Align ANSI.color() with ANSI.red() formatting

[bougueil]

Align ANSI.color() with ANSI.red() ansi formatting for disabling and append_reset

• Disabling ansi color strings:

IO.ANSI.format(IO.ANSI.color(0), false) |> to_string == ""

• Add append_reset for color strings:

[IO.ANSI.format([IO.ANSI.color(3), " color(3) "]), " (no color)"] |> to_string ==
"#{IO.ANSI.color(3)} color(3) #{IO.ANSI.reset()} (no color)"

This change shows a 15% slowness for this bench:

Mix.install([{:benchee, "~> 1.5"}])

data = [:red, "message "]

Benchee.run(
  %{"IO.ANSI.format/1" => &IO.ANSI.format/1},
  inputs:   %{
    "100 ansi code messages" => List.duplicate(data, 100)
    },
  warmup: 2,
  time: 5,
  memory_time: 2
)

[josevalim]

Apologies, but I don't understand what this pull request is doing and which bug it is meant to fix.

[bougueil]

Apologies,

I had issue with current elixir with these 2 statements:

IO.ANSI.format(IO.ANSI.color(0), false) |> to_string 

which returns "\e[38;5;0m" but "" is expected.

[IO.ANSI.format([IO.ANSI.color(3), " color(3) "]), " (no color)"] |> to_string 

which returns "\e[38;5;3m color(3) (no color)" but "#{IO.ANSI.color(3)} color(3) #{IO.ANSI.reset()} (no color)" is expected.

[josevalim]

I see. Thank you.

We only skip the coloring for atoms. We would need to support {:color, ...} notation for you to achieve what you want (or you check for ansi_enabled? yourself, probably preferable).

[bougueil]

yeah, it's possible to add afterwards ANSI.reset and avoid color leakage appearing in the 2nd statement :

[IO.ANSI.format([IO.ANSI.color(3), " color(3) "]), " (no color)"] |> IO.puts

but ...