This is a quick and dirty document to get you on your way to writing custom URI filters for your own URL filtering needs. Why would you want to write a URI filter? If you need URIs your users put into HTML to magically change into a different URI, this is exactly what you need!
Any URI filter you make will be a subclass of HTMLPurifier_URIFilter
.
The scaffolding is thus:
class HTMLPurifier_URIFilter_NameOfFilter extends HTMLPurifier_URIFilter { var $name = 'NameOfFilter'; function prepare($config) {} function filter(&$uri, $config, &$context) {} }
Fill in the variable $name
with the name of your filter, and
take a look at the two methods. prepare()
is an initialization
method that is called only once, before any filtering has been done of the
HTML. Use it to perform any costly setup work that only needs to be done
once. filter()
is the guts and innards of our filter:
it takes the URI and does whatever needs to be done to it.
If you've worked with HTML Purifier, you'll recognize the $config
and $context
parameters. On the other hand, $uri
is something unique to this section of the application: it's a
HTMLPurifier_URI
object. The interface is thus:
class HTMLPurifier_URI { var $scheme, $userinfo, $host, $port, $path, $query, $fragment; function HTMLPurifier_URI($scheme, $userinfo, $host, $port, $path, $query, $fragment); function toString(); function copy(); function getSchemeObj($config, &$context); function validate($config, &$context); }
The first three methods are fairly self-explanatory: you have a constructor,
a serializer, and a cloner. Generally, you won't be using them when
you are manipulating the URI objects themselves.
getSchemeObj()
is a special purpose method that returns
a HTMLPurifier_URIScheme
object corresponding to the specific
URI at hand. validate()
performs general-purpose validation
on the internal components of a URI. Once again, you don't need to
worry about these: they've already been handled for you.
As a URIFilter, we're interested in the member variables of the URI object.
Scheme | The protocol for identifying (and possibly locating) a resource (http, ftp, https) |
---|---|
Userinfo | User information such as a username (bob) |
Host | Domain name or IP address of the server (example.com, 127.0.0.1) |
Port | Network port number for the server (80, 12345) |
Path | Data that identifies the resource, possibly hierarchical (/path/to, ed@example.com) |
Query | String of information to be interpreted by the resource (?q=search-term) |
Fragment | Additional information for the resource after retrieval (#bookmark) |
Because the URI is presented to us in this form, and not
http://bob@example.com:8080/foo.php?q=string#hash
, it saves us
a lot of trouble in having to parse the URI every time we want to filter
it. For the record, the above URI has the following components:
Scheme | http |
---|---|
Userinfo | bob |
Host | example.com |
Port | 8080 |
Path | /foo.php |
Query | q=string |
Fragment | hash |
Note that there is no question mark or octothorpe in the query or fragment: these get removed during parsing.
With this information, you can get straight to implementing your
filter()
method. But one more thing...
You may have noticed that the URI is being passed in by reference. This means that whatever changes you make to it, those changes will be reflected in the URI object the callee had. Do not return the URI object: it is unnecessary and will cause bugs. Instead, return a boolean value, true if the filtering was successful, or false if the URI is beyond repair and needs to be axed.
Let's suppose I wanted to write a filter that de-internationalized domain
names by converting them to Punycode.
Assuming that punycode_encode($input)
converts $input
to
Punycode and returns false
on failure:
class HTMLPurifier_URIFilter_ConvertIDNToPunycode extends HTMLPurifier_URIFilter { var $name = 'ConvertIDNToPunycode'; function filter(&$uri, $config, &$context) { if (is_null($uri->host)) return true; if ($uri->host == utf8_decode($uri->host) { // is ASCII, abort return true; } $host = punycode_encode($uri->host); if ($host === false) return false; $uri->host = $host; return true; } }
Notice I did not return $uri;
.
Having a filter is all well and good, but you need to tell HTML Purifier to use it. Fortunately, this part's simple:
$uri =& $config->getDefinition('URI'); $uri->addFilter(new HTMLPurifier_URIFilter_NameOfFilter());
If you want to be really fancy, you can define a configuration directive for your filter and have HTML Purifier automatically manage whether or not your filter gets loaded or not (this is how internal filters manage things):
HTMLPurifier_ConfigSchema::define( 'URI', 'NameOfFilter', false, 'bool', 'What your filter does.' ); $uri =& $config->getDefinition('URI', true); $uri->registerFilter(new HTMLPurifier_URIFilter_NameOfFilter());
Now, your filter will only be called when %URI.NameOfFilter is set to true.
Check the URIFilter directory for more implementation examples, and see the new directives proposal document for ideas on what could be implemented as a filter.