Skip to content

Latest commit

 

History

History
384 lines (293 loc) · 17.8 KB

File metadata and controls

384 lines (293 loc) · 17.8 KB

Build - main Branch Swift 5.6, 5.7 and 5.8 Tested Join the Smoke Server Side community on gitter Apache 2

SmokeFrameworkApplicationGenerate

Code generator to generate SmokeFramework-based applications from service models.

Generate a SmokeFramework application package from an Open API 3.0/Swagger 2.0 specification file

Step 1: Check out this repository

Clone this repository to your local machine.

Step 2: Create a new directory in which to create your service.

You will point this code generator to this directory to output the generated code.

Step 3: Create a model describing your service using Open API 3.0 or Swagger 2.0

Follow the Open API spec or Swagger spec to create an API specification for your service. Place this file in the directory you have created. The following steps assume you have called this file OpenAPI30.yaml but it can be called anything you require.

Step 4: Create a configuration file for the code generator

Create a smoke-framework-codegen.json file in the directory you have created with the following content-

{
  "baseName" : "EmptyExample",
  "modelFilePath" : "OpenAPI30.yaml",
  "modelFormat" : "OPENAPI3_0",
  "generationType" : "serverUpdateWithPlugin",
  "initializationType": "STREAMLINED",
  "testDiscovery": "ENABLED",
  "mainAnnotation": "ENABLED",
  "asyncAwait": {
    "clientAPIs": "ENABLED",
    "asyncOperationStubs": "ENABLED",
    "asyncInitialization": "ENABLED"
  },
  "operationStubGenerationRule" : {
    "mode" : "allFunctionsWithinContext"
  }
}

This JSON file can contain the following fields-

  • modelFilePath: Specifies the absolute or relative (to the base directory path) file path to the Swagger model. Required.
  • modelFormat: Specifies the format of the provided model. OPENAPI3_0 indicates Open API 3.0 and SWAGGER indicates a Swagger 2.0 model. Optional, defaults to Swagger.
  • baseName: A base name for your service (without the "Service" postfix). Required.
  • applicationSuffix: The suffix that is combined with the baseName to create the service's executable name. Defaults to Service.
  • generationType: server to generate a new service; serverUpdateWithPlugin to preserve changes to existing operation handlers and generate model and client files as part of the build process. Required.
  • applicationDescription: A description of the application. Optional.
  • modelOverride: A set of overrides to apply to the model. Optional.
  • initializationType: STREAMLINED is recommended and uses a code generated initializer protocol to reduce the manual setup required. ORIGINAL requires additional manual initialization. Optional; defaulting to ORIGINAL for legacy applications.
  • testDiscovery: ENABLED is recommended and will not generate LinuxMain.swift and associated TestCase lists. DISABLED will code generate these. Optional; defaulting to DISABLED for legacy applications.
  • mainAnnotation: ENABLED is recommended and will not generate main.swift and will instead specify the @main annotation on the application initializer. DISABLED will code generate main.swift. Optional; defaulting to DISABLED for legacy applications.
  • asyncAwait: Specifies if 1) async client APIs will be generated, 2) operation handler stubs will be code generated as async methods and 3) if the initialization and shutdown methods can be async methods (and will code generated as such if they don't exist). Optional; by default clientAPIs and asyncOperationStubs will be enabled but asyncInitialization will be disabled for legacy applications.
  • eventLoopFutureOperationHandlers: ENABLED will support the use of EventLoopFuture-returning operation handlers. Integration for these types of handlers are contained in the SmokeAsyncHTTP1 product of the smoke-framework package. A dependency on this product will need to be added to the \(applicationBaseName)OperationsHTTP1 product of the code generated application if this option is enabled. Optional, disabled by default.
  • httpClientConfiguration: Configuration for the generated http service clients. Optional.
  • operationStubGenerationRule: How operation stubs are generated and expected by the application. It is recommended that new applications use allFunctionsWithinContext which will generate operation stubs within extensions of the Context type[1].

The schemas for the modelOverride and httpClientConfiguration fields can be found here - https://github.com/amzn/service-model-swift-code-generate/blob/main/Sources/ServiceModelEntities/ModelOverride.swift.

An example configuration - including modelOverride configuration - can be found here - https://github.com/amzn/smoke-framework-examples/blob/612fd9dca5d8417d2293a203aca4b02672889d12/PersistenceExampleService/smoke-framework-codegen.json.

[1] Existing smoke-framework based applications may have operation handlers that are standalone functions. You can use the allStandaloneFunctions value for operationStubGenerationRule to continue this style or migrate the operation handlers - either at once or one-by-one - to being functions on the Context type. See the Migration section at the end of this README for more information.

Step 4: Run the code generator

From within your checked out copy of this repository, run this command-

swift run -c release SmokeFrameworkApplicationGenerate \
  --base-file-path <the path to where you want the service to be generated>

An example command would look like this-

swift run -c release SmokeFrameworkApplicationGenerate \
  --base-file-path /Volumes/Workspace/smoke-framework-examples/PersistenceExampleService

An example service based on the command above can be found here.

Step 5: Modify the stubbed service generated

The code generator will produce a Swift Package Manager repository with the following directory structure-

- Package.swift
- .swiftlint
- .gitignore
- Sources
  - (base-name)Client[1]
  - (base-name)Model[1]
  - (base-name)Operations
  - (base-name)OperationsHTTP1[1]
  - (base-name)Service
- Tests
  - LinuxMain.swift      
  - <base-name>OperationsTests

The following three sections of the repository provides initial stubs and can be filled out as required for the service.
A generationType of serverUpdate will not overwrite changes in these sections-

  • (base-name)Operations: Stub implementations for each operation; should be modified to fullfill the services's logic.
  • (base-name)OperationsTests: Stub test implementations for each operation; should be modified to test the services's logic.
  • (base-name)Service: Operations context initialization and shutdown code; should be modified to create the context for the current environment.

The following three section contain code generated code to help the service operate. These sections will be empty if this generator is configured as an SPM plugin and otherwise should not be manually modified. A generationType of serverUpdate will overwrite changes in these sections-

  • (base-name)Client: APIGateway and mock clients for the service; should not be manually modified.
  • (base-name)Client: Input and output structures and types for the service; should not be manually modified.
  • (base-name)Client: Operation selection and input/output type handling specific to HTTP1; should not be manually modified.

If you add additional operations at a later date, the code generator can be re-run manually to generate operation and test stubs for these new operations.

[1] In the default configuration, the full contents of these packages will be generated during the build process.

Migration to using smoke-framework-codegen.json

smoke-framework-codegen.json generation

Existing smoke-framework based applications may not have been generated by using a smoke-framework-codegen.json file. Moving to a single configuration file is intended to simplify the code generation process and to prepare the way for code generation as part of the build process.

You can easily generate the configuration file by adding the --generate-code-gen-config true flag to the command previously used to generate the service.

swift run -c release SmokeFrameworkApplicationGenerate \
  --base-file-path <the path to where you want the service to be generated> \
  --base-name <a base name for your service (without the "Service" postfix)> \
  --model-path <the path to the Swagger model you created> \
  --generate-code-gen-config true \
  --generation-type [server: to generate a new service|serverUpdate: to preserve changes to existing operation handlers]
 [--model-override-path <optionally the path to a json file that specifies various overrides to the model>]

An example command would look like this-

swift run -c release SmokeFrameworkApplicationGenerate \
  --base-file-path /Volumes/Workspace/smoke-framework-examples/PersistenceExampleService \
  --base-name PersistenceExample \
  --model-path /Volumes/Workspace/smoke-framework-examples/PersistenceExampleService/Swagger.yaml \
  --generate-code-gen-config true \
  --generation-type server \
  --model-override-path /Volumes/Workspace/smoke-framework-examples/PersistenceExampleService/modelOverride.json

The generated smoke-framework-codegen.json file will specify an operationStubGenerationRule of allFunctionsWithinContextExceptForSpecifiedStandaloneFunctions with all current operations listed under operationsWithStandaloneFunctions. The means that by default the migration will require no change to any existing operation handlers but any new operations will have handlers generated within extensions of the Context type.

Operation stub generation

It is also possible to change where the generated application expects an operation handler to be - either a standalone function or a function on the Context type. Migration of operation handlers is entirely optional and can be done if you find this style more convenient.

To migrate an existing operation with a standalone handler function to the Context type, remove the operation from the operationsWithStandaloneFunctions list and manually move the operation handler function to an extension of the Content type, also removing the explicit context parameter to the function.

For example, an existing operation handler-

public func handleGetCustomerDetails(
        input: EmptyExampleModel.GetCustomerDetailsRequest,
        context: EmptyExampleOperationsContext) throws -> EmptyExampleModel.CustomerAttributes {
    ...
}

would be migrated to-

extension EmptyExampleOperationsContext {
    public func handleGetCustomerDetails(input: EmptyExampleModel.GetCustomerDetailsRequest) throws
    -> EmptyExampleModel.CustomerAttributes {
        ...
    }
}

Migration to using the generator as an SPM Plugin

Starting with Swift 5.6, this generator can be used during the build process to avoid having to check the fully code-generated model, client and http1 integration files into the source repository. This section describes the steps to migrate a Smoke-framework based application to using this generator as a plugin.

This generator can still be run manually from the command line to generate operation and test stubs after adding operations to the model file. Changes to the model file that only change payload of existing operations will no longer require this generator to be run manually.

Step 1: Use build tools 5.6

In order to use this generator as an SPM plugin, the application will have to use Swift tools version 5.6 and at compile with at least the Swift 5.6 compiler. At the top of the application's Package.swift manifest, make sure the correct Swift tools version is specified.

// swift-tools-version:5.6

Step 2: Add a dependency on this generator

Still in the Package.swift manifest, add a dependency on this package.

.package(url: "https://github.com/amzn/smoke-framework-application-generate", from: "3.0.0-beta.3")

Step 3: Specific plugs for the Model, Client and Http1 Integration packages

Still in the Package.swift manifest, add plugin declarations for the model, client and Http1 Integration packages. For example

For the Model package-

.target(
    name: "EmptyExampleModel", dependencies: [
        .product(name: "SmokeOperations", package: "smoke-framework"),
        .product(name: "Logging", package: "swift-log"),
    ],
    plugins: [
        .plugin(name: "SmokeFrameworkGenerateModel", package: "smoke-framework-application-generate")
    ]
),

For the Client package-

.target(
    name: "EmptyExampleClient", dependencies: [
        .target(name: "EmptyExampleModel"),
        .product(name: "SmokeOperationsHTTP1", package: "smoke-framework"),
        .product(name: "AWSHttp", package: "smoke-aws-support"),
    ],
    plugins: [
        .plugin(name: "SmokeFrameworkGenerateClient", package: "smoke-framework-application-generate")
    ]
),

For the Http1 Integration package-

.target(
    name: "EmptyExampleOperationsHTTP1", dependencies: [
        .target(name: "EmptyExampleOperations"),
        .product(name: "SmokeOperationsHTTP1", package: "smoke-framework"),
        .product(name: "SmokeOperationsHTTP1Server", package: "smoke-framework"),
    ],
    plugins: [
        .plugin(name: "SmokeFrameworkGenerateHttp1", package: "smoke-framework-application-generate")
    ]
),

Step 4: Delete existing files in these packages

Delete the previously generated files in the Model, Client and Http1 Integration packages. These will now be generated at compile time using this generator.

Step 5: Update the generationType in the configuration file

Update the generationType specified in the smoke-framework-codegen.json file to either serverUpdateWithPlugin or serverWithPlugin.

  "generationType" : "serverUpdateWithPlugin",

Step 6: Manually run this generator

Manually run the generator. This should create a placeholder file in each of the Model, Client and Http1 Integration packages. Due to a current limitation of the SPM plugins for code generators, a placeholder Swift file is required in each package to avoid the package as being seen as empty. These files need to be a Swift file but doesn't require any particular contents.

Using the generator as an SPM Plugin with an external model

You can use the SPM plugin with a model defined in a seperate package by declaring that package as a dependency of the targets using the generator.

Step 1: Prepare the model package

Setup the model package as a Swift package that declares the model file as a resource.

// swift-tools-version: 5.6
// The swift-tools-version declares the minimum version of Swift required to build this package.

import PackageDescription

let package = Package(
    name: "MyModelPackage",
    products: [
        .library(
            name: "MyModelPackage",
            targets: ["MyModelPackage"]),
    ],
    targets: [
        .target(
            name: "MyModelPackage",
            dependencies: [],
            path: "configuration/api",
            resources: [.copy("Swagger.yaml")]),
    ]
)

Note: If the target in this package doesn't define any Swift files, you will need to create an empty Swift file due to a current limitation in SwiftPM.

Step 2: Add the model package as a dependency

Add this model package as a dependency of the package with your service code.

    .package(url: "https://github.com/DonnaNoble/my-model-package.git", from: "1.0.0"),

and then as a dependency of any targets you are using the SPM plugin for.

.target(
    name: "EmptyExampleClient", dependencies: [
        .target(name: "EmptyExampleModel"),
        .product(name: "SmokeOperationsHTTP1", package: "smoke-framework"),
        .product(name: "AWSHttp", package: "smoke-aws-support"),
        .product(name: "MyModelPackage", package: "my-model-package"),
    ],
    plugins: [
        .plugin(name: "SmokeFrameworkGenerateClient", package: "smoke-framework-application-generate")
    ]
),

Step 3: Update the code generation configuration file

Finally, in the smoke-framework-codegen.json configuration file, add a modelsLocation block to specify where the model is located, removing the modelFilePath field at the top level.

{
    "baseName" : "EmptyService",
    "modelLocations": {
        "default": {
            "modelProductDependency": "MyModelPackage",
            "modelFilePath" : "Swagger.yaml"
        }
    },
    ...
}

If you need to specify a separate model path for a particular target, you can also add this in the modelsLocation block.

{
    "baseName" : "EmptyService",
    "modelLocations": {
        "EmptyServiceExternalModel": {
            "modelProductDependency": "MyModelPackage2",
            "modelFilePath" : "Swagger.yaml"
        }
        "default": {
            "modelProductDependency": "MyModelPackage",
            "modelFilePath" : "Swagger.yaml"
        }
    },
    ...
}

By default, the name specified by the modelProductDependency field will be used for both the Product and the Target used by that Product where the model file is located. If the Target has a different name, this can be specified with the modelTargetDependency field.

License

This library is licensed under the Apache 2.0 License.