*
 * <p>{@code ByteSource} provides two kinds of methods:
 *
 * <ul>
 *   <li><b>Methods that return a stream:</b> These methods should return a <i>new</i>, independent
 *       instance each time they are called. The caller is responsible for ensuring that the
 *       returned stream is closed.
 *   <li><b>Convenience methods:</b> These are implementations of common operations that are

 *
 * <p>{@code CharSource} provides two kinds of methods:
 *
 * <ul>
 *   <li><b>Methods that return a reader:</b> These methods should return a <i>new</i>, independent
 *       instance each time they are called. The caller is responsible for ensuring that the
 *       returned reader is closed.
 *   <li><b>Convenience methods:</b> These are implementations of common operations that are

 * Iterator#remove} operations. This utility takes the brute-force approach of trying <i>all</i>
 * possible sequences of these operations, up to a given number of steps. So, if the caller
 * specifies to use <i>n</i> steps, a total of <i>3^n</i> tests are actually performed.
 *
 * <p>For instance, if <i>steps</i> is 5, one example sequence that will be tested is:
 *
 * <ol>
 *   <li>remove();

   * write.
   *
   * 2. visibility of the writes to an afterDone() call triggered by cancel():
   *
   * Since these fields are non-final that means that TimeoutFuture is not being 'safely published',
   * thus a motivated caller may be able to expose the reference to another thread that would then
   * call cancel() and be unable to cancel the delegate.
   * There are a number of ways to solve this, none of which are very pretty, and it is currently

   *   <li>If many such tasks are chained together (such as with {@code
   *       future.transform(...).transform(...).transform(...)....}), they may overflow the stack.
   *       (In simple cases, callers can avoid this by registering all tasks with the same {@link
   *       MoreExecutors#newSequentialExecutor} wrapper around {@code directExecutor()}. More

    otherwise. Do not persist these and assume they can be read by a future
    version of the library.

5.  Our classes are not designed to protect against a malicious caller. You
    should not use them for communication between trusted and untrusted code.

6.  For the mainline flavor, we test the libraries using OpenJDK 8, 11, and 17

  }

  /**
   * Loads {@link WillBeUnloadedException} in a separate {@code ClassLoader}, calls {@code
   * getChecked(future, WillBeUnloadedException.class)}, and returns the loader. The caller can then
   * test that the {@code ClassLoader} can still be GCed. The test amounts to a test that {@code
   * getChecked} holds no strong references to the class.
   */

   * Internal constructor. Users should use {@link PairedStatsAccumulator#snapshot}.
   *
   * <p>To ensure that the created instance obeys its contract, the parameters should satisfy the
   * following constraints. This is the callers responsibility and is not enforced here.
   *
   * <ul>
   *   <li>Both {@code xStats} and {@code yStats} must have the same {@code count}.
   *   <li>If that {@code count} is 1, {@code sumOfProductsOfDeltas} must be exactly 0.0.

    return containsKey(key) ? removeFromBothMaps(key) : null;
  }

  @CanIgnoreReturnValue
  @ParametricNullness
  private V removeFromBothMaps(@CheckForNull Object key) {
    // The cast is safe because the callers of this method first check that the key is present.
    V oldValue = uncheckedCastNullableTToT(delegate.remove(key));
    removeFromInverseMap(oldValue);
    return oldValue;
  }

     * <p>For example, if horizon is {@code [[a, b], [c, d], [e]]}, {@code visitNext()} will return
     * {@code [a, b, null, c, d, null, e, null]} sequentially, encoding the topological structure.
     * (Note, however, that the callers of {@code visitNext()} often insert additional iterators
     * into {@code horizon} between calls to {@code visitNext()}. This causes them to receive
     * additional values interleaved with those shown above.)
     */