Support generating mirror repositories with comments removed

infra/annotate-pipeline.yml

@@ -8,7 +8,7 @@ tasks:
     minute: {$eval: 'now[14:16]'}
     second: {$eval: 'now[17:19]'}
   in:
-    - ID: microannotate-generate
+    - ID: microannotate-generate-tokenize
       created: {$fromNow: ''}
       deadline: {$fromNow: '72 hours'}
       expires: {$fromNow: '1 month'}
@@ -22,6 +22,8 @@ tasks:
         command:
           - bugbug-microannotate-generate
           - /cache/
+          - https://github.com/marco-c/gecko-dev-wordified
+          - --tokenize
         artifacts:
           public/done:
             path: /done
@@ -38,12 +40,11 @@ tasks:
         - notify.email.release-mgmt-analysis@mozilla.com.on-failed"
         - notify.irc-channel.#bugbug.on-failed
       metadata:
-        name: bugbug microannotate repository generator
-        description: bugbug microannotate repository generator
+        name: bugbug microannotate tokenized repository generator
+        description: bugbug microannotate tokenized repository generator
         owner: release-mgmt-analysis@mozilla.com
         source: https://github.com/mozilla/bugbug/raw/master/annotate-pipeline.yml
 
-
     - ID: regressor-finder
       created: {$fromNow: ''}
       deadline: {$fromNow: '118 hours'}

scripts/microannotate_generator.py

@@ -15,8 +15,11 @@ logger = getLogger(__name__)
 
 
 class MicroannotateGenerator(object):
-    def __init__(self, cache_root):
+    def __init__(self, cache_root, repo_url, tokenize, remove_comments):
         self.cache_root = cache_root
+        self.repo_url = repo_url
+        self.tokenize = tokenize
+        self.remove_comments = remove_comments
 
         assert os.path.isdir(cache_root), f"Cache root {cache_root} is not a dir."
         self.repo_dir = os.path.join(cache_root, "mozilla-central")
@@ -29,22 +32,21 @@ class MicroannotateGenerator(object):
         git_user = get_secret("GIT_USER")
         git_password = get_secret("GIT_PASSWORD")
 
-        repo_url = "https://github.com/marco-c/gecko-dev-wordified"
-        repo_push_url = (
-            f"https://{git_user}:{git_password}@github.com/marco-c/gecko-dev-wordified"
+        repo_push_url = self.repo_url.replace(
+            "https://", f"https://{git_user}:{git_password}@"
         )
-        git_repo_path = os.path.basename(repo_url)
+        git_repo_path = os.path.basename(self.repo_url)
 
         retry(
             lambda: subprocess.run(
-                ["git", "clone", repo_url, git_repo_path], check=True
+                ["git", "clone", self.repo_url, git_repo_path], check=True
             )
         )
 
         try:
             retry(
                 lambda: subprocess.run(
-                    ["git", "pull", repo_url, "master"],
+                    ["git", "pull", self.repo_url, "master"],
                     cwd=git_repo_path,
                     capture_output=True,
                     check=True,
@@ -55,7 +57,13 @@ class MicroannotateGenerator(object):
             if b"Couldn't find remote ref master" in e.stdout:
                 pass
 
-        done = generator.generate(self.repo_dir, git_repo_path, limit=10000)
+        done = generator.generate(
+            self.repo_dir,
+            git_repo_path,
+            limit=10000,
+            tokenize=self.tokenize,
+            remove_comments=self.remove_comments,
+        )
 
         with open("done", "w") as f:
             f.write(str(1 if done else 0))
@@ -77,9 +85,21 @@ def main():
     parser = argparse.ArgumentParser(description=description)
 
     parser.add_argument("cache-root", help="Cache for repository clones.")
+    parser.add_argument("repo-url", help="Mirror repository URL.")
+    parser.add_argument(
+        "--tokenize", help="Enable word-level tokenization.", action="store_true"
+    )
+    parser.add_argument(
+        "--remove-comments", help="Enable comment removal.", action="store_true"
+    )
 
     args = parser.parse_args()
 
-    generator = MicroannotateGenerator(getattr(args, "cache-root"))
+    generator = MicroannotateGenerator(
+        getattr(args, "cache-root"),
+        getattr(args, "repo-url"),
+        args.tokenize,
+        args.remove_comments,
+    )
 
     generator.generate()