Consider supporting FORCE_COLOR environmental variable

[matklad]

This is more of a braindump rather than a well-considered feature request, as I am not an expert in the domain. The TL;DR is that there's FORCE_COLOR env variable, which is the opposite of NO_COLOR, and it might be nice to support it.

Background: I am writing a thing which spawns other processes (like cargo) via std::process::Command. The thing is definitely not a terminal emulator, so the subprocess stdout is normal pipe, rather than a TTY. My thing is also capable of parsing ANSI escapes, so it would very much like to receive colored output from subprocess. I wold love to have a general, cross-platform way to tell subprocesses "give me ANSI colors!" and I would love this not involving ioctl to the kernel (the way ttys work). The FORCE_COLOR feels like this solution.

However, there's a problem: unlike NO_COLOR, FORCE_COLOR isn't a "real standard". I think this is just a thing that node is doing. However, it seems like it at least becomes a de-facto standard. See, for example, the list of references in python/mypy#13806 (but, of course, node and python interpret this value differently: FORCE_COLOR=0 force-disables colors for node, and force-enables colors for Python).

[BurntSushi]

The problem is that there are multiple answers to this. For example, #41. And in practice, I'm not sure I really think the goal of "having one true way to force colors for applications" is probably just not going to happen unfortunately.

I resisted #41 because I don't really want to be in the business of just supporting a bunch of different random environment variables. I for example wonder whether it was right to add NO_COLOR support, but I'm not sure I want to expand on what I think might have been a mistake.

[matklad]

Yeah, it does seem that at this point this is just a "random environment variable". The momentum in the python ecosystem seems strong and recent though, so I wouldn't be surprised if this ends up a standard in the future, but it's wrong for termcolor to be a pioneer here. Probably best to close this issue (which is still valuable as a record of prior art!).

[matklad]

For posterity, good recent summary of the state of things in this area: hashicorp/terraform#23708 (comment)

[BurntSushi]

The momentum in the python ecosystem seems strong and recent though, so I wouldn't be surprised if this ends up a standard in the future

Yeah... I would have said that about CLICOLOR too. Hell, there is even a Rust crate for it: https://github.com/mitsuhiko/clicolors-control

And the terraform folks hit the same issue I had when considering whether to add more environment variables: how do they interact with one another? Which one wins?