Fireschema

• Strong type safety for Firestore - Automatically provide type information to nested documents without unsafe type assertions, from the simple schema. Also support data decoding.
• Security rules generation - Generate firestore.rules file including data type validation and access control from the schema.
• React Hooks - Get realtime updates with React Hooks.
• Type safety for Cloud Functions
  • Automatically provide type information to snapshot data on Firestore Trigger Function based on the path string.
  • Guard HTTPS callable function's request/response data type both on compile time and runtime.

• TypeScript (>= 4.4)

yarn add fireschema firebase firebase-admin firebase-functions zod
yarn add -D typescript ts-node

🎉 Since Fireschema v5, you no longer need to compile codes via custom transformer.

Schema Transformation

1. Define schema

The schema definition must be default exported.

import { Merge } from 'type-fest'
import { z } from 'zod'

import { DataModel, FirestoreModel, rules, timestampType } from 'fireschema'

export const UserType = z.object({
  name: z.string(),
  displayName: z.union([z.string(), z.null()]),
  age: z.number().int(),
  timestamp: timestampType(),
  options: z.object({ a: z.boolean() }).optional(),
})

type User = z.infer<typeof UserType>
/* => {
  name: string
  displayName: string | null
  age: number
  timestamp: FTypes.Timestamp
  options?: { a: boolean } | undefined
} */

type UserDecoded = Merge<User, { timestamp: Date }>

const UserModel = new DataModel({
  schema: UserType,
  decoder: (data: User): UserDecoded => ({
    ...data,
    timestamp: data.timestamp.toDate(),
  }),
})

const PostType = z.object({
  authorUid: z.string(),
  text: z.string(),
  tags: z.object({ id: z.number().int(), name: z.string() }).array(),
})

const PostModel = new DataModel({
  schema: PostType,
  selectors: (q) => ({
    byTag: (tag: string) => [
      q.where('tags', 'array-contains', tag),
      q.limit(20),
    ],
  }),
})

export const firestoreModel = new FirestoreModel({
  'function isAdmin()': `
    return exists(${rules.basePath}/admins/$(request.auth.uid));
  `,

  'function requestUserIs(uid)': `
    return request.auth.uid == uid;
  `,

  collectionGroups: {
    '/posts/{postId}': {
      allow: {
        read: true,
      },
    },
  },

  '/users/{uid}': {
    model: UserModel,
    allow: {
      read: true, // open access
      write: rules.or('requestUserIs(uid)', 'isAdmin()'),
    },

    '/posts/{postId}': {
      'function authorUidMatches()': `
        return request.resource.data.authorUid == uid;
      `,

      model: PostModel,
      allow: {
        read: true,
        write: rules.and('requestUserIs(uid)', 'authorUidMatches()'),
      },
    },
  },
})

export default firestoreModel

Write rules are combined with the rules automatically generated from zod schema.

2. Generate firestore.rules

yarn fireschema rules <path-to-schema>.ts

Environment variable TS_NODE_PROJECT is supported.

3. Read/write collections and documents

The Firestore interface of Fireschema supports both Web SDK and Admin SDK.

import { initializeApp } from 'firebase/app' // or firebase-admin
import { initializeFirestore } from 'firebase/firestore'

import { TypedFirestoreWeb } from 'fireschema'
import { firestoreModel } from './1-1-schema.js'

const app = initializeApp({
  // ...
})
const firestoreApp = initializeFirestore(app, {
  ignoreUndefinedProperties: true,
})

/**
 * Initialize TypedFirestore
 */
export const $web: TypedFirestoreWeb<typeof firestoreModel> =
  new TypedFirestoreWeb(firestoreModel, firestoreApp)

/**
 * Reference collections/documents and get snapshot
 */
const usersRef = $web.collection('users') // TypedCollectionRef instance
const userRef = usersRef.doc('userId') // TypedDocumentRef instance

const postsRef = userRef.collection('posts')
const postRef = postsRef.doc('123')
const techPostsQuery = postsRef.select.byTag('tech') // selector defined in schema

await userRef.get() // TypedDocumentSnap<User>
await userRef.getData() // User | undefined
await userRef.getDataOrThrow() // User

await postRef.get() // TypedDocumentSnap<PostA | PostB>
await postsRef.get() // TypedQuerySnap<PostA | PostB>
await postsRef.getData() // (PostA | PostB)[]
await techPostsQuery.get() // TypedQuerySnap<PostA | PostB>

/**
 * Get child collection of retrived document snapshot
 */
const snap = await usersRef.get()
const firstUserRef = snap.docs[0]!.ref

await firstUserRef.collection('posts').get()

/**
 * Reference parent collection/document
 */
const _postsRef = postRef.parentCollection()
const _userRef = postsRef.parentDocument()

/**
 * Reference collections groups and get snapshot
 */
const postsGroup = $web.collectionGroup('posts')
const techPostsGroup = postsGroup.select.byTag('tech')

await postsGroup.get() // TypedQuerySnap<PostA | PostB>
await techPostsGroup.get() // TypedQuerySnap<PostA | PostB>

/**
 * Write data
 */
await userRef.create(({ serverTimestamp }) => ({
  name: 'test',
  displayName: 'Test',
  age: 20,
  timestamp: serverTimestamp(),
  options: { a: true },
}))
await userRef.setMerge({
  age: 21,
})
await userRef.update({
  age: 21,
})
await userRef.delete()

/**
 * Transaction
 */
await $web.runTransaction(async (tt) => {
  const snap = await tt.get(userRef)
  tt.update(userRef, {
    age: snap.data()!.age + 1,
  })
})

Write methods of Fireschema's document reference

• create() - Create a document. (_createdAt / _updatedAt fields are added)
  • Web - Call JS SDK's set() internally. It fails if the document already exists because overwriting _createdAt is denied by the automatically generated security rules.
  • Admin - Call Admin SDK's create() internally. It fails if the document already exists.
• setMerge() - Call set(data, { merge: true }) internally. (_updatedAt field is updated)
• update() - Call update() internally. (_updatedAt field is updated)

set() is not implemented on fireschema because it cannot determine whether _createdAt should be included in update fields without specifying it is a new creation or an overwrite.

4. React Hooks

import React, { Suspense } from 'react'

import { useTypedCollection, useTypedDoc } from 'fireschema/hooks'
import { $web } from './1-3-typed-firestore.js'

/**
 * Get realtime updates of collection/query
 */
export const PostsComponent = () => {
  const userRef = $web.collection('users').doc('user1')
  const postsRef = userRef.collection('posts')

  const posts = useTypedCollection(postsRef)
  const techPosts = useTypedCollection(postsRef.select.byTag('tech'))

  return (
    <Suspense fallback={'Loading...'}>
      <ul>
        {posts.data.map((post, i) => (
          <li key={i}>{post.text}</li>
        ))}
      </ul>
    </Suspense>
  )
}

/**
 * Get realtime updates of document
 */
export const UserComponent = ({ id }: { id: string }) => {
  const user = useTypedDoc($web.collection('users').doc(id))

  return (
    <Suspense fallback={'Loading...'}>
      <span>{user.data?.displayName}</span>
    </Suspense>
  )
}

1. Create functions

import * as functions from 'firebase-functions'
import { z } from 'zod'

import { TypedFunctions } from 'fireschema/admin'
import { UserType, firestoreModel } from './1-1-schema.js'

/**
 * Initialize TypedFunctions
 */
const timezone = 'Asia/Tokyo'
const typedFunctions = new TypedFunctions(firestoreModel, timezone)
const builder = functions.region('asia-northeast1')

/**
 * functions/index.ts file
 */
export const UserJsonType = UserType.extend({ timestamp: z.string() })
export const callable = {
  createUser: typedFunctions.callable({
    schema: {
      input: UserJsonType, // schema of request data (automatically validate on request)
      output: z.object({ result: z.boolean() }), // schema of response data
    },
    builder,
    handler: async (data, context) => {
      console.log(data) // UserJson

      return { result: true }
    },
  }),
}

export const firestoreTrigger = {
  onUserCreate: typedFunctions.firestoreTrigger.onCreate({
    builder,
    path: 'users/{uid}',
    handler: async (decodedData, snap, context) => {
      console.log(decodedData) // UserDecoded (provided based on path string)
      console.log(snap) // QueryDocumentSnapshot<User>
    },
  }),
}

export const http = {
  getKeys: typedFunctions.http({
    builder,
    handler: (req, resp) => {
      if (req.method !== 'POST') {
        resp.status(400).send()
        return
      }
      resp.json(Object.keys(req.body))
    },
  }),
}

export const topic = {
  publishMessage: typedFunctions.topic('publish_message', {
    schema: z.object({ text: z.string() }),
    builder,
    handler: async (data) => {
      data // { text: string }
    },
  }),
}

export const schedule = {
  cron: typedFunctions.schedule({
    builder,
    schedule: '0 0 * * *',
    handler: async (context) => {
      console.log(context.timestamp)
    },
  }),
}

2. Call HTTPS callable function

Automatically provide types to request/response data based on passed functions module type.

import { initializeApp } from 'firebase/app'
import { getFunctions } from 'firebase/functions'
import React from 'react'

import { TypedCaller } from 'fireschema'

type FunctionsModule = typeof import('./2-1-typed-functions.js')

const app = initializeApp({
  // ...
})
const functionsApp = getFunctions(app, 'asia-northeast1')

export const typedCaller = new TypedCaller<FunctionsModule>(functionsApp)

const Component = () => {
  const createUser = async () => {
    const result = await typedCaller.call('createUser', {
      name: 'test',
      displayName: 'Test',
      age: 20,
      timestamp: new Date().toISOString(),
      options: { a: true },
    })

    if (result.error) {
      console.error(result.error)
      return
    }
    console.log(result.data)
  }

  return <button onClick={createUser} />
}