Ici, nous utilisons [`time.perf_counter()`](https://docs.python.org/3/library/time.html#time.perf_counter) au lieu de `time.time()` car cela peut être plus précis pour ces cas d’usage. 🤓

///

## Ordre d’exécution de plusieurs middlewares { #multiple-middleware-execution-order }

///

## Ausführungsreihenfolge bei mehreren Middlewares { #multiple-middleware-execution-order }

      }

      @Override
      public long estimateSize() {
        return delegate.estimateSize();
      }

      @Override
      public int characteristics() {
        return Spliterator.ORDERED
            | Spliterator.SIZED
            | Spliterator.SUBSIZED
            | extraCharacteristics;
      }

      @Override
      public @Nullable Comparator<? super T> getComparator() {

   *
   * <p>Although all tasks are immediately executed in the thread that submitted the task, this
   * {@code ExecutorService} imposes a small locking overhead on each task submission in order to
   * implement shutdown and termination behavior.
   *
   * <p>Because of the nature of single-thread execution, the methods {@code scheduleAtFixedRate}

/// tip

這裡我們使用 [`time.perf_counter()`](https://docs.python.org/3/library/time.html#time.perf_counter) 而不是 `time.time()`，因為在這些用例中它可能更精確。🤓

///

## 多個中介軟體的執行順序 { #multiple-middleware-execution-order }

當你使用 `@app.middleware()` 裝飾器或 `app.add_middleware()` 方法加入多個中介軟體時，每個新的中介軟體都會包裹應用程式，形成一個堆疊。最後加入的中介軟體位於最外層，最先加入的位於最內層。

在請求路徑上，最外層的中介軟體最先執行。

在回應路徑上，它最後執行。

例如：

```Python

     */
    public String jobLogging;

    /**
     * Whether the scheduled job is available.
     */
    public String available;

    /**
     * Sort order of the scheduled job.
     */
    public String sortOrder;

    /**
     * The user who created the scheduled job.
     */
    public String createdBy;

    /**

      h1 = fmix64(h1);
      h2 = fmix64(h2);

      h1 += h2;
      h2 += h1;

      return HashCode.fromBytesNoCopy(
          ByteBuffer.wrap(new byte[CHUNK_SIZE])
              .order(ByteOrder.LITTLE_ENDIAN)
              .putLong(h1)
              .putLong(h2)
              .array());
    }

    private static long fmix64(long k) {
      k ^= k >>> 33;
      k *= 0xff51afd7ed558ccdL;

import java.util.Arrays;
import org.jspecify.annotations.NullMarked;
import org.jspecify.annotations.Nullable;

/**
 * {@code ObjectCountLinkedHashMap} is a subclass of {@code ObjectCountHashMap} with insertion
 * iteration order, and uses arrays to store key objects and count values. Comparing to using a
 * traditional {@code LinkedHashMap} implementation which stores keys and count values as map

    @cached_property
    def oauth_scopes(self) -> list[str]:
        scopes = self.parent_oauth_scopes.copy() if self.parent_oauth_scopes else []
        # This doesn't use a set to preserve order, just in case
        for scope in self.own_oauth_scopes or []:
            if scope not in scopes:
                scopes.append(scope)
        return scopes

    @cached_property

// Trivial collisions are avoided, but this is by no means a strong hash.
func hashDeterministicString(m map[string]string) uint64 {
	// Seed (random)
	crc := uint64(0xc2b40bbac11a7295)
	// Xor each value to make order independent
	for k, v := range m {
		// Separate key and value with an individual xor with a random number.
		// Add values of each, so they cannot be trivially collided.