Week 5

In our last meeting, my mentors and I talked about breaking up this summer's work into multiple major pull requests, as opposed to last year's enormous pull request which was merged toward the very end of the summer. It'll be nice to do this in pieces just to make sure everything gets into the master branch of Astropy as intended, so we're planning on getting a PR in very soon (we discussed a deadline of 1-2 weeks past last Wednesday's meeting). The idea is to have working code that handles changes to Table indices when the Table itself is modified, and after this PR we can focus more on speeding up the index system and adding more functionality.

With that in mind, I mostly spent this week working on previous #TODO's, fixing bugs, and generally getting ready for a PR. Having previously ignored some of the subtleties of Table and Column copying, I found it pretty irritating to ensure that indices are preserved/copied/deep copied as appropriate when doing things like constructing one Table from another, slicing a Table by rows, etc. -- mostly because there are some intricacies involved in subclassing `numpy.ndarray` that I wasn't aware of before running across them. Also, while I managed to get this working correctly, there might end up being relevant time bottlenecks we need to take into consideration.

I also moved the relevant tests for Table indices to a new file `test_index.py` (adding some new tests), and fixed a couple issues including a bug with `Table.remove_rows` when the argument passed is a slice object. For the actual indexing engine, I found a library called bintrees which provides C-based binary tree and red-black tree classes, so for now I'm using this as the default engine (with the optional bintrees dependency, and falling back on my pure-Python classes if the dependency isn't present). I'm looking forward to figuring out the plan for a PR at this Wednesday's meeting, and from there moving on to optimization and increasing functionality.