Ignore PRs once repo is archived

metrics/github/api.py

@@ -114,6 +114,7 @@ def _iter_pull_requests(org, date_range):
             ... on PullRequest {
               createdAt
               closedAt
+              mergedAt
               author {
                 login
               }
@@ -122,6 +123,7 @@ def _iter_pull_requests(org, date_range):
                 owner {
                   login
                 }
+                archivedAt
               }
             }
           }
@@ -139,8 +141,10 @@ def _iter_pull_requests(org, date_range):
         yield {
             "created": date_from_iso(pr["createdAt"]),
             "closed": date_from_iso(pr["closedAt"]),
+            "merged": date_from_iso(pr["mergedAt"]),
             "author": pr["author"]["login"],
             "repo": pr["repository"]["name"],
+            "repo_archived_at": date_from_iso(pr["repository"]["archivedAt"]),
             "org": pr["repository"]["owner"]["login"],
         }
 

metrics/github/backfill.py

@@ -1,7 +1,6 @@
-import sqlite3
-import subprocess
+import json
+import textwrap
 from datetime import date, timedelta
-from pathlib import Path
 
 import click
 import structlog
@@ -10,56 +9,152 @@
 from ..timescaledb import TimescaleDBWriter
 from ..timescaledb.tables import GitHubPullRequests
 from ..tools.dates import date_from_iso, iter_days, previous_weekday
-from .prs import process_prs
+from .api import session
+from .prs import drop_archived_prs, process_prs
 
 
 setup_logging()
 
 log = structlog.get_logger()
 
 
-def get_data(db, org):
-    subprocess.check_call(["github-to-sqlite", "repos", db, org])
-
-    con = sqlite3.connect(db)
-    cur = con.cursor()
+def get_query_page(*, query, session, cursor, **kwargs):
+    """
+    Get a page of the given query
 
-    result = cur.execute(
-        "SELECT name FROM repos WHERE full_name LIKE ?", (f"{org}%",)
-    ).fetchall()
-    repo_names = [r[0] for r in result]
+    This uses the GraphQL API to avoid making O(N) calls to GitHub's (v3) REST
+    API.  The passed cursor is a GraphQL cursor [1] allowing us to call this
+    function in a loop, passing in the responses cursor to advance our view of
+    the data.
 
-    for repo in repo_names:
-        subprocess.check_call(
-            ["github-to-sqlite", "pull-requests", db, f"{org}/{repo}"]
+    [1]: https://graphql.org/learn/pagination/#end-of-list-counts-and-connections
+    """
+    # use GraphQL variables to avoid string interpolation
+    variables = {"cursor": cursor, **kwargs}
+    payload = {"query": query, "variables": variables}
+
+    log.debug(query=query, **variables)
+    r = session.post("https://api.github.com/graphql", json=payload)
+
+    if not r.ok:  # pragma: no cover
+        print(r.headers)
+        print(r.content)
+
+    r.raise_for_status()
+    results = r.json()
+
+    # In some cases graphql will return a 200 response when there are errors.
+    # https://sachee.medium.com/200-ok-error-handling-in-graphql-7ec869aec9bc
+    # Handling things robustly is complex and query specific, so here we simply
+    # take the absence of 'data' as an error, rather than the presence of
+    # 'errors' key.
+    if "data" not in results:
+        msg = textwrap.dedent(
+            f"""
+            graphql query failed
+
+            query:
+            {query}
+
+            response:
+            {json.dumps(results, indent=2)}
+        """
         )
+        raise RuntimeError(msg)
+
+    return results["data"]
+
 
+def get_query(query, path, **kwargs):
+    def extract(data):
+        result = data
+        for key in path:
+            result = result[key]
+        return result
 
-def get_prs(db):
-    sql = """
-    SELECT
-      date(pull_requests.created_at) as created,
-      date(pull_requests.closed_at) as closed,
-      date(pull_requests.merged_at) as merged,
-      authors.login as author,
-      repos.name as repo,
-      owners.login as org
-    FROM
-      pull_requests
-      LEFT OUTER JOIN repos ON (pull_requests.repo = repos.id)
-      LEFT OUTER JOIN users owners ON (repos.owner = owners.id)
-      LEFT OUTER JOIN users authors ON (pull_requests.user = authors.id)
-    WHERE
-      draft = 0
+    more_pages = True
+    cursor = None
+    while more_pages:
+        page = extract(
+            get_query_page(query=query, session=session, cursor=cursor, **kwargs)
+        )
+        yield from page["nodes"]
+        more_pages = page["pageInfo"]["hasNextPage"]
+        cursor = page["pageInfo"]["endCursor"]
+
+
+def iter_repos(org):
+    query = """
+    query repos($cursor: String, $org: String!) {
+      organization(login: $org) {
+        repositories(first: 100, after: $cursor) {
+          nodes {
+            name
+            archivedAt
+          }
+          pageInfo {
+              endCursor
+              hasNextPage
+          }
+        }
+      }
+    }
+    """
+    for repo in get_query(query, path=["organization", "repositories"], org=org):
+        yield {
+            "name": repo["name"],
+            "archived": repo["archivedAt"],
+        }
+
+
+def iter_repo_prs(org, repo):
+    query = """
+    query prs($cursor: String, $org: String!, $repo: String!) {
+      organization(login: $org) {
+        repository(name: $repo) {
+          pullRequests(first: 100, after: $cursor) {
+            nodes {
+              author {
+                login
+              }
+              number
+              createdAt
+              closedAt
+              mergedAt
+            }
+            pageInfo {
+              endCursor
+              hasNextPage
+            }
+          }
+        }
+      }
+    }
     """
-    con = sqlite3.connect(db)
-    con.row_factory = sqlite3.Row
-    cur = con.cursor()
-    return list(cur.execute(sql))
+    for pr in get_query(
+        query,
+        path=["organization", "repository", "pullRequests"],
+        org=org,
+        repo=repo["name"],
+    ):
+        yield {
+            "org": org,
+            "repo": repo["name"],
+            "repo_archived_at": date_from_iso(repo["archived"]),
+            "author": pr["author"]["login"],
+            "created": date_from_iso(pr["createdAt"]),
+            "closed": date_from_iso(pr["closedAt"]),
+            "merged": date_from_iso(pr["mergedAt"]),
+        }
+
+
+def iter_prs(org):
+    for repo in iter_repos(org):
+        yield from iter_repo_prs(org, repo)
 
 
 def open_prs(prs, org, days_threshold):
-    earliest = date_from_iso(min([pr["created"] for pr in prs]))
+    earliest = min([pr["created"] for pr in prs])
     start = previous_weekday(earliest, 0)  # Monday
     mondays = list(iter_days(start, date.today(), step=timedelta(days=7)))
 
@@ -73,8 +168,11 @@ def open_on_day(pr, start, end):
         Checks whether a PR is open today and if it's been open for greater or
         equal to the threshold of days.
         """
-        closed = date_from_iso(pr["closed"]) or today
-        opened = date_from_iso(pr["created"])
+        closed = pr["closed"] or today
+        opened = pr["created"]
+
+        if pr["repo_archived_at"] and start > pr["repo_archived_at"]:
+            return False
 
         open_today = (opened <= start) and (closed >= end)
         if not open_today:
@@ -100,39 +198,32 @@ def open_on_day(pr, start, end):
 
 
 def pr_throughput(prs, org):
-    start = date_from_iso(min([pr["created"] for pr in prs]))
+    start = min([pr["created"] for pr in prs])
     days = list(iter_days(start, date.today()))
 
     with TimescaleDBWriter(GitHubPullRequests) as writer:
         for day in days:
-            opened_prs = [pr for pr in prs if date_from_iso(pr["created"]) == day]
+            opened_prs = [pr for pr in prs if pr["created"] == day]
             log.info("%s | %s | Processing %s opened PRs", day, org, len(opened_prs))
             process_prs(writer, opened_prs, day, name="prs_opened")
 
-            merged_prs = [
-                pr for pr in prs if pr["merged"] and date_from_iso(pr["merged"]) == day
-            ]
+            merged_prs = [pr for pr in prs if pr["merged"] and pr["merged"] == day]
+            merged_prs = drop_archived_prs(merged_prs, key="merged")
             log.info("%s | %s | Processing %s merged PRs", day, org, len(merged_prs))
             process_prs(writer, merged_prs, day, name="prs_merged")
 
 
 @click.command()
 @click.argument("org")
-@click.option("--pull-data", is_flag=True, default=False)
-@click.option("--db-path", type=str, default="github.db")
 @click.pass_context
-def backfill(ctx, org, pull_data, db_path):
+def backfill(ctx, org):
     """Backfill GitHub data for the given GitHub ORG"""
-    if pull_data:
-        # clean up existing db
-        Path(db_path).unlink(missing_ok=True)
-
-        # pull all data down to make backfilling quicker
-        get_data(db_path, org)
-
-    prs = get_prs(db_path)
+    prs = list(iter_prs(org))
 
     org_prs = [pr for pr in prs if pr["org"] == org]
+    # remove all PRs opened after a repo was archived, we don't expect to ever
+    # want these.
+    org_prs = drop_archived_prs(org_prs)
     log.info("Backfilling with %s PRs for %s", len(org_prs), org)
 
     open_prs(org_prs, org, days_threshold=7)

metrics/github/cli.py

@@ -8,7 +8,7 @@
 from ..tools.dates import previous_weekday
 from . import api
 from .backfill import backfill
-from .prs import process_prs
+from .prs import drop_archived_prs, process_prs
 
 
 log = structlog.get_logger()
@@ -51,6 +51,7 @@ def open_prs(ctx, org, date, days_threshold):
         for pr in prs
         if (pr["closed"] - pr["created"]) >= timedelta(days=days_threshold)
     ]
+    open_prs = drop_archived_prs(open_prs)
 
     log.info("%s | %s | Processing %s PRs", date, org, len(open_prs))
     with TimescaleDBWriter(GitHubPullRequests) as writer:
@@ -69,10 +70,12 @@ def pr_throughput(ctx, org, date):
 
     with TimescaleDBWriter(GitHubPullRequests) as writer:
         opened_prs = api.prs_opened_on_date(org, date)
+        opened_prs = drop_archived_prs(opened_prs)
         log.info("%s | %s | Processing %s opened PRs", date, org, len(opened_prs))
         process_prs(writer, opened_prs, date, name="prs_opened")
 
         merged_prs = api.prs_merged_on_date(org, date)
+        merged_prs = drop_archived_prs(merged_prs, key="merged")
         log.info("%s | %s | Processing %s merged PRs", date, org, len(merged_prs))
         process_prs(writer, merged_prs, date, name="prs_merged")
 

metrics/github/prs.py

@@ -1,3 +1,23 @@
+def drop_archived_prs(prs, key="created"):
+    """
+    Drop PRs where the given key happened before the repo's archival date
+
+    By default this removes PRs opened after the point at which a repo was
+    archived, but can be configured for other checks, eg merging.
+    """
+
+    def predicate(pr, key):
+        if not pr["repo_archived_at"]:
+            return True
+
+        if pr[key] < pr["repo_archived_at"]:
+            return True
+
+        return False
+
+    return [pr for pr in prs if predicate(pr, key)]
+
+
 def process_prs(writer, prs, date, name=""):
     """
     Given a list of PRs, break them down in series for writing

pyproject.toml

@@ -14,7 +14,6 @@ classifiers = [
 requires-python = ">=3.11"
 dependencies = [
   "click",
-  "github-to-sqlite",
   "greenlet",
   "requests",
   "slack-bolt",
@@ -40,7 +39,7 @@ source = [
 ]
 
 [tool.coverage.report]
-fail_under = 69
+fail_under = 67
 skip_covered = true
 show_missing = true