On 5/15/06, Barton Willis <willisb at unk.edu> wrote:
> Plan A:
Well, I think rectform ought to respect featurep (., real) rather
than the other way around.
> Plan B:
I think I like this plan better.
> realvalued and mapsrealstoreals aren't features, but they could be.
mapsrealstoreals is OK by me, butdowereallyhavetosmashthewordstogether.
I don't understand why realvalued is needed. If foo is a real-valued
function, why can't declare (foo, real) be enough? (Aside from not
working as expected -- that, we can fix.) I know there is already
a declaration "integervalued" and I have the same question about it.
> The functions realpart, imagpart, ... would be declared to be realvalued.
I'd rather declare ([realpart, imagpart], real) ...
> We'd need to declare %i to be imaginary and infinity to be complex.
OK by me on both counts.
> Condition (3) makes featurep(x[%i], real) --> true. Oh whatever....
If featurep (x, real) = true, then featurep (x[foo], real) = true
for any subscript foo is just what we want, isn't it?
I don't think foo = %i is problematic here.
Plan C is to separate the simple declaration-testing stuff from
the stuff we really want, which is something like is (X in A)
where A is a symbolic set. featurep has evolved in that direction,
because it's useful; at some point we should finish the job.
best,
Robert