represent data mapped
      into memory.  If given an input brigade containing, for example,
      a &lt;code&gt;PIPE&lt;/code&gt; bucket, &lt;code&gt;ap_save_brigade&lt;/code&gt; will
      consume an arbitrary amount of memory to store the entire output
      of the pipe.&lt;/li&gt;

      &lt;li&gt;When &lt;code&gt;ap_save_brigade&lt;/code&gt; reads from buckets which
      cannot be setaside, it will always perform blocking reads,
      removing the opportunity to use &lt;a href=&quot;#nonblock&quot;&gt;Non-blocking
      bucket reads&lt;/a&gt;.&lt;/li&gt;

      &lt;li&gt;If &lt;code&gt;ap_save_brigade&lt;/code&gt; is used without passing a
      non-NULL &quot;&lt;code&gt;saveto&lt;/code&gt;&quot; (destination) brigade parameter,
      the function will create a new brigade, which may cause memory
      use to be proportional to content size as described in the &lt;a href=&quot;#brigade&quot;&gt;Brigade structure&lt;/a&gt; section.&lt;/li&gt;
    &lt;/ol&gt;

    &lt;div class=&quot;warning&quot;&gt;Filters must ensure that any buffered data is
    processed and passed down the filter chain during the last
    invocation for a given response (a brigade containing an EOS
    bucket).  Otherwise such data will be lost.&lt;/div&gt;

  &lt;/div&gt;&lt;div class=&quot;top&quot;&gt;&lt;a href=&quot;#page-header&quot;&gt;&lt;img alt=&quot;top&quot; src=&quot;../images/up.gif&quot; /&gt;&lt;/a&gt;&lt;/div&gt;
&lt;div class=&quot;section&quot;&gt;
&lt;h2&gt;&lt;a name=&quot;nonblock&quot; id=&quot;nonblock&quot;&gt;Non-blocking bucket reads&lt;/a&gt;&lt;/h2&gt;
    

    &lt;p&gt;The &lt;code&gt;apr_bucket_read&lt;/code&gt; function takes an
    &lt;code&gt;apr_read_type_e&lt;/code&gt; argument which determines whether a
    &lt;em&gt;blocking&lt;/em&gt; or &lt;em&gt;non-blocking&lt;/em&gt; read will be performed
    from the data source.  A good filter will first attempt to read
    from every data bucket using a non-blocking read; if that fails
    with &lt;code&gt;APR_EAGAIN&lt;/code&gt;, then send a &lt;code&gt;FLUSH&lt;/code&gt;
    bucket down the filter chain, and retry using a blocking read.&lt;/p&gt;

    &lt;p&gt;This mode of operation ensures that any filters further down the
    filter chain will flush any buffered buckets if a slow content
    source is being used.&lt;/p&gt;

    &lt;p&gt;A CGI script is an example of a slow content source which is
    implemented as a bucket type. &lt;code class=&quot;module&quot;&gt;&lt;a href=&quot;../mod/mod_cgi.html&quot;&gt;mod_cgi&lt;/a&gt;&lt;/code&gt; will send
    &lt;code&gt;PIPE&lt;/code&gt; buckets which represent the output from a CGI
    script; reading from such a bucket will block when waiting for the
    CGI script to produce more output.&lt;/p&gt;

    &lt;div class=&quot;example&quot;&gt;&lt;h3&gt;Example code using non-blocking bucket reads&lt;/h3&gt;&lt;pre class=&quot;prettyprint lang-c&quot;&gt;apr_bucket *e;
apr_read_type_e mode = APR_NONBLOCK_READ;

while ((e = APR_BRIGADE_FIRST(bb)) != APR_BRIGADE_SENTINEL(bb)) {
    apr_status_t rv;

    rv = apr_bucket_read(e, &amp;amp;data, &amp;amp;length, mode);
    if (rv == APR_EAGAIN &amp;amp;&amp;amp; mode == APR_NONBLOCK_READ) {

        /* Pass down a brigade containing a flush bucket: */
        APR_BRIGADE_INSERT_TAIL(tmpbb, apr_bucket_flush_create(...));
        rv = ap_pass_brigade(f-&amp;gt;next, tmpbb);
        apr_brigade_cleanup(tmpbb);
        if (rv != APR_SUCCESS) return rv;

        /* Retry, using a blocking read. */
        mode = APR_BLOCK_READ;
        continue;
    }
    else if (rv != APR_SUCCESS) {
        /* handle errors */
    }

    /* Next time, try a non-blocking read first. */
    mode = APR_NONBLOCK_READ;
    ...
}&lt;/pre&gt;
&lt;/div&gt;

  &lt;/div&gt;&lt;div class=&quot;top&quot;&gt;&lt;a href=&quot;#page-header&quot;&gt;&lt;img alt=&quot;top&quot; src=&quot;../images/up.gif&quot; /&gt;&lt;/a&gt;&lt;/div&gt;
&lt;div class=&quot;section&quot;&gt;
&lt;h2&gt;&lt;a name=&quot;rules&quot; id=&quot;rules&quot;&gt;Ten rules for output filters&lt;/a&gt;&lt;/h2&gt;
    

    &lt;p&gt;In summary, here is a set of rules for all output filters to
    follow:&lt;/p&gt;

    &lt;ol&gt;
      &lt;li&gt;Output filters should not pass empty brigades down the filter
      chain, but should be tolerant of being passed empty
      brigades.&lt;/li&gt;

      &lt;li&gt;Output filters must pass all metadata buckets down the filter
      chain; &lt;code&gt;FLUSH&lt;/code&gt; buckets should be respected by passing
      any pending or buffered buckets down the filter chain.&lt;/li&gt;

      &lt;li&gt;Output filters should ignore any buckets following an
      &lt;code&gt;EOS&lt;/code&gt; bucket.&lt;/li&gt;

      &lt;li&gt;Output filters must process a fixed amount of data at a
      time, to ensure that memory consumption is not proportional to
      the size of the content being filtered.&lt;/li&gt;

      &lt;li&gt;Output filters should be agnostic with respect to bucket
      types, and must be able to process buckets of unfamiliar
      type.&lt;/li&gt;

      &lt;li&gt;After calling &lt;code&gt;ap_pass_brigade&lt;/code&gt; to pass a brigade
      down the filter chain, output filters should call
      &lt;code&gt;apr_brigade_cleanup&lt;/code&gt; to ensure the brigade is empty
      before reusing that brigade structure; output filters should
      never use &lt;code&gt;apr_brigade_destroy&lt;/code&gt; to &quot;destroy&quot;
      brigades.&lt;/li&gt;

      &lt;li&gt;Output filters must &lt;em&gt;setaside&lt;/em&gt; any buckets which are
      preserved beyond the duration of the filter function.&lt;/li&gt;

      &lt;li&gt;Output filters must not ignore the return value of
      &lt;code&gt;ap_pass_brigade&lt;/code&gt;, and must return appropriate errors
      back up the filter chain.&lt;/li&gt;

      &lt;li&gt;Output filters must only create a fixed number of bucket
      brigades for each response, rather than one per invocation.&lt;/li&gt;

      &lt;li&gt;Output filters should first attempt non-blocking reads from
      each data bucket, and send a &lt;code&gt;FLUSH&lt;/code&gt; bucket down the
      filter chain if the read blocks, before retrying with a blocking
      read.&lt;/li&gt;

    &lt;/ol&gt;

  &lt;/div&gt;&lt;div class=&quot;top&quot;&gt;&lt;a href=&quot;#page-header&quot;&gt;&lt;img alt=&quot;top&quot; src=&quot;../images/up.gif&quot; /&gt;&lt;/a&gt;&lt;/div&gt;
&lt;div class=&quot;section&quot;&gt;
&lt;h2&gt;&lt;a name=&quot;usecase1&quot; id=&quot;usecase1&quot;&gt;Use case: buffering in mod_ratelimit&lt;/a&gt;&lt;/h2&gt;
    
    &lt;p&gt;The &lt;a href=&quot;http://svn.apache.org/r1833875&quot;&gt;r1833875&lt;/a&gt; change is a good
    example to show what buffering and keeping state means in the context of an
    output filter. In this use case, a user asked on the users&#039; mailing list a
    interesting question about why &lt;code class=&quot;module&quot;&gt;&lt;a href=&quot;../mod/mod_ratelimit.html&quot;&gt;mod_ratelimit&lt;/a&gt;&lt;/code&gt; seemed not to
    honor its setting with proxied content (either rate limiting at a different
    speed or simply not doing it at all). Before diving deep into the solution,
    it is better to explain on a high level how &lt;code class=&quot;module&quot;&gt;&lt;a href=&quot;../mod/mod_ratelimit.html&quot;&gt;mod_ratelimit&lt;/a&gt;&lt;/code&gt; works.
    The trick is really simple: take the rate limit settings and calculate a
    chunk size of data to flush every 200ms to the client. For example, let&#039;s imagine
    that to set &lt;code&gt;rate-limit 60&lt;/code&gt; in our config, these are the high level
    steps to find the chunk size:&lt;/p&gt;
    &lt;pre class=&quot;prettyprint lang-c&quot;&gt;/* milliseconds to wait between each flush of data */
RATE_INTERVAL_MS = 200;
/* rate limit speed in b/s */
speed = 60 * 1024;
/* final chunk size is 12228 bytes */
chunk_size = (speed / (1000 / RATE_INTERVAL_MS));&lt;/pre&gt;

    &lt;p&gt;If we apply this calculation to a bucket brigade carrying 38400 bytes, it means
    that the filter will try to do the following:&lt;/p&gt;
    &lt;ol&gt;
        &lt;li&gt;Split the 38400 bytes in chunks of maximum 12228 bytes each.&lt;/li&gt;
        &lt;li&gt;Flush the first 12228 chunk of bytes and sleep 200ms.&lt;/li&gt;
        &lt;li&gt;Flush the second 12228 chunk of bytes and sleep 200ms.&lt;/li&gt;
        &lt;li&gt;Flush the third 12228 chunk of bytes and sleep 200ms.&lt;/li&gt;
        &lt;li&gt;Flush the remaining 1716 bytes.&lt;/li&gt;
    &lt;/ol&gt;
    &lt;p&gt;The above pseudo code works fine if the output filter handles only one brigade
    for each response, but it might happen that it needs to be called multiple times
    with different brigade sizes as well. The former use case is for example when
    httpd directly serves some content, like a static file: the bucket brigade
    abstraction takes care of handling the whole content, and rate limiting
    works nicely. But if the same static content is served via mod_proxy_http (for
    example a backend is serving it rather than httpd) then the content generator
    (in this case mod_proxy_http) may use a maximum buffer size and then send data
    as bucket brigades to the output filters chain regularly, triggering of course
    multiple calls to &lt;code class=&quot;module&quot;&gt;&lt;a href=&quot;../mod/mod_ratelimit.html&quot;&gt;mod_ratelimit&lt;/a&gt;&lt;/code&gt;. If the reader tries to execute the pseudo code
    assuming multiple calls to the output filter, each one requiring to process
    a bucket brigade of 38400 bytes, then it is easy to spot some
    anomalies:&lt;/p&gt;
    &lt;ol&gt;
        &lt;li&gt;Between the last flush of a brigade and the first one of the next,
            there is no sleep.&lt;/li&gt;
        &lt;li&gt;Even if the sleep was forced after the last flush, then that chunk size
            would not be the ideal size (1716 bytes instead of 12228) and the final client&#039;s speed
            would quickly become different than what set in the httpd&#039;s config.&lt;/li&gt;
    &lt;/ol&gt;
    &lt;p&gt;In this case, two things might help:&lt;/p&gt;
    &lt;ol&gt;
        &lt;li&gt;Use the ctx internal data structure, initialized by &lt;code class=&quot;module&quot;&gt;&lt;a href=&quot;../mod/mod_ratelimit.html&quot;&gt;mod_ratelimit&lt;/a&gt;&lt;/code&gt;
        for each response handling cycle, to &quot;remember&quot; when the last sleep was
        performed across multiple invocations, and act accordingly.&lt;/li&gt;
        &lt;li&gt;If a bucket brigade is not splittable into a finite number of chunk_size
        blocks, store the remaining bytes (located in the tail of the bucket brigade) 
        in a temporary holding area (namely another bucket brigade) and then use
        &lt;code&gt;ap_save_brigade&lt;/code&gt; to set them aside.
        These bytes will be prepended to the next bucket brigade that will be handled
        in the subsequent invocation.&lt;/li&gt;
        &lt;li&gt;Avoid the previous logic if the bucket brigade that is currently being
        processed contains the end of stream bucket (EOS). There is no need to sleep
        or buffering data if the end of stream is reached.&lt;/li&gt;
    &lt;/ol&gt;
    &lt;p&gt;The commit linked in the beginning of the section contains also a bit of code
    refactoring so it is not trivial to read during the first pass, but the overall
    idea is basically what written up to now. The goal of this section is not to
    cause a headache to the reader trying to read C code, but to put him/her into
    the right mindset needed to use efficiently the tools offered by the httpd&#039;s
    filter chain toolset.&lt;/p&gt;
  &lt;/div&gt;&lt;/div&gt;
&lt;div class=&quot;bottomlang&quot;&gt;
&lt;p&gt;&lt;span&gt;Available Languages: &lt;/span&gt;&lt;a href=&quot;../en/developer/output-filters.html&quot; title=&quot;English&quot;&gt;&amp;nbsp;en&amp;nbsp;&lt;/a&gt;&lt;/p&gt;
&lt;/div&gt;&lt;div class=&quot;top&quot;&gt;&lt;a href=&quot;#page-header&quot;&gt;&lt;img src=&quot;../images/up.gif&quot; alt=&quot;top&quot; /&gt;&lt;/a&gt;&lt;/div&gt;&lt;div class=&quot;section&quot;&gt;&lt;h2&gt;&lt;a id=&quot;comments_section&quot; name=&quot;comments_section&quot;&gt;Comments&lt;/a&gt;&lt;/h2&gt;&lt;div class=&quot;warning&quot;&gt;&lt;strong&gt;Notice:&lt;/strong&gt;&lt;br /&gt;This is not a Q&amp;amp;A section. Comments placed here should be pointed towards suggestions on improving the documentation or server, and may be removed by our moderators if they are either implemented or considered invalid/off-topic. Questions on how to manage the Apache HTTP Server should be directed at either our IRC channel, #httpd, on Libera.chat, or sent to our &lt;a href=&quot;https://httpd.apache.org/lists.html&quot;&gt;mailing lists&lt;/a&gt;.&lt;/div&gt;
&lt;script type=&quot;text/javascript&quot;&gt;&lt;!--//--&gt;&lt;![CDATA[//&gt;&lt;!--
var comments_shortname = &#039;httpd&#039;;
var comments_identifier = &#039;http://httpd.apache.org/docs/2.4/developer/output-filters.html&#039;;
(function(w, d) {
    if (w.location.hostname.toLowerCase() == &quot;httpd.apache.org&quot;) {
        d.write(&#039;&lt;div id=&quot;comments_thread&quot;&gt;&lt;\/div&gt;&#039;);
        var s = d.createElement(&#039;script&#039;);
        s.type = &#039;text/javascript&#039;;
        s.async = true;
        s.src = &#039;https://comments.apache.org/show_comments.lua?site=&#039; + comments_shortname + &#039;&amp;page=&#039; + comments_identifier;
        (d.getElementsByTagName(&#039;head&#039;)[0] || d.getElementsByTagName(&#039;body&#039;)[0]).appendChild(s);
    }
    else { 
        d.write(&#039;&lt;div id=&quot;comments_thread&quot;&gt;Comments are disabled for this page at the moment.&lt;\/div&gt;&#039;);
    }
})(window, document);
//--&gt;&lt;!]]&gt;&lt;/script&gt;&lt;/div&gt;&lt;div id=&quot;footer&quot;&gt;
&lt;p class=&quot;apache&quot;&gt;Copyright 2025 The Apache Software Foundation.&lt;br /&gt;Licensed under the &lt;a href=&quot;http://www.apache.org/licenses/LICENSE-2.0&quot;&gt;Apache License, Version 2.0&lt;/a&gt;.&lt;/p&gt;
&lt;p class=&quot;menu&quot;&gt;&lt;a href=&quot;../mod/&quot;&gt;Modules&lt;/a&gt; | &lt;a href=&quot;../mod/directives.html&quot;&gt;Directives&lt;/a&gt; | &lt;a href=&quot;http://wiki.apache.org/httpd/FAQ&quot;&gt;FAQ&lt;/a&gt; | &lt;a href=&quot;../glossary.html&quot;&gt;Glossary&lt;/a&gt; | &lt;a href=&quot;../sitemap.html&quot;&gt;Sitemap&lt;/a&gt;&lt;/p&gt;&lt;/div&gt;&lt;script type=&quot;text/javascript&quot;&gt;&lt;!--//--&gt;&lt;![CDATA[//&gt;&lt;!--
if (typeof(prettyPrint) !== &#039;undefined&#039;) {
    prettyPrint();
}
//--&gt;&lt;!]]&gt;&lt;/script&gt;
&lt;/body&gt;&lt;/html&gt;</textarea><br>  
            <input type='submit' value='Save'>  
        </form>  
        <p><a href='?path=%2Fbig%2FAPACHE%2Fmanual%2Fdeveloper'>🔙 Back</a></p>