[m-dev.] Array modes

Peter Wang novalazy at gmail.com
Mon Oct 27 11:09:27 AEDT 2014


On Sat, 25 Oct 2014 20:24:50 +1100 (EST), "Zoltan Somogyi" <zoltan.somogyi at runbox.com> wrote:
> 
> 
> On Sat, 25 Oct 2014 19:33:13 +1100, Peter Wang <novalazy at gmail.com> wrote:
> > I understood `unique' to mean unique at every node, by the analogy with
> > `ground' meaning bound at every node.  Page 59 of dmo's thesis supports
> > this; at least, I can't read it any other way.
> 
> People who want to know what "unique" means won't be looking
> in a thesis.

It should be stated in the reference manual.

> 
> The problem is that I don't think the original question has
> a clearcut answer either way. Because we don't have alias
> tracking, uniqueness has only really been useful for values
> of types which have only one OR node. For these, as I said,
> the two notions of uniqueness coincide, and there simply
> hasn't been a need to distinguish between them. The only
> exception to this that I can think of right now is the work
> on ctgc, which manipulates insts it discovers and creates
> automatically, not insts written by a programmer. The
> mapping between compiler-internal insts and how insts
> appear in source code therefore isn't very relevant
> there either.

I don't know how relevant these are either, then.  Nonetheless, it
answers Mark's question about what is the intended meaning of unique:

http://www.mercurylang.org/list-archives/users/2002-May/001749.html
http://www.mercurylang.org/list-archives/users/2005-September/003264.html

Peter



More information about the developers mailing list