Double-Checked Locking: Two Identical `if`s, Two Different Jobs

Double-checked locking uses two identical if statements that solve different problems. Remove either one and lazy initialization breaks under concurrency.

Double-Checked Locking: Two Seemingly Identical Checks, Different Jobs

Lazily initializing shared state at minimal locking cost means writing two identical-looking ifs. They look redundant, but they solve different problems. Remove either one and it breaks.

Check 1 (no lock): the hot path’s fast exit. If the value is already cached and fresh, return immediately without acquiring any lock → concurrent readers don’t need to queue.
Check 2 (while holding the lock): covers the gap between “check 1 failed” and “acquiring the lock.” Another thread may have already built the value during that window, so checking again avoids a redundant rebuild.

def login(self):
    if self._fresh():            # check 1: lock-free fast exit
        return self.session
    with self._login_lock:
        if self._fresh():        # check 2: someone else may have already built it
            return self.session
        self._do_login()         # actually rebuild
        return self.session

Which one can be removed?

Remove #1 → “single-checked locking”: still correct, just acquires the lock on every call.
Remove #2 → deduplication breaks: two threads that both missed will rebuild and overwrite each other.

When is it actually worth it? Double-checked locking only pays off when the hot read path has real contention and the critical section isn’t cheap. If every call is already dominated by a DB query (~ms), the lock’s cost (~ns) is just noise. At that point DCL is convention, not a performance win, and a single check is simpler.

Duplicate-looking code isn’t automatically a smell. First ask whether it’s solving different problems.

References:

Related: see how with lock: guarantees release on exception, or go back to the four-lesson overview on making login thread-safe.

Tags #concurrency #python