Numpy triangles array

[sccolbert]

This adds method to export the full triangles vector as a numpy array via a single memcopy. This avoids the overhead of converting to a list of python objects, and the overhead of iterating them into a numpy array. Since the lib already accepts buffer input, this is analogue for getting buffer output.

I also updated the binding the release the GIL where possible. This allows multiple independent triangulations to computed across Python threads.

[artem-ogre]

Hi @sccolbert and thanks for the PR. 🙂

I've done it slightly different, please take a look and let me know what you think.
There are three ways of accessing the data now:

              count        list        copy        view
vertices     200003   56.439 ms    0.118 ms    0.001 ms
triangles    400001  127.027 ms    0.398 ms    0.001 ms

• copy version is as you proposed, but there's also zero-copy read-only view which is as fast as it gets.
• I also refactored GIL lock release and made it safer by keeping a lock on a triangulation.

Let me know if you have any feedback, and I'll merge it otherwise.

[sccolbert]

Hey there @artem-ogre, thanks for looking! The reason I went with the copy approach instead of a view, is because I am using the output before erasing the super triangle, and further edits to the triangulation would invalidate the view. The reason I don't erase is that the internal book-keeping of the erase is slow compared to what I need from the vertices after accounting for the presence of the of the super triangle.

[sccolbert]

I think your updates are good. My only worry would be taking a view, doing something that would reallocate the internal vector memory, then reading from the view again, causing a corrupted view or a crash.

[sccolbert]

Oh, and now that I have your attention, I have a WASM build of the core C++ lib that I could contribute, if you like?

[artem-ogre]

I wouldn't maintain WASM build but would be interested to see it if you decide to put it in public. I've myself experimented with an idea a while ago.

[artem-ogre]

FYI @sccolbert I added you to the contributors list in #13

[sccolbert]

Thanks! I appreciate that.