Fixed alloc_matmul to handle more than 3 dimensions.

[nhipsman]

Simple issue with the way Julia allocates arrays of higher dimension. Did this to fix #151 on JuliaMath/Interpolations.jl.

My first pull request (ever), so if I've broken any official or unofficial rules of contributing, I'd be appreciative if you let me know.

[codecov-io]

Codecov Report

Merging #7 into master will not change coverage.
The diff coverage is 100%.

@@           Coverage Diff           @@
##           master       #7   +/-   ##
=======================================
  Coverage   99.37%   99.37%           
=======================================
  Files           4        4           
  Lines         159      159           
=======================================
  Hits          158      158           
  Misses          1        1

Impacted Files	Coverage Δ
src/AxisAlgorithms.jl	85.71% <100%> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6cbd6b9...530dec4. Read the comment docs.

[timholy]

Thanks @nhipsman! Adding a test would be great, but for your first pull request this seems good enough to me. I'm honored!

[timholy]

I was looking into writing a test before merging this, and on julia 0.5 this already seems to work. Just to make sure I'm not missing something, are you using Julia 0.4?

[nhipsman]

Using 0.5.1. I get the error (that I mentioned in #151 on Interpolations.jl) running it on either Linux or macOS (haven't tried Windows yet). I have also recently updated my packages, and am using Interpolations v 0.6.0. Hmmm...

Regarding tests - I figured that would be a good idea, but I'm not sure how that's usually done - can you point me somewhere that explains the proper form for such a test?

[timholy]

Check out https://github.com/timholy/AxisAlgorithms.jl/blob/master/test/tridiag.jl. What I usually do is add the test first (so I know it's catching the error) and then make the fix. In this case I guess the easiest would be to copy/paste the error you reported in JuliaMath/Interpolations.jl#151. We'd need to add a file test/REQUIRE with the contents

Interpolations

so Interpolations.jl gets installed on Travis when we run the tests.

If this is confusing, I can merge this and add the test myself, but I'll give you a chance to decide for yourself.

[timholy]

Or perhaps the test/woodbury.jl file. Looks like if you just make the test matrix 4d rather than 3d, you trigger the same error. I was confused initially because I didn't realize it was specific for Woodbury matrices. EDIT: if you do that, you don't necessarily have to invoke Interpolations.

[nhipsman]

I think that's what you were looking for? Thanks for the guidance!

[timholy]

@nhipsman, I owe you a sincere apology. I missed that you had finished this off. I should have merged this ages ago. The new version makes this unnecessary, I'm afraid, so I will have to close this. I regret not merging your first PR!

[timholy]

Actually, in #13 I decided to commit the test component of this, and credited it to you.