*/
@GwtCompatible
@ElementTypesAreNonnullByDefault
public class UnmodifiableCollectionTests {

  public static void assertMapEntryIsUnmodifiable(Entry<?, ?> entry) {
    try {
      // fine because the call is going to fail without modifying the entry
      @SuppressWarnings("unchecked")
      Entry<?, @Nullable Object> nullableValueEntry = (Entry<?, @Nullable Object>) entry;
      nullableValueEntry.setValue(null);

 * try-with-resources. To get the equivalent of that, you must wrap the above code in <i>another</i>
 * try block in order to catch any exception that may be thrown (including from the call to {@code
 * close()}).
 *
 * <p>This pattern ensures the following:
 *
 * <ul>
 *   <li>Each {@code Closeable} resource that is successfully registered will be closed later.

Der springende Punkt ist, dass eine Abhängigkeit aufrufbar („callable“) sein sollte.

Ein „**Callable**“ in Python ist etwas, das wie eine Funktion aufgerufen werden kann („to call“).

Wenn Sie also ein Objekt `something` haben (das möglicherweise _keine_ Funktion ist) und Sie es wie folgt aufrufen (ausführen) können:

```Python
something()
```

oder

```Python

			opts.BaseDir[:len(opts.BaseDir)-1]+globalDirSuffix,
			xlStorageFormatFile))
		diskHealthCheckOK(ctx, err)
		if err == nil {
			// if baseDir is already a directory object, consider it
			// as part of the list call, this is AWS S3 specific
			// behavior.
			if err := send(metaCacheEntry{
				name:     opts.BaseDir,
				metadata: metadata,
			}); err != nil {
				return err
			}
		} else {

    }

    /*
     * We need to be able to mock identityHashCode() calls for tests, because it
     * can take 1-10 seconds to find colliding objects. Mocking frameworks that
     * can do magic to mock static method calls still can't do so for a system
     * class, so we need the indirection. In production, Hotspot should still
     * recognize that the call is 1-morphic and should still be willing to
     * inline it if necessary.

 * tailored to that use case.
 *
 * <p>This uses the same efficient implementation as {@link Ordering#leastOf(Iterable, int)},
 * offering expected O(n + k log k) performance (worst case O(n log k)) for n calls to {@link
 * #offer} and a call to {@link #topK}, with O(k) memory. In comparison, quickselect has the same
 * asymptotics but requires O(n) memory, and a {@code PriorityQueue} implementation takes O(n log

    public void remove() {
      if (modCount != expectedModCount) {
        throw new ConcurrentModificationException();
      }
      if (toRemove == null) {
        throw new IllegalStateException("no calls to next() since the last call to remove()");
      }
      delete(toRemove);
      expectedModCount = modCount;
      toRemove = null;
    }

    abstract T output(BiEntry<K, V> entry);
  }

  @Override

	encryptedData, err := madmin.EncryptData(cred.SecretKey, data)
	if err != nil {
		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
		return
	}

	writeSuccessResponseJSON(w, encryptedData)

	// Call hook for cluster-replication if the service account is not for a
	// root user.
	if newCred.ParentUser != globalActiveCred.AccessKey {
		replLogIf(ctx, globalSiteReplicationSys.IAMChangeHook(ctx, madmin.SRIAMItem{

        requires all optimizer variables to be created at the first
        `apply_gradients()` or `minimize()` call. If your workflow calls
        optimizer to update different parts of model in multiple stages,
        please call `optimizer.build(model.trainable_variables)` before the
        training loop.
    *   **Performance regression on `ParameterServerStrategy`.** This could be

  }

  // TODO: IdentityHashMap, AbstractMap

  private static Map<String, String> toHashMap(Entry<String, String>[] entries) {
    return populate(new HashMap<String, String>(), entries);
  }

  // TODO: call conversion constructors or factory methods instead of using
  // populate() on an empty map
  private static <T, M extends Map<T, String>> M populate(M map, Entry<T, String>[] entries) {
    for (Entry<T, String> entry : entries) {