double[] gauss ops don't run on sample data

[hinerm]

When testing the Ops use case I tried running other filter.gauss signatures.

When running a signature with an sigma array, it goes to double[] (instead of number in the single-sigma variants?) and those then do not work on the input test image (embryos). I am assuming this is an issue with type matching..?

[gselzer]

@hinerm what's the error? Are you using enough dimensions in your sigmas array? The underlying Op uses one sigma per dimension, so you need the length of the sigma list to be (at least) 3 if you're using the embryos test image.

I tried replicating with focal1 from the embryos data, and as long as I passed through a sigma array of 3 values it worked fine for me...

[hinerm]

It's unfortunate that there is no indication that a sigma for each dimension is needed - I guess a symptom of exposing the underlying ops directly instead of an abstraction that matches ops to the params.

Also, unfortunate that no feedback is given in the UI that the Op failed.

[hinerm]

Essentially duplicated by #234

[ctrueden]

@hinerm I do think we should improve the ArrayIndexOutOfBoundsException currently thrown by multi-sigma gauss ops, to instead throw IllegalArgumentException with an explanation that the gauss length needs to match number of image dimensions. Users might at least then understand how they can address the problem. But this is a change needed in imagej-ops. Do you agree? If so: you wanna do it, or shall I?

[ctrueden]

@hinerm There is an already an issue for this in the imagej-ops issue tracker: imagej/imagej-ops#646