Practical: Testing and Linting

Initial Due Date: 2024-09-24 11:59PM Final Due Date: 2024-10-25 4:00PM

Github Classroom Gradescope
In the last practical, you created your first npm package. Today, we will actually use the module we installed (Jest) to do some unit testing and we will add a linter called ESLint to perform static analysis.
Learning Goals:
  • Implement unit tests
  • Use a linter to write more consistent, more maintainable code
Submission: Successfully submitting your assignment is an ordered two-step process:
  1. Commit and push your changes to GitHub
  2. Submit your repository to the Gradescope assignment

I encourage you to treat this (and other practical exercises) as a tutorial that you are trying to learn from, rather than an assignment you are trying to complete as quick as possible. Take your time, read the notes thoroughly and don’t hesitate to ask questions.

Prerequisite

  1. Create the git repository for your practical by accepting the assignment from GitHub Classroom. This will create a new repository for you with a skeleton npm package already set up for you.
  2. Clone the repository to you computer with 💻 git clone (get the name of the repository from GitHub). Read more about the process of creating and cloning your repository here.
  3. Open up the package.json file and add your name as the author of the package.
  4. Install package dependencies by executing 💻 npm install in the terminal.

Setting up unit testing

We want to add automated unit tests for our functions. Unit testing typically requires 1) a test runner to automatically run all the tests and report the results, and 2) an assertion library for implementing expectations about the behavior of the code under test. We will use the Jest unit testing package, which provides both. Jest is one of many possible unit testing libraries; it is not necessarily the best (a matter of opinion) but is widely used and integrated into a number of tools we use throughout the semester.

Install Jest by running 💻 npm install -D jest @swc/jest in the shell.

As a reminder, the -D option specifies that you want to update package.json with this dependency, and that it is a “development” dependency. You only need Jest when developing this module (when you would run the tests) and not in production (when you are using the package). The second package is a tool for transpiling JavaScript so that we can use a consistent, modern, feature set across our projects regardless of the target. In this case, we want to use ES modules which are only partially supported Jest. The latter tool helps us work around that limitation. To configure it create a file named jest.config.js in the root of your project. This inserts the transpilation into the jest workflow (in the future the skeleton will typically already include this configuration file).

export default {
  transform: {
    "^.+\\.(tsx|jsx|js)?$": ["@swc/jest"],
  },
};

Notice that the package.json file now specifies this new dependency (your version for this package and others may be slightly different):

"devDependencies": {
  "@swc/jest": "^0.2.24",
  "jest": "^29.0.2"
}

Now that you have a testing library, you want to update the “test” script specified in the package.json file to run Jest. To do so, edit your package.json file to include:

"scripts": {
  "test": "jest"
},

You can now run Jest with 💻 npm test or 💻 npm run test. However, since you don’t have any tests yet, you will get an error.

Example: validSong()

We are going to write a function to check if a string contains a song (or at the very least, a valid sequence of notes). Our song will be defined as a string containing notes separated by spaces. The notes will be a single upper case letter in the set [A-G]. They can optionally be modified by a sharp (‘#’) or a flat (‘b’). For reasons we will not get into, there is no ‘B#’, ‘Cb’, ‘E#’ or ‘Fb’.

> validSong("C C G G A A G")
true
> validSong("Ab Bb C C# Db D  G")
true
> validSong("Ab Cb")
false
> validSong("Ab Z")
false

Writing tests

We are going to practice test driven development (TDD) to create this function.

In index.js, put in the function declaration, with no body other than a simple return statement.

const validSong = (song) => {

  return false;
}

export default validSong;

Now we are going to pick one feature of this function, and write a test to test it. We will start by testing that it accepts valid, unmodified notes. Open the file called index.test.js (jest will automatically run any files ending in test.js) and create a new test suite and a test.

import validSong from "./index";

describe("Testing validSong()", ()=>{
  test("validSong: accepts valid notes", ()=>{

  });
});

Jest provides the test(string, fn) function. This is a basic test comprising a string description that will be printed when the test is run and a function that will provide the body of the test. We have wrapped that test in the describe function, which helps group tests that are related or share setup and tear down functionality. The “setup” and “tear down” is code that must run before the test and after the test is complete (e.g., to make the tests repeatable and independent).

The test function should contain one or more assertions, i.e., tests of state or values in your code. The expect(value) function takes as an argument a value generated by your code in some way and returns an “expectation object”. To turn this into a test, you apply a matcher to test that value. There are a number of different matchers, but we can stick with toBeTruthy and toBeFalsy since our function returns a Boolean value.

Since we have a fixed number of valid notes, we can check them all. Add this to the body of your test:

const validNotes = ["A","B","C","D","E","F","G"];

validNotes.forEach((note)=>{
  expect(validSong(note)).toBeTruthy();
});

The above assertions only test if the function accepts single notes, we also need to make sure that it can handle strings with multiple notes. Add a second test and name it “validSong: accepts compound strings”. Test the function on the string "A B C D E F G".

Run the tests with 💻 npm test. They should fail since we have not yet implemented the function.

Satisfying tests

Now, we need to write the minimal amount of code to make sure that our tests pass. In this case, we can do that by changing the return value of the function to true. Run the tests again. They should now pass.

Iterate

Clearly these were insufficient tests of the behavior we were targeting. When we specified that the function accepted a certain set of letters as valid notes, we really mean “uniquely”. So, we need to test the “sad path” as well, what happens when we give the function invalid input.

We could test all other letters (or symbols!), but that is on the verge of overkill. At a certain point, we need to acknowledge that we are probably not learning more and we are just wasting time (remember that tests should be Fast). So, we want to focus on boundary cases, places near the valid cases. In this case, we can think of a couple of candidates:

  • ‘H’: The letter immediately after the last valid note is a good candidate
  • ‘Z’: This is a “belt and suspenders” test. It probably isn’t necessary, but it is a boundary on the alphabet AND a random additional letter
  • ‘0’: This is a representative number, and also a boundary
  • ‘a’: This would be a valid note if it was uppercase, by testing this, we would be adding a firm requirement that only uppercase letters are accepted (test as specification)

Write a test called “validSong: rejects invalid characters”. Note that each one of those cases needs to be in a separate assertion (i.e., a separate call to validSong), otherwise the first bad character would mask all of the others.

In addition to these, we should test ‘AB’, which would codify the requirement that notes are separated by spaces. Write a fourth test called “validSong: notes must be separated by spaces”. Note that this time we must use two valid notes so we are only testing the spacing.

Run the tests with npm test. They should fail again, which means that we need to return to the implementation.

Here is a basic implementation that checks if each note in the song is valid. Note that we are using another higher-order function: (every).

const validSong = (song) => {
  const validNotes = ["A", "B", "C", "D", "E", "F", "G"];
  // helper to test individual notes
  const validNote = (note) => {
    return validNotes.includes(note);
  }

  // convert the string to an array of notes  
  const songList = song.split(" "); 

  // every returns true if the passed in function returns true for every value
  const valid = songList.every(validNote);

  return valid;
}

Iterate again

Now we need to add in the sharps and flats. Because of the earlier tests, we only need to test if sharps and flats work with valid notes. We also don’t need to test all possible combinations.

Tests we should perform:

  • Are “A#” and “Ab” valid (we will assume that if it works for A, it will work for the others)
  • Are the special cases invalid (“B#”, “Cb”, “E#”, and “Fb”)

Write two more tests: “validSong: sharps and flats are accepted” and “validSong: special cases are rejected”. Remember that we can group multiple inputs that we expect to pass together as any single failure will cause the assertion to fail (though we lose some feedback back where the error might have happened), but we can’t group inputs we expect to return false as they will mask each other (i.e., the first false means the others are never checked).

Run npm test again to make sure it fails (though the special character will not yet).

Let’s update our function to satisfy these tests. After copying in the code below, run npm test again. Everything should pass.

const validSong = (song) => {
  const validNotes = ["A", "B", "C", "D", "E", "F", "G"];
  const badNotes = ["B#", "Cb", "E#", "Fb"];
  // helper to test individual notes
  const validNote = (note) => {
    // make sure the first character is valid
    let valid = validNotes.includes(note[0]);
    
    // check the second character
    if (note.length === 2) {
      // make sure the second character is  "#" or "b"
      valid = valid && (note[1] === "#" || note[1] === "b");

      // make sure it isn't a bad note
      valid = valid && ! badNotes.includes(note);
    }

    return valid;
  }

  // convert the string to an array of notes  
  const songList = song.split(" "); 

  // every returns true if the passed in function returns true for every value
  const valid = songList.every(validNote);

  return valid;
}

Fixing bugs

One of our users is using the function, forgets to put spaces in, and is surprised when our function accepts it. You can (and should) try this too.

> validSong('A#BB')
true

Add a test that tests this case and demonstrate that it does fail. Then fix the function (you should be able to do this with a simple else if in the note validator).

Check the coverage

You can evaluate how comprehensive your test suites are with Jest’s built-in coverage reports. Run 💻 npx jest --coverage. Your one function should be 100% covered! But as we discussed in class, coverage alone is a limited measure of test quality. A high quality test suite will have high coverage but a high coverage test suite does not guarantee high quality.

Some other Jest hints

When multiple tests are failing it can sometimes be hard to narrow down the problem (is it the code, are the tests interfering with each other, etc.) To run just a single test at a time, append .only to the test function, e.g., test.only(() => { ... }). Hopefully you can better localize the problems.

Running a linter

Linters help us identify “programming errors, bugs, stylistic errors, and suspicious constructs”. For this practical we will use ESLint and the AirBnB ESLint configuration. You and I may not agree with all of AirBnB’s (opinionated) settings, but they provide a good starting point. It is OK for us to deviate from their recommendations, but we should do so as a considered decision.

Install ESLint and the AirBnB configuration as a development dependency by running the following command in the root directory of your package (the directory that contains the package.json file):

💻 npm install -D eslint eslint-config-airbnb-base eslint-plugin-import

To configure ESLint you need to create a new file named .eslintrc.json in the root directory of your package with the following contents. Note that the file name is important as ESLint will look for a file with that exact name.

{
  "extends": ["airbnb-base"],
  "env": {
    "node": true,
    "jest": true
  },
  "rules": {
    "quotes": ["warn", "double"]
  }
}

This configuration specifies that you want to use the AirBnB base configuration and that the Node.js and Jest global variables should be predefined.

I’ve also added a rules section to show how we can customize and build on rule sets like the AirBnB base. We are adding a rule here that says all strings should use double quotes, and the linter should issue a warning if they do not.

To prevent ESLint from trying to analyze the files you created as part of the coverage analysis you will want to also create a file named .eslintignore file with the following list of directories (or files) to be ignored. As with .eslintrc.json, this file should be created in the root directory of your package.

# testing
/coverage

Just as we did for testing, you want to add a script entry point to run the linter. Add the following entry to the scripts section of your package.json file:

"lint" : "eslint ."

That section of package.json should now look like:

"scripts": {
  "test": "jest",
  "lint": "eslint ."
},

Running the linter

Run the linter with 💻 npm run lint (which is equivalent to 💻 npx eslint .). I suspect you may have some errors! ESLint can fix many of the formatting errors automatically by running 💻 npm run lint -- --fix (the first -- indicates that the following options should be passed to the underlying command, i.e., eslint, not interpreted by npm). Other errors will require you to manually refactor your code. To learn more about a particular error, Google the rule name, e.g. no-console, or use the links integrated into your IDE. As pedantic as the formatting requirements may seem, enforcing a consistent style is very helpful in a team context. It is much easier to read your teammate’s code when everyone uses the same style.

You will probably be able to eliminate all of the errors using --fix. However, sometimes there will be linting errors that we can’t eliminate. For example, most linting rule require console.log() calls be removed from production code, but you may be writing something that requires it. If, after very careful consideration you decide that you the rule shouldn’t apply, there are ways to disable certain rules. You can do so in a variety of ways, including globally (in .eslintrc.json), for an entire file (with a comment at the top) and for a single line (with an inline comment). For example to turn off the warnings about the console add the following comment to the top of your index.test.js file.

/* eslint-disable no-console */

Alternately you can add // eslint-disable-line to the offending line to disable ESLint on that line.

Use this power as little as possible and in the most targeted way possible. On most assignments a requirement is that your code passes the linter without errors or warnings. That shouldn’t be true just because you disabled the linter…

Install Prettier

Individuality is great, but not always when you are working on a team. The linter goes a long way towards making sure your team is writing consistent code, a formatter like Prettier can help as well. A formatter isn’t looking at the content of your code like ESlint, just the visual formatting (primarily how white spaces and line breaks are used). Under the hood it tokenizes your code and then outputs a new version with its formatting.

Install Prettier with:

💻 npm install -D prettier eslint-config-prettier

You can run Prettier on your file with

💻 npx prettier index.js

Note that this doesn’t actually update your file, it just dumps a formatted version to the terminal. Try it out. Add some extra new lines, change the spacing of your expression. Just make your code ugly (without changing any of the content). Then run Prettier again. You should still see a nicely formatted output.

Prettier can conflict with the formatting rules enforced by ESLint. To prevent the two tools from fighting each other, we also installed an Prettier ESLint configuration that disables all the ESLint rules that conflict with Prettier. Append that configuration to the extends array in .eslintrc so it takes precedence, e.g.,

{
  "extends": ["airbnb-base", "prettier"],
  ...  
}

Automate linting and formatting

The linter and the formatter are great tools, but they only work if you use them. VSCode can be configured to run both in real time and give you feedback as you work. It probably is already showing you the lint errors live. I don’t recommend turning on its support for Prettier however, I found it interrupted my workflow too much (I like using white space to separate piece of code while I’m working, and Prettier wants to collapse them).

Instead, we are going to set them up so that they are run when you commit your code (that way every commit is always nicely formatted). To do so we will install another tool named husky.

💻 npm install -D husky lint-staged

Husky will add “hooks” to steps in your git commands, and in those hooks it will call the other package you installed (lint-staged). Note that recent updates to Husky (version 9+) have changed the setup procedures. These instructions assume that new version (9+). To initialize Husky, execute 💻 npx husky init in the shell. This command will initialize Husky by creating a .husky directory with example hooks and add a “prepare” script to your package.json file to setup Husky when the package is installed in the future.

Then add the following to package.json at the top level of the object. This configuration specifies that we would like to run every JavaScript file through the linter, and every file through Prettier.

"lint-staged": {
  "*.js": "eslint --cache --fix",
  "*": "prettier --ignore-unknown --write"
}

The last step is to create the hook. Run the following in the shell to create or overwrite the pre-commit hook.

💻 echo "npx lint-staged" > .husky/pre-commit

That tells husky that we would like the lint-staged command that we just configured to be run before anything is committed to git. Note that if the linter finds an error and fails, the commit will fail and you will have to fix it before you can proceed.

Go ahead and commit your work to git now (recall from last time it is a two step process, you have to stage your changes them commit them). You should see the lint-staged tasks run at the start of the commit.

There is a catch. If there are linter errors, we can no longer successfully commit. While this is generally a good thing, sometimes we want to commit partial work to share with a teammate (or the instructor) to get help. In that context we want to commit despite the errors. To do so, add the --no-verify option to your commit command to bypass the checks (i.e., 💻 git commit --no-verify -m "My pithy commit messages").

Finishing up

Commit any changes you may have made since the commit at the end of the last section and then push your changes to GitHub.

You should then submit your repository to Gradescope.

Grading

  • Create a npm package
  • Implement validSong
  • Implement tests with 100% coverage
  • Pass all tests
  • Pass all ESLint checks

Recall that the Practical exercises are evaluated as “Satisfactory/Not yet satisfactory”. Your submission will need to implement all of the required functionality (i.e., pass all the tests) to be Satisfactory (2 points).

© Laura Biester, Michael Linderman, and Christopher Andrews 2019-2024.