I went on vacation before I had time to finish my stream updates, but I thought it might be useful to split out the scanf implementation and post it separately. Don't expect much of a response from me for another week or so as I'm busy overeating and swimming in the ocean, but here are a few details...  This is a (mostly tested) C99 compliant scanf implementation.  It adds %C and %S for D strings, as (AFAIK) the typeinfo currently isn't good enough to distinguish between C pointers and D arrays.  The only feature that is not implemented is %p because I wanted to see what Walter came up with for the printf side.  The declaration for wscanf is commented out because there was some issue finding the wide version of a function it uses.  I would also like to add complex number support, but again some coordination should be done with printf.

Functionally, the implementation is a good demonstration of things that aren't possible in C++.  Alias template parameters and inner functions made things pretty clean and efficient--C++ would have required function pointers or wrapper classes to do the same thing--and it was a neat discovery that D functions can require interfaces of their own.  I did use goto a bit for clarity (yes clarity) and exceptions for flow control.  And this same implementation can be used as a mixin in the stream classes so there's no unnecessary code duplication or odd machinations.

The unit tests are basically nonexistent at the moment and will be expanded.  I had some problems with .94 so I've commented out the %s test (it was giving an access violation) for now--it worked in .93 so I'm going to assume it still does.

The link is here:

http://home.f4.ca/sean/d/scanf.d


Sean

Just a quick update.  I added support for %p and fixed an issue with C pointers. The implementation is now complete unless there are any other bugs lurking.  The file contains fscanf, sscanf, scanf, and the wchar versions of each.  The unit tests have examples of string use but nothing else yet.  I'll improve that within the next week.


Sean

Looks like the implementation still has problems with typeid.  I'm posting a small update that includes some modified unit tests, but the updates will continue for the next week or two as I get the problems worked out.  Right now I'm defaulting to C strings for the scanset type and D strings don't work for that parameter type.  This version also includes a fix for the suppress flag.


Sean