Batching and retries

Bound the work and expose retry exhaustion

batches is a generator: calling it creates an iterator; each iteration reaches yield and supplies the next slice. range(0, len(items), size) advances by the batch size, and items[start:start + size] keeps the last partial batch. Empty input yields no batches. Use a positive non-boolean integer size; the exercises reject invalid sizes with ValueError when consumed.

The mock fetch raises RuntimeError twice and then succeeds. A three-attempt loop must break after success, print Retry only when another attempt remains, and re-raise the last failure on exhaustion. Catch this declared mock error, not every exception: unrelated ValueError must remain visible. An exhausted loop is not a successful empty result.

This exercise makes no network calls and waits zero real seconds. Real retries need a service-specific transient-error policy, capped backoff and a check that repeating the operation is safe; do not blindly retry every RuntimeError or repeat a non-idempotent write. The complete helper receives a callable, allowing immediate success and exhaustion to be tested without a network.

# Re-raise the last mock failure rather than hiding it.
def retry(fetch, attempts=3):
    for attempt in range(1, attempts + 1):
        try:
            return fetch()
        except RuntimeError:
            if attempt == attempts:
                raise

def unavailable():
    raise RuntimeError("still unavailable")

print(retry(lambda: "ready"))
try:
    retry(unavailable)
except RuntimeError as error:
    print("Exhausted:", error)

The helper assumes attempts is a positive integer. Validate configurable limits in a production boundary; the coding task uses exactly three.

Official reference for this lesson