regular expression unicode character class trouble

[Diez B. Roggisch]

Hi,

I need in a unicode-environment the character-class

set("\w") - set("[0-9]")

or aplha w/o num. Any ideas how to create that? And what performance 
implications do I have to fear? I mean I guess that the characterclasses 
aren't implementet as sets, but as comparison-function that compares a 
value with certain well-defined ranges.

Regards,

Diez