[replit] The lesser of two evils

Currently the replit library has a very gross quirk: it has a global in
`replit.database.default_db.db`, and the mere action of importing this
library causes side effects to run! (connects to the database, starts a
thread to refresh the URL, and prints a warning to stdout, adding insult
to injury).

So we're trading that very gross quirk with a gross workaround to
preserve backwards compatibility: the modules that somehow end up
importing that module now have a `__getattr__` that _lazily_ calls the
code that used to be invoked as a side-effect of importing the library.
Maybe in the future we'll deploy a breaking version of the library where
we're not beholden to this backwards-compatibility quirck.

src/replit/__init__.py

@@ -2,10 +2,12 @@
 
 """The Replit Python module."""
 
+from typing import Any
+
 from . import web
 from .audio import Audio
 from .database import (
-    db,
+    LazyDB,
     Database,
     AsyncDatabase,
     make_database_proxy_blueprint,
@@ -23,3 +25,12 @@ def clear() -> None:
 
 
 audio = Audio()
+
+# Previous versions of this library would just have side-effects and always set
+# up a database unconditionally. That is very undesirable, so instead of doing
+# that, we are using this egregious hack to get the database / database URL
+# lazily.
+def __getattr__(name: str) -> Any:
+    if name == "db":
+        return LazyDB.get_instance().db
+    raise AttributeError(name)

src/replit/database/__init__.py

@@ -1,6 +1,8 @@
 """Interface with the Replit Database."""
+from typing import Any
+
 from .database import AsyncDatabase, Database, DBJSONEncoder, dumps, to_primitive
-from .default_db import db, db_url
+from .default_db import LazyDB
 from .server import make_database_proxy_blueprint, start_database_proxy
 
 __all__ = [
@@ -14,3 +16,15 @@
     "start_database_proxy",
     "to_primitive",
 ]
+
+
+# Previous versions of this library would just have side-effects and always set
+# up a database unconditionally. That is very undesirable, so instead of doing
+# that, we are using this egregious hack to get the database / database URL
+# lazily.
+def __getattr__(name: str) -> Any:
+    if name == "db":
+        return LazyDB.get_instance().db
+    if name == "db_url":
+        return LazyDB.get_instance().db_url
+    raise AttributeError(name)

src/replit/database/default_db.py

@@ -1,41 +1,64 @@
 """A module containing the default database."""
-from os import environ, path
+import logging
+import os
+import os.path
 import threading
-from typing import Optional
+from typing import Any, Optional
 
 
 from .database import Database
 
 
-def get_db_url() -> str:
+def get_db_url() -> Optional[str]:
     """Fetches the most up-to-date db url from the Repl environment."""
     # todo look into the security warning ignored below
     tmpdir = "/tmp/replitdb"  # noqa: S108
-    if path.exists(tmpdir):
+    if os.path.exists(tmpdir):
         with open(tmpdir, "r") as file:
-            db_url = file.read()
-    else:
-        db_url = environ.get("REPLIT_DB_URL")
+            return file.read()
 
-    return db_url
+    return os.environ.get("REPLIT_DB_URL")
 
 
-def refresh_db() -> None:
-    """Refresh the DB URL every hour."""
-    global db
-    db_url = get_db_url()
-    db.update_db_url(db_url)
-    threading.Timer(3600, refresh_db).start()
+class LazyDB:
+    """A way to lazily create a database connection."""
 
+    _instance: Optional["LazyDB"] = None
 
-db: Optional[Database]
-db_url = get_db_url()
-if db_url:
-    db = Database(db_url)
-else:
-    # The user will see errors if they try to use the database.
-    print("Warning: error initializing database. Replit DB is not configured.")
-    db = None
+    def __init__(self) -> None:
+        self.db: Optional[Database] = None
+        self.db_url = get_db_url()
+        if self.db_url:
+            self.db = Database(self.db_url)
+            self.refresh_db()
+        else:
+            logging.warning(
+                "Warning: error initializing database. Replit DB is not configured."
+            )
 
-if db:
-    refresh_db()
+    def refresh_db(self) -> None:
+        """Refresh the DB URL every hour."""
+        if not self.db:
+            return
+        self.db_url = get_db_url()
+        if self.db_url:
+            self.db.update_db_url(self.db_url)
+        threading.Timer(3600, self.refresh_db).start()
+
+    @classmethod
+    def get_instance(cls) -> "LazyDB":
+        if cls._instance is None:
+            cls._instance = LazyDB()
+        return cls._instance
+
+
+# Previous versions of this library would just have side-effects and always set
+# up a database unconditionally. That is very undesirable, so instead of doing
+# that, we are using this egregious hack to get the database / database URL
+# lazily.
+def __getattr__(name: str) -> Any:
+    if name == "db":
+        return LazyDB.get_instance().db
+    if name == "db_url":
+        return LazyDB.get_instance().db_url
+    raise AttributeError(name)

src/replit/database/server.py

@@ -4,7 +4,7 @@
 
 from flask import Blueprint, Flask, request
 
-from .default_db import db
+from .default_db import LazyDB
 
 
 def make_database_proxy_blueprint(view_only: bool, prefix: str = "") -> Blueprint:
@@ -20,17 +20,23 @@ def make_database_proxy_blueprint(view_only: bool, prefix: str = "") -> Blueprin
     app = Blueprint("database_proxy" + ("_view_only" if view_only else ""), __name__)
 
     def list_keys() -> Any:
-        user_prefix = request.args.get("prefix")
+        db = LazyDB.get_instance().db
+        if db is None:
+            return "Database is not configured", 500
+        user_prefix = request.args.get("prefix", "")
         encode = "encode" in request.args
-        keys = db.prefix(prefix=prefix + user_prefix)
-        keys = [k[len(prefix) :] for k in keys]
+        raw_keys = db.prefix(prefix=prefix + user_prefix)
+        keys = [k[len(prefix) :] for k in raw_keys]
 
         if encode:
             return "\n".join(quote(k) for k in keys)
         else:
             return "\n".join(keys)
 
     def set_key() -> Any:
+        db = LazyDB.get_instance().db
+        if db is None:
+            return "Database is not configured", 500
         if view_only:
             return "Database is view only", 401
         for k, v in request.form.items():
@@ -44,12 +50,18 @@ def index() -> Any:
         return set_key()
 
     def get_key(key: str) -> Any:
+        db = LazyDB.get_instance().db
+        if db is None:
+            return "Database is not configured", 500
         try:
             return db[prefix + key]
         except KeyError:
             return "", 404
 
     def delete_key(key: str) -> Any:
+        db = LazyDB.get_instance().db
+        if db is None:
+            return "Database is not configured", 500
         if view_only:
             return "Database is view only", 401
         try:

src/replit/web/__init__.py

@@ -9,6 +9,15 @@
 from .app import debug, ReplitAuthContext, run
 from .user import User, UserStore
 from .utils import *
-from ..database import AsyncDatabase, Database, db
+from ..database import AsyncDatabase, Database, LazyDB
 
 auth = LocalProxy(lambda: ReplitAuthContext.from_headers(flask.request.headers))
+
+# Previous versions of this library would just have side-effects and always set
+# up a database unconditionally. That is very undesirable, so instead of doing
+# that, we are using this egregious hack to get the database / database URL
+# lazily.
+def __getattr__(name: str) -> Any:
+    if name == "db":
+        return LazyDB.get_instance().db
+    raise AttributeError(name)

src/replit/web/user.py

@@ -5,9 +5,7 @@
 import flask
 
 from .app import ReplitAuthContext
-from ..database import Database, db as real_db
-
-db: Database = real_db  # type: ignore
+from ..database import LazyDB
 
 
 class User(MutableMapping):
@@ -32,9 +30,15 @@ def set_value(self, value: str) -> None:
         Args:
             value (str): The value to set in the database
         """
+        db = LazyDB.get_instance().db
+        if db is None:
+            raise RuntimeError("database not configured")
         db[self.db_key()] = value
 
     def _ensure_value(self) -> Any:
+        db = LazyDB.get_instance().db
+        if db is None:
+            raise RuntimeError("database not configured")
         try:
             return db[self.db_key()]
         except KeyError:
@@ -103,6 +107,9 @@ def __getitem__(self, name: str) -> User:
         return User(username=name, prefix=self.prefix)
 
     def __iter__(self) -> Iterator[str]:
+        db = LazyDB.get_instance().db
+        if db is None:
+            raise RuntimeError("database not configured")
         for k in db.keys():
             if k.startswith(self.prefix):
                 yield self._strip_prefix(k)