Sage Development Process¶
This section is a concise overview of the Sage development process. In it, we will see how to make changes to the Sage source code and record them in the git revision control system.
In the following section on Collaborative Development with Git-Trac we will look at communicating these changes back to the Sage project. We also have a handy one-page “cheat sheet” of commonly used git commands that you can print out and leave on your desk. We have some recommended references and tutorials as well.
You can alternatively fork and create a merge request at GitLab which will automatically fetch your code and open a ticket on our trac server.
One way or another,
git is what Sage uses for tracking changes.
So first, open a shell (for instance, Terminal on Mac) and check that
[[email protected]]$ git usage: git [--version] [--help] [-C <path>] [-c name=value] ... The most commonly used git commands are: add Add file contents to the index ... tag Create, list, delete or verify a tag object signed with GPG 'git help -a' and 'git help -g' lists available subcommands and some concept guides. See 'git help <command>' or 'git help <concept>' to read about a specific subcommand or concept.
Don’t worry about the giant list of subcommands. You really only need a handful for effective development, and we will walk you through them in this guide. If you got a “command not found” error, then you don’t have git installed. Now is the time to install it; see Setting Up Git for instructions.
Because we also track who does changes in Sage with git, you must tell git how you want to be known. This only needs to be done once:
If you have multiple accounts / computers use the same name on each of them. This name/email combination ends up in commits, so do it now before you forget!
Obtaining the Sage Source Code¶
Obviously one needs the Sage source code to develop. You can use your local installation of Sage, or (to start without Sage) download it from github which is a public read-only mirror (=faster) of our internal git repository:
[[email protected] ~]$ git clone git://github.com/sagemath/sage.git Cloning into 'sage'... [...] Checking connectivity... done.
This creates a directory named
sage containing the sources for the
current stable and development releases of Sage. You next need to switch
to the develop branch (latest development release):
You will then need to compile Sage in order to use it. If you cloned, you will need to remain on the internet for it to download various packages of Sage:
If your system supports multiprocessing and you want to use multiple processors to build Sage, replace the last line above by:
[[email protected] sage]$ MAKE='make -jNUM' make
to tell the
make program to run
NUM jobs in parallel when
Mac OS X allows changing directories without using exact capitalization.
Beware of this convenience when compiling for OS X. Ignoring exact
capitalization when changing into
SAGE_ROOT can lead to build
errors for dependencies requiring exact capitalization in path names.
For the experts, note that the repository at git.sagemath.org is where development actually takes place.
In order to start modifying Sage, we want to make a branch of Sage. A branch is a copy (except that it doesn’t take up twice the space) of the Sage source code where you can store your modifications to the Sage source code and which you can upload to trac tickets.
To begin with, type the command
git branch. You will see the following:
[[email protected]]$ git branch * develop master
The asterisk shows you which branch you are on. Without an argument,
git branch command displays a list of all local branches
with the current one marked by an asterisk.
It is easy to create a new branch; first make sure you are on the branch from
which you want to branch out. That is, if you are not currently on the
develop branch, type the command
git checkout develop:
[[email protected] sage]$ git checkout develop Switched to branch 'develop' Your branch is up-to-date with 'origin/develop'.
Then use the
git branch command to create a new branch, as follows:
[[email protected] sage]$ git branch last_twin_prime
Also note that
git branch creates a new branch, but does not switch
to it. For this, you have to use
[[email protected] sage]$ git checkout last_twin_prime Switched to branch 'last_twin_prime'
Now if you use the command
git branch, you will see the following:
[[email protected]]$ git branch develop * last_twin_prime master
Note that unless you explicitly upload (“push”) a branch to a remote git repository, the branch is a local branch that is only on your computer and not visible to anyone else.
To avoid typing the new branch name twice you can use the shortcut
git checkout -b my_new_branch to create and switch to the new
branch in one command.
It is always a good idea to check that you are making your edits on the version that you think you are on. The first one shows you the topmost commit in detail, including its changes to the sources:
[[email protected] sage]$ git show
To dig deeper, you can inspect the log:
[[email protected] sage]$ git log
By default, this lists all commits in reverse chronological order.
If you find your branch to be in the wrong place, see the Reset and Recovery section.
Many programs are available to help you visualize the history tree better.
tigis a very nice text-mode such tool.
Editing the Source Code¶
Once you have your own branch, feel free to make any changes as you like. Subsequent chapters of this developer guide explain how your code should look like to fit into Sage, and how we ensure high code quality throughout.
Status is probably the most important git command. It tells you which files changed, and how to continue with recording the changes:
[[email protected] sage]$ git status On branch last_twin_prime Changes not staged for commit: (use "git add <file>..." to update what will be committed) (use "git checkout -- <file>..." to discard changes in working directory) modified: some_file.py modified: src/sage/primes/all.py Untracked files: (use "git add <file>..." to include in what will be committed) src/sage/primes/last_pair.py no changes added to commit (use "git add" and/or "git commit -a")
To dig deeper into what was changed in the files you can use:
[[email protected] sage]$ git diff some_file.py
to show you the differences.
Once you have made any changes you of course want to build Sage and
try out your edits. As long as you only modified the Sage library
(that is, Python and Cython files under
src/sage/...) you just
have to run:
[[email protected] sage]$ ./sage -br
to rebuild the Sage library and then start Sage. This should be quite fast. If you made changes to third-party packages, then you have to run
as if you were installing Sage from scratch. However, this time only packages which were changed (or which depend on a changed package) will be recompiled, so it should be much faster than compiling Sage the first time.
If you have pulled a branch from trac,
it may depend on changes to third-party packages, so
may fail. If this happens (and you believe the code in this branch
should compile), try running
Rarely there are conflicts with other packages, or with the already-installed older version of the package that you changed, in that case you do have to recompile everything using:
[[email protected] sage]$ make distclean && make
If you switch between branches based on different releases, the timestamps of modified files will change. This triggers recythonization and recompilation of modified files on subsequent builds, whether or not you have made any additional changes to files. To minimize the impact of switching between branches, install ccache using the command
[[email protected] sage]$ ./sage -i ccache
Recythonization will still occur when rebuilding, but the recompilation stage first checks whether previously compiled files are cached for reuse before compiling them again. This saves considerable time rebuilding.
Whenever you have reached your goal, a milestone towards it, or just feel like you got some work done you should commit your changes. A commit is just a snapshot of the state of all files in the repository (the program you are working on).
Unlike with some other revision control programs, in git you first need to stage the changed files, which tells git which files you want to be part of the next commit:
[[email protected] sage]$ git status # On branch my_branch # Untracked files: # (use "git add <file>..." to include in what will be committed) # # src/sage/primes/last_pair.py nothing added to commit but untracked files present (use "git add" to track) [[email protected] sage]$ git add src/sage/primes/last_pair.py [[email protected] sage]$ git status # On branch my_branch # Changes to be committed: # (use "git reset HEAD <file>..." to unstage) # # new file: src/sage/primes/last_pair.py #
Once you are satisfied with the list of staged files, you create a new
snapshot with the
git commit command:
[[email protected] sage]$ git commit ... editor opens ... [my_branch 31331f7] Added the very important foobar text file 1 file changed, 1 insertion(+) create mode 100644 foobar.txt
This will open an editor for you to write your commit message. The commit message should generally have a one-line description, followed by an empty line, followed by further explanatory text:
Added the last twin prime This is an example commit message. You see there is a one-line summary followed by more detailed description, if necessary.
You can then continue working towards your next milestone, make
another commit, repeat until finished. As long as you do not
git checkout another branch, all commits that you make will be part of
the branch that you created.