Home

Awesome

Git for Type Designers

What is Version Control?

Version control is keeping snapshots of a project at arbitrary points in time. There are multiple options to consider for version control: The simplest approach is just duplicating and dating files, or adding version numbers, but that can get quite messy and confusing.

Version control is a very common workflow for anyone working with code, and those people use dedicated version control systems like Git. Some aspects of Git make is similar to a backup system like TimeMachine, but with much more actual control.

Any creative job on a computer could make use of version control, but type design especially is an ideal candidate for version control; since type designers work with UFO files.

The UFO file format is perfect for version control, since it is text-based. A UFO is basically just a folder with many little XML-files in it. A UFO file could theoretically be created and modified in a text editor. This opens the field of digital type design for version control tools normally used only by software developers.

The UFO format is the native file format for the Robofont font editor, but (of course) version control can be used with any other file format as well:
Glyphs uses its own, text-based .glyphs file format, which has the same human-readable qualities as a UFO file, and therefore can be compared (diffed) from one revision to the other.
FontLab Studio creates binary .vfb files, which only FontLab can understand. Those files can be stored in a Git repository for the purpose of creating backup data, but the capability for diffing is lost.

Glossary

If we imagine our whole project as a single drawing on paper, the mechanisms of version control (and Git) can easily be explained:

Benefits of a Version Control System

Version Control with Git adds the following aspects:

About this Guide

A condensed version of this guide was first presented at the TypeLab of the Typographics conference in New York. For the sake of better presentability and easier updates, it has been moved to a Git repository itself.

This guide is in no way complete. There are many general guides on git out there, and this guide cannot possibly compete with any of them. Instead, this guide tries to focus on the simple aspects of Git which are needed by a type designer.

Many UI applications are available for Git. Those are likely to change, and this guide would have a hard time keeping up. Therefore, the focus is on the bare-bones terminal commands.

If you have never used a terminal before: don’t worry – it’s easy!

Using Git locally

Start a New Git Repository

Version Control can start at any point in your project. No matter if just starting out, or if you have worked on a project for years. In the root level of your project folder, simply type:

git init

This command will initialize a new repository; and make the current folder (and its subfolders) ready for version control with Git.

Add files to the Repository

Git uses a three-level approach of looking at files.

When files in the project folder are modified, git recognizes they have changed. Git also notices when files are added or removed. To make those files eligible for the next commit, you have to add them to the stage like this:

git add fileName.txt
# add one specific file to stage

git add .
#add all files of the current folder (and its subfolders) to stage

git add -A
#add all files in the project to stage (even when not in root folder)

Adding files to the stage is usualy immediately followed by a commit:

Commit to the Repository

A commit to a repository can mean “This completes a logical step (add first iteration of Small Caps, for instance), let’s make a note of it.”. Alternatively, it can also be “I am scared of what comes next! Let’s save what we have, just in case anything breaks.”

Commiting is crucial to keeping a history and log of your project. I usually commit after logical steps, several times during the day. The commit frequency depends on the working style, but it usually does not hurt making a commit to many. There have been experiments with commiting after every save (for the purpose of making a video), but for a day-to-day workflow this is probably too extreme.

git commit -m 'New Small Caps'  
# the -m option allows adding the commit message immediately

git commit -am 'DIAGONALS!!'
# the -a option adds all modified files to the commit
# new files are not added; only works with files that have been commited earlier

Commit messages: I keep them brief but specific. I also add the commit message right with the commit command, because otherwise an external editor opens. The external editor is Vim; many casual users will be confused by Vim.

Helpers

Those commands all exist for use on the command line. In reality, I use them very rarely. That’s where a UI tool (like SourceTree) comes in handy. Looking at the log of a specific file; seeing which files are currently changed; reverting a file to a specific commit. Nevertheless, here they are:

git status
# see which files have been changed, which have been added, or deleted.

git status --short
# same as above, but in a more compact notation

git log
# show the commit history with time stamps and commit messages

git log specificFile.txt
# show the commit for specificFile.txt

git diff
# see what actually has changed

Using Git with a Remote Server

Git repositories can be mirrored to a server very easily. Hosting Git repositories on a server has been popularized by services like GitHub and Bitbucket.

Both hosting services require creating a user account. Once the account is created, repositories can either be started directly on the site, and then “cloned” to the local machine; or an already existing, local repository can be connected to the service.

For the sake of simplicity, the username assumed here is user; the repository is called repo.

git remote add origin https://github.com/user/repo.git
# add connection to remote server, Github-style

git remote add origin https://user@bitbucket.org/user/repo.git
# add connection to remote server, Bitbucket-style

# both commands connect to the URL of a remote git repository
# the server name "origin" is standard, but we could also write:
git remote add myFavoriteServer https://github.com/user/repo.git

To copy (backup) our work to the remote repository, the push command (and an internet connection) is needed:

git push origin master
# push the latest commits
# from ”master” branch to the server called “origin” 

If the contents of a hosted git repository are needed, you can either download the .zip file from the site (boring!) or just clone the repo with git clone. The cloned repository will include all its history, logs, and previous revisions of itself.

git clone https://github.com/user/repo.git
# get a repo from github

To get the latest non-local changes of a repository, and make the local version equal to what is available online:

git pull
# update local repository with latest changes from server

Little Helpers

The .gitignore file

It might be a good idea to exclude some files from ever being added to a Git repository. This is done with an invisible .gitignore file, which can contain single folder-or file names; and also wildcard search strings. An example:

*.vfbak
*.pyc
.DS_Store

test/*

The repository containing this .gitignore file will not track .vfbak (FontLab backup) and .pyc (compiled Python modules) files. Also, the .DS_Store file (created by Finder) is ignored. Finally, there might be a folder test, of which the contents are not important. It will not be added to the commit.

The .gitignore file will be visibile as a new file the next time git status is run. One can decide if it is useful to have it be part of the repository or not – in this case, the .gitignore file itself can also be ignored.

The --amend flag

Sometimes, mistakes happen. Typo in the commit message. Forgotten file.

git commit --amend -m 'Superior Petite Caps IPA'
# Replaces previous commit with a new commit (and a new commit message).
# In case there are any staged files, those are also committed.

Advanced Git Operations

As mentioned above, this guide on Git is very simple. It only covers the necessities of someone who just wants versioning and a backup.

One very powerful operation which is not mentioned here is branching. Branching is great, good for collaboration or new departures; but it comes with its own set of problems. None of those problems I feel equipped to cover here. Look it up when you get there.