How can I call 'git pull' from within Python?

This is a sample recipe, I've been using in one of my projects. Agreed that there are multiple ways to do this though. :)

>>> import subprocess, shlex
>>> git_cmd = 'git status'
>>> kwargs = {}
>>> kwargs['stdout'] = subprocess.PIPE
>>> kwargs['stderr'] = subprocess.PIPE
>>> proc = subprocess.Popen(shlex.split(git_cmd), **kwargs)
>>> (stdout_str, stderr_str) = proc.communicate()
>>> return_code = proc.wait()

>>> print return_code
0

>>> print stdout_str
# On branch dev
# Untracked files:
#   (use "git add <file>..." to include in what will be committed)
#
#   file1
#   file2
nothing added to commit but untracked files present (use "git add" to track)

>>> print stderr_str

The problem with your code was, you were not passing an array for subprocess.Popen() and hence was trying to run a single binary called git pull. Instead it needs to execute the binary git with the first argument being pull and so on.

subprocess.Popen expects a list of the program name and arguments. You're passing it a single string, which is (with the default shell=False) equivalent to:

['git pull']

That means that subprocess tries to find a program named literally git pull, and fails to do so: In Python 3.3, your code raises the exception FileNotFoundError: [Errno 2] No such file or directory: 'git pull'. Instead, pass in a list, like this:

import subprocess
process = subprocess.Popen(["git", "pull"], stdout=subprocess.PIPE)
output = process.communicate()[0]

By the way, in Python 2.7+, you can simplify this code with the check_output convenience function:

import subprocess
output = subprocess.check_output(["git", "pull"])

Also, to use git functionality, it's by no way necessary (albeit simple and portable) to call the git binary. Consider using git-python or Dulwich.

Have you considered using GitPython? It's designed to handle all this nonsense for you.

import git 

g = git.cmd.Git(git_dir)
g.pull()

https://github.com/gitpython-developers/GitPython

The accepted answer using GitPython is little better than just using subprocess directly.

The problem with this approach is that if you want to parse the output, you end up looking at the result of a "porcelain" command, which is a bad idea

Using GitPython in this way is like getting a shiny new toolbox, and then using it for the pile of screws that hold it together instead of the tools inside. Here's how the API was designed to be used:

import git
repo = git.Repo('Path/to/repo')
repo.remotes.origin.pull()

If you want to check if something changed, you can use

current = repo.head.commit
repo.remotes.origin.pull()
if current != repo.head.commit:
    print("It changed")