I've added a wysiwyg editor (CKEditor) on a guestbook I'm building, which among other things has insertable smiley images. When validating a submited entry form I run strip_tags() on the message. Ideally I would like to only allow <b>, <i> and <a> tags but I'm forced to allow <img> tags in order to not strip away smileys.

So, after running strip_tags(), I would like to go through the message to also strip away any <img> tags that is not a smiley. That is, strip away all images that have a src that doesn't point to the local folder where all the images are.

I recon this can be done with preg_replace() or similar functions, but regexp really isn't my strongest side

You could use DOM ( http://us2.php.net/manual/en/book.dom.php ) to parse out any images where the href attribute points to an external site.

Why not just use the native BBCode extension or the Pear HTML_BBCodeParser??

Thanks for your replies. I looked in to the DOM and the native BBCode extension, but I found them too big to dig into just for this small, one-time application. And I'd prefer not to drag a big framework like Pear into this, again, just for this small application.

Nevertheless, after several hours of trial and error I managed to get it working just the way I wanted using some regex functions. I used preg_replace_callback() with the pattern "|<img[^>]*>|i" to find all <img> tags, and a callback function in which I used preg_find_all() with the pattern "|([a-z]+)=\"[^\"]*\"|i" to find all attributes. I went through all the attributes and stripped away those that I will not allow and also checked the src attribute, to see if it points to a local file. And if it doesn't I'll return and empty string which will then replace the entire <img> tag.