I like this a lot better than the first.
http://www.artima.com/weblogs/viewpost.jsp?thread=86641
Now, the base objective seems to be to incorporate PyChecker functionality into the root. This in turn requires type inference, which in turn strongly suggests type annotations to help the inferencer out over rough spots.
I like this approach a lot.
There's also an explicit recognition that there are a lot of use cases for being able to access type information at run time.
I also like this. I'm kind of neutral to the idea of using it for actual type checking: I don't think that it catches that many errors if the developer is doing a decent job of testing. However, since it's going to be supplemental to type inference, it will be a lot less intrusive.
There's a new approach to interfaces that looks intriguing. One nice part is the ability to include Design By Contract type preconditions with the interface. It will be even better if they can be checked at compile time, or by the PyChecker functionality.
I have to say this turned my attitude around on the subject: I'm quite in favor of the direction Guido seems to be going.
John Roth
-- http://mail.python.org/mailman/listinfo/python-list
