Hello, Hartmut! You wrote to "'Boost Users mailing list'" : ??>> What is "proper way" of implementing result value of parser, ??>> have i implement new parser or can (and how?) i use grammar ??>> construct with user defined result value that can be passed ??>> to attached action (for ex. matched GUID returning)? HK> Grammars may have return values! By convention the first member of a HK> closure associated with the grammar is used as its return value. See HK> the docs for details. Thanks. Found that. But, because of mine little knowledge of FP i got many new problems. I have struct member declared something like that: 'unsigned char Data4[ 8 ];' and 'bind(&GUID::Data4[0])(self.val) = arg1' in action call produces compile time error :( And it seams that i can't apply 'bind(&GUID::Data1)(self.val) = arg1' in subrules, got compiler error. ??>> What is performance penalty for using vectors (they need ??>> realloc, don't they) in ast trees? HK> Not known so far. Try it out and report it to the list. I'm sure this HK> would be interesting to know for many of us. As soon as i can go throught the 'hell' of C++ MP&FP and finally write parser :) ??>> And is there anyway to store scanner defined result in tree, ??>> besides using user-specified value in node_val_data and type ??>> casting it to specific type depending of node id? HK> What do you mean with 'scanner defined result'? My mistake, should be 'parser defined result'. -- With best regards, GIE.