experimental use of vcrpy

.gitattributes

@@ -0,0 +1,2 @@
+# Recorded GitHub API responses, re-generated rather than edited by hand
+tests/cassettes/** linguist-generated=true

.github/workflows/live-api.yaml

@@ -0,0 +1,35 @@
+# Run the test suite against the live GitHub API once a week, to catch
+# changes on GitHub's side that the recorded cassettes can't show.
+name: live-api
+
+on:
+  schedule:
+    # Mondays at 06:00 UTC
+    - cron: "0 6 * * 1"
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  tests:
+    if: github.repository == 'executablebooks/github-activity'
+    runs-on: ubuntu-24.04
+    env:
+      GITHUB_ACCESS_TOKEN: "${{ github.token }}"
+      # Identical requests are only sent once, so the run fits within the
+      # GITHUB_TOKEN rate limit, see tests/conftest.py
+      GITHUB_ACTIVITY_LIVE_TESTS: "1"
+
+    steps:
+      - uses: actions/checkout@v7
+      - uses: actions/setup-python@v7
+        with:
+          python-version: "3.14"
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -e ".[testing]"
+
+      - name: Run tests
+        run: pytest

.github/workflows/tests.yaml

@@ -41,15 +41,9 @@ jobs:
 
   tests:
     runs-on: ubuntu-24.04
-    env:
-      GITHUB_ACCESS_TOKEN: "${{ github.token }}"
     strategy:
-      # max-parallel declared to reduce rate limitation issues
-      max-parallel: 1
       matrix:
         include:
-          # Only test oldest supported and latest python version to reduce
-          # GitHub API calls, as they can get rate limited
           - python-version: 3.x
           - python-version: "3.10"
 
@@ -63,8 +57,9 @@ jobs:
           python -m pip install --upgrade pip
           pip install -e ".[testing]"
 
+      # GitHub API responses are replayed from tests/cassettes/
       - name: Run tests
-        run: pytest
+        run: pytest --block-network
 
   docs:
     runs-on: ubuntu-24.04

docs/contribute.md

@@ -11,6 +11,25 @@ This will install the local version of the package and run the test suite.
 nox -s test
 ```
 
+The tests don't call the GitHub API.
+Instead they replay responses recorded with [`pytest-recording`](https://github.com/kiwicom/pytest-recording), which are stored in `tests/cassettes/`.
+
+If you change the requests github-activity makes, the tests will fail with `CannotOverwriteExistingCassetteException`.
+Re-record the cassettes with a GitHub token (read-only access to public repositories is enough) and commit them:
+
+```bash
+GITHUB_ACCESS_TOKEN=... nox -s test -- --record-mode=rewrite
+```
+
+Tokens are filtered out of the recordings.
+
+Because the cassettes can't show changes on GitHub's side, the `live-api` workflow also runs the tests against the real GitHub API once a week.
+To do the same locally, run:
+
+```bash
+GITHUB_ACCESS_TOKEN=... GITHUB_ACTIVITY_LIVE_TESTS=1 nox -s test
+```
+
 ## Build the documentation
 
 The easiest way to build the documentation locally is using `nox`.

github_activity/cli.py

@@ -171,12 +171,12 @@ def load_config_and_defaults(args):
             setattr(args, argname, config.get(configname, ARG_DEFAULTS.get(configname)))
 
 
-def main():
+def main(argv=None):
     if not _git_installed_check():
         print("git is required to run github-activity", file=sys.stderr)
         sys.exit(1)
 
-    args = parser.parse_args()
+    args = parser.parse_args(argv)
     if args.target and args._target:
         raise ValueError(
             "target cannot be passed as both a positional and keyword argument"

github_activity/github_activity.py

@@ -212,25 +212,29 @@ def get_activity(
     since_dt_str = f"{since_dt:%Y-%m-%dT%H:%M:%SZ}"
     until_dt_str = f"{until_dt:%Y-%m-%dT%H:%M:%SZ}"
 
+    # GitHub App requests using user access tokens return no nodes for mixed
+    # issue/PR searches, so we search for each kind separately.
+    kinds = ["issue", "pr"]
     if kind:
-        allowed_kinds = ["issue", "pr"]
-        if kind not in allowed_kinds:
-            raise ValueError(f"Kind must be one of {allowed_kinds}, got {kind}")
-        search_query += f" type:{kind}"
+        if kind not in kinds:
+            raise ValueError(f"Kind must be one of {kinds}, got {kind}")
+        kinds = [kind]
 
     # Query for both opened and closed issues/PRs in this window
     print(f"Running search query:\n{search_query}\n\n", file=sys.stderr)
     query_data = []
     all_bot_users = set()
     for activity_type in ["created", "closed"]:
-        ii_search_query = (
-            search_query + f" {activity_type}:{since_dt_str}..{until_dt_str}"
-        )
-        qu = GitHubGraphQlQuery(ii_search_query, auth=auth)
-        qu.request()
-        query_data.append(qu.data)
-        # Collect bot users from each query
-        all_bot_users.update(qu.data.attrs.get("bot_users", set()))
+        for ii_kind in kinds:
+            ii_search_query = (
+                search_query
+                + f" type:{ii_kind} {activity_type}:{since_dt_str}..{until_dt_str}"
+            )
+            qu = GitHubGraphQlQuery(ii_search_query, auth=auth)
+            qu.request()
+            query_data.append(qu.data)
+            # Collect bot users from each query
+            all_bot_users.update(qu.data.attrs.get("bot_users", set()))
 
     query_data = (
         pd.concat(query_data).drop_duplicates(subset=["id"]).reset_index(drop=True)

github_activity/graphql.py

@@ -322,6 +322,11 @@ def is_bot(user_dict):
         self.data = pd.DataFrame(self.issues_and_or_prs)
         self.data.attrs["bot_users"] = bot_users
 
+        # Issues don't have the PR only fields, so make sure the columns always exist
+        for column in ["mergedBy", "mergeCommit", "baseRefName", "reviews", "commits"]:
+            if column not in self.data:
+                self.data[column] = None
+
         # Add some extra fields
         def get_login(user):
             return user["login"] if pd.notna(user) else user

pyproject.toml

@@ -25,6 +25,7 @@ Source = "https://github.com/executablebooks/github-activity"
 testing = [
     "pytest",
     "pytest-cov",
+    "pytest-recording",
     "pytest-regressions",
 ]
 sphinx = [