Re: refactoring

Stefan Seefeld wrote:
> 
> I agree. I was actually thinking of traces of all the decisions a 
> compiler does implicitely when looking up symbols, instantiating 
> templates (SFINAE !), etc., because without these a generated error 
> may be quite misleading.
> 
> Concept checks would be nice, too. I wonder how Richard Kelly's work
> with synopsis is going...

It's been going fairly well, but of course not as rapidly as I always
hope it will.  I have produced a demo version of the basic program, and
I've been demonstrating that to the people around here.  We're also
trying to get funding for a full-time development effort, but getting
development money takes a while.  The people around here are really
excited about the program, and they really want to use it, but that
doesn't necessarily mean they want to fund it.  In the meantime, I work
on it in my off time, so progress is slower.

The demo program is not nearly as smart (nor as bulletproof) as the real
program will be, but it does let me show people how helpful such a
concept tool would be to us.  I also examined some of our existing code
and was able to estimate how much of that code the concept tool would be
able to generate, error-check, and/or maintain for us: about 20% of our
entire source code base!

I think that 20% number will be different for different developers,
depending on factors like the domain they work in, the general style of
their code, the clarity of their design, and more.  So for some
developers the concept tool would handle less than 20%, and for others
the tool would handle more.  Personally, I'm hopeful that for the great
majority of C++ developers, the tool would handle at least 10-15% of
their source code base.

Rick
-- 
Richard M Kelly                             [email protected]
[PGP  =>  D5 C3 CC D2 B7 D0 A9 B0  D5 90 B2 55 5A 80 23 B6  FC AB 67 B4]