I find the autoformatting that we currently use in cirq to be less than ideal. There are three main issues:

• yapf formatting is often bad, particular with things like nested tuples but also due to mixing hanging versus 4-space indentation styles (here's an example of mixed indentation)
• 80 character lines seem to be too short, particularly with heavy use of type annotations as we have in our code.
• The incremental formatting is line-based, so we have lots of old code that has never been formatted.

Describe the solution you'd like

• Switch formatter from yapf to black. Black's formatting algorithm is simpler and the resulting formatting is more predictable and regular than yapf's, IME.
• Increase the line length to something like 100 characters. I think this would cut down on a lot of unnecessary line breaking and improve readability of the code.
• Switch the incremental autoformatting to be file-based instead of line-based. As part of doing this, we would have a one-time change to format the entire code base, which would be a massive diff but entirely automated. There's also a bit of work to update PRs that are in-flight at the time of the big formatting change, but this too is manageable.

Describe alternatives/workarounds you've considered

• Could increase line length without switching from yapf to black.
• Could tweak various yapf knobs to possibly improve the formatting (I prefer black's approach of having fewer knobs in the first place so there's less to argue about, but YMMV).
• Could switch to file-based incremental formatting and pay the one-time cost of formatting all old code without switching to black.

What is the urgency from your perspective for this issue? Is it blocking important work?

P3 - I'm not really blocked by it, it is an idea I'd like to discuss / suggestion based on principle