Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!elroy.jpl.nasa.gov!sdd.hp.com!wuarchive!rex!ames!uhccux!munnari.oz.au!brolga!uqcspe!batserver.cs.uq.oz.au!farrell
From: farrell@batserver.cs.uq.oz.au (Friendless)
Newsgroups: comp.lang.functional
Subject: Re: Type Inference in ML (LML, actually)
Message-ID: <6809@uqcspe.cs.uq.oz.au>
Date: 17 Jan 91 06:53:38 GMT
References: <54161@eerie.acsu.Buffalo.EDU> <4353@undis.cs.chalmers.se> <6784@uqcspe.cs.uq.oz.au> <3603@bruce.cs.monash.OZ.AU>
Sender: news@uqcspe.cs.uq.oz.au
Reply-To: farrell@batserver.cs.uq.oz.au
Lines: 14

In <3603@bruce.cs.monash.OZ.AU> mmcg@bruce.cs.monash.OZ.AU (Mike McGaughey) writes:
>	let x = hd o hd in x
>will typecheck properly - to (List (List *a))->*a) - whereas
>	(\x.x o x) hd
>will not.  The problem with the second expression seems to be that the
>inference algorithm erroneously expects both of the 'x' values in the
>x o x to have the same type signature, 

  This is the correct behaviour of Milner typing - parameters only have one
type. I was surprised too when I found it out, but I think it complicates the
algorithm a lot to fix it.


Friendless