julia - Interpolations.jl throws "ERROR: InexactError()" for vector, not for point

Question

I'm interpolating over a 2d matrix J:

J_itp = interpolate(J, (BSpline(Linear()), NoInterp()), OnGrid())

As you see, I'm only interpolating along the first dimension. The 2nd dimension is required to be on the grid. I have a vector of points I'd like to interpolate at.

julia> Np_args[1,67:73]
1x7 Array{Float64,2}:
1.0  1.0  1.0  1.0  1.01394  1.03275  1.0517

I call the interpolation function like so:

J_itp[ Np_args[1,67:73], 1]

and get the following error

ERROR: InexactError()
in trunc at /Applications/Julia-    0.4.3.app/Contents/Resources/julia/lib/julia/sys.dylib
[inlined code] from /Users/btengels/.julia/v0.4/Interpolations/src/b-splines/linear.jl:6
in getindex at /Users/btengels/.julia/v0.4/Interpolations/src/b-splines/indexing.jl:39

I can, however, interpolate using the lower half of my inputs where equal to one. So it can't be a type issue (as far as I understand).

julia> J_itp[ Np_args[1,67:70], 1]
4-element Array{Float64,1}:
0.75366
0.75366
0.75366
0.75366

Things crash when I include any of the points greater than 1.0, as in the first example. I can, however, interpolate at one of those trouble points by itself.

julia> J_itp[ Np_args[1,71], 1]
0.753702904230203

But if in an array as in J_itp[ Np_args[1,71]*ones(3,1), 1], things crash as before.

Is this a bug or am I doing something wrong? Recently installed with Pkg.add("Interpolations"), OSX - yosemite. Julia version 0.4.3

Answer 1

It's a little hard to say exactly, because I don't know what element type J has, and therefore it's not quite enough information to reconstruct your problem. However, I was able to trigger an InexactError from

julia> J = rand(1:10, 100,100);

julia> J_itp = interpolate(J, (BSpline(Linear()), NoInterp()), OnGrid());

julia> J_itp[rand(5)+10, 1]
ERROR: InexactError()
 in to_index at deprecated.jl:454
 in _unsafe_getindex at multidimensional.jl:192
 in getindex at abstractarray.jl:488

but this works:

Float64[J_itp[x, 1] for x in rand(5)+10]

(You need the Float64 if you're running this in global scope; if you put it in a function, you shouldn't need it.)

At least in the case I tested, the error is being generated by julia, not Interpolations; the issue is that indexing by vectors is supported in julia, and it wants you to use integer-valued indexes. This may (or may not) change in julia-0.5. This bypasses vector-indexing by using a comprehension.