git best practices for quickly switching between branches

Question

I have multiple active branches that I need to work on at the same time. Clearly I can create two working directories with a distinct branch per directory. Is that the only way to do it without having to "commit" and "checkout" in order to switch from one branch to another?

Answer 1

If you are temporarily switching branches git stash is useful, however, remember that commits don’t need to persist forever; you may make temporary commits to roll back later.

So my recommendation is, if it is a many hours long switch, to do a git commit instead, because, depending on your memory, stashes can be easy to forget/lose/etc.

[In MyBranch]
>$ git commit -m "WIP: Stuff I was working on."
>$ git checkout AnotherBranch
[Do Stuff]
>$ git checkout MyBranch
>$ git reset HEAD^
[Continue]

And since this is a question about best practices, remember to give your stash a useful message using git stash save otherwise it can be difficult to find later.

Answer 2

Yes, though you can use git stash instead of commit if you're not ready to finalize your current work in progress.

Answer 3

git clone, through the local protocol, is a good alternative to be able to work on multiple branches at the same time.

I usually clone one local bare repo into multiple copies (one for each active branch), and use that bare repo as a central integration repo (since I can push easily to a bare repo, versus not being able to push to non-bare repo).

Answer 4

I got tired of switching between branches and so I wrote a smarter git checkout. Insert the following into your ~/.bash_profile, source it, and then simply use gch to switch to the last branch you were on.

current_git_branch() {
    git branch | grep \* | awk '{ print $2 }'
}
# a smart git checkout, with no args it switches to last branch.
gch() {
    if [ -n "$1" ]; then 
        echo `current_git_branch` >"/tmp/last_git_branch_used.txt"
        git checkout "$@"
    else
        if [ ! -f "/tmp/last_git_branch_used.txt" ]; then echo >&2 "ERROR: Please run gch with 1 argument first."
        else
            echo `current_git_branch` >"/tmp/last_git_branch_used.temp"
            git checkout `cat /tmp/last_git_branch_used.txt`
            mv "/tmp/last_git_branch_used."{temp,txt}
        fi
    fi
}

Answer 5

If you are doing what is called branch-per-feature development as explained here:

http://martinfowler.com/bliki/FeatureBranch.html

you might want to also ensure that you switch the database schemas. Git can help in this by means of smudge and clean. Managing multiple databases locally is then possible. When you checkout a new branch, you smudge the connection string to annotate the database name with the branch name. Should the configuration file be committed at any point, it is cleaned by removing the name of the branch from the connection.

For more information, take a look at the Pro Git book.

Answer 6

I have a bash function like this:

function gitredocommit {
  lastcomment=`git log | grep Date -A 2 -m 1 | tail -1 | sed -e 's/^ *//' -e 's/ *$//' | grep -v Merge`
  if [ -n "$lastcomment"  ]; then
    git reset --soft HEAD^; git add ../; git commit -m"$lastcomment"
  else
    echo "last commit was a merge, won't redo it"
  fi
}

You create a new branch, make a first (and last) commit and then with this you can do new stuff and overwrite this commit. If you need to update from the master you do it with

git rebase master

of course, so your commit is always on top in the branch. This works as long as you don't merge the branch in master.