@ava/cooperate
AVA 4 plugin to enable cooperation between test files.
Install this as a development dependency alongside AVA itself:
npm install --save-dev @ava/cooperate
Usage
Cooperation takes place within a shared context:
import {SharedContext} from '@ava/cooperate';
const context = new SharedContext('my-context');
Across all test files, contexts with the same ID (here: my-context
) are shared.
Locks
You can create a lock within a context:
const lock = context.createLock('my-lock');
A lock needs to be acquired. This is asynchronous:
const release = await lock.acquire();
Release the lock when you no longer need it:
release();
Locks are released automatically once your tests are done.
Use acquireNow()
to either acquire the lock, or fail:
const release = await lock.acquireNow();
If the lock cannot be acquired this will throw with a LockAcquisitionError
:
try {
await lock.acquireNow();
} catch (error) {
}
Reservations
You can reserve primitive values like big integers, numbers and strings. Once reserved, no other test file can reserve these same values (if they use the correct shared context). Reserved values are released when your tests are done.
const reserved = await context.reserve(1, 2, 3);
Semaphores
You can create a counting semaphore within a shared context:
const initialValue = 3;
const semaphore = context.createSemaphore('my-semaphore', initialValue);
Within the same context, semaphores with the same ID must be created with the same initial value. Semaphores created with a different value are unusable. Their methods will reject with a SemaphoreCreationError
.
Semaphores have two methods: acquire()
and acquireNow()
. Use acquire()
to decrement the semaphore's value. If the semaphore's value would become negative, instead acquire()
waits until the semaphore's value is high enough.
const semaphore = context.createSemaphore('my-semaphore', 3);
const release = await semaphore.acquire();
acquire()
returns a function, release()
, which increments the semaphore's value by the same amount as was acquired.
The semaphore is managed: if you don't call release()
, it'll be run automatically when the test worker exits. Any pending acquire()
calls will also be removed from the queue at this time.
acquireNow()
works like acquire()
, except that if the semaphore can't be decremented immediately, acquireNow()
rejects with a SemaphoreDownError
rather than wait.
Semaphores are weighted. acquire()
and acquireNow()
accept a non-negative integer amount, defaulting to 1
, by which to decrement or increment the value:
await semaphore.acquire(0);
await semaphore.acquireNow(2);
You can also pass an amount to release()
to release just part of the acquisition at a time:
const release = await semaphore.acquire(3);
release(1);
release();
acquire()
calls resolve in FIFO order. If the current value is 1
, and a call tries to acquire 2
, subsequent acquire()
calls have to wait, even if they want to acquire just 1
.
acquireNow()
skips the queue and decrements immediately if possible.
Lower-level, unmanaged semaphores
You can create a lower-level, unmanaged semaphore which doesn't have any auto-release behavior. Instead you need to increment the semaphore in code.
const initialValue = 3;
const semaphore = context.createUnmanagedSemaphore('my-semaphore', initialValue);
Unmanaged semaphores mustn't use the same ID as a managed semaphore, within the same context. Semaphores with the same ID must be created with the same initial value. Mismatched managed and unmanaged semaphores, or those created with different values are unusable. Their methods will reject with a SemaphoreCreationError
.
Unmanaged semaphores have three methods. down()
and downNow()
decrement the value and up()
increments:
await semaphore.down(0);
await semaphore.downNow(2);
await semaphore.up();
Like the acquire()
and acquireNow()
methods of managed semaphores, down()
waits for the semaphore's value to be at least the requested amount, while downNow()
rejects with SemaphoreDownError
if the value cannot be decremented immediately.
These unmanaged semaphores do not release the "acquired" amount when a test worker exits.