def get_modified_datasets(
*,
since: datetime | None = None,
until: datetime | None = None,
) -> set[str]:
"""Retrieve the modified datasets.
Requires to set GITHUB_ACCESS_TOKEN as an environment variable.
Parameters
----------
since : Optional[datetime], optional
Start date from which commits are retrieved, by default date of the previous crawl, if never crawled set to
one week ago.
until : Optional[datetime], optional
Latest date at which commit are retrieved, by default `now`
Returns
-------
set[str]
Path of the dataset folders.
"""
now = datetime.now().astimezone()
if since is None:
if os.path.exists(".conp-archive"):
with open(".conp-archive") as fin:
since = datetime.fromisoformat(fin.read().rstrip("\n"))
else:
since = now - timedelta(weeks=1)
if until is None:
until = now
try:
gh_access_token = os.environ.get("GITHUB_ACCESS_TOKEN", None)
if gh_access_token is None:
raise OSError("GITHUB_ACCESS_TOKEN is not defined.")
except OSError as e:
# The program is not stopped since GitHub allows 60 query per hours with
# authentication. However the program will most likely fail.
logger.critical(e)
logger.info(f"Retrieving modified datasets since {since}")
repo = Github(gh_access_token).get_repo("CONP-PCNO/conp-dataset")
commits = repo.get_commits(since=since, until=until)
with open(".conp-archive", "w") as fout:
fout.write(now.isoformat())
modified_datasets: set[str] = {
os.path.basename(file_.filename)
for commit in commits for file_ in commit.files
if file_.filename.startswith("projects/")
}
return modified_datasets
with open(FILE_CONTRIBUTIONS_COMMITS, "r") as fh:
checked_commits = set(json.load(fh))
except Exception:
checked_commits = set()
for branch in repo.get_branches():
# Ignore branches from `IGNORE_BRANCHES` list
if branch.name in IGNORE_BRANCHES:
print("Skipping branch: {}".format(branch.name))
continue
print("Reading branch: {}".format(branch.name))
for commit in repo.get_commits(branch.name,
since=READ_SINCE,
until=READ_UNTIL):
message = commit.commit.message.strip().lower()
# Ignore commits with messages starting with `merge` or
# ending with `(#digits)` (considered to be pull request)`
if (message.startswith("merge") or re.search(r"\(#\d+\)$", message)):
continue
sha = commit.commit.tree.sha
if sha in checked_commits:
continue
checked_commits.add(sha)
def refresh_leaderboard_commits(id):
# Populates a leaderboard with commits, also works if the leaderboard is out of date
leaderboard = get_object_or_404(models.Leaderboard, id=id)
if leaderboard.closed:
return False
# Use the leaderboard's token to access github
token = SocialToken.objects.filter(
account__user=leaderboard.owner, account__provider="github"
).values_list("token")
if token:
token = token[0][0]
else:
token = ""
repo_str = urlparse(leaderboard.repo_url).path[1:]
if repo_str[-1] == "/":
repo_str = repo_str[:-1]
if token:
repo = Github(token).get_repo(repo_str)
else:
repo = Github().get_repo(repo_str)
github_commits = models.Commit.objects.filter(leaderboard=leaderboard).order_by(
"-timestamp"
)
if github_commits.exists():
latest_commit = github_commits[0]
else:
latest_commit = None
github_commits = []
# Stop adding new commits when we see the latest commit of our table in github's response
for commit in repo.get_commits():
if not latest_commit or commit.sha != latest_commit.nodeid:
github_commits.append(commit)
updated = 0
for commit in github_commits:
if models.Commit.objects.filter(
nodeid=commit.sha, leaderboard=leaderboard
).exists():
# This if statement might be pointless because of the latest_commit check above
# Someone needs to test this and remove it if that is the case
continue # skip if commit object already exists
# Missing author bug for some reason
if not commit.author:
break
models.Commit.objects.create(
leaderboard=leaderboard,
user=commit.author.login,
nodeid=commit.sha,
message=commit.commit.message,
url=commit.url,
html_url=commit.html_url,
timestamp=commit.commit.author.date,
)
updated += 1
return {"total": len(github_commits), "new": updated}
