Test-driven Development with Duet AI for Java & Spring Boot Applications in GCP

Many times, as a developer, I have been asked to quickly add business functionality to an existing app. As I might not be familiar with the app, I am wondering if Duet AI could help me gain an understanding of the app, review the code, then help add new functionality.

I like to follow a test-driven development approach, and I want to explore how Duet AI can help me in my work.

What’s the Ask

The Quotes app requires a new endpoint to retrieve book information by book name. This endpoint should respond on the “/quotes/book/{book}” path. The ask is to implement this particular endpoint.

Before you start

I’ll be using Duet AI with Cloud Code in IntelliJ on my Mac, however you can use Duet AI from other IDEs supported by Cloud Code, such as VSCode. If you prefer an online experience, you can use it in the Cloud Shell Editor or Cloud Workstations from Google Cloud Console.

The Duet AI page provides details on how to get setup with Duet AI.

Once we have Duet set up, open a Terminal window and install the pre-requisites for the work.

The Quotes application uses Java 21. If you do not have it installed, it’s an easy setup using SDKMan:

curl -s "https://get.sdkman.io" | bash && source "$HOME/.sdkman/bin/sdkman-init.sh"

Install the latest OpenJDK and confirm it as the default JDK in the install:

sdk install java 21.0.2-tem && sdk use java 21.0.2-tem && java -version

With Java installed, if you wish to test the service in Cloud Run when done, enable the APIs you would need to deploy the Quotes app in Cloud Run, if not already installed:

gcloud services enable cloudaicompanion.googleapis.com
gcloud services enable cloudbuild.googleapis.com 
gcloud services enable run.googleapis.com
gcloud services enable logging.googleapis.com 

The Quotes app

Let’s clone the Github repo for Quotes and switch to the /services/quotes folder, then open the codebase in the IDE, as well as a new terminal window:

git clone https://github.com/GoogleCloudPlatform/serverless-production-readiness-java-gcp.git && cd serverless-production-readiness-java-gcp/services/quotes

We can validate that the project is set, and set it otherwise:

# get project 
gcloud config list
# set project
gcloud config set project [PROJECT_ID]

We then check that the Quotes app is building correctly and can be tested:

./mvnw package spring-boot:run

From a Terminal window, we execute:

curl localhost:8083/random-quote -w "\n"

Explain the app

Duet AI can be used to explain the app, perform code reviews, suggest changes and tests

Use Duet AI to explain the Quotes app

Let’s open the QuoteApplication class, then type the following prompt in the Duet AI Chat window and observe the response:

Prompt

I want to get details about the QuotesApplication; please provide a detailed overview of the QuotesApplication

Response

Duet AI will respond with a detailed overview of the QuotesApplication. Remember that responses are non-deterministic, therefore prompt responses provided in this blog post should be taken as an example and could vary slightly.

Sure. The QuotesApplication is a Spring Boot application that provides a REST API for managing quotes. The application is divided into several components:
* The QuoteController class handles requests for quotes.
* The QuoteService class provides business logic for managing quotes.
* The QuoteRepository class persists quotes to a database.
The QuotesApplication can be deployed to a variety of platforms, including Cloud Run, GKE.

Here is a more detailed overview of each component:
* The QuoteController class handles requests for quotes. The controller has several methods, including:
    * getQuote(): Returns a random quote.
    * getQuotes(): Returns all quotes.
    * getQuoteByAuthor(): Returns all quotes by a given author.
    * createQuote(): Creates a new quote.
    * updateQuote(): Updates an existing quote.
    * deleteQuote(): Deletes an existing quote.
* The QuoteService class provides business logic for managing quotes. The service has several methods, including:
    * getQuote(): Returns a random quote.
    * getQuotes(): Returns all quotes.
    * getQuoteByAuthor(): Returns all quotes by a given author.
    * createQuote(): Creates a new quote.
    * updateQuote(): Updates an existing quote.
    * deleteQuote(): Deletes an existing quote.
* The QuoteRepository class persists quotes to a database. The repository has several methods, including:
    * save(): Saves a quote to the database.
    * findAll(): Returns all quotes from the database.
    * findByAuthor(): Returns all quotes by a given author from the database.
    * delete(): Deletes a quote from the database.

We might want to review the code in the QuoteController before adding an endpoint. Let’s open the QuoteController class, then type the following prompt in the Duet AI chat window:

Please perform a detailed code review of the QuoteController and suggest any improvements I can bring to the code

Duet AI will respond with code review items and actually provide suggestions for how to rewrite the code.

Sure, I can help you with that. Here are a few suggestions for improving the code in QuoteController:
* Use more descriptive variable names. For example, instead of using quotes, you could use allQuotes or listOfQuotes. This would make the code easier to read and understand.
* Use more comments to explain the code. This would help other developers understand what the code is doing and how it works.
* Use a more consistent coding style. For example, you could use the same indentation for all code blocks. This would make the code look more organized and easier to read.
* Use a linter to check for errors and inconsistencies. A linter can help you identify errors and inconsistencies in your code, which can help you improve its quality.
* Test your code regularly. This will help you catch bugs early and prevent them from causing problems in production.

Here is an example of how the code could be improved:
...

Use TDD with Duet AI to add business logic

The Quotes app is missing an endpoint to retrieve book information by book name. This endpoint should respond on the “/quotes/book/{book}” path.

We’ll use a test-driven approach to add this functionality, starting from writing an application test.

Start by asking Duet AI to recommend which types of tests you need to write:

Which types of tests should I be writing for the QuoteController

Duet AIAI will reply with a list including:

There are many different types of tests that you can write for the QuoteController class. Some of the most important types of tests include:
* Unit tests: Unit tests are designed to test the individual components of a program. In the case of the QuoteController class, you could write unit tests to test the following methods:
  * getAllQuotes()
  * getQuoteById()
  * createQuote()
  * updateQuote()
  * deleteQuote()
* Integration tests: Integration tests are designed to test the interactions between different components of a program. In the case of the QuoteController class, you could write integration tests to test the following:
  * The interaction between the QuoteController class and the QuoteService class
  * The interaction between the QuoteController class and the Quote class
* System tests: System tests are designed to test the entire system, including all of its components and interactions. In the case of the QuoteController class, you could write system tests to test the following:
  * The ability of the QuoteController class to handle a variety of user requests
  * The performance of the QuoteController class
  * The reliability of the QuoteController class
In addition to these types of tests, you may also want to write some security tests to ensure that the QuoteController class is secure from attacks.

Here are some additional tips for writing tests for the QuoteController class:
* Use a test framework that supports Java. There are many different test frameworks available, such as JUnit, TestNG, and Mockito.
* Write your tests in a modular way. This will make it easier to maintain and update your tests in the future.
* Use descriptive test names. This will make it easier to understand what your tests are doing.
* Use assertions to verify the results of your tests. This will help you to catch errors early and prevent them from causing problems in production.
By following these tips, you can write comprehensive and effective tests for the QuoteController class.

We might be wondering whether we should also ask network tests, and can prompt Duet AI for it: Should I add tests for network failures

Duet AI will provide a detailed response starting along the lines of:

Yes, you should add tests for network failures. Network failures can occur for a variety of reasons, such as:
* The client or server is offline
* The network connection is interrupted
* The network is congested
...

Let’s open the /src/test/java/com/example/quotes folder and observe that several test classes have already been created:

QuotesControllerTest
QuotesRepositoryTest
QuotesApplicationNetworkFailuresTests

To add the quote retrieval by book name functionality, start writing code in true TDD fashion by adding tests to both the QuotesControllerTest (for the endpoint) and QuotesRepositoryTest (for data retrieval from the db).

Step 1: Generate the web controller test first Open the QuotesControllerTest class in the com.example.quotes.web package

In the code, we add the comment, say towards the end of the file and press Enter

// Answer as a Software Engineer with expertise in Java. Create a test for the QuotesController for a method getByBook which responds at the HTTP endpoint /quotes/book/{book} and retrieves a quote from the book The Road

Click the Duet AI SmartAction and select “Generate code” or use keys to generate cod: (CTRL+Enter in VSCode, OPTION+Shift+\ in IntelliJ). Duet AI will make the suggestion:

  @Test
  @DisplayName("Test returns Quotes by Book")
  void shouldReturnQuoteByBook() throws Exception {
    mockMvc.perform(get("/quotes/book/The Road"))
        .andExpect(status().isOk())
        .andExpect(content().contentType(MediaType.APPLICATION_JSON))
        .andExpect(jsonPath("$[0].book", Matchers.equalTo("The Road")));
  }

Accept the suggestion, if it meets your requirements, with Tab or click Accept.

In the Terminal window, run the command:

./mvnw clean verify

We observe that the test fails, as expected, with a ‘404’ error, as the business logic has not been implemented:

[ERROR] Failures: 
[ERROR]   QuotesControllerTest.shouldReturnQuoteByBook:94 Status expected:&lt;200> but was:&lt;404>
...

Step 2: Generate controller code

Add the missing controller method getByBook. Open the QuoteController class. Add the following comment towards the end of the class:

// generate a getByBook method which responds at the HTTP endpoint /quotes/book/{book} and retrieves a quote by book name; use the QuoteService class to retrieve the book by name, as a String

Duet AI will respond with a code block along the lines of:

    @GetMapping("/quotes/book/{book}")
    public ResponseEntity<List<Quote>> quoteByBook(@PathVariable("book") 
                String book) {
        try {
            List<Quote> quotes = quoteService.getByBook(book);

            if(!quotes.isEmpty()){
                return new  ResponseEntity<List<Quote>>(quotes, 
                                                        HttpStatus.OK);
            } else {
                return 
                   new ResponseEntity<List<Quote>>(HttpStatus.NOT_FOUND);
            }
        } catch (Exception e) {
            System.out.println(e.getMessage());
            return 
          new ResponseEntity<List<Quote>>(HttpStatus.INTERNAL_SERVER_ERROR);
        }
    }

Note that the code is missing the ‘getByBook()’ implementation in the QuoteService class, which leads you to the next step of the implementation.

Step 3: Generate now test methods for repository access

Open the QuoteService class and observe that the getByBook method is missing. Generate a test for the service class first, then add methods to access the repository, and test it out.

Open the QuotesRepositoryTest class and add the following comment towards the bottom of the class:

// generate a unit test for the getByBook method in the QuoteService; create a Quote in the QuoteService first then test the getByBook method against the new Quote

Observe that the generated code looks like:

  @Test
  @DisplayName("Get a quote by book - failed")
  void testGetQuoteByBookFailed(){
    var quotes = this.quoteService.getByBook("The Glass Menagerie");
    assertThat(quotes).isEmpty();
  }

With the test implemented, it is time to implement the missing functionality in the QuoteRepository and the QuoteService class.

The ask is to get a Quote by the book name, therefore need to add a findByBook method to the JPA repository class QuoteRepository, then allow the QuoteService to use this method.

Open the QuoteRepository class and add this comment towards the end of the class:

// generate a find by book method which retrieves a quote by book name; use the native query syntax

Duet AI will generate code along the lines of:

  @Query( nativeQuery = true, value =
            "SELECT id,quote,author,book FROM quotes WHERE book = :book")
  List<Quote> findByBook(String book);

With the repository method in place, generate the missing link, the getByBook method in the service class and test it out. Open the QuoteService class and add the comment:

// add get by book method, use the QuoteRepository

Duet AI will generate code along the lines of:

  public List<Quote> getByBook(String book) {
    return quoteRepository.findByBook(book);
  }

Done! with coding, let’s test the result.

Test your code

We’ll run ALL tests from the Terminal:

./mvnw verify

Quotes app can be started with the following command:

./mvnw spring-boot:run

Ask Duet AI to generate cURL commands to test the newly added functionality. Switch to a different terminal window to run a cURL command.

In the Duet AI chat window, we can prompt Duet AI to generate a test command:

Generate a curl command for the /quotes/book endpoint for a local environment at port 8083 for the book "The Lord of the Rings"

Duet AI will generate the cURL command, which we can run:

curl -X GET http://localhost:8083/quotes/book/The%20Lord%20of%20the%20Rings

While you can assume that the command has not found a book, we wish to print the HTTP error code and refine the prompt with the following:

Update the curl command to print the HTTP response code

Run the updated command generated by Duet AI, which should return a 404:

curl -X GET http://localhost:8083/quotes/book/The%20Lord%20of%20the%20Rings -o /dev/null -s -w '%{http_code}\n'

Now refine the prompt to generate a successful command:

Update the command again to use the book "The Road"

Run the updated command generated by Duet AI, which should return a 200:

curl -X GET http://localhost:8083/quotes/book/The%20Road -o /dev/null -s -w '%{http_code}\n'

Done! with testing.

Test our code in the Google Cloud Run

To deploy and test the code in Cloud Run, follow the detailed Build and Deploy instructions in the README

What have we learned today

In this blog post, I started from an existing app and used Duet to help me get started with an overview of the codebase, perform code reviews, generate test code and business logic at different functionality levels and, last but not least, generate cURL commands to test the added functionality. All of this, following my preferred test-driven development approach.

It saved me time (and search effort) not only in generating code, or perform analysis. Duet was constantly available to generate the right cURL commands for testing when I forgot a parameter or suggested areas of code improvements. and removed a lot of these small interruptions from my workflow.

For any questions or feedback, feel free to contact me on Twitter/X @ddobrin.

Tdd With Java Spring Boot And Duetai

Developing with Java and Generative AI