Fix method ambiguity error of NaNMath.pow
#690
Conversation
|
Hmm seems not sufficient... |
src/methods.jl
Outdated
| @@ -98,6 +98,9 @@ end | |||
| @number_methods(BasicSymbolic{<:Number}, term(f, a), term(f, a, b), skipbasics) | |||
| @number_methods(BasicSymbolic{<:LiteralReal}, term(f, a), term(f, a, b), onlybasics) | |||
|
|
|||
| # Fix method ambiguity issue in NaNMath >= 1.0.2 | |||
| NaNMath.pow(x::BasicSymbolic{<:Number}, y::Integer) = x^y | |||
There was a problem hiding this comment.
| NaNMath.pow(x::BasicSymbolic{<:Number}, y::Integer) = x^y | |
| NaNMath.pow(x::BasicSymbolic{<:Number}, y::Integer) = maketerm(typeof(x), NaNMath.pow, [x, y], metadata(x)) |
There was a problem hiding this comment.
We want to avoid NaNMath in that case in NaNMath < 1.0.2 and on NaNMath >= 1.0.2 an Integer as second argument falls back to ^ anyway. So I think using ^ is the better and safer approach, regardless of the NaNMath version.
There was a problem hiding this comment.
Then can we add this to @number_methods?
There was a problem hiding this comment.
I did but I guess incorrectly. The original error is gone but it seems it causes https://github.com/JuliaSymbolics/SymbolicUtils.jl/actions/runs/12708708876/job/35426212172?pr=690#step:6:885.
|
In fact, this should instead be a change to |
|
Actually, never mind 😅 Symbolics is the only one with the ambiguity, right? Because |
|
I think a couple of the tests need to be updated |
NaNMath 1.0.2 introduced a method ambiguity error.