If you are using a Ruby sandboxing system like RVM or rbenv (either or which I would recommend) then simply do:

If you are not using RVM or rbenv, you will likely need to precede that with “sudo”.

Some older operating systems (such as OSX 10.6) are using an old version of RubyGems, which can cause installation problems. Do “gem update --system” to fix.

1. Install Ruby (if you have not done so already) from http://rubyinstaller.org/

   • If it complains about not being able to compile native code, install DevKit.

   • See this StackOverflow for help.

2. Open a command prompt and type gem install git-process

3. THERE IS A KNOWN PROBLEM WITH HELP ON WINDOWS (GH-120).

Currently tested and maintained against Ruby 2.0 and Ruby 2.3

To get full git help and manpage support, do:

1. User creates new local branch for focused work.

2. User pushes local branch to remote (as feature branch) by merging/rebasing with the integration branch, then pushing to the branch to remote.

3. User initiates GitHub "pull request" to ease collaboration.

4. User closes local branch by rebasing integration branch first, then pushing local to integration.

• git new-fb - Create a new feature branch based on the integration branch.

• git sync - Gets the latest changes that have happened on the integration branch and remote feature branch, then pushes your changes to a feature branch on the server.

• git pull-req - Create or get a Pull Request for the current branch.

• git to-master - Rebase against the integration branch, then pushes to it. Knows how to deal "intelligently" with pull-requests.

All commands are well documented within themselves: Use the "git help" to see the full documentation. (e.g., “git help sync”)

(See Notes for more details)

• gitProcess.integrationBranch : The name of the integration branch. Defaults to master, but can be set to develop or other.

• gitProcess.keepLocalIntegrationBranch : Controls asking about removing the local integration branch. Defaults to 'false' (i.e., do not assume the branch should be there).

• gitProcess.remoteName : Explicitly sets the remote server name to use.

• gitProcess.defaultRebaseSync: Should git sync default to using rebase instead of merge? Defaults to 'true' (i.e., Sync using rebase.)

Jim is working on "my_project" and needs to start work on a new feature.

He does lots of work. Checkin, checkin, checkin.

A sudden new brilliant idea happens.

After creating a Sheering class and tests, he commits his changes.

Time to get back to work on "save_the_planet".

Do more work. Commit. Commit. Commit.

Liking to have a clean history, he squashes and edits the commits to hide the evidence of false starts and stupid ideas so that anyone who sees the code in the future will think he was simply a genius.

Time to release to a grateful world.

John, Alice, Bill and Sally are working on "big_monies." Alice and John are pairing and need to start work on a new feature.

They do lots of work. Checkin, checkin, checkin. It has a lot of steps…​

Meanwhile Bill has been working on his great idea:

He creates his "Laaaaame" class and checks it in, with a pull request asking Sally to do a code review.

Sally sees the email. After looking at it in the web interface, she wants to test it.

After verifying that the tests still work and "it’s all good" she promotes the code to integration.

Over lunch Alice gets a brainstorm ("a duck and rubber hose!") and rushes off to her computer:

She makes her changes, syncs back up with the server, and heads over to pair with John again.

John, meanwhile, had made some changes of his own.

At this point, his local branch has Alice’s change as well as Bill and Sally’s A.W.E.S.O.M-O 4000 enhancements.

After confirming with Alice and Bill that everything looks good, he pushes his changes up for integration.

Profit!!

"git-flow" is designed around having a very strongly defined process around keeping new development, hotfixes, release process changes, etc. all clearly separated. The problem I have with it is that it’s too much "process" for not enough gain. (It has a waterfall feel to it, very much against the more modern Continuous Delivery approach.)

"GitHub Flow" is a lot cleaner, but relies too heavily (IMHO) on web-based tools and on merging instead of rebasing. It is also focussed very tightly on a Continuous Deployment process, which is great for them, but not practical for everyone.

Branches are extremely powerful tools that allow for clean organization/modularization of development.

• Branches make it easy to sandbox changes while they are in a state of flux, while at the same time allowing you to be very fearless about making potentially breaking changes.

  • For example, I commit "green to green": Doing TDD, I commit every time I have a newly passing test case. So, assuming I’m in a regular development flow, I’m committing my changes every five minutes or so. Tiny commits, but lots of them. What that means is that if I make a "less than wise choice" at some point, it’s trivial to rewind to before I’d made the mistake, potentially keep the throw-away code in another branch while I do my cleanup, and generally use the full power of a revision control system to make my life safer and easier. The branch(es) are pretty chaotic, but that’s not a problem because before integrating with the mainline, I take a moment to cleanup: Squash related commits together, write clearer commit messages (since now I know what "the answer" is), and generally move from my drafts to a more finished result. (See below on objections related to "lying with rebase.") That may just be me, though, because I’m very paranoid when it comes to computers. I tend to automatically hit Cmd/Ctl-S every time I type a period when I’m writing, or when I close a block when I’m programming. I have a minimum of three copies/backups around the world of all my important documents. And I “git sync” frequently to make sure my machine isn’t the only place where all my hard work is being stored. Have I mentioned I don’t trust computers?

• Branches allow for focused collaboration. Because a branch is about exactly one thing, it means that a team can collaborate around a feature/bug (especially when used in conjunction with a "pull request"), and keep such changes sandboxed until such time that they are ready to bring a larger audience into the mix.

  • Branches encourage being less "shy" about your code. I have heard, on a number of occasions, developers say "I’m not ready to push this to the server yet because [it’s still rough (and embarrassing)]/[it may break other people]/etc." All of those reasons for "hoarding" code are moot with branches.

Jez Humble, a brilliant Principle at ThoughtWorks Studios, talks a lot about how "branches are evil." Unfortunately, people hear that, know how smart he is, and simply repeat it without really understanding what his objections are. Fortunately, he posted clarification about what’s really meant by that. He essentially says that the problem is that developers abuse branches by not merging with mainline (i.e., "master") on a regular basis. Not constantly getting changes from mainline makes life rough when it comes time to integrate. Not putting your changes into mainline means that your changes are not being validated (via Continuous Integration, or - better - with Continuous Delivery). Both are, in fact, sins akin to not doing automated testing.

Making it "easier to do things right than wrong" (i.e., using branches and keeping them synced with mainline) was the primary motivation for this project. Every command here is focused on making it trivial to use branches that stay in sync with mainline and encourage collaboration.

Like any powerful tool, git rebase is "dangerous" if used incorrectly, just like rm/del. You simply need to know when and how to use it safely. And in the world of version control systems, “rebasing” is easily one of the most *useful* tools to come around since the commit command.

A famous article that people have been parroting in various forms for a while makes the case that rebasing (and its various forms, such as squashing, amending commits, etc.) is a "lie." As with so many things, context is everything.

You almost certainly should not rebase things that you have “published.” Generally this really means “Don’t rebase the 'master' branch!” Fortunately, these scripts make it impossible to rebase the mainline by accident.

Rebasing “your” code is an extremely useful way of communicating clearly. In the "green to green" scenario above about branches, a lot of noise is generated. If someone wants to review my code, or cherry-pick in my changes, it’s too much of a mess to effectively do so. Also, as part of the process of squashing, I have the opportunity to write clearer commit message based upon my newly enhanced understanding. The intermediate commits were my "drafts" and I’m now submitting my cleaned up copy.

If you have ever seen an "active" project that uses a process like “git-flow” that encourages a lot of branching and merging, you’ve seen how hard it can be to follow a particular line of development. Branch lines are flying around everywhere, and half the commits are pretty much pure noise. (e.g., "Merge branch 'master' of …​ into master".) It’s also hard to follow the order in which commits actually impacted the mainline. In many ways, in practice merges turn into “a truth effectively being a lie” (because it’s buried in the noise) versus rebases that are "a lie (changed from it’s 'original' form) to tell an effective truth" (clean and very clear about its impact).

One significant advantage of using automation like this is that it lets you have the best of both worlds. For example, git sync uses “rebase” instead of “merge” in a way to is completely safe for collaboration on the same branch. As long as the other people are also using git sync, it will make sure that changes are automatically incorporated with and brought in line. (See the extensive test suite in sync_spec.rb if you want to see how this works.)

This project is trying to promote clear communication about reality as it applies to the code, over micro-management over no-longer-relevant history. Thus rational for the judicious use of rebase.