Re: Major change proposal

Mark Hahn wrote:
>...
> 
> Does anyone know what the current state of the performance analysis tools
> for the CLR runtime is?  Am I going to have to be as good as Jim Hugunin to
> get Prothon to run fast?

One hopes that now that they are on the second or third version of 
Visual Studio.NET that there are good profilers...but I'm no expert. 
Maybe you should join a mono mailing list: they'll be happy to hear 
about your current thoughts and would love to help you.

> Of course I have one big advantage over ActiveState.  I have an existence
> proof.  I know it can be done.

And as you say later: the code is open source. Rip off ideas.

> At first I thought Python would have an unfair advantage over Prothon in
> implementation since Python is class-oriented and CLR is class-oriented, but
> then I realized that the same phenomenom that caused me to discover that
> Python can do anything that Prothon can do over the last four months is
> going to cause Python to have the same implementation difficulties that
> Prothon will have in the CLR.  This assumes that IronPython will implement
> the full Python capabilities, metaclasses and all.

IronPython's test suite includes a lot of Dynamic Python stress test 
stuff. Here's some pretty nasty dynamic stuff:

def instrumentClass(cls):
     cname = cls.__name__ + '.'
     for name in cls.__dict__:
         if name == '__dict__':
             continue
         descr = cls.__dict__[name]
         if hasattr(descr, '__get__'):
             setattr(cls, name, instrumentDescriptor(cname+name, descr))

  Paul Prescod