This repo allows you to generate a blog from GitHub issues on a repo. It powers the OneGraph Product Updates blog, Stepan Parunashvili's blog, bdougie.live, and more.
All of the posts are stored as issues on the repo (e.g. OneGraph/onegraph-changelog).
When you visit the page at onegraph.com/changelog, a GraphQL query fetches the issues from GitHub via OneGraph's persisted queries and renders them as blog posts.
The persisted queries are stored with authentication credentials for GitHub that allows them to make authenticated requests. Persisting the queries locks them down so that they can't be made to send arbitrary requests to GitHub.
You can learn more about persisted queries in the docs.
Use an existing OneGraph app or sign up sign up at OneGraph to create a new app.
Copy /.env.example to /.env and set the following environment variables.
| Environment Variable | Description |
|---|---|
GITHUB_TOKEN |
A GitHub token, which you can get from https://github.com/settings/tokens. The token only needs the "public_repo" scope. |
NEXT_PUBLIC_TITLE |
The title of your site |
NEXT_PUBLIC_DESCRIPTION |
A short description of your site. |
NEXT_PUBLIC_GITHUB_REPO_OWNER |
The owner of the repo that we should pull issues from (e.g. linus in linus/oneblog). If you're using the Vercel deploy button, you don't need to provide this. |
NEXT_PUBLIC_GITHUB_REPO_NAME |
The name of the repo that we should pull issues from (e.g. oneblog in linus/oneblog). If you're using the Vercel deploy button, you don't need to provide this. |
Remove the generated files (they're tied to the OneGraph app they were generated with):
yarn relay:clean
# which runs rm -r src/__generated__
(Note: any time you change the variables in .env, it's a good idea to stop the relay compiler, remove the files in src/__generated__, and restart the compiler)
Install dependencies
yarn install
This project uses Relay as its GraphQL client because of its high-quality compiler and great support for persisted queries.
In another terminal window, start the relay compiler
yarn relay --watch
You may need to install watchman, a file watching service. On mac, do brew install watchman. On Windows or Linux, follow the instructions at https://facebook.github.io/watchman/docs/install.html.
Now that we've generated the relay files, we can start the server.
yarn start
The project will load at http://localhost:3000.
The project comes with setups for deploying to Google's Firebase, Zeit's Now, Netlify, and Fly.io.
Use the deploy button to set up a new repo:
If you've already set up the repo, just run the vercel command.
# If not installed
# npm i -g vercel
vercel
If you see an error when you visit the site, make sure the site's origin is listed in the CORS origins for your app on the OneGraph dashboard.
Please open an issue if you'd like help deploying with Firebase.
Please open an issue if you'd like help deploying with Netlify.
Please open an issue if you'd like help deploying with Fly.io
The client is an ordinary React app. The best to place to start is /src/App.js.
It uses Grommet as the UI library. Visit https://v2.grommet.io/components to view the documentation for Grommet.
It uses Relay as the GraphQL client. https://relay.dev/docs/en/graphql-in-relay has a good introduction to Relay.
To refresh the GraphQL schema, run yarn fetch-schema. That will fetch the schema from OneGraph and add some client-only directives that we use when we persist the queries to OneGraph.
The persistFunction for the relay compiler is set to /scripts/persistQuery.js. Every time a GraphQL query in the project changes, the relay compiler will call that function with the new query.
That function will parse the query and pull out the @persistedQueryConfiguration directive to determine if any auth should be stored alongside the query. In the changelog, the queries for fetching posts use persisted auth, but the mutations for adding reactions require the user to log in with OneGraph and use their auth.
The @persistedQueryConfiguration directive is stripped from the query and a next api route is generated to execute the query.
The server uses Next.js to allow us to render the content on the server. This helps with SEO and allows people to view the blog with Javascript turned off.
When a request comes in to the server, the server creates a mock Relay environment and prefetches the query for the route using fetchQuery from relay-runtime. This populates the record source that Relay uses to render.
React renders the app to a string, which is sent to the client.
On the client, React rehydates the app. To prevent Relay from showing a loading state, we inject the serialized record source with getStaticProps. That data is stored in the environment before Relay makes its first query. The fetchPolicy opt is set to "store-and-network" so that it uses the data from the store instead of showing a loading state.