private final boolean allowNulls;

  // Package-private so that it can be extended.
  MinimalCollection(Class<? super @NonNull E> type, boolean allowNulls, E... contents) {
    // TODO: consider making it shuffle the contents to test iteration order.
    this.contents = Platform.clone(contents);
    this.type = type;
    this.allowNulls = allowNulls;

    if (!allowNulls) {
      for (Object element : contents) {

  private final boolean allowNulls;

  // Package-private so that it can be extended.
  MinimalCollection(Class<? super @NonNull E> type, boolean allowNulls, E... contents) {
    // TODO: consider making it shuffle the contents to test iteration order.
    this.contents = Platform.clone(contents);
    this.type = type;
    this.allowNulls = allowNulls;

    if (!allowNulls) {
      for (Object element : contents) {

 * randomly ordered data (the probability decreases faster than exponentially in N), but if you are
 * passing in unsanitized user data then a malicious user could force it. A light shuffle of the
 * data using an unpredictable seed should normally be enough to thwart this attack.
 *
 * <p>The time taken to compute multiple quantiles on the same dataset using {@link Scale#indexes

    *   `tf.data.Dataset.zip` now supports Python-style zipping, i.e. `Dataset.zip(a, b, c)`.
    * `tf.data.Dataset.shuffle` now supports `tf.data.UNKNOWN_CARDINALITY` When doing a "full shuffle" using  `dataset = dataset.shuffle(dataset.cardinality())`. But remember, a "full shuffle" will load the full dataset into memory so that it can be shuffled, so make sure to only use this with small datasets or datasets of small objects (like filenames).

*   `tf.math`