Can you provide some examples of why it is hard to parse XML and HTML with a regex?

Here's some fun valid XML for you:

<!DOCTYPE x [ <!ENTITY y "a]>b"> ]>
<x>
    <a b="&y;>" />
    <![CDATA[[a>b <a>b <a]]>
    <?x <a> <!-- <b> ?> c --> d
</x>

And this little bundle of joy is valid HTML:

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd" [
    <!ENTITY % e "href='hello'">
    <!ENTITY e "<a %e;>">
]>
    <title>x</TITLE>
</head>
    <p id  =  a:b center>
    <span / hello </span>
    &amp<br left>
    <!---- >t<!---> < -->
    &e link </a>
</body>

Not to mention all the browser-specific parsing for invalid constructs.

Good luck pitting regex against that!

EDIT (Jörg W Mittag): Here is another nice piece of well-formed, valid HTML 4.01:

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
  "http://www.w3.org/TR/html4/strict.dtd"> 
<HTML/
  <HEAD/
    <TITLE/>/
    <P/>

Actually

<img src="imgtag.gif" alt="<img>" />

is not valid HTML, and is not valid XML either.

It is not valid XML because the '<' and '>' are not valid characters inside attribute strings. They need to be escaped using the corresponding XML entities &lt; and &gt;

It is not valid HTML either because the short closing form is not allowed in HTML (but is correct in XML and XHTML). The 'img' tag is also an implicitly closed tag as per the HTML 4.01 specification. This means that manually closing it is actually wrong, and is equivalent to closing any other tag twice.

The correct version in HTML is

<img src="imgtag.gif" alt="&lt;img&gt;">

and the correct version in XHTML and XML is

<img src="imgtag.gif" alt="&lt;img&gt;"/>

The following example you gave is also invalid

<
tag
attr="5"
/>

This is not valid HTML or XML either. The name of the tag must be right behind the '<', although the attributes and the closing '>' may be wherever they want. So the valid XML is actually

<tag
attr="5"
/>

And here's another funkier one: you can actually choose to use either " or ' as your attribute quoting character

<img src="image.gif" alt='This is single quoted AND valid!'>

All the other reasons that were posted are correct, but the biggest problem with parsing HTML is that people usually don't understand all the syntax rules correctly. The fact that your browser interprets your tagsoup as HTML doesn't means that you have actually written valid HTML.

Edit: And even stackoverflow.com agrees with me regarding the definition of valid and invalid. Your invalid XML/HTML is not highlighted, while my corrected version is.

Basically, XML is not made to be parsed with regexps. But there is also no reason to do so. There are many, many XML parsers for each and every language. You have the choice between SAX parsers, DOM parsers and Pull parsers. All of these are guaranteed to be much faster than parsing with a regexp and you may then use cool technologies like XPath or XSLT on the resulting DOM tree.

My reply is therefore: not only is parsing XML with regexps hard, but it is also a bad idea. Just use one of the millions of existing XML parsers, and take advantage of all the advanced features of XML.

HTML is just too hard to even try parsing on your own. First the legal syntax has many little subtleties that you may not be aware of, and second, HTML in the wild is just a huge stinking pile of (you get my drift). There are a variety of lax parser libraries that do a good job at handling HTML like tag soup, just use these.


I wrote an entire blog entry on this subject: Regular Expression Limitations

The crux of the issue is that HTML and XML are recursive structures which require counting mechanisms in order to properly parse. A true regex is not capable of counting. You must have a context free grammar in order to count.

The previous paragraph comes with a slight caveat. Certain regex implementations now support the idea of recursion. However once you start adding recursion into your regex expressions, you are really stretching the boundaries and should consider a parser.

Tags:

Html

Xml

Regex

Recent Posts