From: gabriele renzi <surrender_it@...1.vip.ukl.yahoo.com>
Date: 2004-03-25T19:54:26+09:00
Subject: About type inference for ruby

Hi gurus and nubys,

I'm going to put some memes toghether running away in few minutes, to
see what the community thinks.

I just run across this paper[1] from latest PyCon
that explains a way to compile Python to actually efficient C++ 
using automatic type inference. 

I remember matz and others were someway interested in Type/interface
inference, (see this[2] and this[3])
and I, FWIW, would love to see type inference in ruby.

It seem that in the thread referenced by [3] matz pointed out that
stuff like this:

 def whatiwould(x, flags)
    if flags
      x = Regex.new(x.join("|"))
    end
    x.match(@str)
  end

would be actually too hard to analyze for interface/duck signature
inference. 

And this relates to blogging from PragDave[4] and Chad Fowler[5]
I which appears the interesting meme that "we don't want to check
variable usage anytime, we would just check that once a variable has a
type, it does not change substantially".

Actually I believe too, that you don't need and should not do
x=12
x='str'

Would it make sense to have a flag to say to the interpreter "raise an
exception if this variable gets a new assignment" and then do type
inference?



[1]
http://www.python.org/pycon/dc2004/papers/1/paper.pdf
[2]
http://blade.nagaokaut.ac.jp/cgi-bin/vframe.rb/ruby/ruby-talk/1437?1215-1527+split-mode-vertical
[3]
that's a *huge* thread.. 
http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-talk/85890
[4]
http://www.pragprog.com/pragdave/Tech/Ruby/LazyType.rdoc
[5]
http://www.chadfowler.com/index.cgi/Computing/Programming/Ruby/TypeWatching.rdoc,v