Consider

axiom

binomial(n,2)

**Type: **Expression(Integer)

axiom

binomial(n,n-2)

**Type: **Expression(Integer)

which should, by the definition of the binomial coefficient, give identical results.

The attached patch fixes this.

There is, however, a more fundamental issue to consider: Often, it is desirable that such expressions are not "expanded", but some thought needs to go into this. In my current application (in SPAD), I know a priori that a given `ex`

in `EXPR INT`

(possibly containing binomial coefficients or expressions of the form ) can always be retracted to a `FRAC POLY INT`

. So, in this case, I really need "full" expansion.

I think, MuPAD? and Maple have a `convert`

function for this. In Axiom, it really would be the responsibility of `retract`

/`retractIfCan`

. But that would mean to have at least one other property attached to operators, that tells Axiom how to `retract`

to various domains. I guess that possibly different functions would have to be called for different domains we would like to retract to...

Martin

binomial.patch