Cory Nelson wrote:
I'm attempting to learn xpressive by writing a simple IRC regex.
The IRC protocol goes like: :prefix numericreplyorcommandstring param1 param2 :trailing param
The regex looks like this (ignore comments and newlines): ^ (?::(\S+)\s)? // prefix (?:(\d+)|(\S+)) // command (?:\s([^:]\S*))* // parameters (?:\s:(.+))? // trailing parameter $
And my static xpressive looks like this: bos >> optional(':' >> (s1=+~_s) >> _s) >> // prefix ((s2=+_d)|(s3=+~_s)) >> // command *(_s >> (s4 = ~as_xpr(':') >> *~_s)) >> // parameters optional(_s >> ':' >> (s5=+_)) >> // trailing parameter eos
I'm not sure how to retrieve the multiple matches for s4 - any help?
In this regard, xpressive behaves like Perl and like Boost.Regex. For captures in repeats, only the last iteration is remembered. If you want to remember all the places s4 matches, you should use a nested regex instead: sregex param = ~as_xpr(':') >> keep(*~_s); sregex protocol = bos >> optional(':' >> (s1=+~_s) >> _s) >> // prefix ((s2=+_d)|(s3=+~_s)) >> // command *(_s >> parameter) >> // parameters optional(_s >> ':' >> (s5=+_)) >> // trailing parameter eos; Now, when you match "protocol", the match_results struct will contain nested matches for the "parameter" regex, which you can access with match_results<>::nested_results(). (You have nested quantifiers in your pattern. I tried to make it more efficient with the keep() directive.) HTH, -- Eric Niebler Boost Consulting www.boost-consulting.com