[FEATURE REQUEST] Fallbacks when true color or ANSI 256 is not supported about ansis

[webdiscus]

Hallo @RobinTail,

Thanks for the feature request!

1. I will check the issue with default color in the next output.
2. Yes, it is very important to have a fallback for a terminal supported 256 colors only.
   I will implement it in the next version.

Grüß aus Köln

[RobinTail]

Yes, it is very important to have a fallback for a terminal supported 256 colors only.
I will implement it in the next version.

Thank you, @webdiscus
Ideally it should have fallbacks down to the minimal supported palette.

True color —> 256 colors —> 16 colors —> no colors.

For example, doing hex("#5BCEFA") in environment having 16 colors only, it should fall back approximately to cyan.

[webdiscus]

@RobinTail
I have implemented the detection of color spaces: truecolor, 256 colors, 16 colors, mono.
The fallback truecolor —> 256 colors —> 16 colors —> no colors works too.
I'm by testing and refactoring.

[webdiscus]

@RobinTail

the new version 3.0.0 is released with fallback to supported color space.

[RobinTail]

Awesome! I'm going to try it today, @webdiscus

[RobinTail]

Dear @webdiscus ,
It works perfectly! Thank you so much for all the effort you applied to this feature and your attention to details.
I highly appreciate that you measured the performance and updated documentation.
Fallbacks, named imports, CommonJS compatibility — everything works well.
I tested it in iTerm, Terminal of MacOS and GitHub CI.

[webdiscus]

@RobinTail

Thank You for the very useful feature request!
This makes the library even better.

I have invested a time to optimize performance and size for the new features.
New features adds in bundle only ~600 bytes and the code bundle is still < 4 KB.

[webdiscus]

vitest now also has benchmarking tools: https://vitest.dev/api/#bench

Thanks @RobinTail for the info.

Yes, I know, and I will switch the benchmark to vitest/bench, but in next step. First I wanted to release the feature and then optimize the benchmark.

Vitest is a very cool tools. This allow to test ESM modules "out of the box". Jest doesn't work with ESM in JetBrains IDEA, so I switched to vitest.

[RobinTail]

New features adds in bundle only ~600 bytes and the code bundle is still < 4 KB.

That's a great result, @webdiscus !
I highly appreciate your passion.
I think the issue is resolved, so I'm going to close it.
Thank you so much for all the support :)

[webdiscus]

@RobinTail

I have create some simple tests suits using vitest bench.

I am saddened :-(

1. it's still very raw and is not yet released:

   Benchmarking is an experimental feature.
   Breaking changes might not follow SemVer, please pin Vitest's version when using it.
   

2. it's very very buggy and does not work stably. Using only 4 suits occurs fatal error:

   Allocation failed - JavaScript heap out of memory
   

3. test results are false and cannot be trusted, e.g., the simple bench for chalk.red('foo') and ansis.red('foo') have results:

   chalk  7.000.000
   ansis 23.000.000 (x3 faster than chalk - this is unreal/FALSE result)
   

   To compare with the result of the benchmark:

   chalk 73.000.000
   ansis 69.000.000 (this is TRUE result)
   

So, we cannot trust the results of the vitest bench.
We need to wait until the first release appears.

[RobinTail]

Alright, you seem to made a deep research on that topic.
Perhaps it's not yet ready for the project of your scale.
Regarding the accuracy of the measurements.
I think it's different because vitest runs the tests against the transformed code (using esbuild and rollup under the hood). Perhaps it does that transformation differently for ansis and chalk.
I hope it will get better. Thanks for letting me know, @webdiscus

[RobinTail]

I'm glad to read it, @webdiscus , great news!
Let me know if you need any help.

[RobinTail]

By the way, @webdiscus
I noticed you also switched to vitest, and I'd like to inform you that vitest now also has benchmarking tools:
https://vitest.dev/api/#bench

So that you can have a single dependency both for testing and benchmarking.
I use it myself and I like how it works.