Context managers in Python: guaranteeing clean-up

What it is

A context manager is an object with __enter__ and __exit__ methods. with manager as value: calls __enter__, binds its result, runs the block, and always calls __exit__(type, value, traceback) afterwards – on normal completion, on return, and on an exception. contextlib.contextmanager builds one from a generator function that yields exactly once; the code before yield is the set-up, the code after it (inside try/finally) is the clean-up.

Why it matters

Resources that must be released – file handles, locks, database sessions, temporary directories – are released at a predictable point without repeating try/finally at every call site. Errors inside the block propagate unless __exit__ returns a true value, which makes suppression explicit rather than accidental.

How to apply

• Prefer the standard library's context managers: open(), threading.Lock, tempfile.TemporaryDirectory, contextlib.suppress, contextlib.ExitStack for a dynamic number of resources.
• For your own resources write a class or a generator-based manager; keep __exit__ short and never let it raise a new exception that hides the original one.
• Use ExitStack when the set of resources is only known at run time (for example, a list of files).

Pitfalls

A generator-based manager that forgets try/finally skips clean-up on exceptions. Returning True from __exit__ swallows every exception type, including KeyboardInterrupt in some designs; suppress narrowly. Holding a lock across an await in asynchronous code needs the async variants (async with, contextlib.asynccontextmanager).