WIP: Use "remotes" package

[krlmlr]

Had to remove/disable a few tests. A first test of devtools::install_github("hadley/tibble@r-1.2") went well, but it'll need to sit for a while until it's good. Also, need to migrate the missing remote types to "remotes".

CC @gaborcsardi.

[gaborcsardi]

Note that remotes does have some important open bugs, e.g. compiling on windows.

[krlmlr]

It's really not much more than a proof of concept now. I'll use this on my machine and update the branch as necessary. Feel free to beta-test, too.

[hadley]

I'd like to have a plan to get this merged into the next release of devtools. Is there a good list of missing functionality?

[gaborcsardi]

Here is a list:

• Only install from if a different sha from current (Only install from if a different sha from current #903)
• Install sha of git repos (Install sha (second attempt) #1027)
• install_bioc()
• install_cran()
• Always look up remote SHA f0d6fae
• Move lazy download logic to install_remote 88f8f9b
• Add update_remote functionality to update_packages 00f3bef
• Refactor remote and CRAN dependencies to use a common path for both cf2fbbf
• install_git: Support passthrough of credentials to git2r 91fddbf
• install_git: Support detached HEAD b260b71
• Fully specify host url for install_github 4630239
• install_min
• Install binary packages from zip files directly on windows 4d8e4e1
• local installs use SHA1 if using git, otherwise use version numbers 7fcd8b9
• install_svn updates: https://github.com/hadley/devtools/commits/master/R/install-svn.r?author=lev-kuznetsov

Various fixes. These might or migth not be relevant:

• 5b20fe7
• 3be5877
• 276ae93
• 5cd2d80

Plus a bunch of stuff potentially in install.r, I didn't go over all of these:
https://github.com/hadley/devtools/commits/master/R/install.r

So, yeah, a lot of stuff....

[hadley]

Yeah, that's a lot of stuff.

I think we also need to do quite a bit of refactoring with an eye to making it all easier to test (i.e. separating out a data structure that describes a set of need actions and performing those actions). I'm not sure how much of that you did in remotes?

[gaborcsardi]

I'm not sure how much of that you did in remotes?

Not much. It is basically the code from (old) devtools, fixed up a bit, and some extra to get rid of dependencies or make dependencies optional.

I would not mind refactoring it.

[hadley]

Let's close this for now. I think we'll aim for a release of devtools that incorporates the existing diaspora about from remotes. And then we'll aim for another release fairly soon after that aims to switch over to remotes.