Topher Cawlfield

I'm using meirumeiru's clone, which appears to work well with 1.4. See the pull request.

I added support for `angle` along with `real` and `imag` in the branch 2917-npreal-npimag-npround-nparound-cant-take-an-awkward-array-as-input. I have not implemented `real_if_close` yet, which has the same issues. Looking into `round` and `around`....

I added a unit test and did some preliminary characterization. See https://github.com/scikit-hep/awkward/blob/9c66fb23e436d5631b794ba7c5e3f96b181aebc4/tests/test_3098_ak_where_with_arrays_containing_optionals.py.

Spark's use of backticks seems convenient, apparently borrowing from Markdown. But it brings up a number of related issues and a host of complexities that would be time-consuming to document...

Okay, I like the idea that if we provide a list of lists of strings, the user is being explicit and not using patterns. Patterns are a convenience only, and...

Working on the code, I found that it would take a much larger change to insist that column-names-as-a-tuple (or list) will not use pattern-matching. Although I prefer that approach from...

About escapes: Yes, I think that's another valid approach. Special characters `.*?[]{}\` could all be escaped, or maybe even all punctuation a-la Markdown. I suppose that's easy to do with...

> It's looking good! The next step would be to make `ak.to_arrow_table` and `ak.from_arrow_table` (and, by extension, `ak.to_parquet` and `ak.from_parquet`) do this automatically. Then a lot of the existing tests...

I'm marking this as ready for review, even though there's failures on Python 3.8 still. I'll look into what it takes to reproduce this, but so far I'm hitting errors...

Okay I think this is all ready for review now, with tests passing! > Is the difficulty of making an environment with pyarrow 7 related to conda versus pip venv...