Fortune Firestore Adapter

This is a Firestore adapter for Fortune. It uses Google’s official Node.js client for Google Cloud Firestore.

It is ported from fortunejs/mongodb and incoporates some design ideas from the third-party Google Cloud Datastore adapter.

Usage

Install the fortune-firestore package with yarn or npm:

$ yarn add fortune-firestore

$ npm install fortune-firestore

Then use it with Fortune:

import fortune from "fortune";
import firestoreAdapter from "fortune-firestore";

const store = fortune({ ... }, {
  adapter: [
    firestoreAdapter,
    { // options
      projectId: "my firebase/google cloud project id",
      keyFilename: "path/to/some/keyFile.json",
      credentials: {
        client_email:
        "[email protected]",
        private_key: "some long private key"
      }
    }
  ]
});

Connecting to Firestore

For the adapter to connect to Firestore, as it’s implemented, one of the following three combinations of options must be present, as above:

1. keyFilename as a JSON file.
2. projectId, credentials.client_email, and credentials.private_key, as likely harvested from the keyFile.
3. keyFilename as .pem or .p12 file, projectId, and credentials.client_email

Getting a keyFile

Both Firebase and Google Cloud Platform use the GCP interface to manage service accounts, so you should log in to the GCP console.

1. Select the appropriate project for the adapter.
2. Navigate to “IAM & Admin” and then “Service accounts.”
3. Click on an appropriate service account or create a new one. If you create a new one, it should have the “Cloud Datastore User” role at a minimum.
4. Generate a key by clicking on “+ Create Key.” This will give you the option of downloading a JSON keyFile or a .p12 keyFile.

This keyFile, by itself, is sufficient to connect to Firestore. However, seeing as you should never commit this file, it may make more sense to harvest a few keys from it and use them as environment variables. The properties to harvest into your environment are project_id, private_key, and client_email.

Note that private_key is likely to have special characters that may break with some continuous integration solutions like Travis. My solution in spec/helpers.js is to encode the private key as a hex string, paste that blob into Travis, and then convert that back to ascii within the script.

Options

In addition to the credentials indicated above, the adapter anticipates four options:

• typeMap: An object that maps type names (keys) to collection names (values). For example: {user: "users"}. If unset, it assumes the type passed from Fortune has the same name as the collection in Firestore.
• bufferEncoding: Fortune provides for saving Buffer-type blobs in records. Firestore does not, without using their custom Blob type. Instead, the adapter converts all fields indicated as type: Buffer in the record types to strings. This option sets the encoding for that conversion. The default is "base64".
• convertTimestamps: Firestore receives Dates and converts them to Timestamps. By default, this option is set to true, meaning that when records are fetched, the adapter converts the Timestamps Firestore has saved to Dates.
• nullUndefinedFields: Firestore doesn’t know about the schema you have registered with Fortune. It will make a record with as many fields as are passed in the create() method. Fortune, however, expects undefined fields to be set as null or [] (if they are arrays). This parameter, which defaults to true can be deactivated, keeping the skinny records.

Contributing, etc.

Please do. I ported this by writing to the tests that Fortune provides, but I have not yet tested the adapter in the wild.

License

This software is licensed under the MIT License.