[m-dev.] for review: changes to library/map.m

[Zoltan Somogyi]

> The fact that you have to recompile the program in a grade that
> supports stack traces in order to get a stack trace is an implementation
> detail of the current implementation.  It is not necessarily
> true of all implementations -- for example, if we ever develop a Java
> backend, then it won't be true.  We shouldn't base the design of
> the standard library on those kinds of things.

You are proposing instead that we base the design of the standard library
on something we *might* do in the future?

> > If anything, I think that map__lookup etc should be changed to take
> > an error message argument.  We can keep the exiting versions for backward
> > compatibility.
> 
> The transition cost would be very high.  I don't think this is worth it.

The cost would SMOP. A lot of SMOP, but just SMOP. The potential benefit
is significant, and I for one think it would be worthwhile.

> And adding an error message argument to all those predicates
> (there are about 30 of them in the standard library)
> would be a major change which is not an appropriate thing to do
> while we have a feature freeze on.

I was not proposing doing it now.

Zoltan.