Lambda the Ultimate

inactiveTopic Type inference a good thing?
started 7/14/2002; 8:12:26 AM - last post 7/18/2002; 1:32:34 PM
Ehud Lamm - Type inference a good thing?  blueArrow
7/14/2002; 8:12:26 AM (reads: 661, responses: 4)
Type inference a good thing?
A comp.lang.functional thread.

Nothing new in this thread. Still, this is an important question so feel free to share your views.

Posted to general by Ehud Lamm on 7/14/02; 8:13:20 AM

Ehud Lamm - Re: Type inference a good thing?  blueArrow
7/14/2002; 8:39:24 AM (reads: 671, responses: 0)
When I wrote my Component Libraries and Language Features paper I was of the opinion that type inference should be kept within modules. Module interfaces should by explicit.

Now I think that this approach may be too restrictive. I'd be happy to hear about the experience of others.

Patrick Logan - Re: Type inference a good thing?  blueArrow
7/14/2002; 7:46:22 PM (reads: 631, responses: 0)
Unfortunately I could only come up with a gut reaction that is not very thoughtful. But it's how I feel, and it has the advantage of being somewhat nostalgic.

Wouter van Oortmerssen - Re: Type inference a good thing?  blueArrow
7/15/2002; 7:48:22 AM (reads: 603, responses: 0)
I agree that module boundaries should be explicit. The major issue is that TI hurts an open-world compilation model, it is very easy for a type to generalise accidentally because of implementation changes.

Otherwise, I love TI, unlike some others I really feel code becomes more readable without types everywhere.

Ehud Lamm - Re: Type inference a good thing?  blueArrow
7/18/2002; 1:32:34 PM (reads: 459, responses: 0)
While we are discussing TI...

A student of mine came across some confusing remarks on the Net about the role of type checking in the presence of TI (e.g., in Haskell). I explained the interaction (and sent him to reread EOPL2...) but I was wondering if someone knows of a good/short explanation specifically of this (possibly confusing) point.