/// 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

import org.apache.maven.artifact.resolver.ArtifactNotFoundException;
import org.apache.maven.artifact.resolver.ArtifactResolutionException;
import org.apache.maven.execution.MavenExecutionRequest;
import org.apache.maven.execution.MavenSession;
import org.apache.maven.model.Repository;
import org.apache.maven.model.building.ModelBuildingException;
import org.apache.maven.model.building.ModelBuildingRequest;

            options.addOption(Option.builder(VERBOSE)
                    .longOpt("verbose")
                    .desc("Produce execution verbose output")
                    .get());
            options.addOption(Option.builder(SHOW_ERRORS)
                    .longOpt("errors")
                    .desc("Produce execution error messages")
                    .get());
            options.addOption(Option.builder(BATCH_MODE)

import java.util.Objects;
import java.util.stream.Collectors;

/**
 * TeamCity helper script that prints the list of commits in the current PR (one SHA per line).
 *
 * Usage (Java 11+ single-file source execution):
 *   java .teamcity/scripts/FindCommits.java <target_branch_name>
 *
 * Commit range logic:
 * - Uses origin/<target_branch_name> as the target

#include "tensorflow/core/framework/types.pb.h"
#include "tensorflow/core/platform/status.h"

namespace tensorflow {

// Abstract interface to an operation.
// This interface allows building and executing an operation in either
// tracing or immediate execution mode.
class AbstractOperation {
 protected:
  enum AbstractOperationKind {
    kGraph,
    kMlir,
    kEager,
    kTfrt,
    kTape,
    kOpHandler
  };

   *       hash(x}} consistently return the same value provided {@code x} remains unchanged
   *       according to the definition of the equivalence. The hash need not remain consistent from
   *       one execution of an application to another execution of the same application.
   *   <li>It is <i>distributable across equivalence</i>: for any references {@code x} and {@code

/// tip

这里我们使用 [`time.perf_counter()`](https://docs.python.org/3/library/time.html#time.perf_counter) 而不是 `time.time()`，因为在这类场景中它可能更精确。🤓

///

## 多个中间件的执行顺序 { #multiple-middleware-execution-order }

当你使用 `@app.middleware()` 装饰器或 `app.add_middleware()` 方法添加多个中间件时，每个新中间件都会包裹应用，形成一个栈。最后添加的中间件是“最外层”的，最先添加的是“最内层”的。

在请求路径上，最外层的中间件先运行。

在响应路径上，它最后运行。

例如：

```Python

  AbstractTensorHandle* ZerosLike() const;

  AbstractTensorHandle* GetHandle() const;

 private:
  AbstractTensorHandle* handle_;
};

// A tracing/immediate-execution agnostic tape.
//
// Gradient functions defined for this tape must support handling null incoming
// gradients.
class Tape : protected eager::GradientTape<AbstractTensorHandle,

        }, "Should throw UnknownHostException for non-existent NetBIOS name with type");

        // Verify the exception message indicates name resolution failure
        assertTrue(exception.getMessage().contains("NONEXISTENT") || exception.getMessage().contains("unknown")
                || exception.getMessage().contains("not found"), "Exception message should indicate name resolution failure");
    }

    @Test

   * AtomicIntegerFieldUpdater}, which will prevent us from selecting the {@code SafeAtomicHelper}
   * strategy.
   *
   * <p>Stashing this in a static field avoids loading it over and over again and speeds up test
   * execution significantly.
   */
  private static final ClassLoader NO_ATOMIC_FIELD_UPDATER =
      getClassLoader(
          ImmutableSet.of(
              AtomicIntegerFieldUpdater.class.getName(),