def test_app_lifespan_state(state: State) -> None:
    @asynccontextmanager
    async def lifespan(app: FastAPI) -> AsyncGenerator[None, None]:
        state.app_startup = True
        yield
        state.app_shutdown = True

    app = FastAPI(lifespan=lifespan)

    @app.get("/")
    def main() -> Dict[str, str]:
        return {"message": "Hello World"}

    assert state.app_startup is False

				}
				m[v.Int()] = true
				i++
			}
			if i != 3 {
				t.Fatalf("should loop three times")
			}
		}},
		{"func", ValueOf(func(yield func(int) bool) {
			for i := range 4 {
				if !yield(i) {
					return
				}
			}
		}), func(t *testing.T, s iter.Seq[Value]) {
			i := int64(0)
			for v := range s {
				if v.Int() != i {

                                         island_body.begin(),
                                         std::prev(island_body.end()));
    // Forward island fetches (tf_executor.yield operands) to island op result
    // uses.
    for (auto result :
         llvm::zip(island_op.getOutputs(), island_op.GetYield().getFetches()))
      std::get<0>(result).replaceAllUsesWith(std::get<1>(result));
  }

              ),
          ]
      )
      def add(self, x, y):
        res = math_ops.add(x, y)
        return {'output': res}

    def data_gen():
      for _ in range(255):
        yield {
            'x': ops.convert_to_tensor(
                np.random.uniform(size=(10)).astype('f4')
            ),
            'y': ops.convert_to_tensor(
                np.random.uniform(size=(10)).astype('f4')

    %4 = "tf.LayerNorm"(%a1, %a2, %a3, %a4) {_tfl_quant_trait = "fully_quantizable", device = ""} : (tensor<128x128xf32>, tensor<1xf32>, tensor<1xf32>, tensor<1xi32>) -> tensor<128x128xf32>
   "tfl.yield"(%4) : (tensor<128x128xf32>) -> ()
  }) {_tfl_quant_trait = "fully_quantizable", device = ""} : (tensor<128x128xf32>, tensor<1xf32>, tensor<1xf32>, tensor<1xi32>) -> tensor<128x128xf32>

//   func @my_fn(%argi...) -> (result_t) {
//     %results:[[n_args]] = tf_executor.graph {
//        %island_results:[[nargs + 1]] = tf_executor.island {
//          ... original ops ...
//          tf_executor.yield %results...
//        }
//        tf_executor.fetch %island_results#...
//      }
//      return %graph_results#...
//    }

#define GEN_PASS_DEF_FUNCTIONALTOEXECUTORDIALECTCONVERSIONPASS

     *
     * Unfortunately, [symbol] for the `action("")` call will be pointing to the `Function1<P1, R>.invoke(p1: P1): R`, because we
     * intentionally unwrap use-site substitution overrides. Because of this, `symbol.name` will yield `"p1"`, and not `"bar"`.
     *
     * To overcome this problem, [name] property is introduced: it allows to get the intended name of the parameter,
     * with respect to `@ParameterName` annotation.
     *

)


private
val Class<*>.relevantFields: Sequence<Field>
    get() = declaredFields.asSequence()
        .filterNot { field ->
            field.isStatic
                || field.isTransient
                || Workarounds.isIgnoredBeanField(field)
        }.filter { field ->
            isNotAbstractTaskField(field)
                || field.name in abstractTaskRelevantFields
        }

        "tf.saved_model.save(tf_module, saved_model_dir)\n",
        "\n",
        "def calibration_dataset():\n",
        "  rng = np.random.default_rng(seed=1235)\n",
        "  for _ in range(2):\n",
        "    yield {\n",
        "        'lhs_operand': rng.uniform(low=-1.0, high=1.0, size=input_shape).astype(\n",
        "            np.float32\n",
        "        )\n",
        "    }\n",

* It can do something to that **response** or run any needed code.
* Then it returns the **response**.

!!! note "Technical Details"
    If you have dependencies with `yield`, the exit code will run *after* the middleware.

    If there were any background tasks (documented later), they will run *after* all the middleware.

## Create a middleware