setindex! for Fac

[SoongNoonien]

I'm a bit confused by the behavior of setindex! for Fac. Currently I see no advantage over mulpow!. I'd expect to be able to change the exponents of factors already included, especially considering the docstring:

If $b$ is a factor of $a$, the corresponding entry is set to $c$.

What's also a bit weird, is that with setindex! non-positive exponents can be set. Resulting in interesting outputs:

julia> fac=factor(q^2-1)
1 * (q - 1) * (q + 1)

julia> setindex!(fac, 0, q);

julia> fac
1 * (q - 1) * q^0 * (q + 1)

julia> setindex!(fac, -1, q-2);

julia> fac
1 * (q - 1) * (q - 2)^-1 * q^0 * (q + 1)

julia> evaluate(fac)
ERROR: DomainError with -1:
Exponent must be non-negative
Stacktrace:
 [1] ^(x::QQPolyRingElem, y::Int64)
   @ Nemo ~/.julia/packages/Nemo/y7UoI/src/flint/fmpq_poly.jl:166
 [2] evaluate(a::Fac{QQPolyRingElem})
   @ AbstractAlgebra ~/.julia/packages/AbstractAlgebra/HxIHB/src/Factor.jl:56
 [3] top-level scope
   @ REPL[67]:1

[thofma]

In Hecke:

julia> factor(ZZ, QQ(1//100))
1 * 5^-2 * 2^-2

Regarding the setindex! and mulpow!. Yes, there is some overlapping functionality, but it is quite "natural" to have it given the other functionality for Fac.

[SoongNoonien]

Ok, but zero as exponent should never be useful, except setting zero removes the factor all together. And why is redefining of exponents not allowed?

[thofma]

Because someone thought it was a good idea when it was implemented. But I guess we could change it, if there is some interest.

[fieker]

the way it was used orignally: redefining an exponent meant a logic flaw in the code...

[fingolfin]

I understand this desire to prevent accidental changes to a Fac struct, but in that case wouldn't it make even more sense to disallow setindex! completely? To stop users from accidentally modifying the Fac at all?