새로운 버전을 시작해야 하거나, 그냥 실행을 멈추고 싶을 수도 있습니다. 🤷

///

### Lifespan 함수

먼저 주목할 점은, `yield`를 사용하여 비동기 함수(async function)를 정의하고 있다는 것입니다. 이는 `yield`를 사용한 의존성과 매우 유사합니다.

{* ../../docs_src/events/tutorial003.py hl[14:19] *}

함수의 첫 번째 부분, 즉 `yield` 이전의 코드는 애플리케이션이 시작되기 **전에** 실행됩니다.

그리고 `yield` 이후의 부분은 애플리케이션이 완료된 후 **나중에** 실행됩니다.

### 비동기 컨텍스트 매니저

함수를 확인해보면, `@asynccontextmanager`로 장식되어 있습니다.

}, {
  "scenario" : "org.gradle.performance.regression.buildcache.TaskOutputCachingJavaPerformanceTest.clean assemble with empty remote http cache",
  "durations" : [ {
    "testProject" : "largeJavaMultiProject",
    "linux" : 1159
  }, {
    "testProject" : "largeMonolithicJavaProject",
    "linux" : 1313
  } ]
}, {

   */
  static void assertInvariants(HashFunction hashFunction) {
    int objects = 100;
    Set<HashCode> hashcodes = Sets.newHashSetWithExpectedSize(objects);
    Random random = new Random(314159);
    for (int i = 0; i < objects; i++) {
      int value = random.nextInt();
      HashCode hashcode1 = hashFunction.hashInt(value);
      HashCode hashcode2 = hashFunction.hashInt(value);

            .put(Range.closedOpen(1, 5), 1)
            .put(Range.closedOpen(7, 9), 2)
            .put(Range.closedOpen(9, 14), 0)
            .put(Range.closedOpen(14, 15), 2)
            .put(Range.closedOpen(17, 20), 3)
            .put(Range.closedOpen(22, 25), 4)
            .build(),
        rangeMap.asMapOfRanges());
  }


  public void testSubRangeMapExhaustive() {

				10: {vObj},             // disk 10
				11: {vDelMarker, vObj}, // disk 11
				12: {vDelMarker, vObj}, // disk 12
				13: {vDelMarker, vObj}, // disk 13
				14: {vDelMarker, vObj}, // disk 14
				15: {vDelMarker, vObj}, // disk 15
			},
			quorum:      7,
			reqVersions: 0,
			want:        []xlMetaV2ShallowVersion{vDelMarker, vObj},
		},
	}
	for _, test := range testCases {

* 🌐 Update Russian translation for `docs/ru/docs/tutorial/security/first-steps.md`. PR [#13159](https://github.com/fastapi/fastapi/pull/13159) by [@Yarous](https://github.com/Yarous).

- Support for parsing more v1.Taint forms has been added. For example, `key:effect`, `key=:effect-` are now accepted. ([#74159](https://github.com/kubernetes/kubernetes/pull/74159), [@dlipovetsky](https://github.com/dlipovetsky))

### Cloud Provider

* Move hardPodAffinitySymmetricWeight to scheduler policy config ([#44159](https://github.com/kubernetes/kubernetes/pull/44159), [@wanghaoran1988](https://github.com/wanghaoran1988))

* Align Extender's validation with prioritizers. ([#45091](https://github.com/kubernetes/kubernetes/pull/45091), [@k82cn](https://github.com/k82cn))

* Move hardPodAffinitySymmetricWeight to scheduler policy config ([#44159](https://github.com/kubernetes/kubernetes/pull/44159), [@wanghaoran1988](https://github.com/wanghaoran1988))
* AWS: support node port health check ([#43585](https://github.com/kubernetes/kubernetes/pull/43585), [@foolusion](https://github.com/foolusion))

*   Adds `beta` hyperparameter to
    [FTRL](https://www.tensorflow.org/api_docs/python/tf/keras/optimizers/Ftrl)
    optimizer classes (Keras and others) to match
    [FTRL paper](https://research.google.com/pubs/archive/41159.pdf).
*   `Optimizer.__init__` now accepts a `gradient_aggregator` to allow for
    customization of how gradients are aggregated across devices, as well as