I wanted to write this up earlier but, as so frequently happens all my dormant work projects decided it was time to rise up and demand instant attention.
When we left off, we had students pulling code and submitting homework via github.
Form here, it's an easy step to have them working on a small project in their own repo. In fact, sometimes, I'll jump right to a two person project where one person creates the repo and invites the other (and me) as a collaborator.
Once here, it's important to emphasize the normal workflow:
- edit stuff with an occasional
git rm, or
- and an occasional
Regardless of how much you emphasize the workflow you're now going to have to cover merges as well how to recover from problems by going to an earlier revision.
For going back to earlier versions or otherwise recovering from disasters, I don't show them the "right" way of doing things. I show them a way that's easy to remember and reliably works without tons of searching for answers.
First, given some repo, I have them clone a version under another name:
git clone firstname.lastname@example.org:hunterdaedalus/classcode.git newdirname
Now, they have the original repo and a clone under the directory **newdirname**.
I have them go into that new copy and use
git log to go through the
logs to find an earlier version that they want to go back to.
The log will look something like this:
<p> commit 67eceb5e0a01ca5f5fb54ace65a4fe134f71edae </p> <p> Author: Mike Zamansky <email@example.com> </p> <p> Date: Sat Feb 25 12:32:26 2017 -0500 </p> <p> updated Makefile </p> <p> commit 7efed10eb6015276b0cb82874ce786dc68a683ae </p> <p> Author: Mike Zamansky <firstname.lastname@example.org> </p> <p> Date: Sat Feb 25 08:45:34 2017 -0500 </p> <p> broke out main --> main and shapes </p> <p> commit e8b5c240123a7cb17920d52b4aba9cf5787ddab2 </p> <p> Author: Mike Zamansky <email@example.com> </p> <p> Date: Sat Feb 25 08:36:29 2017 -0500 </p> <p> added lab3 code </p> <p> commit 6d5bcf866306334ddc5c6a48e8f49fb39ddbcb18 </p> <p> Author: Some other coder <firstname.lastname@example.org> </p> <p> Date: Sun Feb 19 19:29:51 2017 -0500 </p>
Find the hash that they want to get back to and checkout that version (you normally only have to use the first few characters of the hash):
git checkout 7efed10
Now they can grab what they want, copy it into the working copy of the repo, commit the changes and push them back.
For Merging, I start by having them do it manually. I'll have them load the file in question into their editor and look for the chunks that look like this:
<p> <<<<<<< A </p> <p> lines in file A </p> <p> <code>=====</code> </p> <p> lines in file B </p> <p> >>>>>>> B </p>
They'll manually make the changes, then save, commit and push back.
Later, I'll show them a couple of tools to help along the way. I usually show ediff in emacs along with meld.
Once they've been doing this for a while, it's easy to expand group sizes. After that, I'll introduce issues via the github interface.
The last big topic is branches.
First I show basic branches and merging. Here's a pretty good run through. Afterwards, I'll show how branching works in conjunction with github. This is also a good time to emphasize that students have to actually read the messages that git gives them when things don't go well. In most cases, the message will contain the exact thing to type to fix the problem (such as linking a branch with a branch up on github).
Now is also the time to introduce pull requests and how to use them to support code review. The general setup I recommend is one branch per group member and a main "deployment" branch with group members creating more branches as they see fit.
That's about it. I've followed this process, more or less, over a semester, over a year, and over multiple years. It's worked for me, I hope some of this helps you as well.
In the next and probably last git/github post I'll talk about the educational benefits that I've discovered along the way. After that, I'll get back to some emacs videos, other SIGCSE stuff and my normal rants.Tweet