Add note regarding MSVC

[filmor]

This would have saved me quite a bit of time :)

[codecov-io]

Codecov Report

Merging #266 into master will not change coverage. The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #266   +/-   ##
=======================================
  Coverage   93.11%   93.11%           
=======================================
  Files          34       34           
  Lines        1903     1903           
  Branches      239      239           
=======================================
  Hits         1772     1772           
  Misses         97       97           
  Partials       34       34

---

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 11e9d66...f56e1e4. Read the comment docs.

[stefanv]

This should not be true; can you expand on the problem? We may be able to fix it quite easily.

[filmor]

Well, it is :)

The Lomb-Scargle code (e.g. https://github.com/cesium-ml/cesium/blob/master/cesium/features/_lomb_scargle.h#L81, but there are a lot of other places both in _lomb_scargle.h and _eigs.h) uses Variable Length Arrays which are explicitly not supported by MSVC (see https://stackoverflow.com/questions/5246900) as they are not valid C++.

If someone wanted to clean that up, the correct function to use instead would be https://msdn.microsoft.com/en-us/library/5471dc8s.aspx. Of course, if the size of those arrays could be limited in some way, you could just use staticly sized buffers instead.

[stefanv]

I believe you :)

I don't think there's any reason why we can't just allocate some memory here, instead of using the heap. Are you interested in helping out, or should I take a pass at it? This is the oldest part of the code base (came from a previous project), and hasn't seen much attention recently I'm afraid.

[filmor]

I currently don't have the time to work on this, a note in your README is all I can do right now ;)

If you are looking into this, consider dirching the implementation and instead using only scipy's or gatspy's. You depend on those anyway and this would make your package pure Python.

[stefanv]

We've thought about replacing this implementation, but it still offers some features not found in the others. But we'd certainly be open to that, eventually.