gws

gws is a KISS, bash, colorful helper to manage workspaces composed of Git repositories.

SHOW ME PICTURES!

Here are some screen captures of gws:

Installation

Requirements: git, bash > 4.0

• Arch Linux: Install the gws package from AUR

• openSUSE: Install the gws package from the devel:tools repository

• Mac:

  • On Mac OS X, it may be necessary to upgrade bash to have a version > 4.0. It could be done with: brew install bash.

  • There is currently a bug in version 0.1.8. The script uses some options that are specific to GNU sed and cut, which are not available to OS X versions. A workaround is to install coreutils and gnu-sed with brew (brew install gnu-sed coreutils) and then define the following alias in your ~/.bashrc:

  alias gws="PATH=/usr/local/opt/coreutils/libexec/gnubin:usr/local/opt/gnu-sed/libexec/gnubin:$PATH gws"

• Other Linux: You simply need to have the src/gws bash script somewhere in your $PATH:

  • If someone made a package for your distribution you are lucky.

  • You can put it directly in /bin as root user, but it is complicated to keep it up-to-date.

  • It is also possible to put it in your home folder, for example in ~/.local/bin. You have to be sure that this folder is in your $PATH. For bash you can include any directory on your $PATH by including export PATH="$PATH:/path/to/scripts/dir" in your ~/.bashrc.

On a side note, I could also suggest you to have a look at peru which lets you keep files from different sources up to date with one command.

Quick Start

• Create a file named .projects.gws in a desired workspace folder (e.g. ~/dev) and fill it with project definitions (see Syntaxes below):

  # Work related
  work/tools/q | https://github.com/harelba/q.git
  
  # Other
  contrib/gws  | https://github.com/StreakyCobra/gws.git
  contrib/peru | https://github.com/buildinspace/peru

or

• Let gws detect existing repositories and create the .projects.gws for you:

  $ cd path/to/your/workspace
  $ gws init

and then

• Clone all missing repositories with gws update, or some specific ones with gws clone.

• Do some hacking.

• Show the status of the workspace with gws. It reveals which repositories are clean, which ones have uncommited changes, and even which ones are not up-to-date with origin.

But better

Let's say you made a ~/dev/ workspace folder and you created your .projects.gws list in it. Then your workspace became really easy to replicate! Just make this ~/dev folder a Git repository, add two files and commit them: .projects.gws and the following .gitignore:

# Ignore everything, so all repositories in our case
*

# But not these files
!.projects.gws
!.gitignore

Now, when you need to get your workspace on another computer, just clone the dev repository, for instance again to the ~/dev folder. Go into it and do a gws update. Everything is cloned and ready to be hacked!

You want to add a new project into your workspace? Add it to the .projects.gws list, do a gws update to get it. Then commit and push the .projects.gws file, so when you arrive at work for instance, you just need to git pull on the ~/dev folder and then gws update to get the same workspace structure that you had at home.

Why?

If you are, like me, a Linux programmer/hacker/coder who uses Git a lot, you certainly have a directory in your home folder named dev, workspace, code or something else that contains all the projects you are working on. For instance my current organisation is:

dev
├── archlinux
│   ├── aur
│   └── habs
├── perso
│   ├── gws
│   ├── imaxplore
│   └── teafree
└── config

where aur, habs, gws, imaxplore, teafree, config are Git repositories.

Since I use at least three different computers - one laptop, one at home and one at work - I like to have the same folder structure on all of them. Of course remembering which project was added recently on other computers and in which folder is tedious.

So I started to think about using Git submodules to register all projects on which I am working and syncing them with Git between the two computers. But clearly Git submodules are not usable because they are work with specific commits and not by following branches.

No worry. The problem is pretty trivial, so I decided to start write a little bash (YOLO) script that reads a simple list of repositories, and clones them if they don't exist. And then, commit by commit, the script as grown to finally become a helper to sync, monitor and check workspaces.

I thought it can be useful to other people, so I made a little cleanup, wrote some small documentation, and there it is. I hope you will enjoy it!

Features

This tool offers some features, including:

• It uses a list of projects, named .projects.gws, containing many projects described by their names, their repository URLs, and optionaly an upstream URL (mapped as a Git remote named upstream), like:

  work/theSoftware | git@github.com:You/theSoftware.git
  perso/gws        | git@github.com:You/gws.git         | git@github.com:StreakyCobra/gws.git

• It can use an ignore list, named .ignore.gws, containing regular expressions which discard some specific projects, for instance to disable on your home computer the work-related projects.

  ^work/

• It can detect already existing repositories and create the projects list from that.

  $ gws init

• It can clone missing repositories from the projects list (but not delete ones removed from the list, you have to do that manually for safety. Note that there is the check command to identify unlisted repositories).

  $ gws update

  update accepts the --only-changes option. If present, repos that have at least one remote and are unaffected by the update will not be printed.

• It can also clone a specified selection of missing repositories from the projects list, if you don't need all of them right now.

  $ gws clone work/theSoftware

• It can monitor all listed repositories in one command, showing uncommitted changes, untracked changes and branches not synced with origin.

  $ gws status

  or simply

  $ gws

  gws status and gws accept the --only-changes option. If present, missing repos as well as repos that have at least one remote and only clean branches will not be shown.

• It can fetch the modifications from origin for all repositories. It is useful to make sure you have the latest modifications, for instance before getting on a train with no internet connection:

  $ gws fetch

  fetch accepts the --only-changes option, which has the same effect as for status.

• It can also (for the same reasons) pull the modifications from origin for all repositories (but fast-forward only). Same as gws fetch, but also does fast-forward merges.

  $ gws ff    # Mnemonic: ff=fast-forward

  ff accepts the --only-changes option, which has the same effect as for status.

• It can check the workspace for all repositories (known, unknown, ignored, missing). Note: This command can be quite slow in large repositories (e.g. home folder), because it needs to search the entire space for unknown repositories. Mainly used from time to time to check workspace consistency:

  $ gws check

Syntaxes

.projects.gws

One project per line. Must be of the form

<any/folder/path> | <remote_url1> <remote_name1> [ | <remote_url2> <remote_name2> [ |  ... ]]

where

• the <remote_name1> can be skipped and origin will be used instead.

• the <remote_name2> can be skipped and upstream will be used instead.

• there must be at least one <remote_name> mapping to origin.

• there can also be blank lines, comments or inline comments. Comments start with # and continue to the end of the line.

• the folder path can be any valid linux folder path not containing |, # or spaces.

• the remote names can be any string not containing |, # or spaces.

• the remote URLs are passed to Git as-is, so they can be anything accepted by Git but must not contain |, # or spaces. For instance if you have SSH aliases in your config they are accepted.

.ignore.gws

One regular expression per line. The regular expression will be matched against each project's folder path. Some examples:

• Ignore the folder work and all its subfolders:

  ^work/

• Ignore all repositories ending with -work:

  -work$

• Ignore all repos containing an a inside:

  a

This function is really useful for locally ignoring some projects that are not needed or not accessible.

Theme file

You can customise the color scheme by placing a shell script at one of the following locations:

1. ./.git/theme.gws
2. ${HOME}/.theme.gws
3. ${HOME}/.config/gws/theme

The first existing file in the above order, if any, will be sourced into the main gws script if standard output is a terminal. In the theme script you can redefine the color codes set as C_* near the beginning of the main script.

-Why can't I put this in the projects directory with .projects.gws and .ignore.gws?

Because then if you sync your .projects.gws via Git you could inadvertently pull a commit that writes a theme file containing arbitrary shell code, which gws would then blindly run. You're still owned in other ways if an attacker has arbitrary write access in your home directory, but this way the write can't be done remotely over Git.

Other thoughts

• Except for cloning repositories, this script does not have as a goal to interact with your repositories. So no pull all, push all, delete all unused, features will be implemented (except fast-forward). This will imply too much checking to prevent data loss. Instead, just look at the status of the repositories and perform any needed actions manually on regular basis.

• You can use the commands from any subfolder of the workspace (as git does for instance).

• The file .projects.gws can easily be verisoned to sync the list of projects between different computers.

• The file .ignore.gws allows for keeping the same .projects.gws list on all computers, but to locally disable some projects (for instance work-related projects at home because they are unneeded or even not accessible from there).

• gws check can be quite slow (for instance if the workspace is the home folder) because it searches all existing Git projects recursively.

Future

This project aims to stay simple. I will try to keep the project as alive as possible for a bash script, by correcting bugs and maybe some improvements. Also, maybe one day, if I have time, I will rewrite all of this script in Python or other more maintainable language.

Edit: It seems there will be a continuation of this idea, and it will be written in OCaml! I'll let you know if the project matures! In any case this project will stay here for users who want something simple and portable.

Maintainers

The project is currently maintained by Emil Lundberg (emlun), after being started by Fabien Dubosson (StreakyCobra).

Contributors

Many thanks to these people for contributions:

• Frédéric Mahé
• Blount
• Alex Sanchez
• Antoine Belvire
• Pascal Obry
• Thomas Pérennou
• Emil Lundberg
• François Fleur
• Thomas Lemaire