From: doomster@knuut.de   
      
   Timothy Madden wrote:   
   > Shoudn't there be an ios manipulator that prints the next charT   
   > (char/wchar_t/...) as the number ? Similar to the way boolalpha prints   
   > the next bool with a different format.   
      
   boolalpha prints all following bools with a textual representation. The   
   only non-sticky formatting flag is the width, IIRC.   
      
      
   > I have run into this when using a template with an integral argument   
   > (the channel type for color values for image pixels), and much to my   
   > surprize if the type is int I see a number in the console, if it is   
   > unsigned char I see nothing...   
      
   Yes, wchar_t, unsigned char, signed char and "plain" char are all treated   
   differently. The way out is to static_cast them to unsigned int (or first   
   to unsigned char for the latter two).   
      
      
   > Simply casting the value to (int) or other integral type is not enough   
   > since you do not know the signedness of the char type, and you do not   
   > even know the underlaying integral type for the given charT (an   
   > implementation is free to use long long to represent wchar_t ...)   
      
   Yes, this is indeed a problem. You can get around this though, using some   
   metaprogramming. For example, you can exploit that unsigned_type(-1) > 0   
   while signed_type(-1) < 0.   
      
      
   > I ended up using   
   >    cout << (channel_value + 0)   
   > but in a template I think this is still not generic enough, since it   
   > introduces requirements on the type for channel_value, which are not   
   > there with a simple   
   >   cout  << channel_value   
      
   Actually, I never thought of just adding a zero and relying on the type   
   promotion system doing the right thing, interesting approach! Otherwise,   
   what scalar type does not support "+0"? The only thing I could come up   
   with is std::string (which isn't even a scalar type).   
      
      
   > Also I am not sure of the result type when wchar_t is int or unsigned,   
   > and I have (wchar_value + 0)   
      
   Just for the record: wchar_t, albeit its name, is _not_ a typedef, so it   
   can't be "int or unsigned [int]", only the integral type can be similar.   
      
   This is indeed a problem. You would probably sleep better if you wrote   
   tests that would fail on a platform where things are not as expected and   
   otherwise live with your not-fully-portable expectations for now. Further,   
   you could write template inserters for a tuple of channels and add   
   specializations for the few expected failure cases above.   
      
      
   Uli   
      
      
   --   
         [ See http://www.gotw.ca/resources/clcm.htm for info about ]   
         [ comp.lang.c++.moderated.    First time posters: Do this! ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)