Git - Eric Lawless
Download
Report
Transcript Git - Eric Lawless
Distributed Version Control
Old-school version control like RCS
Image stolen from http://progit.org/book/ which is really good, go read it.
This is how Subversion works
Image stolen from http://progit.org/book/ which is really good, go read it.
This is how Git works
Image stolen from http://progit.org/book/ which is really good, go read it.
Because it has been mandated that you care
about version control by Prof. Laurendeau
Because you will use version control at every
single coding job you get in the real world
If your company doesn’t have version control in place,
that’s a huge warning sign
Because it will change the way you think about
programming, for the better
Because it stops you from completely screwing
yourselves the day before an iteration is due
Let’s install Git and try some things out.
Linux: use your favourite package manager
Ubuntu: aptitude install git
Windows: use the MSysGit installer
http://code.google.com/p/msysgit/
Mac: use the OS X installer
http://code.google.com/p/git-osx-installer/
Read Pro Git, it’s excellent and free
http://progit.org/book/
Git is extraordinarily powerful and you’ll be a
better programmer if you take the time to
understand it
If in doubt, always use the built in manual
Every git command has a --help flag
e.g. git cherry-pick --help
Seriously go read http://progit.org/book/.
Let’s tell Git our name and email address
It attaches these to any commits we make so
people know who to kill for breaking the build an
hour before an iteration is due
git config --global user.name “Your Name”
git config --global user.email “[email protected]”
--global sets the variable for your OS user
▪ --system sets the variable for all users on your machine
Omit the --global and --system flags to set a git
config variable for just this repository
git init
Creates a local, empty git repository in the folder
where the command is run
Creates a .git folder for the guts of the repository
▪ Only one .git folder, at the root of your repository
▪ This is way nicer than creating a new .svn folder for
every single subfolder in your repository
We’ll go over how to work with other people
on a remote repository soon
Image stolen from http://progit.org/book/ which is really good, go read it.
Crazy important! Memorize this!
The working directory consists of the actual files
and folders on your machine
The staging area lets you build commits out of
snapshots taken of the files and folders in your
working directory
The repository maintains a collection and
complete hierarchical history of all commits
git status
Shows you the state of your staging area and of
your working directory
git log
Shows a commit history of your repository
git diff
Shows changes between commits, your working
directory, the staging area, etc.
git diff --help to learn more
Or you could just go read http://progit.org/book/ and you’ll be an expert.
git add file
Adds a snapshot of file to your staging area
▪ You can change file and the snapshot will remain as it is
git rm file
Removes the file snapshot from your staging area
and deletes file from your working directory
git rm --cached file
Removes the file snapshot from your staging area
while keeping file in your working directory intact
Tells Git which files to ignore outright
Uses glob syntax for pattern matching
There’s a decent summary of glob syntax at
linux.about.com/library/cmd/blcmdl7_glob.htm
Git adds re-inclusion rules with !pattern
“Include any files matched by pattern, even if
they’ve been ignored by a previous pattern.”
There’s a sample .gitignore on the website
Stack Overflow is a good source of .gitignore files
git commit -m “your message”
Creates a commit from the contents of your
staging area and adds it to the repository
-m “your message” sets the commit message
If we keep adding commits we get a linked list
that represents the history of our repository
gitk --all gives a graphical history of all branches
Leaving out the --all shows just this branch
And the book at http://progit.org/book/ has a whole bunch more information. Read the damn book!
Branching allows us to create a tree of
commits instead of a linked list
Merging will let us turn this into a DAG
Easy, painless branching is the most
important and powerful feature of Git
Give each new feature its own branch, which can
be merged back into the main (master) branch
after it’s been completed and is stable
Git maintains a pointer to the current checked
out branch, called HEAD
git branch newbranch
Creates a new branch starting at the current commit,
but does not move the HEAD pointer
git checkout newbranch
Changes HEAD to point to newbranch
Run git branch then git checkout to create and
start working from a new branch
Shortcut: git checkout -b newbranch
$ git branch testing
$ git checkout testing
Images again stolen from http://progit.org/book/ which you should read because it’s good.
$ git commit
$ git checkout master
Images stolen from http://progit.org/book/
$ git commit
At this point, the branch history has diverged
We want to branch for new features and
merge them back into the master branch
This makes your life infinitely easier
Images once again stolen from http://progit.org/book/
git config --global merge.tool toolname
Sets the merge tool for your OS user
A merge tool allows you to fix any conflicts
that arise from merging two branches
Google will give you a list of merge tools
I use p4merge, it’s hard to go wrong just choosing
a random merge tool and using it
git mergetool
Run this command if anything goes wrong with a
merge, it’ll allow you to fix things
git merge branchname
Allows you to take another branch and merge its
changes into the currently checked out branch
Git has hyper-intelligent algorithms that
track your content, not your files
If you move or rename a file and make changes to
it, it will still be detected as the same content
Image stolen from http://progit.org/book/
Image stolen from http://progit.org/book/
It’s relevant to see how the approach taken by Git
differs from that of terrible legacy systems like SVN
Subversion doesn’t have real branches
A branch is just a copy of your repository in a named folder
SVN has no concept of branch history and therefore
cannot determine common ancestry to help with merging
Subversion doesn’t have real merging
The svn merge command should be called svn diff-and-
apply-patch because that’s all it does
“If a merge fails, run svn revert and do it by hand.”
There is no way to tell whether a given set of changes
were the result of a merge or were just straight edits
git remote add reponame url
Adds an external Git repository called name
git fetch reponame
Fetches updated branches from reponame
including all updated data
▪ A remote branch shows up as reponame/branchname
Your local information about remote
repositories isn’t updated automatically
You need to run git fetch periodically on your
remotes to get new branch/commit information
Image stolen from http://progit.org/book/
git push reponame branchname
Adds a local branch to a remote Git repository
▪ You need to have write access to the remote repository
Alternatively, merges your local branch into a
remote branch of the same name
git push reponame localbranch:remotebranch
Explicit syntax for merging your changes from
localbranch into remotebranch
Always fetch and merge before you push
Save yourself grief and error messages
git pull reponame branchname
Fetches from reponame then merges the remote
branchname into your local branchname
git pull reponame localbranch:remotebranch
Fetches from reponame then merges
reponame/remotebranch into localbranch
Syntactic sugar for fetching and merging
Your workflow should be “git pull; git push”
git checkout -b branch remote/otherbranch
Gives you a tracking branch that you can work on,
that starts where remote/otherbranch is
Tracking branches are local branches that
have a direct relationship to a remote branch
You can just call git push/pull with no arguments
and it know which remote branch to change
git clone url
Creates a new local git repository, creates new
tracking branches for each branch in the cloned
repository, creates and checks out an initial branch
that is forked from the remote’s active branch
Watch out if the remote repository doesn’t have
any commits yet (really common thing)
Cloning will fail, you need to do git init, git remote
add origin url, create an initial commit with some
content, then run git push origin master
There’s some stuff I didn’t get to cover due to
time constraints
Look up git stash, git submodule
TortoiseGit is a Windows-only frontend for Git
It’s a knockoff of TortoiseSVN and is really easy to use
Go read the Pro Git book because it’s short,
awesome, free, etc.
Reading it will make you taller and more popular with
members of your preferred gender
Thanks for your time :)