Matlab spline function in Julia

Question

I am trying to interpolate two points in Julia, using the same approach of Matlab ( https://uk.mathworks.com/help/matlab/ref/spline.html ). I have tried the Interpolations ( https://github.com/tlycken/Interpolations.jl ) library, but I am having several issues in creating a working script.

I have a DataArrays.DataArray{Float64,1} with two points (let's say 1.5 and 10.5) and 5 NA between them:

using DataFrames
using Interpolations

a = @data([1.5, NA, NA, NA, NA, NA, 10.5]);

In Matlab it would be sufficient to run the spline function. In Julia, the interpolate function allows cubic interpolations. However, it seems not to be working with NAs. How can I do it? Also, do they use the same / an analogous algorithm for interpolate those points?

Answer 1

If you make the assumption that they are evenly spaced, then you are assuming a linear interpolation for which you can use linspace. You just need the start, the end, and the number of values inbetween:

linspace(a[1],a[end],sum(isna(a)))

More generally, to do an interpolation between any NAs, you can find the non-NA values with

idxs = find(~isna(a))

and then do

for i in 1:length(idxs)-1
  tmpidxs = idxs[i]:idxs[i+1]
  a[idxs[i]+1:idxs[i+1]-1] = linspace(a[idxs[i]],a[idxs[i+1]],length(tmpidxs))[2:end-1]
end

You can clean that up or put it in a function if you want. It's an odd assumption though to assume that between each known value it's supposed to be linearly spaced.