How to do equivalent of an irfft?

[OliverDudgeon]

Hi, thanks for your work on this package. I was wondering if it's possible, how I could do an irfft. For example, currently for an equidistantly spaced signal I do the following:

using FFTW
irfft(data, 2length(data)-1)

What would be the closest analogue with NFFT.jl? data will eventually have a non-equidistant spacing.

[tknopp]

The inverse NFFT does not exist analytically. But there is an approximation that is outlined here:
https://juliamath.github.io/NFFT.jl/dev/tools/#Sampling-Density
Programing-wise you would first calculate the density weights, then apply them and finally do the adjoint NFFT.

We do this in MRIReco:
https://github.com/MagneticResonanceImaging/MRIReco.jl/blob/master/src/Reconstruction/DirectReconstruction.jl#L35
and the line afterwards.