Removing CDATA close tag ( ]]> ) unbalances the CDATA block

Couldn't we just make this a filter that is applied for RSS feeds but not e.g. admins, like kses?

I was the one who pointed out the reason for that line. The JS-replacement plugin is probably saving your feeds from the problem that line was added to fix.

I think we can come up with a better solution for this. If there is a good reason to wrap some of your post content as CDATA, perhaps you would be better served by adding a filter to your plugin to wrap the_content as CDATA.

Has anyone else run into this?

I would also appreciate an update to make difference between feed / content, my plugin has exact the same problem, cause when I want the XHTML validation, I need the CDATA implementation, cause the "socalled" flashvars for a Flash video plugin are diveded by ampersand's. At the moment it's possible to use <!-- around CDATA to covers the CDATA problematic in the conent, but it seems to be deprecate from W3C.

So more and More Flash integration trap into this problematic

Andy, I agree that the filter is saving me from needing that line. But, embedded JS is a valid thing to do in the content. That replace function should only really happen when we are processing a feed. The global is_feed() (I think) method should just wrap that line at a minimum.

Checking is_feed does not get to the root of the problem, which is that we are trying to nest CDATA sections where we should not. I don't know enough about XML but I think we ought to stop using CDATA markers in feed templates and use a function that can intelligently wrap any content in CDATA as needed, removing any nested markers.

Is there really any reason to encode them rather than simply remove them?

So what I propose, and what should be seen by somebody with XML chops, is this unintelligent function:

function wp_cdata($data) {

// Maybe test for conditions necessitating // CDATA markers and return if no need

$data = str_replace('<![CDATA[', , $data); $data = str_replace(']]>', , $data);

// Sometimes a trailing square bracket // can mess up XML parsers: if ( substr($data, -1) == ']' )

$data .= ' ';

return "<![CDATA[$data]]>';

}

function wp_cdata($data) {
 // Maybe test for conditions necessitating
 // CDATA markers and return if no need

 $data = str_replace('<![CDATA[', '', $data);
 $data = str_replace(']]>', '', $data);

 // Sometimes a trailing square bracket
 // can mess up XML parsers:
 if ( substr($data, -1) == ']' )
  $data .= ' ';

 return "<![CDATA[$data]]>';
}

Does wordpress support PHP < 4.0.5? If not, you could use arrays in str_replace:

$data = str_replace(array('<![CDATA[', ']]>'), array('', ''), $data);

Not sure how helpful this is, but it is kind of possible to nest CDATA sections. Technically I suppose it's double-encoding rather than nesting, but regardless, this is 100% valid and should work with any compliant XML parser:

<![CDATA foo <![CDATA bar ]]]><![CDATA[]> baz ]]>

It's actually two separate CDATA sections concatenated together.

When decoded by an XML parser, this becomes:

foo <![CDATA bar ]]> baz

Here's a PHP snippet to do it:

return '<![CDATA.str_replace(?]>', ']]]><![CDATA[]>', $str).']]>';

Whoops, here's that code again:

return '<![CDATA['.str_replace(']]>', ']]]><![CDATA[]>', $str).']]>';

Any updates, Andy?

I'm still a little unclear as to the purpose of this line of code. I keep running into this problem, so I'd like to see a fix (I'd be happy to submit a patch), but I don't understand why we are keeping it at all. If it's just needed for feeds, lets limit it to affecting feeds. If someone could clearly explain to me why it's here (when it's needed), I'll put a patch together.

This ticket has sat here forever with nothing getting done, so I added a patch to get the ball rolling. The patch will fix two problems. It will fix one expressed above by only replacing ]]> with ]]&gt; when it is processing a feed.

There was another problem because we so often use:

//<![CDATA[
alert('test');
//]]>

and strip_tags() leaves the leading // there, which looks ugly in the feeds as well. I added a simple statement in two places to handle that: $text = preg_replace('|//\s*<!\[CDATA\[|', '<![CDATA[', $text);