Owo color

[stefnotch]

Motivation

Fix #2759

Also update docs to fix https://github.com/tokio-rs/tracing/issues/2268

Solution

Replaced nu-ansi-term with owo-colors.

I then ran all the unit tests cargo test, and tested the cargo run --example sloggish and cargo run --example fmt-pretty examples. I did my best to verify that the output hasn't changed.

Questions

During development, rust-analyzer always shows me bogus error messages and I'm not entirely sure what I'm doing wrong. I frequently ended up having to work with cargo check and a bit of guesswork.

Old fmt-pretty output

New, visually unchanged fmt-pretty output

[stefnotch]

@mladedav Thank you for the excellent review!

To answer your question, yes, it is possible to get the prefix-suffix pattern to work. I had that, and removed it in this commit https://github.com/tokio-rs/tracing/pull/2914/commits/51758ee3e5f83e7857f32e0ae3f60fb2ae18735a That approach basically (ab)used a dummy struct to call the relevant owo-colors functions.

There's also an alternative approach, along the lines of

struct FormatPair<'a, 'b, A, B>(&'a A, &'b B, Style);

impl fmt::Display for FormatPair {
  // print the style prefix
  // print a
  // print b
  // print the style suffix
}

I could absolutely add one or more of those approaches, if the extra complexity is worth it.

[stefnotch]

In case a reviewer is curious, I also tried out an alternative design, where a Style class also manages the is_ansi flag. https://github.com/tokio-rs/tracing/compare/master...stefnotch:tracing:owo-color-alternate-design#diff-8f6911488e5b3b7a50b97db5ab61c50af439560e330a54015889651c63a0dc52

[stefnotch]

Is there any chance of getting this merged anytime soon? Or would a different direction/dependency be more useful for the tracing project?