Long-running calls

Generating large or expert puzzles can take a while. These options bound the work, stop a call early and run calls in threads.

Limits

Every function takes a limits argument:

pp.generate("kakuro", (16, 16), difficulty="expert",
            limits=pp.Limits(work=10_000_000, attempts=100))

Limit	Default	Bounds	Raises
work	100,000,000	The computation in one call	BudgetExceededError
attempts	10,000	The candidate puzzles generate tries	AttemptsExhaustedError

A result never depends on the limits: a call either returns the same result as with higher limits, or raises.

Cancellation

To stop a call from another thread, pass a pp.CancellationToken as cancel and call its cancel() method. The call raises CancelledError.

import threading

token = pp.CancellationToken()
threading.Timer(5.0, token.cancel).start()
try:
    pp.generate("kakuro", (16, 16), difficulty="expert", cancel=token)
except pp.CancelledError:
    print("gave up after 5 seconds")

A token stays cancelled; use a new one for the next call. token.cancelled tells whether it was cancelled.

Threads

The engine runs without holding the GIL, so other Python threads keep running during a call, and calls in separate threads run in parallel. Ctrl-C during a call raises KeyboardInterrupt.