github
/
docs
зеркало из https://github.com/github/docs.git
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность
docs/contributing/search.md

8.2 KiB
Исходник Ответственный История

Search

Overview

This site's search functionality is powered by Algolia, a third-party service.

To see all existing search-related issues and pull requests, visit github.com/github/docs/labels/search.

search-screenshot

How to search

The site search is part of every version of docs.github.com. On any page, you can use the search box to search the documents we've indexed. You can also query our search endpoint directly at: https://docs.github.com/search?language=en&version=dotcom&query=jekyll This endpoint responds in JSON format, and fronts Algolia and Lunr. We recommend using this endpoint over directly integrating with Algolia or Lunr, as the endpoint will be more stable.

Production deploys

A GitHub Actions workflow triggered by pushes to the main branch syncs the search data to Algolia. This process generates structured data for all pages on the site, compares that data to what's currently on Algolia, then adds, updates, or removes indices based on the diff of the local and remote data, being careful not to create duplicate records and avoiding any unnecessary (and costly) indexing operations.

The Actions workflow progress can be viewed (by GitHub employees) in the Actions tab of the repo.

Because the workflow runs after a branch is merged to main, there is a slight delay for search data updates to appear on the site.

Manual sync from a checkout

It is also possible to manually sync the indices to Algolia from your local checkout of the repo, before your branch is merged to main.

Prerequisite: Make sure the environment variables ALGOLIA_APPLICATION_ID and ALGOLIA_API_KEY are set in your .env file. You can find these values on Algolia.

Build without sync (dry run)

To build all the indices without uploading them to Algolia's servers (this takes about an hour):

npm run sync-search-dry-run

To build indices for a specific language and/or version (this is much faster):

VERSION=<PLAN@RELEASE> LANGUAGE=<TWO-LETTER CODE> npm run sync-search-dry-run

You can set VERSION and LANGUAGE individually, too.

Substitute a currently supported version for <PLAN@RELEASE> and a currently supported two-letter language code for <TWO-LETTER-CODE>.

Build and sync

To build all the indices and sync them to Algolia (this also takes about an hour):

npm run sync-search

To build indices for a specific language and/or version and sync them to Algolia:

VERSION=<PLAN@RELEASE LANGUAGE=<TWO-LETTER CODE> npm run sync-search

You can set VERSION and LANGUAGE individually, too.

Substitute a currently supported version for <PLAN@RELEASE> and a currently supported two-letter language code for <TWO-LETTER-CODE>.

Label-triggered Actions workflow

Docs team members can use an Actions workflow on GHES release PRs by applying a label in this format:

sync-english-index-for-<PLAN@RELEASE>

This label will run a workflow on every push that builds and uploads ONLY the English index for the specified version. This means:

  • The GHES content will be searchable at the same time the release PR is shipped, with no delay.
  • The GHES content will be searchable on staging throughout content creation.
  • No manual steps (unless you want to do a dry run test).

Why do we need this? For our daily shipping needs, it's tolerable that search updates aren't available for up to an hour after the content goes live. But GHES releases are more time-sensitive, and writers have a greater need to preview search data on staging.

Files

Actions workflow files

Code files

Indices

There's a separate search index for each combination of product and language. Some examples:

Index Name Description
github-docs-dotcom-cn GitHub.com Chinese
github-docs-dotcom-en GitHub.com English
github-docs-dotcom-es GitHub.com Spanish
github-docs-dotcom-ja GitHub.com Japanese
github-docs-2.18-cn GitHub Enterprise 2.18 Chinese
github-docs-2.18-en GitHub Enterprise 2.18 English
github-docs-2.18-es GitHub Enterprise 2.18 Spanish
github-docs-2.18-ja GitHub Enterprise 2.18 Japanese
github-docs-2.17-cn GitHub Enterprise 2.17 Chinese
github-docs-2.17-en GitHub Enterprise 2.17 English
github-docs-2.17-es GitHub Enterprise 2.17 Spanish
github-docs-2.17-ja GitHub Enterprise 2.17 Japanese

Records

Each record represents a section of a page. Sections are derived by splitting up pages by their headings. Each record has a title, intro (if one exists in the frontmatter), body content (in text, not HTML), a url, and a unique objectID that is currently just the permalink of the article. Here's an example:

{
  objectID: '/en/actions/creating-actions/about-actions#about-actions',
  url: 'https://help.github.com/en/actions/creating-actions/about-actions#about-actions',
  slug: 'about-actions',
  breadcrumbs: 'GitHub Actions / Creating actions / About actions',
  heading: 'About actions',
  title: 'About actions',
  content: "You can create actions by writing custom code that interacts with your repository in any way you'd like..."
}

Notes

  • It's not strictly necessary to set an objectID as Algolia will create one automatically, but by creating our own we have a guarantee that subsequent invocations of this upload script will overwrite existing records instead of creating numerous duplicate records with differing IDs.
  • Algolia has typo tolerance. Try spelling something wrong and see what you get!
  • Algolia has lots of controls for customizing each index, so we can add weights to certain attributes and create rules like "title is more important than body", etc. But it works pretty well as-is without any configuration.
  • Algolia has support for "advanced query syntax" for exact matching of quoted expressions and exclusion of words preceded by a - sign. This is off by default but we have it enabled in our browser client. This and many other settings can be configured in Algolia.com web interface. The settings in the web interface can be overridden by the search endpoint. See middleware/search.js.