xexpy(0, 1000) should be zero

[Jorge Fernandez-de-Cossio-Diaz]

julia> xexpy(0, 1000)
NaN

I think xexpy can handle this better and return zero?

[Jorge Fernandez-de-Cossio-Diaz]

Fixed by https://github.com/JuliaStats/LogExpFunctions.jl/pull/80/files

[David Widmann]

Why should it be 0? At least the current implementation agrees with the docstring.

[Jorge Fernandez-de-Cossio-Diaz]

Mathematically, 0 * exp(1000) = 0. Why should it not be zero?

[Tamas K. Papp]

I also think that we should make xexpy finite if y is finite and x is zero. The purpose of writing a function is precisely to handle these kind of cases.

[David Widmann]

Fixed by #80.