os: add .homedir()

[robertkowalski]

In almost every cli-client that I write I need the current homedir
at a certain point.

While we have a function for getting the directory for storing
temporary files (callled tmpdir), a function for receiving the
location of the home directory is still missing.

[trevnorris]

What happens if process.env.{HOME,USERPROFILE} isn't set?

[Fishrock123]

I think this could be useful if there is a reliable and robust implementation.

[fivdi]

See also user-home

[robertkowalski]

oh good points, i just took a look at user-home

maybe it makes sense to take the code from https://github.com/sindresorhus/user-home/blob/master/index.js and add a notice t our LICENSE file (it is MIT)

[Fishrock123]

Then again, if userland can do this fine, why should it be in core?

[domenic]

I am definitely +1 on including something like this, although indeed it seems this might not be robust enough of an implementation.

[ChALkeR]

Should it not throw an Error instead of returning null or undefined (user-home returns null, your code returns undefined) when the home directory couldn't be determined?

[cjihrig]

I don't think this should be done using environment variables. There are existing C APIs for this, right?

[vkurchatkin]

prior discussion: #682

[sindresorhus]

I strongly feel this should have a robust implementation in libuv instead of just checking process.env => libuv/libuv#11

The ecosystem use the home dir for so many different things, but the current userland ways of inferring it are just too fragile.

[rvagg]

As per the previous discussion in #682 (my comment specifically: #682 (comment)) and echoing @sindresorhus' comment above; I'm -1 on this unless we can come up with a solid cross-platform implementation that doesn't involve a lot of guesswork. The main reason the previous 2 persistent-history for REPL PRs didn't make it in is because they came bundled with implementations like this and @chrisdickinson found a good compromise of just removing the need for a homedir completely--it'd be nice if we could move to using a homedir but the current solution(s) are unsatisfactory. When we expose an API like this to users we are suggesting that it's reliable, but it's absolutely not (sadly!).

[piscisaureus]

I think an API to retrieve the home directory would be an acceptable addition to libuv.

The implementation is also not that hard - on unix the proper way would be to to call getpwnam(3) and return the pw_dir member from the returned struct passwd*.

On Windows the implementation would use SHGetKnowFolderPath and to retrieve the FOLDERID_Profile value.

[cjihrig]

@piscisaureus I have already done that. Would you mind looking over https://github.com/cjihrig/userinfo and letting me know if that would be suitable for libuv?

[piscisaureus]

@cjihrig

That looks pretty good in general.
I could comment a bit (use GetUsernameW), but reviewing it in depth makes little sense since it's not a libuv patch.

[cjihrig]

OK, thanks. I'll work on something for libuv.

[trevnorris]

@piscisaureus What if the home directory doesn't exist?

[domenic]

Well I think the precedent for other os APIs has been to return 0 in that case ... ;)

[Fishrock123]

Landed in libuv: libuv/libuv@a62c2d5 \o/

[cjihrig]

I'll work on the io.js bits soon and PR it once it's in deps/libuv.

[robertkowalski]

i'll just keep this open for reference / further discussion, please close it once @cjihrig has a superseeding PR

[Fishrock123]

Closing in favor of #1791