Always use the CommitCache, and make it a singleton (#78203)

Pull Request resolved: https://github.com/pytorch/pytorch/pull/78203
Approved by: https://github.com/soulitzer

scripts/release_notes/.gitignore

@@ -0,0 +1 @@
+results/*

scripts/release_notes/categorize.py

@@ -1,12 +1,12 @@
 import argparse
 import os
 import textwrap
-from common import categories, topics, CommitDataCache
+from common import categories, topics, get_commit_data_cache
 from commitlist import CommitList
 
 class Categorizer:
     def __init__(self, path, category='Uncategorized'):
-        self.cache = CommitDataCache()
+        self.cache = get_commit_data_cache()
         self.commits = CommitList.from_existing(path)
 
         # Special categories: 'Uncategorized'

scripts/release_notes/commitlist.py

@@ -2,9 +2,10 @@ import argparse
 from common import run, topics, get_features
 from collections import defaultdict
 import os
+from pathlib import Path
 import csv
 import pprint
-from common import CommitDataCache
+from common import get_commit_data_cache, features_to_dict
 import re
 import dataclasses
 from typing import List
@@ -68,6 +69,7 @@ class CommitList:
     def write_to_disk(self):
         path = self.path
         rows = self.commits
+        os.makedirs(Path(path).parent, exist_ok=True)
         with open(path, 'w') as csvfile:
             writer = csv.writer(csvfile)
             for commit in rows:
@@ -81,7 +83,8 @@ class CommitList:
 
     @staticmethod
     def gen_commit(commit_hash):
-        features = get_features(commit_hash, return_dict=True)
+        feature_item = get_commit_data_cache().get(commit_hash)
+        features = features_to_dict(feature_item)
         category, topic = CommitList.categorize(features)
         a1, a2, a3 = (features["accepters"] + ("", "", ""))[:3]
         return Commit(commit_hash, category, topic, features["title"], features["author"], a1, a2, a3)
@@ -253,7 +256,7 @@ def to_markdown(commit_list, category):
             return commit.title
         return match.group(1)
 
-    cdc = CommitDataCache()
+    cdc = get_commit_data_cache()
     lines = [f'\n## {category}\n']
     for topic in topics:
         lines.append(f'### {topic}\n')

scripts/release_notes/common.py

@@ -1,5 +1,5 @@
 from collections import namedtuple
-from os.path import expanduser
+from pathlib import Path
 import locale
 import subprocess
 import re
@@ -132,7 +132,7 @@ def parse_pr_number(body, commit_hash, title):
 
 def get_ghstack_token():
     pattern = 'github_oauth = (.*)'
-    with open(expanduser('~/.ghstackrc'), 'r+') as f:
+    with open(Path('~/.ghstackrc').expanduser(), 'r+') as f:
         config = f.read()
     matches = re.findall(pattern, config)
     if len(matches) == 0:
@@ -190,7 +190,7 @@ def github_data(pr_number):
     return labels, author, accepters
 
 
-def get_features(commit_hash, return_dict=False):
+def get_features(commit_hash):
     title, body, files_changed = (
         commit_title(commit_hash),
         commit_body(commit_hash),
@@ -202,16 +202,25 @@ def get_features(commit_hash, return_dict=False):
     if pr_number is not None:
         labels, author, accepters = github_data(pr_number)
     result = Features(title, body, pr_number, files_changed, labels, author, accepters)
-    if return_dict:
-        return features_to_dict(result)
     return result
 
-class CommitDataCache:
-    def __init__(self, path='results/data.json'):
+
+_commit_data_cache = None
+
+def get_commit_data_cache(path='results/data.json'):
+    global _commit_data_cache
+    if _commit_data_cache is None:
+        _commit_data_cache = _CommitDataCache(path)
+    return _commit_data_cache
+
+class _CommitDataCache:
+    def __init__(self, path):
         self.path = path
         self.data = {}
         if os.path.exists(path):
             self.data = self.read_from_disk()
+        else:
+            os.makedirs(Path(path).parent, exist_ok=True)
 
     def get(self, commit):
         if commit not in self.data.keys():