| permalink | eleventyExcludeFromCollections |
|---|---|
false |
true |
Content for the 18F website is written using Markdown, and pages are styled based on the layouts which are stored in templates/layouts. Individual posts and pages are customized using front matter, which is the set of key-value pairs you see at the top of many pages and posts. Front matter is written in YAML format and sets up some configuration for the page.
This document is a work in progress. If you don't see the information you're looking for, please open a new issue.
For debugging, prepend your command with DEBUG=Eleventy:{scope}, where scope is the subset of errors you want to see in the log. To see everything, use DEBUG=Eleventy:*, and the log is tagged with all the error types.
We've found the following command to be useful. It writes any exception logs to a file called debug.log while running npm run dev.
$ DEBUG=Eleventy:EleventyErrorHandler npm run dev > debug.log 2>&1$ DEBUG=Eleventy:* npm run dev > debug.log 2>&1Every pull request will trigger a build on Cloud.gov pages. Additionally, we have a github workflow in place that performs a number of tests on every pull request:
- Automated accessbility test with
pa11y-ci - Code linting with
eslint - HTML validation with
html-validate - Internal link checking with
check-html-links
Additionally, we manually use prettier for code formatting.
We use pa11y-ci is used to scan for accessibility issues. The scan runs as part of
our CI setup (see the pull-request.yml workflow)
on every pull request, but it can also be run locally. To run locally, type:
npm run test:pa11y-ci
Note that running pa11y-ci inside the docker container may not always work.
In certain cases we may need pa11y-ci to ignore an element. For example, in the accessibility guide there are elements that violate a11y rules on purpose. We know those will fail and don't want to fix them because they are showing an example of a bad practice, and so we want pa11y-ci to ignore them. To do so we can the data attribute data-pa11y-ignore to the element that should be ignored.
Example:
<span style = "color:#58AA02" class="exampleFailure" data-pa11y-ignore>This text fails.</span>We have a command to locally run all the checks that happen in each build, so you can catch errors before you commit and push:
npm run precommit
This will run code linting, HTML validation, and link checking.
We use eslint for code linting:
npx eslint . --fix
html-validate will check for valid HTML. It is configured in .htmlvalidate.json.
check-html-links will test internal links on the site. The internal link check tests whether a target link file exists in the _site folder at the expected location. Because the current version of check-html-link does not return an error value when it finds broken links, the npm script for this check includes an additional grep search for a "✅" which would appear only if there are no broken links. With this (hopefully) temporary fix in place, github actions will report a failure if there are broken links.
If you'd like to run these locally you could run npm run test:links. Alternatively you could use npm run test:links-internal, which will run the test with colorized output if you find that helpful, but note that it will not return an accurate exit code.
If there is a link that is still to be deteremined as we are moving guides, you can use '/TODO/' as the URL. This will visually highlight the link as TODO, and the link will be ignored in the link test.
We use Prettier for code formatting. You can run prettier manually with
npx prettier . --write
Note that this will overwrite files in place. See npx prettier --help for more CLI options.
An easier way to use prettier is to integrate it into your IDE/editor. For example, integration exists for VS Code such that prettier runs on a file every time you save it.
You can also add prettier as a git commit hook, but you will need to set up the script yourself. For example, you can symlink this template file into .git/hooks/pre-commit
We want to avoid commiting the assetPaths.json file, but need to keep it out of the project .gitignore in order to allow eleventy to rebuild when it is changed. One way to resolve this issue is to add assetPaths.json to the git exclude list:
- Open up
.git/info/exclude - Add
assetPaths.jsonto that file
If that doesn't work, type in git update-index --assume-unchanged data/assetPaths.json into the terminal.
Any link in the contents of the guides (i.e. not part of a layout or page component), will be tested to determine if it's an external link and if the access to the linked resource is restricted to 18F (e.g google docs, murals, etc...). We are using the patterns developed as part of work on the UX guide.
An external link is defined as any link that is not a federal .gov or .mil website. However, as there does not seem to be a programmatic way to distinguish between a federal and non-federal .gov domain, state and local-domains need to be marked as such manually. In order to mark a link as external we can add the USWDS usa-link and usa-link--external classes. To do so in markdown we can utilize the installed markdown-it-attrs plugin and append the class to the link using curly brackets ({ }). For example: [external link](example.com){.usa-link .usa-link--external}.
Private or restricted links are determined by comparing against the list of links in config/privateLinksList.js. If there are other links that are restricted you should add them to this list.
The content for all of the pages and posts are in the content/pages and content/posts.
To add a blog post: Create new file in content/posts with the file format YYYY-MM-DD-title-slug.md. Copy the latest blog post's frontmatter to get started, then edit for your specific post.
All the data that's accessible sitewide is kept in data/.
Anything we display or iterate over that isn't content goes here. This includes things like all our author data, agencies whose logos we display on the homepage and on "Work With Us", etc.
De-duplicate data wherever possible. In the previous site we had "agencies" and "featured agencies" — now we have one agencies file, and featured agencies are marked with featured: true.
The data file names should be semantically meaningful. As a counterexample, "featured" has a key "case studies" which list agency names, not case studies. This introduces confusion: are these agencies or are they case studies? Ideally, the relationships should be clearer than this.
Eleventy uses a "plugins" system to manage site-specific configuration.
Because we have A LOT of configuration to customize our site, it's important that this configuration remain organized.
To add configuration:
-
Create a subfolder inside
config/and add an index.js file, for exampleconfig/inflectors/index.js -
If your configuration is short and sweet, like it is with
config/browsersync/index.js, you can keep the config all in theindex.jsfile. -
If you are defining multiple functions or classes, add a file per function or per class, then
requireand export them inindex.js. For example, you might haveconfig/inflectors/singularize.jsandconfig/inflectors/pluralize.js.
Then, in index.js, you'd write:
const singularize = require('./singularize')
const pluralize = require('./pluralize')
const inflectorsPlugin = (eleventyConfig) => {
// ^^^^^^^^^^ remember to change the plugin name here and at the bottom
eleventyConfig.addInflector('singularize', singularize)
eleventyConfig.addInflector('pluralize', pluralize)
}
module.exports = inflectorsPlugin;- Add the plugin to the main site-wide plugin in
config/index.js. For example:
// ... other config ...
const inflectors = require('./inflectors')
module.exports = function EighteenF(eleventyConfig) {
// ... other plugins ...
eleventyConfig.addPlugin(inflectors)
}Never add configuration directly to:
.eleventy.jsconfig/index.js- a file inside
config/— configuration should always be in a folder
11ty uses “collections” to create content groupings. We can create a distinct collection for each guide, which allows us to group relevant content together. Site pages can be added to a collection simply by adding a tag to the front matter with the appropriate guide name as the value. The tag name is used throughout the site to refer to each guide (for example to determine the guide’s title).
Examples:
- De-risking guide content would have the front matter
tags: derisking - UX guide pages would have
tags: ux-guide
The data/navigation.yaml file is used to define the primary navigation for each guide. The guide’s tag is used as a key which maps to its list of link names and urls.
Example:
agile:
- name: Home
url: /agile/
By default, the page's <title> tag will use the title set in the page's front matter.
You can also set a custom page title using seo_title in the front matter, to improve the experience for people skimming search results. Reasons to write a custom page title include:
- The
titleis more than 30-35 characters long - The
titleis too similar to titles on other guides. (Examples are "Introduction" or "Planning.")
By default, the title front matter will be rendered as an h1 element. There are two additional front matter options that control the markup for the title:
page_title_tag: When you need the title of the page to be something other than H1, use this. This takes the name of the tag only, likeh2ordiv— don't set the full tag like<h3>.hidden_guide_title: If added, this will take the value ofhidden_guide_titleand render a screen reader onlyh1element before thepage_title_tag. This option is meant to be used together with thepage_title_tag.
Example usage:
title: Questions to ask
page_title_tag: h2
hidden_guide_title: State Software Budgeting Handbook
Which will render the following html:
<h1 class="usa-sr-only">State Software Budgeting Handbook</h1>
<h2 class="page-title derisking">Questions to ask</h2>
Eleventy does not use {{site.baseurl}} to refer to other pages. When linking to another page on the site, use Eleventy's url filter as such:
- For the home
index.mdpage, use{{ '[Markdown filename]' | url }}. - For any other page in
content/[guide], use{{ '../[Markdown file name]/' | url }}(remember the trailing slash!) - For pages in their own section within each guide, use
{{ '../../[Markdown file name]/' | url }}(remember the trailing slash!). An example is in the Engineering Hiring guide, where there are several pages incontent/eng-hiring/interviews/. Any page within theinterviewsfolder needs to use../../to link to other pages incontent/eng-hiring/
Many guides used Jekyll's redirect_from and redirect_to frontmatter keys to redirect old pages to current ones. In order to preserve past redirects from old guides, we ported over these frontmatter keys and implemented an 11ty version of this functionality, found in /content/redirect.html.
Note that paths listed as values for these keys should NOT contain the beginning forward slash (like is needed for permalinks):
redirect_from:
- content-guide/foobar/
redirect_from:
- /content-guide/foobar/
TK
TK
This project uses Github's Dependabot to keep the NPM dependencies up to date.
Dependabot takes care of noticing version updates and proposing the updates to
package.json and package-lock.json as pull requests (PRs). A human
developer needs to review these Dependabot PRs and merge them into the main
branch.
If Dependabot PRs go un-reviewed for too long, they can have merge conflicts and complex interactions with other code changes. The best practice is to review and merge Dependabot PRs as soon as possible. The automated test suite for this project is quite good, so this is fairly straightforward:
-
Comment
@dependabot recreateon the PR to request Dependabot to update the changes to the current state ofmain. -
After Dependabot updates the PR in accordance with the re-create request, review the changes that are included in the PR. There should be changes only to the
package.jsonandpackage-lock.jsonfiles and ideally they would be small changes, obviously connected to version updates. After reviewing the changes, you should approve the PR.Semantic versioning practices suggest that for dependencies where the "major" version changes (the first number, e.g. 3.x.y), breaking changes might be present and additional testing might be warranted. The reviewer could pull the Git branch and test the site build and function locally.
-
When the automated tests have completed successfully and the PR is reviewed and approved, go ahead and merge the PR.
This information was relevant during the replatforming effort to merge all 18F guides into this repo, but may not continue to be relevant after replatforming.