The value '<!\[CDATA\[.*?\]\]>|[^<>&]*' of the facet 'pattern' is not a valid regular expression

Question

I'm trying to use a regex to validate a field in my xml using xsd. I came up with the regex to do what I want which is to disallow special characters unless the text is wrapped in CDATA tags. This is the regular expression I came up with:

<!\[CDATA\[.*?\]\]>|[^<>&]*

Works great when I test it on http://regexr.com/ to match my pattern. The problem is when I try to then plug it into a simpleType pattern restriction I'm getting an error saying its not a valid regular expression.

The value '<!\[CDATA\[.*?\]\]>|[^<>&]*' of the facet 'pattern' is not a valid regular expression.

Here is my xsd code (note I had to replace &<> in the regular expression with < > and & so it would be valid xml):

<xs:element name="description" minOccurs="1" maxOccurs="1">
    <xs:simpleType>
        <xs:restriction base="xs:string">
            <xs:pattern value="&lt;!\[CDATA\[.*?\]\]&gt;|[^&lt;&gt;&amp;]*"/>          
        </xs:restriction>
    </xs:simpleType>
</xs:element>

So I assume there's something about the way regex works in xsd patterns that I'm not getting.

Lachrymose

Answer 1

By process of elimination I was able to find a fix. First I determined that the problem somewhere in the <!\[CDATA\[.*?\]\]> part of the expression. So I tore this out of my expression and slowly adding bits back until I got the same error. The character that what causing the problem was the ?. The expression still works without the ? even if there is nothing inside the CDATA.

I'm not sure 1) why the person I got that pattern from included it, and 2) why doesn't xsd allow it in my pattern?

Lachrymose

It's the non-greedy operator, that selects the minimal, rather than the maximal, match. I don't know what regex engine your XSD parser uses. I don't think all engines have the ?. The standard Java engine does, however – Overwrought

Thanks for the clarification. This is why regex are so frustrating. Everything depends on your environment. In my case it works fine without the non-greedy so I'll stick with it. – Lachrymose

.*? is invalid Regex all around actually. .* means it's optional already. (i.e. you're looking for zero or more matches, and you can't get any more optional than that!). -- You should also note that regexes are GREEDY -- so if you have more than one CDATA, this expression will begin matching at the start of the first one, and not end until the second one (because that's what .* does). – Blum