BooleanPolynomial.variables() consider "1" as a variable

[Bouillaguet]

This is in contradiction with other functions (BooleanPolynomial?.nvariables), and with the behavior of the "normal" class MPolynomial.

Example:

sage: R.<x,y,z> = BooleanPolynomialRing()
sage: p = R(1)
sage: p.variables()
(1,)
sage: p.nvariables()
0

Which differs from:

sage: R.<x,y,z> = GF(2)[]
sage: p = R(1)
sage: p.variables()
()
sage: p.nvariables()
0

[Bouillaguet]

patch that fixes the problem

[malb]

Hi, can you add a doctest that shows that the issue is resolved? I also CCed the PolyBoRi? developers to make sure there isn't some obscure reason for this behaviour.

[Bouillaguet]

Hi,

I actually did, in fact. I modified the existing docstring to verify that:

sage: P.one_element().variables()  
()

Should I add anything else?

[malb]

Sorry, my bad!

[AlexanderDreyer]

p.variables() is a Sage extension (see #1715), the corresponding p.vars_as_monomial() () already behaves like this. (I think in early versions of PolyBoRi p.vars_as_monomial() could become 0, but this was a design error.) So I can clearly give a positive review.