`Inf` does not work in `quadgk!()`

I tried to use quadgk!(), which is the in-place quadgk(), but encountered problems. Simply put, quadgk!() does not accpet Inf or -Inf in the domain boundary while it is not a problem with quadgk(). The following is a MWE.

julia> using QuadGK, StatsFuns

julia> f(x) = normpdf(x)
f (generic function with 1 method)

julia> f!(y,x) = y .= f(x)
f! (generic function with 1 method)

julia> int_g = zeros(1)
1-element Vector{Float64}:
 0.0

julia> # without `Inf` both are fine

julia> quadgk(f, 0.0, 1.0)
(0.341344746068543, 9.992007221626409e-16)

julia> quadgk!(f!, int_g, 0.0, 1.0)
([0.341344746068543], 9.992007221626409e-16)

julia> # with `Inf` problem arises

julia> quadgk(f, 0.0, Inf)  # fine
(0.4999999999999999, 5.67681333382849e-9)

julia> quadgk!(f!, int_g, 0.0, Inf)  # not work
ERROR: MethodError: objects of type QuadGK.InplaceIntegrand{typeof(f!), Vector{Float64}, Vector{Float64}} are not callable

Is it a bug or a design feature? If it is the former, could it be fixed? Thanks.

It’s a bug; we didn’t handle the infinite-domain case when implementing the in-place API.

It’s fixable, but in the meantime you can just do the coordinate transformation to a finite domain manually.

1 Like

Got it. Thanks!