[CI] Validate scraped push commits via GitHub API

Author: jriv01

llvm-ops-metrics/ops-container/process_llvm_commits.py

@@ -3,6 +3,8 @@
 import logging
 import os
 import git
+from github import Auth
+from github import Github
 from google.cloud import bigquery
 import requests
 
@@ -224,6 +226,56 @@ def query_for_reviews(
   return list(new_commits.values())
 
 
+def validate_push_commits(
+    new_commits: list[LLVMCommitInfo], github_auth: Auth.Token
+) -> None:
+  """Validate that push commits don't have a pull request.
+
+  To address lossiness of data from GitHub Archive BigQuery, we check each
+  commit to see if it actually has an associated pull request.
+
+  Args:
+    new_commits: List of commits to validate.
+    github_auth: The authentication token to access the GitHub API with.
+
+  Returns:
+    None
+  """
+  github_object = Github(auth=github_auth)
+  github_repo = github_object.get_repo("llvm/llvm-project")
+
+  # Validate that each push commit does not have a pull request
+  # Keep track of how many commits were inaccurately marked as push commits
+  inaccuracy_count = 0
+  for commit in new_commits:
+    # Skip pull requested commits
+    if commit.has_pull_request:
+      continue
+
+    # Check if this commit actually has a pull request
+    pull_requests = github_repo.get_commit(commit.commit_sha).get_pulls()
+    if pull_requests.totalCount == 0:
+      continue
+
+    commit.has_pull_request = True
+    commit.pull_request_number = pull_requests[0].number
+
+    # Check if the pull request was reviewed and/or approved
+    reviews = pull_requests[0].get_reviews()
+    commit.is_reviewed = reviews.totalCount > 0
+    commit.is_approved = any(
+        review.state.lower() == "approved" for review in reviews
+    )
+
+    inaccuracy_count += 1
+    logging.info("\tReview data for commit %s amended", commit.commit_sha)
+
+  logging.info(
+      "Found %d commits that were incorrectly marked as push commits",
+      inaccuracy_count,
+  )
+
+
 def upload_daily_metrics(
     grafana_api_key: str,
     grafana_metrics_userid: str,
@@ -271,6 +323,7 @@ def upload_daily_metrics(
 
 
 def main() -> None:
+  github_auth = Auth.Token(os.environ["GITHUB_TOKEN"])
   grafana_api_key = os.environ["GRAFANA_API_KEY"]
   grafana_metrics_userid = os.environ["GRAFANA_METRICS_USERID"]
 
@@ -295,6 +348,9 @@ def main() -> None:
   logging.info("Querying for reviews of new commits.")
   new_commit_info = query_for_reviews(new_commits, date_to_scrape)
 
+  logging.info("Validating new commits.")
+  validate_push_commits(new_commit_info, github_auth)
+
   logging.info("Uploading metrics to Grafana.")
   upload_daily_metrics(grafana_api_key, grafana_metrics_userid, new_commit_info)