A collaborative newsroom documentation site, powered by Google Docs.
Table of Contents
A working (read only) demo of Library is available at https://nyt-library-demo.herokuapp.com. The site contains instructions for how to get the most out of Library.
-
Clone and
cdinto the repo:git clone git@github.com:nytimes/library.git && cd library -
From the Google API console, create or select a project, then create a service account with the Cloud Datastore User role. It should have API access to Drive and Cloud Datastore. Store these credentials in
server/.auth.json.- To use oAuth, you will also need to create oAuth credentials.
- To use the Cloud Datastore API for reading history, you will need to add in your
GCP_PROJECT_ID.
-
Install dependencies:
npm install --no-optional -
Create a
.envfile at the project root. An example.envmight look like
NODE_ENV=development # node environment
# Google oAuth credentials
GOOGLE_CLIENT_ID=123456-abcdefg.apps.googleusercontent.com
GOOGLE_CLIENT_SECRET=abcxyz12345
GCP_PROJECT_ID=library-demo-1234
APPROVED_DOMAINS="nytimes.com,dailypennsylvanian.com" # comma separated list of approved access domains.
SESSION_SECRET=supersecretvalue
# Google drive Configuration
DRIVE_TYPE=team # or folder, if using a folder instead of a team drive
DRIVE_ID=0123456ABCDEF # the ID of your team's drive or shared folder. The string of random numbers and letters at the end of your team drive or folder url.Make sure to remove all comments after the DRIVE_TYPE and DRIVE_ID vars.
-
Start the app:
npm run build && npm run watch
The app should now be running at localhost:3000. Note that Library requires Node v8 or higher.
You can run functional and unit tests, which test HTML parsing and routing logic, with npm test. A coverage report can be generated by running npm run test:cover.
The HTML parsing tests are based on the Supported Formats doc. To download a fresh copy of the HTML after making edits, run node test/utils/updateSupportedFormats.js.
Styles, text, caching logic, and middleware can be customized to match the branding of your organization. This is covered in the customization readme.
A sample customization repo is provided at nytimes/library-customization-example.
Wherever you deploy Library, you'll likely want to set up a Google service account and OAuth 2.0 client Set up your service account with API access to Drive and Cloud Datastore.
If you wish to deploy Library with customizations, create a git repo with the files you would like to include. Set the CUSTOMIZATION_GIT_REPO environment variable to the cloning URL. Files in the repo and packages specified in the package.json will be included in your library installation.
This button can quickly deploy to Heroku:
Set your app's GOOGLE_APPLICATION_JSON, GOOGLE_CLIENT_ID, and GOOGLE_CLIENT_SECRET with values from the service account and Oauth client. Add <your-heroku-app-url>.com as an authorized domain in the general OAuth consent screen setup and then add http://<your-heroku-app-url>.com/auth/redirect as the callback url in the OAuth credential setup itself.
You can also deploy Library to GAE, using the included app.yaml. Note that you will need to enable billing on your GCP project in order to use Google App Engine. More detailed instructions are provided on the demo site.
Library can be used as a base image for deployment using Docker. This allows you
to automate building and deploying a custom version of Library during Docker's
build phase. If you create a repo with the contents of your custom folder, you
could deploy library from that repo with a Dockerfile like the following:
FROM nytimes/library
# copy custom files to library's custom repo
COPY . ./custom/
# move to a temporary folder install custom npm packages
WORKDIR /usr/src/tmp
COPY package*.json .npmrc ./
RUN npm i
# copy node modules required by custom node modules
RUN yes | cp -rf ./node_modules/* /usr/src/app/node_modules
# return to app directory and build
WORKDIR /usr/src/app
RUN npm run build
# start app
CMD [ "npm", "start" ]Library is a standard node app, so it can be deployed just about anywhere. If you are looking to deploy to a standard VPS, Digital Ocean's tutorials are a great resource.
The main entry point to the app is index.js.
This file contains the express server which will respond to requests for docs in the configured team drive or shared folder. Additionally, it contains logic about issuing 404s and selecting the template to use based on the path.
Views (layouts) are located in the layouts folder. They use the .ejs
extension, which uses a syntax similar to underscore templates.
Base styles for the views are in the styles directory containing Sass files.
These files are compiled to CSS and placed in public/css.
Doc HTML fetch and parsing is handled by docs.js. fetchDoc takes the ID of a
Google doc and a callback, then passes the HTML of the document into the
callback once it has been downloaded and processed.
Traversing the contents of the NYT Docs folder is handled by list.js. There
are two exported functions:
-
getTreeis an async call that returns a nested hash (tree) of Google Drive Folder IDs mapped to their children. It is used by the server to determine whether a route is valid or not. -
getMetasynchronously returns a hash of Google Doc IDs to metadata objects that were saved in the course of populating the tree. This metadata includes edit history, document authors, and parent folders.
The tree and file metadata are repopulated into memory on an interval (currently 60s). Calling getTree multiple times will not return fresher data.
Authentication with the Google Drive v3 api is handled by the auth.js file, which exposes a single method getAuth. getAuth will either return an already instantiated authentication client or produce a fresh one. Calling getAuth multiple times will not produce a new authentication client if the credentials have expired; we should build this into the auth.js file later to automatically refresh the credentials on some sort of interval to prevent them from expiring.