podman-desktop/CODE-GUIDELINES.md

Guidelines for Podman Desktop Code

Production code

Imports

Prefer using path aliases for all cross-directory imports to improve readability and maintainability.

Good:

import type { WSL2Check } from '/@/checks/windows/wsl2-check';
import { WinPlatform } from './win-platform';

Bad:

import type { WSL2Check } from '../checks/windows/wsl2-check';

Svelte

On templates of Svelte components, avoid using inline code and arrow functions. Instead, call a function defined in the script part of the component.

✅ Use this pattern:

<script lang="ts">
async function onButtonClicked(): Promise<void> {
  // the code here
}
</script>

<button on:click={onButtonClicked}>

🚫 Instead of:

<button on:click={(): Promise<void> => { /* the code here */ }}>

If values have to be passed from the template to the function, use the bind method on the function to pass the parameter.

✅ Use this pattern:

<script lang="ts">
async function onButtonClicked(object: Object): Promise<void> {
  // the code here
}
</script>

{#each objects as object (object.id)}
  <button on:click={onButtonClicked.bind(undefined, object)}>
{/each}

🚫 Instead of:

{#each objects as object (object.id)}
  <button on:click={(): Promise<void> => onButtonClicked(object)}>
{/each}

Unit tests code

Use vi.mocked, not a generic myFunctionMock

If you define a mock with const myFunctionMock = vi.fn(); its type is Mock<Procedure>, which is a generic type.

For example, do not write this, or Typescript won't be able to detect that you passed an object instead of a string to mockResolvedValue:

const windowMethodMock = vi.fn();

Object.defineProperty(global, 'window', {
  value: {
    windowMethod: windowMethodMock,
  },
});

test('...', () => {
  windowMethodMock.mockResolvedValue({ msg: 'a string' }); // here, Typescript is not able to detect that the type is wrong
});

Instead, you can write vi.mocked(window.windowMethod).mock..., and Typescript will check that you correctly pass a string to mockResolvedValue:

Object.defineProperty(global, 'window', {
  value: {
    windowMethod: vi.fn(),
  },
});

test('...', () => {
  vi.mocked(window.windowMethod).mockResolvedValue('a string');
});

Mock complete modules, spy on parts of module for specific tests

When testing a module, you have to decide for each imported module if you mock the entire module or if you spy on specific functions of the module for specific tests and keep the real implementation for the other functions.

System modules (node:fs, etc) are most generally mocked, so you are sure that unit tests are executed in isolation of the system. For internal modules, it's up to you to decide if you want to mock them or not, depending on the coverage you want for the unit tests.

Mock a complete module

Mock completely an imported module with vi.mock('/path/to/module), and define mock implementation for each test with vi.mocked(function).mock...().

Use vi.resetAllMocks() in the top-level beforeEach to reset all mocks to a no-op function returning undefined before to start each test.

import { existsSync } from 'node:fs';
import { beforeEach, describe, expect, test, vi } from 'vitest';

// completely mock the fs module, to be sure to
// run the tests in complete isolation from the filesystem
vi.mock('node:fs');

beforeEach(() => {
  vi.resetAllMocks();
});

describe('the file exists', () => {
  beforeEach(() => {
    vi.mocked(existsSync).mockReturnValue(true);
  });

  test('file exists', () => {
    // existsSync is mocked to return true
    expect(codeCheckingIfFileExists('/file/not/found')).toBeTruthy();
  });
});

describe('the file does not exist', () => {
  beforeEach(() => {
    vi.mocked(existsSync).mockReturnValue(false);
  });

  test('root does not exists', () => {
    // existsSync is mocked to return false
    expect(codeCheckingIfFileExists('/')).toBeFalsy();
  });
});

test('file existence is not defined', () => {
  // a no-op mock returning undefined is called
  expect(codeCheckingIfFileExists('/file/not/found')).toBeUndefined();
});

Spy on a function for a specific test

When you want to mock only one or a small number of functions of a module (for example a function of the module you are testing, or a function of an helper module from which you want to use real implementation for some functions) for a particular test, you can use vi.spyOn(module, 'function') to mock only function and keep the original implementation for the rest of the module.

To be sure that the spied function is restored to its original implementation for the other tests, use vi.restoreAllMocks() in the top-level beforeEach.

// helpers.ts
export function f1(): boolean {
  return true;
}

// mymodule.ts
import { f1 } from './helpers.js';

export class MyModuleToTest {
  f2(): boolean {
    return f1();
  }
}

// mymodule.spec.ts
import { beforeEach, describe, expect, test, vi } from 'vitest';
import { MyModuleToTest } from './mymodule.js';
import * as helpers from './helpers.js';

let myModuleToTest: MyModuleToTest;

beforeEach(() => {
  myModuleToTest = new MyModuleToTest();

  // restore f1 to its original implementation
  vi.restoreAllMocks();
});

describe('f1 returns false', () => {
  beforeEach(() => {
    vi.spyOn(helpers, 'f1').mockReturnValue(false);
  });

  test('f2 returns false', () => {
    expect(myModuleToTest.f2()).toBeFalsy();
    expect(helpers.f1).toHaveBeenCalledOnce();
  });
});

test('f2 returns true', () => {
  // use the original implementation of f1
  expect(myModuleToTest.f2()).toBeTruthy();
  // this won't work, as f1 is not spied for this test
  // expect(helpers.f1).toHaveBeenCalledOnce();
});

screen.getBy vs screen.queryBy

Calling element = screen.getBy... throws an error if no element is found. For this reason, it is not necessary to call expect(element).toBeInTheDocument(), as the assertion has already been done as part of screen.getBy....

It is necessary to use element = screen.queryBy... followed by expect(element).not.toBeInTheDocument() when checking if a component does NOT exist, as this call does not throw any error, but returns a null value if the element is not found.

Testing style attribute

When we need to ensure a given style is applied to an HTMLElement, we should be using tohavestyle

Examples

const { getByText } = render(<MyComponent>);

const text = getByText('text in the page');
// [Good]
expect(text).toHaveStyle({ color: '#FFFFF'});

waitFor vs waitUntil

Use waitFor (https://vitest.dev/api/vi.html#vi-waitfor) to retry an assertion until it passes, and waitUntil (https://vitest.dev/api/vi.html#vi-waituntil) to wait for a function to return a truthy value.

→ waitFor → needs an exception

→ waitUntil → needs a boolean

Example:

// Use waitFor with an assertion
await waitFor(() => expect(get(providerInfos)).not.toHaveLength(0));

// Use waitUntil with a boolean value
await vi.waitUntil(() => get(imagesInfos).length > 0);

Mocking a sub-component

To test a component in isolation without testing its sub-components, you have the possibility to mock the sub-components. For example:

Compo1.svelte

<script lang="ts">
import Compo2 from './Compo2.svelte';
</script>

<Compo2 />

Compo2.svelte

Some content

Compo1.spec.ts

import { render } from '@testing-library/svelte';
import { expect, test, vi } from 'vitest';

import Compo1 from './Compo1.svelte';
import Compo2 from './Compo2.svelte';

vi.mock(import('./Compo2.svelte'));

test('Compo1 calls Compo2', async () => {
  render(Compo1);
  expect(Compo2).toHaveBeenCalled();
});

When the sub-components have bindable properties

When a sub-component has a bindable property, you may want to test that some operations are performed on this property. But as the property value is returned by the sub-component, which is now mocked, you have to return such an object when you mock the component.

For this, you can mock the implementation of the sub-component constructor, and update the bound property passed as parameter with an object for which you can spy the methods. For example:

Compo1.svelte

<script lang="ts">
import { onMount } from 'svelte';
import type { Obj } from './compo2';
import Compo2 from './Compo2.svelte';

let myobj = $state<Obj>();

onMount(() => {
  myobj?.fct1('a name');
});
</script>

<Compo2 bind:obj={myobj} />

Compo2.svelte

<script lang="ts">
import type { Obj } from './compo2';

interface Props {
  obj?: Obj;
}

let { obj = $bindable() }: Props = $props();
</script>

compo2.ts

export interface Obj {
  fct1: (name: string) => void;
}

Compo1.spec.ts

import { render } from '@testing-library/svelte';
import { expect, test, vi } from 'vitest';

import Compo1 from './Compo1.svelte';
import type { Obj } from './compo2';
import Compo2 from './Compo2.svelte';

vi.mock(import('./Compo2.svelte'));

test('compo1 calls fct1 of obj on mount', async () => {
  // create a mock Obj, with methods you can spy
  const obj: Obj = {
    fct1: vi.fn(),
  };
  vi.mocked(Compo2).mockImplementation((_, props) => {
    props.obj = obj; // update the value of the prop with your mock Obj
    return {};
  });
  render(Compo1);
  expect(Compo2).toHaveBeenCalled();
  expect(obj.fct1).toHaveBeenCalledWith('a name'); // check the method has been called
});

Using Fake Timers with Svelte Components

When testing Svelte components in the packages/renderer package, always enable automatic time advancement by using:

vi.useFakeTimers({ shouldAdvanceTime: true });

Avoid calling vi.useFakeTimers() without options.

If shouldAdvanceTime is not enabled, fake timers will completely freeze time, which can lead to deadlocks when:

• Svelte’s internal async updates wait for the next event loop tick
• Testing Library’s async queries (findBy*, waitFor) continuously poll for elements

By setting shouldAdvanceTime: true, timers will automatically advance during pending async operations. This prevents hangs while still allowing manual time control with vi.advanceTimersByTime().

✅ Use this pattern:

beforeEach(() => {
  vi.useFakeTimers({ shouldAdvanceTime: true });
});

🚫 Instead of:

afterEach(() => {
  vi.useRealTimers();
});