Use state machine to parse directives

[djc]

Intended to fix #3174. Have not done any benchmarks yet -- suggestions on how best to approach that are welcome.

This is a little bit more code, but IMO maybe slightly more readable?

It's correct enough that it passes the tests, there might be edge cases that aren't covered but those should be easy to solve.

[dpc]

> hyperfine --warmup 30 -N ./target/release/dotr-before ./target/release/dotr-after
Benchmark 1: ./target/release/dotr-before
  Time (mean ± σ):       1.2 ms ±   0.1 ms    [User: 0.4 ms, System: 0.7 ms]
  Range (min … max):     1.1 ms …   1.9 ms    2044 runs

Benchmark 2: ./target/release/dotr-after
  Time (mean ± σ):     613.8 µs ±  79.6 µs    [User: 289.1 µs, System: 254.0 µs]
  Range (min … max):   511.5 µs … 979.8 µs    5264 runs

Summary
  ./target/release/dotr-after ran
    2.03 ± 0.34 times faster than ./target/release/dotr-before 

Added true just to have a baseline:

Benchmark 1: ./target/release/dotr-before
  Time (mean ± σ):       1.2 ms ±   0.1 ms    [User: 0.4 ms, System: 0.7 ms]
  Range (min … max):     1.1 ms …   1.8 ms    2535 runs

Benchmark 2: ./target/release/dotr-after
  Time (mean ± σ):     611.7 µs ±  80.3 µs    [User: 295.6 µs, System: 244.1 µs]
  Range (min … max):   518.8 µs … 1663.5 µs    3825 runs

Benchmark 3: true
  Time (mean ± σ):     595.8 µs ±  74.1 µs    [User: 311.4 µs, System: 215.4 µs]
  Range (min … max):   512.3 µs … 1035.5 µs    5362 runs

This app sets RUST_LOG internally, so setting it from the outside doesn't make a difference.

AFAICT you fixed it and it's nearly free now.

[klensy]

After this, regex can be removed from Cargo.toml (for tracing-subscriber), only left in dev-deps?

[djc]

After this, regex can be removed from Cargo.toml (for tracing-subscriber), only left in dev-deps?

Nice catch, yes! Added that into the first commit.

[hds]

Could you please rebase this against master. We merge everything in there first and then David or I will handle backporting to v0.1.x.

Before making this change, I think we need more tests to ensure that the behavior isn't changing. Especially because there are things I'm seeing in the code that are different to what the docs say.

I'm happy to have a look at writing some of those tests, but it may take me a little bit to get to it.

[hds]

Let's keep the comment that explains why we're picking TRACE:

Suggested change

	Err(_) => (LevelFilter::TRACE, Some(level_or_target.to_owned())),
	// Setting the target without the level enables every level for that target
	Err(_) => (LevelFilter::TRACE, Some(level_or_target.to_owned())),

[hds]

This appears to match the previous behavior. But it doesn't match the docs (unless this behavior is performed elsewhere):
https://docs.rs/tracing-subscriber/latest/tracing_subscriber/filter/struct.EnvFilter.html#method.from_default_env

I actually think that before we release this change, we need some more test coverage so we can understand what is really changing. I know this is probably more than you signed up for, so I can try to set some time aside to write some of those tests too if that would help.

[djc]

I actually think that before we release this change, we need some more test coverage so we can understand what is really changing. I know this is probably more than you signed up for, so I can try to set some time aside to write some of those tests too if that would help.

I don't mind generating more test coverage, can you just make more explicit what you think we need more test coverage for? (As detailed as you want.)