Chore: improve performance of indent rule

[not-an-aardvark]

What is the purpose of this pull request? (put an "X" next to item)

[x] Other, please explain:

What changes did you make? (Give an overview)

This updates the indent rule to be more performant by storing information on ranges of tokens rather than performing operations on many individual tokens one-at-a-time.

This results in a 19% end-to-end performance increase (!) according to our npm run perf benchmark.

End-to-end benchmark

master branch:

Single File:
  CPU Speed is 3100 with multiplier 13000000
  Performance Run #1:  6601.645335ms
  Performance Run #2:  6537.859522ms
  Performance Run #3:  6469.447266ms
  Performance Run #4:  6777.504465ms
  Performance Run #5:  6431.605835ms

  Performance budget exceeded: 6537.859522ms (limit: 4193.548387096775ms)

indent-performance branch:

Single File:
  CPU Speed is 3100 with multiplier 13000000
  Performance Run #1:  5287.6544109999995ms
  Performance Run #2:  5284.148783ms
  Performance Run #3:  5326.879172ms
  Performance Run #4:  5338.387146ms
  Performance Run #5:  5340.651033ms

  Performance budget exceeded: 5326.879172ms (limit: 4193.548387096775ms)

Rule-specific benchmark on ESLint codebase

master branch:

$ TIMING=1 eslint lib/ conf/ bin/ tests/ --no-eslintrc --rule 'indent: [error, 4, {SwitchCase: 1}]' --env=es6 --no-inline-config
Rule   | Time (ms) | Relative
:------|----------:|--------:
indent |  5448.737 |   100.0%

indent-performance branch:

$ TIMING=1 eslint lib/ conf/ bin/ tests/ --no-eslintrc --rule 'indent: [error, 4, {SwitchCase: 1}]' --env=es6 --no-inline-config
Rule   | Time (ms) | Relative
:------|----------:|--------:
indent |  4610.891 |   100.0%

Is there anything you'd like reviewers to focus on?

I tried to write descriptive comments to explain what's going on. Is there anything that's unclear or could be improved?

[eslintbot]

LGTM

[mention-bot]

@not-an-aardvark, thanks for your PR! By analyzing the history of the files in this pull request, we identified @vitorbal, @gyandeeps and @Trott to be potential reviewers.

[not-an-aardvark]

I marked this PR as WIP because this module appears to be unmaintained. Its readme also mentions that the immutable approach ends up hurting performance for consumers that don't need immutability. We might be able to get an even larger performance boost if we implement our own balanced BST (or use a different maintained module, but I had trouble finding another balanced BST module that has the features we need).

[not-an-aardvark]

I decided to unmark this as WIP because I looked for more balanced BST libraries and didn't find any good alternatives at the moment. I also tried implementing one myself for awhile, but balanced BSTs are relatively complicated and difficult to get right, so it's probably not worth adding an implementation to the repo if this module is working for us right now.

[eslintbot]

LGTM

[eslintbot]

LGTM

[ilyavolodin]

Very nice perf boost.

[eslintbot]

LGTM

[not-an-aardvark]

I've made a few small improvements -- now this improves end-to-end performance by 20%.

Benchmark

master:

Single File:
  CPU Speed is 3100 with multiplier 13000000
  Performance Run #1:  5877.433702ms
  Performance Run #2:  5805.863323ms
  Performance Run #3:  5738.451947ms
  Performance Run #4:  5712.380049ms
  Performance Run #5:  5754.480289ms

  Performance budget exceeded: 5754.480289ms (limit: 4193.548387096775ms)

indent-performance:

Single File:
  CPU Speed is 3100 with multiplier 13000000
  Performance Run #1:  4577.530063ms
  Performance Run #2:  4641.629504ms
  Performance Run #3:  4603.666983ms
  Performance Run #4:  4684.530936ms
  Performance Run #5:  4614.500893ms

  Performance budget exceeded: 4614.500893ms (limit: 4193.548387096775ms)

(I think my laptop has a different battery level than when I ran the benchmark in the original PR, so these two runs are comparable to each other, but they're not comparable to the original two runs in terms of absolute time.)