Programming Fundamentals › Memory & Runtime

Memory Leak

Memory that's never released, growing until the program slows or crashes.

Also known as: leak, memory growth

A memory leak is memory the program no longer needs but can’t release, usually because something still refers to it. In a garbage-collected language, the collector frees only what’s unreachable. If an object stays reachable through a forgotten reference, it stays in memory, and the process grows over time.

_cache = {}

def handle_request(request_id, payload):
    _cache[request_id] = payload   # never removed: grows with every request
    return process(payload)

Each request adds an entry, and nothing ever removes one. The process works fine in testing and slowly runs out of memory in production. Adding a size limit, an expiry time or removing the entry when the request finishes fixes it.

The trade-off is that caches and history are useful, so you can’t simply avoid keeping data. The question is how long each piece of data should live. Memoization and other caches need an upper bound on size or a time limit.

The classic mistakes are unbounded caches, event listeners that are added and never removed, and closures or timers that hold large objects after they’re finished with. In a browser, a component that subscribes to something but doesn’t unsubscribe on unmount is a common source. Watch memory over a longer run than your tests, and use a profiler to see which objects are growing. The garbage collector can’t free anything the program still holds.