}
  // TODO(b/166669239): This is needed to support AttrBuilder::Get for string
  // attributes. Number type attrs and DataType attrs work fine without this.
  // Consider getting rid of this and making the behavior between number types
  // and string consistent.
  forward_op_->attrs.BuildNodeDef();
  std::unique_ptr<GradientFunction> gradient_fn;
  TF_RETURN_IF_ERROR(registry.Lookup(*forward_op_, &gradient_fn));

 *       unnecessary; for example, {@code copyOf(copyOf(anArrayList))} should copy the data only
 *       once. This reduces the expense of habitually making defensive copies at API boundaries.
 *       However, the precise conditions for skipping the copy operation are undefined.
 *   <li><b>Warning:</b> a view collection such as {@link ImmutableMap#keySet} or {@link

   */
  public final <S extends @Nullable T> Equivalence<Iterable<S>> pairwise() {
    // Ideally, the returned equivalence would support Iterable<? extends T>. However,
    // the need for this is so rare that it's not worth making callers deal with the ugly wildcard.
    return new PairwiseEquivalence<>(this);
  }

  /**
   * Returns a predicate that evaluates to true if and only if the input is equivalent to {@code

thought Alice `without pictures or conversation?'

  So she was considering in her own mind (as well as she could,
for the hot day made her feel very sleepy and stupid), whether
the pleasure of making a daisy-chain would be worth the trouble
of getting up and picking the daisies, when suddenly a White
Rabbit with pink eyes ran close by her.

  There was nothing so VERY remarkable in that; nor did Alice

It has the same fields as `HeroBase`, so it won't include `secret_name`.

Finally, the identity of our heroes is protected! 🥷

It also re-declares `id: int`. By doing this, we are making a **contract** with the API clients, so that they can always expect the `id` to be there and to be an `int` (it will never be `None`).

/// tip

Those two tasks are what **consume time the most**. That's the main work of maintaining FastAPI.

If you can help me with that, **you are helping me maintain FastAPI** and making sure it keeps **advancing faster and better**. 🚀

## Join the chat { #join-the-chat }

Join the 👥 [Discord chat server](https://discord.gg/VQjSZaeJmf) 👥 and hang out with others in the FastAPI community.

/// tip

       * more information, see the comments in that class.
       *
       * We already know that that's how it behaves, and subclasses of Converter can't change that
       * behavior. So there's no sense in making all subclass authors exclude the method from any
       * NullPointerTester tests that they have.
       */
      ignoredMembers.add(Converter.class.getMethod("apply", Object.class));

      int same = 0x0; // bitset for output bits with same values
      int diff = 0x0; // bitset for output bits with different values
      int count = 0;
      // originally was 2 * Math.log(...), making it try more times to avoid flakiness issues
      int maxCount = (int) (4 * Math.log(2 * keyBits * hashBits) + 1);
      while (same != 0xffffffff || diff != 0xffffffff) {
        int key1 = rand.nextInt();

       * more information, see the comments in that class.
       *
       * We already know that that's how it behaves, and subclasses of Converter can't change that
       * behavior. So there's no sense in making all subclass authors exclude the method from any
       * NullPointerTester tests that they have.
       */
      ignoredMembers.add(Converter.class.getMethod("apply", Object.class));

      public Iterator<T> iterator() {
        return Iterators.concat(Iterators.transform(inputs.iterator(), Iterable::iterator));
      }
    };
  }

  /** Concatenates a varargs array of iterables without making a defensive copy of the array. */
  private static <T extends @Nullable Object> FluentIterable<T> concatNoDefensiveCopy(
      Iterable<? extends T>... inputs) {
    for (Iterable<? extends T> input : inputs) {