assertEquals(1, (int) mmHeap.peek());
    assertEquals(1, (int) mmHeap.poll());
    assertEquals(3, (int) mmHeap.peekLast());
    assertEquals(2, (int) mmHeap.peek());
    assertEquals(2, (int) mmHeap.poll());
    assertEquals(3, (int) mmHeap.peekLast());
    assertEquals(3, (int) mmHeap.peek());
    assertEquals(3, (int) mmHeap.poll());
    assertThat(mmHeap.peekLast()).isNull();
    assertThat(mmHeap.peek()).isNull();

Pero al revisar benchmarks y comparaciones, debes tener en cuenta lo siguiente.

## Benchmarks y velocidad { #benchmarks-and-speed }

        socket.sink.use {
          while (true) {
            val action = actions.poll() ?: break
            action(socket)
          }
        }
      }
      return@FutureTask null
    }
  }

  /** Returns once all stream actions complete successfully. */
  fun awaitSuccess() {
    val futureTask =
      results.poll(5, TimeUnit.SECONDS)
        ?: throw AssertionError("no onRequest call received")

Al usar solo `BackgroundTasks` (y no `BackgroundTask`), es posible usarla como un parámetro de *path operation function* y dejar que **FastAPI** maneje el resto por ti, tal como cuando usas el objeto `Request` directamente.

Todavía es posible usar `BackgroundTask` solo en FastAPI, pero debes crear el objeto en tu código y devolver una `Response` de Starlette incluyéndolo.

//
// Returns:
// - A pointer to a PreparedStmtDB instance, which manages prepared statements using the provided connection pool and configuration.
func NewPreparedStmtDB(connPool ConnPool, maxSize int, ttl time.Duration) *PreparedStmtDB {
	return &PreparedStmtDB{
		ConnPool: connPool,                     // Assigns the provided connection pool to manage database connections.

          "expr": "minio_cluster_health_status{job=~\"$scrape_jobs\"}",
          "fullMetaSearch": false,
          "includeNullMetadata": true,
          "interval": "",
          "legendFormat": "Pool: {{pool}} Set: {{set}}",
          "range": true,
          "refId": "A",
          "useBackend": false
        }
      ],
      "title": "Cluster Health Status",
      "type": "stat"
    },
    {

    @Override
    public boolean offer(E o) {
      assertTrue(Thread.holdsLock(mutex));
      return delegate.offer(o);
    }

    @Override
    public @Nullable E poll() {
      assertTrue(Thread.holdsLock(mutex));
      return delegate.poll();
    }

    @Override
    public E remove() {
      assertTrue(Thread.holdsLock(mutex));
      return delegate.remove();
    }

    @Override

{* ../../docs_src/body_nested_models/tutorial009_py39.py hl[7] *}

/// tip | Consejo

Ten en cuenta que JSON solo admite `str` como claves.

Pero Pydantic tiene conversión automática de datos.

Esto significa que, aunque tus clientes de API solo pueden enviar strings como claves, mientras esos strings contengan enteros puros, Pydantic los convertirá y validará.

    return delegate().offer(e, timeout, unit);
  }

  @CanIgnoreReturnValue // TODO(kak): consider removing this
  @Override
  public @Nullable E poll(long timeout, TimeUnit unit) throws InterruptedException {
    return delegate().poll(timeout, unit);
  }

  @Override
  public void put(E e) throws InterruptedException {
    delegate().put(e);
  }

  @Override
  public int remainingCapacity() {

	queue        chan *queueItem        // output from workers in order
	err          error                  // global error state, only touched by Reader.Read
	bufferPool   bpool.Pool[[]byte]     // pool of []byte objects for input
	csvDstPool   bpool.Pool[[][]string] // pool of [][]string used for output
	close        chan struct{}          // used for shutting down the splitter before end of stream