mod.engine.dispose()


def test_crud_app(client: TestClient):
    # TODO: this warns that SQLModel.from_orm is deprecated in Pydantic v1, refactor
    # this if using obj.model_validate becomes independent of Pydantic v2
    with warnings.catch_warnings(record=True):
        warnings.simplefilter("always")
        # No heroes before creating
        response = client.get("heroes/")

        return MultipartBody(boundary, type, parts.toImmutableList())
      }
    }

  companion object {
    /**
     * The "mixed" subtype of "multipart" is intended for use when the body parts are independent
     * and need to be bundled in a particular order. Any "multipart" subtypes that an implementation
     * does not recognize must be treated as being of subtype "mixed".
     */
    @JvmField

That's what we'll solve, let's load the model before the requests are handled, but only right before the application starts receiving requests, not while  the code is being loaded.

## Lifespan { #lifespan }

Now let's go back a bit and understand what is all that.

The `password` "flow" is one of the ways ("flows") defined in OAuth2, to handle security and authentication.

OAuth2 was designed so that the backend or API could be independent of the server that authenticates the user.

But in this case, the same **FastAPI** application will handle the API and the authentication.

So, let's review it from that simplified point of view:

This was changed in FastAPI 0.106.0 with the intention to not hold resources while waiting for the response to travel through the network.

/// tip

Additionally, a background task is normally an independent set of logic that should be handled separately, with its own resources (e.g. its own database connection).

So, this way you will probably have cleaner code.

///

   * This means that dynamically loaded web applications and OSGi bundles can't be unloaded.
   *
   * If the library is loaded in an application class loader, we try to break the cycle by loading
   * Finalizer in its own independent class loader:
   *
   * System class loader -> Application class loader -> ReferenceMap -> FinalizableReferenceQueue ->
   * etc. -> Decoupled class loader -> Finalizer
   *

        // if multiple test tasks run in parallel. We also pass it to the Gradle build as a maximum (maxParallelForks)
        // for each test task, such that we are independent of whatever default value is defined in the build itself.
        "-Dorg.gradle.workers.max=$maxParallelForks",
        "-PmaxParallelForks=$maxParallelForks",
        PLUGINS_PORTAL_URL_OVERRIDE,

 * independent of N (but depends on the quantiles being computed). When calling {@link
 * ScaleAndIndex#computeInPlace computeInPlace} (in {@linkplain ScaleAndIndexes#computeInPlace
 * either form}), only the overhead is required. The number of object allocations is independent of
 * N in both cases.
 *
 * @author Pete Gillin
 * @since 20.0
 */

	switch e.getFunctionName() {
	case aggFnCount, aggFnAvg, aggFnMax, aggFnMin, aggFnSum:
		return e.getAggregate()
	default:
		return e.evalSQLFnNode(r, tableAlias)
	}
}

// evalNode on a literal value is independent of the node being an
// aggregation or a row function - it always returns a value.
func (e *LitValue) evalNode(_ Record) (res *Value, err error) {
	switch {
	case e.Int != nil:

$ python ./scripts/docs.py build-all

Building docs for: en
Building docs for: es
Successfully built docs for: es
```

</div>

This builds all those independent MkDocs sites for each language, combines them, and generates the final output at `./site/`.

Then you can serve that with the command `serve`:

<div class="termy">

```console