Light Dark System

Contributing

Pure UI is an open source project, meaning everyone can use it and contribute to its development. When you join our community, you’ll find a friendly group of enthusiasts at all experience levels who are willing to chat about anything and everything related to Pure UI.

The easiest way to get started contributing is to join the community chat. This is where we hang out, discuss new ideas, ask for feedback, and more!

A common misconception about contributing to an open source project is that you need to know how to code. This simply isn’t true. In fact, there are many ways to contribute, and some of the most important contributions come from those who never write a single line of code. Here’s a list of ways you can make a meaningful contribution to the project:

  • Submitting well-written bug reports
  • Submitting feature requests that are within the scope of the project
  • Improving the documentation
  • Responding to users that need help in the community chat or discussion forum
  • Triaging issues on GitHub
  • Being a developer advocate for the project
  • Sponsoring the project financially
  • Writing tests
  • Sharing ideas
  • And, of course, contributing code!

Please take a moment to review these guidelines to make the contribution process as easy as possible for both yourself and the project’s maintainers.

AI-generated Code

As an open source maintainer, I respectfully ask that you refrain from using AI-generated code when contributing to this project. This includes code generated by tools such as GitHub Copilot, even if you make alterations to it afterwards. While some of Copilot’s features are indeed convenient, the ethics surrounding which codebases the AI has been trained on and their corresponding software licenses remain very questionable and have yet to be tested in a legal context.

I realize that one cannot reasonably enforce this any more than one can enforce not copying licensed code from other codebases, nor do I wish to expend energy policing contributors. I would, however, like to avoid all ethical and legal challenges that result from using AI-generated code. As such, I respectfully ask that you refrain from using such tools when contributing to this project. At this time, I will not knowingly accept any code that has been generated in such a manner.

Using the Issue Tracker

The issue tracker is for bug reports, feature requests, and pull requests.

  • Please do not use the issue tracker for personal support requests. Use the discussion forum instead.
  • Please do not use the issue tracker for feature requests. Use the discussion forum instead.
  • Please do not derail, hijack, or troll issues. Keep the discussion on topic and be respectful of others.
  • Please do not post comments with ”+1″ or ”👍”. Use reactions instead.
  • Please do use the issue tracker for feature requests, bug reports, and pull requests.

Issues that do not follow these guidelines are subject to closure. There simply aren’t enough resources for the author and contributors to troubleshoot personal support requests.

Feature Requests

Feature requests can be added using the discussion forum.

  • Please do search for an existing request before suggesting a new feature.
  • Please do use the voting buttons to vote for a feature.
  • Please do share substantial use cases and perspective that support new features if they haven’t already been mentioned.
  • Please do not bump, spam, or ping contributors to prioritize your own feature.

Bug Reports

A bug is a demonstrable problem caused by code in the library. Bug reports are an important contribution to the quality of the project. When submitting a bug report, there are a few steps you can take to make sure your issues gets attention quickly.

  • Please do not paste in large blocks of irrelevant code
  • Please do search for an existing issue before creating a new one
  • Please do explain the bug clearly
  • Please do provide a minimal test case that demonstrates the bug (e.g. jsfiddle.net or CodePen)
  • Please do provide additional information, when necessary, to replicate the bug

A minimal test case is critical to a successful bug report. It demonstrates that the bug exists in the library and not in surrounding code. Contributors should be able to understand the bug without studying your code, otherwise they’ll probably move on to another bug.

Pull Requests

To keep the project on track, please consider the following guidelines before submitting a PR.

  • Please do not submit a PR without opening an issue first, unless the changes are trivial (e.g. fixing typos or outdated docs). This may prevent you from doing work that won’t be accepted for various reasons (e.g. someone is already working on it, it’s not a good fit for the project’s roadmap, it needs additional planning, etc.)
  • Please do make sure your PR clearly defines what you’re changing. Even if you feel your changes are obvious, please explain them so other contributors can more easily review your works. PRs without detailed descriptions are subject to closure pending more details.
  • Please do open your PR against the next branch.
  • Please do not edit anything in dist/. These files are generated automatically, so you need to edit the source files instead.

The author reserves the right to reject any PR that’s outside the scope of the project or doesn’t meet code quality standards.

Branches

current - This branch reflects the latest release and powers pureui.xyz.

next - This is the branch you should submit pull requests against. It reflects what’s coming in the next release, which can be previewed at next.pureui.xyz.

Developing

To set up a local dev environment, fork the repo on GitHub, clone it locally, and install its dependencies.

git clone https://github.com/YOUR_GITHUB_USERNAME/pure-ui
cd pure-ui
npm install

Once you’ve cloned the repo, run the following command to spin up the dev server.

npm start

After the initial build, a browser will open automatically to a local version of the docs. The documentation is powered by Eleventy and a number of custom plugins.

Cloud-based Development

Alternatively, you can use Gitpod to setup a dev environment in the cloud using only your browser.

Open in Gitpod

Creating New Components

To scaffold a new component, run the following command, replacing p-tag-name with the desired tag name.

npm run create p-tag-name

This will generate a source file, a stylesheet, and a docs page for you. When you start the dev server, you’ll find the new component in the “Components” section of the sidebar.

Dev Sandbox

Component development occurs within the local docs site. I’ve found that offering common variations in the docs is more beneficial for users than segmenting demos and code examples into separate tools such as Storybook. This encourages more thorough documentation, streamlines development for maintainers, and simplifies how the project is built. It also reduces installation and startup times significantly.

There is currently no hot module reloading (HMR), as browsers don’t provide a way to unregister custom elements. However, most changes to the source will reload the browser automatically.

For more information about running and building the project locally, refer to README.md in the project’s root.

Testing

Pure UI uses Web Test Runner for testing. To launch the test runner during development, open a terminal and launch the dev server.

npm start

In a second terminal window, launch the test runner.

npm run test:watch

Follow the on-screen instructions to work with the test runner. Tests will automatically re-run as you make changes.

To run all tests only once:

npm run test

To test a single component, use the component’s basename as shown in the following example.

npm run test:component breadcrumb-item

Documentation

Maintaining good documentation can be a painstaking task, but poor documentation leads to frustration and makes the project less appealing to users. Fortunately, writing documentation for Pure UI is fast and easy!

Most of Pure UI’s technical documentation is generated with JSDoc comments and TypeScript metadata from the source code. Every property, method, event, etc. is documented this way. In-code comments encourage contributors to keep the documentation up to date as changes occur so the docs are less likely to become stale. Refer to an existing component to see how JSDoc comments are used in Pure UI.

Instructions, code examples, and interactive demos are hand-curated to give users the best possible experience. Typically, the most relevant information is shown first and less common examples are shown towards the bottom. Edge cases and gotchas should be called out in context with tips or warnings.

The docs are powered by Eleventy. Check out docs/components/*.md to get an idea of how pages are structured and formatted. If you’re creating a new component, it may help to use an existing component’s markdown file as a template.

If you need help with documentation, feel free to reach out on the community chat.

Pure UI-flavoured Markdown

The Pure UI documentation uses an extended version of markdown-it. Generally speaking, it follows the Commonmark spec while sprinkling in some additional features.

Code Previews

To render a code preview, use the standard code field syntax and append :preview to the language.

```html:preview
[code goes here]
```

You can also append :expanded to expand the code by default, and :no-codepen to disable the CodePen button. The order of these modifiers doesn’t matter, but no spaces should exist between the language and the modifiers.

```html:preview:expanded:no-codepen
[code goes here]
```

This particular syntax was chosen for a few reasons:

  1. It’s easy to remember
  2. It works out of the box with markdown-it
  3. It appears to have the best support across editors and previewers (the language is usually highlighted correctly)

Callouts

Special callouts can be added using the following syntax.

:::tip
This is a tip/informational callout
:::

:::warning
This is a warning callout
:::

:::danger
This is a danger callout
:::

Asides

To place content that’s indirectly related, use the following syntax.

:::aside
This content is indirectly related and will appear in an `<aside>` element.
:::

Details

To provide additional details that can be expanded/collapses, use the following syntax.

:::details Title Here
The details here are expandable.
:::

GitHub Issues

To link to a GitHub issue, PR, or discussion, use the following syntax.

[#1234]

Best Practices

The following is a non-exhaustive list of conventions, patterns, and best practices we try to follow. As a contributor, we ask that you make a good faith effort to follow them as well. This ensures consistency and maintainability throughout the project.

If in doubt, use your best judgment and the maintainers will be happy to guide you during the code review process. If you’d like clarification on something before submitting a PR, feel free to reach out on the community chat.

Accessibility

Pure UI is built with accessibility in mind. Creating generic components that are fully accessible to users with varying capabilities across a multitude of circumstances is a daunting challenge. Oftentimes, the solution to an a11y problem is not written in black and white and, therefore, we may not get it right the first time around. There are, however, guidelines we can follow in our effort to make Pure UI an accessible foundation from which applications and websites can be built.

We take this commitment seriously, so please ensure your contributions have this goal in mind. If you need help with anything a11y-related, please reach out to the community for assistance. If you discover an accessibility concern within the library, please file a bug on the issue tracker.

It’s important to remember that, although accessibility starts with foundational components, it doesn’t end with them. It everyone’s responsibility to encourage best practices and ensure we’re providing an optimal experience for all of our users.

Code Formatting

Most code formatting is handled automatically by Prettier via commit hooks. However, for the best experience, you should install it in your editor and enable format on save.

Please do not make any changes to prettier.config.cjs without consulting the maintainers.

Composability

Components should be composable, meaning you can easily reuse them with and within other components. This reduces the overall size of the library, expedites feature development, and maintains a consistent user experience.

Component Structure

All components have a host element, which is a reference to the <p-*> element itself. Make sure to always set the host element’s display property to the appropriate value depending on your needs, as the default is inline per the custom element spec.

:host {
  display: block;
}

Aside from display, avoid setting styles on the host element when possible. The reason for this is that styles applied to the host element are not encapsulated. Instead, create a base element that wraps the component’s internals and style that instead. This convention also makes it easier to use BEM in components, as the base element can serve as the “block” entity.

When authoring components, please try to follow the same structure and conventions found in other components. Classes, for example, generally follow this structure:

  • Static properties/methods
  • Private/public properties (that are not reactive)
  • @query decorators
  • @state decorators
  • @property decorators
  • Lifecycle methods (connectedCallback(), disconnectedCallback(), firstUpdated(), etc.)
  • Private methods
  • @watch decorators
  • Public methods
  • The render() method

Please avoid using the public keyword for class fields. It’s simply too verbose when combined with decorators, property names, and arguments. However, please do add private in front of any property or method that is intended to be private.

Class Names

All components use a shadow DOM, so styles are completely encapsulated from the rest of the document. As a result, class names used inside a component won’t conflict with class names outside the component, so we’re free to name them anything we want.

Internally, each component uses the BEM methodology for class names. There is no technical requirement to do this — it’s purely the preference of the author to enforce consistency and clarity throughout components. As such, all contributions are expected to follow this pattern.

Boolean Props

Boolean props should always default to false, otherwise there’s no way for the user to unset them using only attributes. To keep the API as friendly and consistent as possible, use a property such as noHeader and a corresponding kebab-case attribute such as no-header.

When naming boolean props that hide or disable things, prefix them with no-, e.g. no-spin-buttons and avoid using other verbs such as hide- and disable- for consistency.

Conditional Slots

When a component relies on the presence of slotted content to do something, don’t assume its initial state is permanent. Slotted content can be added or removed any time and components must be aware of this. A good practice to manage this is:

  • Add @slotchange={this.handleSlotChange} to the slots you want to watch
  • Add a handleSlotChange method and use the hasSlot utility to update state variables for the the respective slot(s)
  • Never conditionally render <slot> elements in a component — always use hidden so the slot remains in the DOM and the slotchange event can be captured

See the source of card, dialog, or drawer for examples.

Dynamic Slot Names and Expand/Collapse Icons

A pattern has been established in <p-details> and <p-tree-item> for expand/collapse icons that animate on open/close. In short, create two slots called expand-icon and collapse-icon and render them both in the DOM, using CSS to show/hide only one based on the current open state. Avoid conditionally rendering them. Also avoid using dynamic slot names, such as <slot name=${open ? 'open' : 'closed'}>, because Firefox will not animate them.

There should be a container element immediately surrounding both slots. The container should be animated with CSS by default and it should have a part so the user can override the animation or disable it. Please refer to the source and documentation for <p-details> and/or <p-tree-item> for details.

Fallback Content in Slots

When providing fallback content inside of <slot> elements, avoid adding parts, e.g.:

<slot name="icon">
  <p-icon part="close-icon"></p-icon>
</slot>

This creates confusion because the part will be documented, but it won’t work when the user slots in their own content. The recommended way to customize this example is for the user to slot in their own content and target its styles with CSS as needed.

Custom Events

Components must only emit custom events, and all custom events must start with p- as a namespace. For compatibility with frameworks that utilize DOM templates, custom events must have lowercase, kebab-style names. For example, use p-change instead of slChange.

This convention avoids the problem of browsers lowercasing attributes, causing some frameworks to be unable to listen to them. This problem isn’t specific to one framework, but Vue’s documentation provides a good explanation of the problem.

Change Events

When change events are emitted by Pure UI components, they should be named p-change and they should only be emitted as a result of user input. Programmatic changes, such as setting el.value = '…' should not result in a change event being emitted. This is consistent with how native form controls work.

CSS Custom Properties

To expose custom properties as part of a component’s API, scope them to the :host block.

:host {
  --color: var(--p-color-primary-500);
  --background-color: var(--p-color-neutral-100);
}

Then use the following syntax for comments so they appear in the generated docs. Do not use the --p- prefix, as that is reserved for design tokens that live in the global scope.

/**
 * @cssproperty --color: The component's text color.
 * @cssproperty --background-color: The component's background color.
 */
export default class SlExample {
  // ...
}

Focusing on Disabled Items

When an item within a keyboard navigable set is disabled (e.g. tabs, trees, menu items, etc.), the disabled item should not receive focus via keyboard, click, or tap. It should be skipped just like in operating system menus and in native HTML form controls. There is no exception to this. If a particular item requires focus for assistive devices to provide a good user experience, the item should not be disabled and, upon activation, it should inform the user why the respective action cannot be completed.

When to use a property vs. a CSS custom property

When designing a component’s API, standard properties are generally used to change the behavior of a component, whereas CSS custom properties (“CSS variables”) are used to change the appearance of a component. Remember that properties can’t respond to media queries, but CSS variables can.

There are some exceptions to this (e.g. when it significantly improves developer experience), but a good rule of thumbs is “will this need to change based on screen size?” If so, you probably want to use a CSS variable.

When to use a CSS custom property vs. a CSS part

There are two ways to enable customizations for components. One way is with CSS custom properties (“CSS variables”), the other is with CSS parts (“parts”).

CSS variables are scoped to the host element and can be reused throughout the component. A good example of a CSS variable would be --border-width, which might get reused throughout a component to ensure borders share the same width for all internal elements.

Parts let you target a specific element inside the component’s shadow DOM but, by design, you can’t target a part’s children or siblings. You can only customize the part itself. Use a part when you need to allow a single element inside the component to accept styles.

This convention can be relaxed when the developer experience is greatly improved by not following these suggestions.

Naming CSS Parts

While CSS parts can be named virtually anything, within Pure UI they must use the kebab-case convention and lowercase letters. Additionally, a BEM-inspired naming convention is used to distinguish parts, subparts, and states.

When composing elements, use part to export the host element and exportparts to export its parts.

render() {
  return html`
    <div part="base">
      <p-icon part="icon" exportparts="base:icon__base" ...></p-icon>
    </div>
  `;
}

This results in a consistent, easy to understand structure for parts. In this example, the icon part will target the host element and the icon__base part will target the icon’s base part.

Dependencies

TL;DR – a component is a dependency if and only if it’s rendered inside another component’s shadow root.

Many Pure UI components use other Pure UI components internally. For example, <p-button> uses both <p-icon> and <p-spinner> for its caret icon and loading state, respectively. Since these components appear in the button’s shadow root, they are considered dependencies of Button. Since dependencies are automatically loaded, users only need to import the button and everything will work as expected.

Contrast this to <p-select> and <p-option>. At first, one might assume that Option is a dependency of Select. After all, you can’t really use Select without slotting in at least one Option. However, Option is not a dependency of Select! The reason is because no Option is rendered in the Select’s shadow root. Since the options are provided by the user, it’s up to them to import both components independently.

People often suggest that Pure UI should auto-load Select + Option, Menu + Menu Item, Breadcrumb + Breadcrumb Item, etc. Although some components are designed to work together, they’re technically not dependencies so eagerly loading them may not be desirable. What if someone wants to roll their own component with a superset of features? They wouldn’t be able to if Pure UI automatically imported it!

Similarly, in the case of <p-radio-group> there was originally only <p-radio>, but now you can use either <p-radio> or <p-radio-button> as child elements. Which component(s) should be auto-loaded dependencies in this case? Had Radio been a dependency of Radio Group, users that only wanted Radio Buttons would be forced to register both with no way to opt out and no way to provide their own customized version.

For non-dependencies, the user should decide what gets registered, even if it comes with a minor inconvenience.

Form Controls

Form controls should support submission and validation through the following conventions:

  • All form controls must use name, value, and disabled properties in the same manner as HTMLInputElement
  • All form controls must have a setCustomValidity() method so the user can set a custom validation message
  • All form controls must have a reportValidity() method that report their validity during form submission
  • All form controls must have an invalid property that reflects their validity
  • All form controls should mirror their native validation attributes such as required, pattern, minlength, maxlength, etc. when possible
  • All form controls must be tested to work with the standard <form> element

System Icons

Avoid inlining SVG icons inside of templates. If a component requires an icon, make sure <p-icon> is a dependency of the component and use the system library:

<p-icon library="system" name="..."></p-icon>

This will render the icons instantly whereas the default library will fetch them from a remote source. If an icon isn’t available in the system library, you will need to add it to library.system.ts. Using the system library ensures that all icons load instantly and are customizable by users who wish to provide a custom resolver for the system library.

Writing tests

What to test for a given component:

  • Start with a simple test that checks that the default version of the component still renders.
  • Add at least one accessibility test (The accessibility check only covers the parts of the DOM which are currently visible and rendered. Depending on the component, more than one accessibility test is required to cover all scenarios.):
const myComponent = await fixture<SlAlert>(html`<p-my-component>SomeContent</p-my-component>`);

await expect(myComponent).to.be.accessible();
  • Try to cover all features advertised in the component’s description

Guidelines for writing tests:

  • Each test should declare its own, hand crafted hml fixture for the component. Do not try to write one big component to match all tests. This helps keeping each test understandable in isolation.
  • Tests should not produce log lines. Note that sometimes this cannot be prevented as the test runner might log errors (e.g. 404s).
  • Try keeping the main test readable: Extract more complicated sets of selectors/commands/assertions into separate functions.
  • Try to aim testing the user facing features of the component instead of the internal workings of the component.
  • Group multiple tests for one feature into describe blocks.