I cannot find anything about it in the documentation. The only thing i found about flex RegExp is the \w (which is all letters) or \W (which is all non-letters) pattern.

The problem is that those have no unicode support (or better: I did not find it yet, might be there are some undocumented patterns or character classes to use).

They afaik only match against character classes like [a-z A-Z]. I need something that is more general and supports non-english letters, which is stated in the unicode standard as a property (which can be matched against in other languages using regexp patterns like \p{L}

No, because I need to match against the class of all unicode characters beeing letters, not only some specific. This would make a really giant regexp if I include all letter characters from any language, won't it?

It might be possible to do it with range, but it's a more high-level regexp property called "character properties".

Before I try to explain, take a look at the page http://www.regular-expressions.info/unicode.html under the heading "Unicode Character Properties".

It is supported by java and .NET obviously, but it is missing in the flex RegExp class.

For my special needs I want to match against \p{L} (which means all non-letter unicode characters) in order to remove them from my string (such as copyright, smileys and other stuff people tend to use somewhere in their nicknames and chat message)

Thanks for your hint, but could you tell me where exactly I have to request such a new feature?

Ok, now I am messing around with regexp ranges and unicode. I cannot say that it is working either.

I am trying to strip all non-cyrillic characters from a string.

To accomplish it, I would like to remove all characters but a specified range using a regexp such as : /[^\u0400-\u050f]/g

I encountered a really strange behaviour in flex though, which can be reproduced by this small example:

            var myString : String = "HelloWorld";
            // first replace (all non-cyrillic characters)
            trace("first replace: " + myString.replace(/[^\u0400-\u050f]/g,""));
            // second replace (all non-cyrillic characters + 1)
            trace("second replace: " + myString.replace(/[^\u0400-\u0510]/g,""));

As you can see, the regexp only differs in the upper bound of the character class (actually the difference is only a single character).

The outcome of both replaces should be that they return an empty string, since there is no cyrillic character in myString.

But nonexpextedly, the output is this:

first replace: 
second replace: HelloWorld

which indicates that actually regexp and unicode are not working as expected.

Can anyone reproduce this problem?

I am pretty stuck on this one ...