successorFunction.successors(node); // Will throw if node doesn't exist
    }
    return copy;
  }

  /**
   * Abstracts away the difference between traversing a graph vs. a tree. For a tree, we just take
   * the next element from the next non-empty iterator; for graph, we need to loop through the next
   * non-empty iterator to find first unvisited node.
   */
  private abstract static class Traversal<N> {

   * throw an exception in response to any of {@code elements} being null or of the wrong type.
   *
   * <p><b>Note:</b> this method does not take into account the occurrence count of an element in
   * the two collections; it may still return {@code true} even if {@code elements} contains several

            boolean isBlankVersion = descriptor.getRequiredMavenVersion() == null
                    || descriptor.getRequiredMavenVersion().trim().isEmpty();

            if (isBlankVersion) {
                // only take value from underlying POM if plugin descriptor has no explicit Maven requirement
                descriptor.setRequiredMavenVersion(artifact.getProperty("requiredMavenVersion", null));
            }

      this.strength = strength;
    }

    @Override
    public void parse(CacheBuilderSpec spec, String key, @CheckForNull String value) {
      checkArgument(value == null, "key %s does not take values", key);
      checkArgument(spec.keyStrength == null, "%s was already set to %s", key, spec.keyStrength);
      spec.keyStrength = strength;
    }
  }

  /** Parse weakValues and softValues */

      // we allow one second worth of work to go in a burst (i.e. take less than a second)
      assertTrue(burst <= 1000);
      long afterBurst = measureTotalTimeMillis(limiter, oneSecWorthOfWork, new Random());
      // but work beyond that must take at least one second
      assertTrue(afterBurst >= 1000);
    }
  }

  /**

    // In this case, both the verify and the report goals are called
    // in a different lifecycle. Though the compiler-plugin has a valid use case, although
    // that seems to work fine. We need to take versions and lifecycle into account.
    public ProjectSorter(Collection<MavenProject> projects) throws CycleDetectedException, DuplicateProjectException {
        graph = new Graph();

                        .makeBehavior(REACTOR_MAKE_DOWNSTREAM)
                        .expectResult(MODULE_C_2),
                scenario("Excluding an also make dependency from selectedProject does take its transitive dependency")
                        .activeRequiredProjects(MODULE_C_2)
                        .inactiveRequiredProjects(MODULE_B)
                        .makeBehavior(REACTOR_MAKE_UPSTREAM)

   * generated using this are byte-wise identical to those created using the C++ version, but note
   * that this uses unsigned integers (see {@link com.google.common.primitives.UnsignedInts}).
   * Comparisons between the two should take this into account.
   *
   * <p>Fingerprint2011() is a form of Murmur2 on strings up to 32 bytes and a form of CityHash for
   * longer strings. It could have been one or the other throughout. The main advantage of the

    throw new NoSuchMethodError(methodName);
  }

  @Override
  public void run() {
    assertSame(this, Thread.currentThread());
    try {
      while (true) {
        Request request = requestQueue.take();
        Object result;
        try {
          result = invokeMethod(request.methodName, request.arguments);
        } catch (ThreadDeath death) {
          return;

 * naturally <i>not</i> views.
 *
 * <h3>Subinterfaces</h3>
 *
 * <p>Instead of using the {@code Multimap} interface directly, prefer the subinterfaces {@link
 * ListMultimap} and {@link SetMultimap}. These take their names from the fact that the collections
 * they return from {@code get} behave like (and, of course, implement) {@link List} and {@link
 * Set}, respectively.
 *