Hello, I have some questions encountered while attempting to extend the boost::filesystem library (BTW, thanks to all involved with this excellent library). I often need to do something with all files matching certain criteria which reside (recursively) in subdirectories matching some (probably different) criteria. E.g., I might want to list all *.hpp files in all subdirectories of 'develop', or search for the first *.cpp file in all subdirectories of "my_proj" excluding "my_proj/bin". Using the boost::iterator library (BTW, thanks to all involved with this excellent library as well), I built a recursive_directory_iterator, parameterized by a functor indicating whether to expand a subdirectory, and a filter iterator, parameterized by a functor indicating whether to retain an entry with a given path, which filters the sequence formed by some other underlying iterator (e.g., recursive_directory_iterator, in this case). This led to two questions: 1. Does the library already include similar functionality? (I couldn't find any;) If not, are there any pitfalls involved? 2. There are (at least) two ways to implement recursive_directory_iterator, which display a memory/time tradeoff: (a) using a stack of boost::filesystem::directory_iterator, and (b) using a boost::filesystem::directory_iterator (for the current subdirectory) and a boost::filesystem::path (for the (parent) subdirectory entry to which to return once the current subdirectory is exhausted). Consequently, I made this too a policy of recursive_directory_iterator. Version (b), however, throws an exception when trying to iterate over a subdirectory "foo!": specifically, I can't assign "foo!" (or any name including '!') to a boost::filesystem::path object, although it seems that my filesystem (on Windows XP) finds "foo!" an acceptable name for a subdirectory. Should I be using boost::filesystem differently? Many Thanks, Ami