Pour en savoir plus, consultez la section <a href="https://fastapi.tiangolo.com/fr/benchmarks/" class="internal-link" target="_blank">Benchmarks</a>.

## Dépendances facultatives

Utilisées par Pydantic:

For example, consider a MinIO deployment behind a proxy `https://minio.example.net`, `https://console.minio.example.net` with rules for forwarding traffic on port :9000 and :9001 to MinIO and the MinIO Console respectively on the internal network. Set `MINIO_BROWSER_REDIRECT_URL` to `https://console.minio.example.net` to ensure the browser receives a valid reachable URL.

Daha fazla bilgi için, bu bölüme bir göz at <a href="https://fastapi.tiangolo.com/tr/benchmarks/" class="internal-link" target="_blank">Kıyaslamalar</a>.

## Opsiyonel Gereksinimler

Pydantic tarafında kullanılan:

// can be read and written in a streaming manner.
// This package aims to cover most variations of the format,
// including those produced by GNU and BSD tar tools.
package tar

import (
	"errors"
	"fmt"
	"internal/godebug"
	"io/fs"
	"maps"
	"math"
	"path"
	"reflect"
	"strconv"
	"strings"
	"time"
)

// BUG: Use of the Uid and Gid fields in Header could overflow on 32-bit

さらに、Pythonが**データサイエンス**、機械学習、特にディープラーニングの主要言語であるという単純な事実により、FastAPIはデータサイエンス/機械学習のWeb APIおよびアプリケーション (他の多くのアプリケーションとの) に非常によく適合しています。

本番環境でこの並列処理を実現する方法については、[デプロイ](deployment/index.md){.internal-link target=_blank}に関するセクションを参照してください。

## `async` と `await`

現代的なバージョンのPythonには、非同期コードを定義する非常に直感的な方法があります。これにより、通常の「シーケンシャル」コードのように見え、適切なタイミングで「待機」します。

// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

// This contains the majority of valid opcode combinations
// available in cmd/internal/obj/ppc64/asm9.go with
// their valid instruction encodings.

#include "../../../../../runtime/textflag.h"

// In case of index mode instructions, usage of
// (Rx)(R0) is equivalent to (Rx+R0)

    CacheBuilder<Object, Object> builder = CacheBuilder.newBuilder().maximumSize(16);
    assertThrows(IllegalStateException.class, () -> builder.maximumWeight(16));
  }

  @GwtIncompatible // digs into internals of the non-GWT implementation
  public void testMaximumSize_largerThanInt() {
    CacheBuilder<Object, Object> builder =
        CacheBuilder.newBuilder().initialCapacity(512).maximumSize(Long.MAX_VALUE);

    CacheBuilder<Object, Object> builder = CacheBuilder.newBuilder().maximumSize(16);
    assertThrows(IllegalStateException.class, () -> builder.maximumWeight(16));
  }

  @GwtIncompatible // digs into internals of the non-GWT implementation
  public void testMaximumSize_largerThanInt() {
    CacheBuilder<Object, Object> builder =
        CacheBuilder.newBuilder().initialCapacity(512).maximumSize(Long.MAX_VALUE);

      .assertLogMatch(Regex("""<-- END HTTP \(\d+ms, 15-byte body\)"""))
      .assertNoMoreLogs()
  }

  private fun request(): Request.Builder = Request.Builder().url(url)

  internal class LogRecorder(
    val prefix: Regex = Regex(""),
  ) : HttpLoggingInterceptor.Logger {
    private val logs = mutableListOf<String>()
    private var index = 0

    fun assertLogEqual(expected: String) =

                      params->body_outputs, num_loop_vars, outputs));
        return absl::OkStatus();
      };

  // Create the while loop using an internal scope.
  tensorflow::Scope scope =
      NewInternalScope(&parent->graph, &status->status, &parent->refiner)
          .NewSubScope(params->name);

  const int first_new_node_id = parent->graph.num_node_ids();