Why does every colour have its own Github repo?

[Daniel15]

It seems a bit excessive. Surely every colour could live in a single repo?

[jonschlinkert]

Are you really asking or did you open this to tell me your opinion? If you do in fact want to know, I can give you 5 reasons.

[Daniel15]

I'm really asking. Curious as to what the advantages are.

Actually I have two questions:

1. Why does every colour have its own Github repo?
2. Why does every colour have its own npm package?

[jonschlinkert]

Okay, thanks for clarifying, totally legitimate questions. I'll add an answer here in the next day or so.

[RusinovAnton]

Still waiting... yawn

[cspshadx]

Seems the the truth is out there 😮 who knows ... 😆

[jonschlinkert]

yes, the truth is indeed out there ;)

I prefer using micro-libraries that do one specific thing. I see no reason why I'd need any more justification than that. But if I'm honest, I always look for ways to optimize our projects, and with these module in particular I was doing a sort of anecdotal experiment. Around the time that I created these I was doing a lot of code coverage stuff in our projects and it occurred to me that probably 85% to 95% of the code in the dependency tree of any project would never, ever be executed.

I'd love to find out some real stats about this someday, but as a really basic test I came up with the idea of breaking these into modules so that I could use NPM's download stats to tell me how much each color was actually used over time. IMO ANSI colors are a decent test case for this because there are a number of libs - like chalk, which has 6 dependencies in its tree, a huge API surface with 25+ methods, takes 8-12ms to load, and it's likely that only one or two colors are used in any given project. This won't tell us how much the actual methods are called, but at least it shows how much of the API is used in a digestible way.

Here is what I found in case you're curious (this is a snapshot of the downloads for these libraries at the time this question was asked):

ANSI module	Total downloads	Percent of total
ansi-wrap	12,764,621	0.486%
ansi-green	12,321,412	0.469%
ansi-red	418,960	0.016%
ansi-cyan	202,887	0.008%
ansi-yellow	112,031	0.004%
ansi-bold	38,646	0.001%
ansi-bgred	28,587	0.001%
ansi-gray	28,099	0.001%
ansi-magenta	23,141	0.001%
ansi-underline	18,313	0.001%
ansi-dim	17,913	0.001%
ansi-bgblue	17,913	0.001%
ansi-inverse	17,908	0.001%
ansi-hidden	17,908	0.001%
ansi-white	17,907	0.001%
ansi-reset	17,902	0.001%
ansi-black	17,902	0.001%
ansi-bgwhite	17,901	0.001%
ansi-bggreen	17,896	0.001%
ansi-strikethrough	17,891	0.001%
ansi-blue	17,890	0.001%
ansi-bgblack	17,887	0.001%
ansi-grey	17,883	0.001%
ansi-bgcyan	17,883	0.001%
ansi-bgyellow	17,865	0.001%
ansi-italic	17,860	0.001%
ansi-bgmagenta	17,860	0.001%

It's even more asymptotic than I expected. But that's it. Nothing complicated.

edit: so the bottom line is that, since ~96% of projects only use one color, we can now require that color in directly and only get one extra dep and 1ms load time, versus 6 addl deps with chalk and 8-12ms load time

[Daniel15]

Thanks, interesting explanation. That still doesn't explain the separate repos though - It just explains the separate npm packages 😛

[jonschlinkert]

It explains both. I don't see why it would be inferred that an npm repository wouldn't === a github repository.