The Stroz

Testing MySQL Applications With Node.js and Testcontainers

Aug 11, 2023
16 minutes

MySQL Node.js TDD Better Tests Testcontainers

Over the years, I have become obsessed with writing code tests. One big pain point with writing tests is testing code that interacts with a database. In this post, we are going to talk about some of the difficulties that may be encountered when writing tests for database interaction and how we can use Testcontainers to mitigate those issues.

The Issues

On one project I worked, the development team (about 15 of us) shared a dev database…yeah, I know…ew.

There are quite a few issues with this setup when writing tests for code that will interact with a database.

  • Running tests that insert data would cause those test inserts to be visible to every other developer.
    • This could lead developers to think there are issues with their code, as they may see unexpected results.
  • A developer working on a new DDL script could ‘break’ the database for everyone else.
    • This happened more times than I care to remember.

The solution put into place before I joined the project was only marginally better - we ran all of our tests for database interactivity against an H2 database. While this addressed many issues we encountered, it caused other issues. For example, the version of H2 we were using did not support stored procedures or custom/stored functions. So, if we needed to write a new stored procedure or custom/stored function, we could only test that a particular error was thrown by the testing framework.

It would have made everyone’s life easier if we ran our tests against the same database system the project used. For various reasons, standing up copies of the database in our development environment was not an option. So, we were left with the less-than-ideal architecture that had been cobbled together over a few iterations of the codebase.

A Solution

If I had known about Testcontainers at the time, I would have lobbied to allow us time to implement a testing architecture that would make everyone’s lives easier. So, what is Testcontainers? The ‘official’ description from the website states:

Testcontainers is an open-source framework for providing throwaway, lightweight instances of databases, message brokers, web browsers, or just about anything that can run in a Docker container.

More simply, Testcontainers allows developers to spin up Docker images for testing purposes. The framework can be used for a variety of scenarios. Still, we will focus on using Testcontainers for testing interaction with MySQL. For this post, we are using the Node MySQL module using the built-in Node Test Runner.

Before We Get Started

Before we get started, you need to have the following installed:

  • Node - I am using version 20.5.1. There are some issues with earlier versions of Node Test Runner, so your mileage may vary.
  • Docker - Or a Docker equivalent such as Rancher or Podman. I am using Rancher on my machine.
    • Check out the documentation on setting up other container runtimes.
  • An IDE
  • Git - This will be the best way to get the demo up and running quickly.

Getting the Code

To get the code in this demo, head over to this GitHub repo and clone it. Right now, there is only a single demo. I plan to expand these demos to include other test runners (such as Jest or Mocha) and languages (such as Java).

The command to clone the repo from the command line over SSH is:

git clone git@github.com:boyzoid/Test-Containers.git

Next, Change the directory for this demo.

cd Test-Containers/Node/Node_Test_Runner

Lastly, we need to install all the Node dependencies by running the following:

npm install

Code Overview

We should now have a directory structure that looks like the following:

Project Directory Structure

Let’s talk about some of these files/directories.

  • The repository directory holds the files responsible for interacting with the database.
    • In this demo, we have only a single file named user-repository.js that is used to interact with the user table.
  • The setup directory contains files that will be run each time we spin up our tests.
    • I named the single file in this directory ddl.js because it is used to run SQL commands that will set up our database schema. In a future post, I plan on integrating database migration scripts to show how we can track database changes more easily.
  • The test directory holds all of our test files - or file in this case.
    • There is a single file named user-repository.test.js that contains our tests for the user repository.
    • By naming the directory test, the Node Test Runner detects that test files are in this directory.
  • The utils directory contains files with code I would reuse often.
    • The testUtils.js file contains code I would often use when running tests - such as a method that will generate a random string of a given length.
    • The dataUtils.js file contains code used to massage the data. In this case, there is a single method that takes a result set from a query to a MySQL database and converts it to JSON.

If you take a peak inside package.json, you will see the following:

 "dependencies": {
    "@mysql/xdevapi": "^8.0.33"
  },
  "devDependencies": {
    "testcontainers": "^9.9.1"
  }

These dependencies show what modules are used for this demo. We can see we are using the X DevAPI - which is the Oracle-supported connector (SDK) for interacting with MySQL in a Node application. Under the devDependencies, we specify the testcontainers module.

Deeper Dive Into The Code

Let’s take a deeper dive into the different pieces of this code.

The Tests

Since it is the focus of this demo, let’s look at user-repository.test.js first.

The file starts off with a few import statements.

import { test, before, after } from 'node:test';
import { strict as assert } from 'node:assert';

These statements import modules from the Node tests Runner. The second one sets up assert to use a strict comparison by default. We’ll see more about this in a little bit.

Next, we import the testcontainers module.

import {MySqlContainer} from "testcontainers";

Lastly, we import the project-specific code.

import { ddl } from "../setup/ddl.js";
import { testUtils } from "../utils/testUtils.js";
import UserRepo from "../repository/user-repository.js";

These import the ddl module to set up our database schema when we run the tests, our testUtils module with helper methods for running tests, and our UserRepo. We will go into more detail on each of these soon.

We then specify our test suite using test(). We can nest sub-suites of tests using test() inside this suite. We also initialize some variables that will be used in our tests.

test('Testing User Repository, async (t) => {
let container;
let userRepo;

Since much of the X DevAPI database interaction is asynchronous, we use async in the test definition. Also, not that we pass an argument named t into the callback function.

When setting up a test suite, we can define code that gets run before the suite is run. In Node Test Runner, we define this code inside before(). Anything in this block is run at the start of the test suite. We use before() in this example to set up our container and create an instance of UserRepo. Finally, we execute some code in the ddl module to set up our database.

before(async ()=>{
        container = await new MySqlContainer().withExposedPorts(3306, 33060).start();  
        userRepo = new UserRepo(
            container.getUsername(),
            container.getUserPassword(),
            container.getHost(),
            container.getMappedPort(33060),
            container.getDatabase()
            )
        await ddl.createUserTable(await userRepo.getSession()); 
    })

When we create our container, we use the withExposedPorts() method because the X DevAPI connector for Node must have access to port 33060 as well as port 3306. Testcontainers handles mapping these ports in the container to ports on our local system when this container is spun up.

When we create an instance of UserRepo, we pass in connection information - username, password, host, port number, and database name. These are all provided by the connector. Note that we use container.getMappedPort(33060) for the port number. This will give the local system port mapped to port 33060 on the container.

The last task we perform in before is to call ddl.createUserTable() to run the SQL query to create our user table and pass in the result of userRepo.getSession().

In a test suite, we use after() to define code that runs after all the tests in the suite have been executed. In our example, we want to stop the container after all the tests run. The code for this would be:

after(async ()=>{
    await container.stop();
})

Our test suite has two tests defined. Let’s look at the first one, designed to test that the container is running and we can connect to it.

await t.test('Container should be running', async (t)=>{
    const queryResult = await container.executeQuery("SELECT 1 as res");
    assert.equal(queryResult,"res\n1\n" )
});

We define individual tests using the t.test() method. This takes two arguments:

  1. The name of the test. This should be descriptive but not overly verbose, as it could be challenging to read.
  2. A callback function that does the actual testing. Note we add t to this function just like we did for test().

On the first line of our test, we use functionality built into our container to run a simple query. On the second line, we use assert.equal() to verify that the query results match what we expect. The test would fail if the call to assert.equal() returned false.

Our second test uses our instance of UserRepo to add a user to the user table and test that the value was added correctly.

await t.test('Should create user', async(t)=>{
        const name = testUtils.generateString(50);
        await userRepo.createUser({name: name});
        const queryResult = await userRepo.getUserByName(name);
        assert.equal(1, queryResult.length);
        assert.equal(name, queryResult[0].name);
    })

In the first line, we use testUtils.generateString() to create a random string that will be used as the name for our new user. We pass this value as part of a JSON object into the userRepo.createUser() method. After we have created a user, we need to check that it was added correctly. In this example, we use userRepo.getUserByName() and pass our randomly generated string.

We have two assertions:

  1. We check that only 1 row is returned.
  2. We check that the value of the name property of the array’s first (and only) row is equal to the value of the name variable we generated.

Usually, I would not use other code in UserRepo to test the functionality of a method in UserRepo, but I did this to make it easier to follow. I plan to put out other blog posts for rules I like to follow concerning this behavior.

The Repository

Now that we have broken down our test let’s look at UserRepo.

At the top of that file, we have imports for the X DevAPI and our dataUtils.

import * as mysqlx from '@mysql/xdevapi'
import dataUtils from '../utils/dataUtils.js'

We define this module as a class that can be instantiated elsewhere. We also define a few private variables and a constructor to set up variables when the class is instantiated.

export default class UserRepo{
    #connectionUrl
    #pool
    constructor(dbUser, dbPassword, dbHost, dbPort, schemaName) {
        this.#connectionUrl =
            `mysqlx://${dbUser}:${dbPassword}@${dbHost}:${dbPort}/${schemaName}`
        this.#pool = mysqlx.getClient(this.#connectionUrl, {
            pooling: {
                enabled: true,
                maxSize: 10,
                maxIdleTime: 20000,
                queueTimeout: 5000
            }
        })
    }

The private variables are named #connectionUrl and #pool. These are used to set up the MySQL connection string and the connection pool, respectively.

In our constructor, we expect arguments for the dbUser, dbPassword, dbHost, dbPort, and schemaName. These values are used to build out our connection string.

Note that we use mysqlx as the protocol in the connection string. This is because we need to connect over the X Protocol to use the X DevAPI.

Next, we set up a connection pool by calling mysqlx.getClient() and passing in #connectionUrl and a configuration block for the connection pool.

We need to use a database session every time we interact with the database, so we add a helper method to grab a session from the connection pool.

async getSession(){
    return await this.#pool.getSession()
}

There are two other methods in UserRepo. Each uses the X Dev API to insert and select data from our database. Let’s take a look.

async createUser(user) {
    const session = await this.getSession();
    const db = session.getSchema();
    const table = db.getTable('user');
    table.insert(['name'])
        .values(user.name)
        .execute();
    session.close();
}

In createUser, we expect a JSON object with a name property. First, we grab a session from the connection pool. Then we call session.getSchema() to get a reference to the schema we want to use. Since we used the schema name in the connection string, this will return the default schema.

Next, we get a reference to the table into which we wish to insert data.

Once we have the table, we call table.insert() and pass in the columns we want to insert into. In this case, we are just inserting into the name column. As you will see soon, there is an id column that is a primary key, but it is an auto increment column. We use method chaining to call values() and pass in the value we want to use and then call execute(). This code will be turned into an INSERT statement and be executed against the database.

The last thing we do is close the session. When we do this, the session is returned to the connection pool.

The other method in UserRepo is getUserByName(), which is defined as:

async getUserByName(name) {
    let ret = [];
    const session = await this.getSession();
    const db = session.getSchema();
    const table = db.getTable('user');
    const rows = await table.select(['id', 'name'])
        .where('name = :nameParam')
        .bind('nameParam', name)
        .execute();
    const data = rows.toArray();
    const columns = rows.getColumns()
    ret = dataUtils.formatData( data, columns )
    session.close();
    return ret;
}

Breaking this down, we see that, again, we first get a session from the connection pool and get references to the schema and table. We use table.select() and pass in an array of column names. Here we want to return id and name. Using method chaining, we then call where() and pass in a WHERE condition for our query. We use :nameParam to parameterize the input and bind() to set the parameter’s value. And we then call execute().

The result of the call to execute() is set into a variable named rows. We call rows.toArray() to turn our result set into an array (which is actually a two-dimensional array). Currently, the column names are not returned in the results set, but we can get that information from rows.getColumns(). We then call dataUtils.formatData() and pass in the array of data, and the column information, and it will return a JSON object where the properties match the column names.

We then close the session and return the value returned from dataUtils.formatData().

The Data Utilities

Let’s look at the code since we were talking about dataUtils. Over time, I might have more functionality in this file, but for now, it is just formatData().

formatData(data, columns){
    let ret = [];
    data.forEach((row) =>{
        let obj = {};
        row.forEach((item,i)=>{
            obj[columns[i].getColumnLabel()] = item;
        })
        ret.push(obj)
    })
    return ret;
}

As I mentioned, the data in array form is a two-dimensional array where each element in the main array is an array of values from the database. The values that make up the columns argument are also in an array, and the order of the column names matches the order of the values. This means we can loop over the data array, and when looping over each element in the sub-array, we can use the same array index to get the column name. Here I use getColumnLabel() to get the column names. There is also a method named getColumnName(), but you should be careful using this if there is a column alias in the call to select() because getColumnName() returns the actual database column name. In contrast, getColumnLabel() will return the defined alias.

As we can see, this code creates a JSON array of objects where each object’s property matches the name of the database column (or alias) in the query. While this fits our needs here, formatData() would likely need to be updated if we needed nested objects in our results.

DDL Setup

As I mentioned, I have a code that runs when the test suite starts that sets up our database. We are going to take a look at that code now.

async createUserTable(session) {
    const sql = "CREATE TABLE IF NOT EXISTS `user` (id INT NOT NULL AUTO_INCREMENT, name VARCHAR(100) NOT NULL, PRIMARY KEY (id))";
    const query = await session.sql(sql).execute();
    session.close();
}

The createUserTable() method expects a session from the connection pool. We create a SQL string that creates a table named user and defines the two columns. We then session.sql() and pass in this string and then call execute(). This is the syntax we would use to run regular queries if we did not want to use the X DevAPI for our CRUD operations.

Once again, we need to close the session.

As an application gets more complex, a formal system of tracking schema changes might become necessary. I plan on covering this topic in a future blog post.

Testing Utilities

When I write tests, I prefer not to use hard-coded values. Instead, I like using randomly created values (when I can) to test database inserts, etc. One way I handle this is by creating an easy-to-use library of helper methods for generating test data. In this example, that testUtils. For this example, I have a single method that returns a random string of the given length.

generateString(length){
    const characters ='ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789';
    let result = '';
    const charactersLength = characters.length;
    for ( let i = 0; i < length; i++ ) {
        result += characters.charAt(Math.floor(Math.random() * charactersLength));
    }
    return result;
}

In the past, I have also had the need to generate random numbers, a random number in a given range, create an array of an arbitrary length, etc. I would add these methods to testUtils so they can be reused easily.

Running the Tests

I hope you are still with me because we are now ready to run our tests.

You can run the tests using a few different methods. The easiest is to execute the following command in a command/terminal window:

node --test

It may take a while to complete the first time you run these tests. That is because the Docker images need to be fetched. When the tests are completed, you should see an output that resembles the image below.

Test result output

We can tell the Node test Runner to watch a directory for changes and run the tests when a file is saved by using the command:

node --test --watch ./test

Alternatively, for this project, you can run the following:

npm run test

The result for each of these will look similar to this screenshot:

Test result with --watch

Now, if a change is made to any of the files used in the tests, the test suite(s) will be re-run.

The Wrap-up

One of the most essential parts of writing code is reliably testing to ensure the code is doing what it is intended to do. When you have code interacting with a database, it can be difficult to effectively test that code. The Testcontainers framework gives developers a tool to make it easier to reliably test that database interactivity.

Image by Darko Stojanovic from Pixabay

Related Entries