Mark Ashley Bell

01 Jul 2020

Local GitHub repository backups with Python

This follows on from my previous post about backing up GitHub repositories. I wrote the original backup script in Powershell, but it had a couple of issues. The script only retrieved the first 100 repositories, so if you had more than 100 it wouldn't back them all up. Secondly, it would only work on repositories belonging to a user, not an organisation.

I could have updated the Powershell version, but I wanted to practice my Python—so I decided to rewrite the script and solve the issues at the same time.


We're still using a Github Personal Access Token to authenticate. If you don't have one, please see the Create a GitHub personal access token guide in the previous post.

I'm also assuming you have a working Python 3.8 install. This script requires the requests and GitPython packages.

Python Script

Again, the script is quite simple:

import os
import sys
import requests

from git import Repo
from string import Template

# Get the command-line arguments
user_type = sys.argv[1]
user_id = sys.argv[2]
github_access_token = sys.argv[3]
url_type = sys.argv[4]
backup_folder_path = sys.argv[5]

# Defines a template for a GraphQL query to retrieve repositories
query = """
query {
    $user_type(login:"$user_id") {
        repositories(first:100,orderBy: {field: NAME, direction: ASC}$after) {
            pageInfo {
            nodes {

query_template = Template(query)

# Creates a new query with the specified 'after' parameter value
def create_query(after=None):
    after_str = f',after:"{after}"' if after else ''
    return query_template.substitute(

# Makes the query request to the GitHub API
def send_query(query): 
    response =
        json={'query': query}, 
        headers={"Authorization": "Token " + github_access_token})
    if response.status_code == 200:
        return response.json()
        raise Exception(f'{response.status_code}: {response.json()}')

# Fetches one page of the results and return the data as a tuple
def fetch_page(after=None):
    q = create_query(after)
    result = send_query(q)

    data = result['data']

    repos = data[user_type]['repositories']['nodes']
    page_info = data[user_type]['repositories']['pageInfo']

    after = page_info['endCursor']
    has_next_page = page_info['hasNextPage']

    return (has_next_page, after, repos)

# Gets a 'friendly' short SHA hash for a commit
def short_hash(ref):
    return ref.object.hexsha[0:8]

after = None

# Keep fetching pages until has_next_page is false
while True:
    (has_next_page, new_after, repos) = fetch_page(after)

    for repo in repos:
        name = repo['name']
        url = repo[url_type]
        mirror_path = os.path.join(

        msg = None

        print(name, end=': ')

        if os.path.exists(mirror_path):
            # We already have a mirror of this repo, so update it
            repo = Repo(mirror_path)
            existing_sha = short_hash(repo.head)
            updated_sha = short_hash(repo.head)
            if existing_sha != updated_sha:
                msg = f"Updated [{existing_sha}->{updated_sha}]"
                msg = f"No changes [{existing_sha}]"
            # This repo has been created since the last backup
            Repo.clone_from(url, mirror_path, multi_options=['--mirror'])
            repo = Repo(mirror_path)
            msg = f"Cloned [{short_hash(repo.head)}]"


    if has_next_page:
        after = new_after

print("All repositories backed up")

The complete script is available here. The script arguments are:

(organisation|user) <github-username> <github-access-token> (url|sshUrl) <backup-folder-path>

Specify organisation or user depending on which you are backing up repositories for. The url/sshUrl argument lets you use either HTTP or SSH to clone and update the repository backups.

So to back up my personal repositories, I can run:

python user markashleybell <MY_TOKEN> sshUrl 'C:\GitHub Backups'

I hope someone else finds this useful!

Questions or comments? Get in touch @markashleybell, or email [Turn On Javascript To View Email Address].

More articles

© Mark Ashley Bell 2018