Bug in summation of toric divisors

[novoselt]

Here is the problem:

sage: P1 = toric_varieties.P1()
sage: print sum([P1.divisor(1)])
V(s) + V(t)

Which I think comes from

sage: print P1.divisor(1).parent()(0)
V(s)

Perhaps, we can allow creation of divisors by indices of rays/variables, but only in the form P1.divisor(0)? Since zero in the group of divisors should be just zero and other integers don't really make sense.

[novoselt]

Hi Volker,

I looked through the code and a possible solution is to catch integers in _element_constructor_ of ToricDivisorGroup instead of passing them down to ToricDivisor. "Catch" means replace 0 with None to construct the trivial divisor and raise an exception for anything else.

It is not very elegant, but keeps current functionality and X.divisor(i) returning the divisor of the i-th ray still seems quite reasonable to me, while X.divisor_group()(i) is questionable.

If you agree, I'll make the change and try to make it clear in the documentation of all the methods involved.

[vbraun]

The whole ToricDivisor(X, int) construction is a bit wonky. Thinking about it, the fact that ToricDivisor(X,0) is not the trivial element in this Abelian group is just confusing. I'd prefer to get rid of integer arguments in it altogether, there are already enough ways to construct a toric divisor. And you can always replace it with ToricDivisorGroup(X).gen(int), so its not like its a big loss. I'd like to keep the convenience method X.divisor(i), mind you. What do you think?

[novoselt]

Sounds good! Will post a patch over the weekend.

[vbraun]

Looks good!

[novoselt]

Why can't this go into 5.0? It is a bugfix, not an enhancement, and it is small and non-invasive.

[jdemeyer]

sage-5.0.beta15 == sage-5.0.rc0