Begin script to fix bad gzip text

apiserver/database.py

@@ -101,6 +101,20 @@ def del_ref(ref):
     finally:
         session.close()
 
+def count_stories():
+    try:
+        session = Session()
+        return session.query(Story).count()
+    finally:
+        session.close()
+
+def get_story_list():
+    try:
+        session = Session()
+        return session.query(Story.sid).all()
+    finally:
+        session.close()
+
 if __name__ == '__main__':
     init()
 

apiserver/fix-stories.py

@@ -0,0 +1,40 @@
+import logging
+import database
+
+import json
+
+database.init()
+
+def fix_gzip_bug(story_list):
+    FIX_THRESHOLD = 150
+
+    count = 1
+    for sid in story_list:
+        try:
+            sid = sid[0]
+            story = database.get_story(sid)
+            full_json = json.loads(story.full_json)
+            text = full_json.get('text', '')
+
+            count = text.count('<EFBFBD>')
+            if not count: continue
+
+            ratio = count / len(text) * 1000
+            print('Bad story:', sid, 'Num ?:', count, 'Ratio:', ratio)
+        except KeyboardInterrupt:
+            raise
+        except BaseException as e:
+            logging.exception(e)
+            breakpoint()
+
+if __name__ == '__main__':
+    num_stories = database.count_stories()
+
+    print('Fix {} stories?'.format(num_stories))
+    print('Press ENTER to continue, ctrl-c to cancel')
+    input()
+
+    story_list = database.get_story_list()
+
+    fix_gzip_bug(story_list)
+

apiserver/reindex.py

@@ -23,28 +23,14 @@ def put_stories(stories):
 def get_update(update_id):
     return search.meili_api(requests.get, 'tasks/{}'.format(update_id))
 
-def count_stories():
-    try:
-        session = database.Session()
-        return session.query(database.Story).count()
-    finally:
-        session.close()
-
-def get_story_list():
-    try:
-        session = database.Session()
-        return session.query(database.Story.sid).all()
-    finally:
-        session.close()
-
 if __name__ == '__main__':
-    num_stories = count_stories()
+    num_stories = database.count_stories()
 
     print('Reindex {} stories?'.format(num_stories))
     print('Press ENTER to continue, ctrl-c to cancel')
     input()
 
-    story_list = get_story_list()
+    story_list = database.get_story_list()
 
     count = 1
     while len(story_list):