Python interview questions and answers

Since 3.7, dicts iterate in insertion order, so APIs that serialize dicts are stable if you insert consistently.

__slots__ declares a fixed set of attributes and typically omits __dict__, saving memory when you allocate millions of objects.

CPython interns some identifiers and literals, so two copies of 'foo' may be the same object.

The GIL made reference counting and C-API thread safety much simpler, which helped CPython grow extensions quickly.

The C API provides macros such as Py_BEGIN_ALLOW_THREADS around blocking or CPU work that does not touch Python objects.

Experimental free-threaded builds can run without a GIL so Python threads use multiple cores, with a performance tax on some single-thread workloads.

A thread can hold a Python lock, block on I/O, and still create lock-order cycles with another thread that needs the GIL and that lock.

Do not run CPU-heavy work on the request thread at all

A decorator that returns a raw function can break method binding because functions are descriptors that inject self.

Definition-time wrapping is bottom-up: the decorator closest to the def wraps first.

method_decorator converts a function decorator into a method decorator that preserves self.

When a generator is closed, Python raises GeneratorExit at the paused yield so finally blocks can run.

Old asyncio used @asyncio.coroutine and yield from.

Generators express custom lazy steps with local state.

A metaclass is the class of a class

A descriptor defines __get__/__set__ on the class so instance.field access is mediated.

Each class calls super() and passes the same signature so C3 walks every class once.

__new__ is a static constructor that allocates and returns the instance (or another object).

Enforce access in the view layer (LoginRequiredMixin, permission_required, DRF permissions) so APIs cannot be bypassed.

Unsafe methods need a CSRF token in a header or form field