>>>>> "Robert" == Robert Dodier <robert.dodier at gmail.com> writes:
Robert> On 2013-08-28, Raymond Toy <toy.raymond at gmail.com> wrote:
>> I think it would be nice if maxima indicated that somehow instead of
>> just returning a noun form. Users should have to trace internal
>> functions to figure that out.
I meant "shouldn't have to", of course.
Robert> Agreed. I wonder if Maxima can distinguish integrands which don't
Robert> evaluate to a number but which might (e.g. unbound variable), from
Robert> those which can't (e.g. floating point error), and print different
Robert> messages for those.
I think this is all buried in coerce-float-fun. For gcl, any errors
cause the function to return T. For other lisps, we do distinguish
between an arithmetic-error and a maxia-$error, but return T in both
cases anyway.
I'm not sure what the best way for coerce-float-fun to distinguish
between the two cases. And in any case, gcl can't so that seems not
an option.
For the quad routines, I think the best we can do now would be to
detect a non-numerical result and signal a maxima error.
Ray