.build();

  /**
   * type → implementation. Inherently mutable interfaces and abstract classes are mapped to their
   * default implementations and are "new"d upon get().
   */
  private static final ConcurrentMap<Class<?>, Class<?>> implementations = Maps.newConcurrentMap();

  private static <T> void setImplementation(Class<T> type, Class<? extends T> implementation) {

          .build();

  /**
   * type → implementation. Inherently mutable interfaces and abstract classes are mapped to their
   * default implementations and are "new"d upon get().
   */
  private static final ConcurrentMap<Class<?>, Class<?>> implementations = Maps.newConcurrentMap();

  private static <T> void setImplementation(Class<T> type, Class<? extends T> implementation) {

   * one of the predefined {@link Policies} or specify a custom implementation. Implementations must
   * be thread-safe.
   *
   * @since 13.0
   */
  public interface Policy {

    /**
     * Called when a potential deadlock is encountered. Implementations can throw the given {@code
     * exception} and/or execute other desired logic.
     *

   * graph implementations, this method should replace {@link #graph} with a new graph that includes
   * this node.
   */
  abstract void addNode(Integer n);

  /**
   * A proxy method that adds the edge {@code e} to the graph being tested. In case of Immutable
   * graph implementations, this method should replace {@link #graph} with a new graph that includes
   * this edge.

   * graph implementations, this method should replace {@link #graph} with a new graph that includes
   * this node.
   */
  abstract void addNode(Integer n);

  /**
   * A proxy method that adds the edge {@code e} to the graph being tested. In case of Immutable
   * graph implementations, this method should replace {@link #graph} with a new graph that includes
   * this edge.

 *       worse, and possibly better, than creating a mutable collection and copying it.
 *   <li>Implementations generally do not cache hash codes. If your element or key type has a slow
 *       {@code hashCode} implementation, it should cache it itself.
 * </ul>
 *
 * <h4>Example usage</h4>
 *
 * <pre>{@code
 * class Foo {

as a servlet container), but still allow each web application to provide its own LogFactory implementation, if it so desires. An instance of this class will then be created, and cached per class loader. The Default LogFactory Implementation The Logging Package APIs include a default LogFactory implementation class ( org.apache.commons.logging.impl.LogFactoryImpl) that is selected if no other implementation class name can be discovered. Its primary purpose is to create (as necessary) and return Log...

 *       worse, and possibly better, than creating a mutable collection and copying it.
 *   <li>Implementations generally do not cache hash codes. If your element or key type has a slow
 *       {@code hashCode} implementation, it should cache it itself.
 * </ul>
 *
 * <h4>Example usage</h4>
 *
 * <pre>{@code
 * class Foo {

   *
   * TODO(cpovirk): Could this be simplified if we modified implementations of LegacyConverter to
   * override methods (probably called "unsafeDoForward" and "unsafeDoBackward") with the same
   * signatures as the methods below, rather than overriding the same doForward and doBackward
   * methods as implementations of normal converters do?
   *

   *
   * TODO(cpovirk): Could this be simplified if we modified implementations of LegacyConverter to
   * override methods (probably called "unsafeDoForward" and "unsafeDoBackward") with the same
   * signatures as the methods below, rather than overriding the same doForward and doBackward
   * methods as implementations of normal converters do?
   *