Allow convolve to work with Real arrays with arbitrary precision (#159)

* Allow `convolve!` with real arrays

* Allow adjoint to work with real vectors too

* Refine type declarations

* To v0.14.2

* Test convolve!

* size checks

* Add convolve_transpose tests and real tests

* Fix size to Ñ in comment

* Multiple dispatch for complex/real mismatch

* Fix

Author: JeffFessler

AbstractNFFTs/src/interface.jl

@@ -225,7 +225,7 @@ Overwrite `R`-dimensional array `fHat`
 (often R=1, and `fHat` has length `p.J`)
 with the result of "convolution" (i.e., interpolation)
 between `D`-dimensional equispaced input array `g`
-(often of size `p.N`)
+(often of size `p.Ñ`)
 and the interpolation kernel in NFFT plan `p`.
 """
 @mustimplement convolve!(p::AbstractNFFTPlan, g::AbstractArray, fHat::AbstractArray)

Project.toml

@@ -1,7 +1,7 @@
 name = "NFFT"
 uuid = "efe261a4-0d2b-5849-be55-fc731d526b0d"
-authors = ["Tobias Knopp <[email protected]>"]
-version = "0.14.1"
+authors = ["Tobias Knopp <[email protected]> and contributors"]
+version = "0.14.2"
 
 [deps]
 AbstractNFFTs = "7f219486-4aa7-41d6-80a7-e08ef20ceed7"