Resolvers
Resolvers dynamically resolve individual properties based on a context, in a Feathers application usually the hook context.
This provide a flexible way to do things like:
- Populating associations
- Returning computed properties
- Securing queries and e.g. limiting requests for a user
- Setting context (e.g. logged in user or organization) specific default values
- Removing protected properties for external requests
- Add read- and write permissions on the property level
- Hashing passwords and validating dynamic password policies
You can create a resolver for any data type and resolvers can also be used outside of Feathers.
Example
Here is an example for a standalone resolver using a custom context:
import { resolve } from '@feathersjs/schema'
type User = {
id: number
name: string
}
type Message = {
id: number
userId: number
likes: number
text: string
user: User
}
class MyContext {
async getUser(id) {
return {
id,
name: 'David'
}
}
async getLikes(messageId) {
return 10
}
}
const messageResolver = resolve<Message, MyContext>({
likes: async (value, message, context) => {
return context.getLikes(message.id)
},
user: async (value, message, context) => {
return context.getUser(message.userId)
}
})
const resolvedMessage = await messageResolver.resolve(
{
id: 1,
userId: 23,
text: 'Hello!'
},
new MyContext()
)
Property resolvers
Property resolvers are a map of property names to resolver functions. A resolver function is an async
function that resolves a property on a data object. If it returns undefined
the property will not be included. It gets passed the following parameters:
value
- The current value which can also beundefined
data
- The initial data objectcontext
- The context for this resolverstatus
- Additional status information like current property resolver path, the properties that should be resolved or a reference to the initial context.
const userResolver = resolve<User, MyContext>({
isDrinkingAge: async (value, user, context) => {
const drinkingAge = await context.getDrinkingAge(user.country)
return user.age >= drinkingAge
},
fullName: async (value, user, context) => {
return `${user.firstName} ${user.lastName}`
}
})
danger
Property resolver functions should only return a value and not have side effects. This means a property resolver should not do things like create new data or modify the data
or context
object. Hooks should be used for side effects.
Virtual property resolvers
Virtual resolvers are property resolvers that do not use the value
but instead always return a value of their own. The parameters are ((data, context, status)
). The above example can be written like this:
import { resolve, virtual } from '@feathersjs/schema'
const userResolver = resolve<User, MyContext>({
isDrinkingAge: virtual(async (user, context) => {
const drinkingAge = await context.getDrinkingAge(user.country)
return user.age >= drinkingAge
}),
fullName: virtual(async (user, context) => {
return `${user.firstName} ${user.lastName}`
})
})
Important
Virtual resolvers should always be used when combined with a database adapter in order to make valid $select queries. Otherwise queries could try to select fields that do not exist in the database which will throw an error.
Options
A resolver takes the following options as the second parameter:
converter
(optional): Aasync (data, context) => {}
function that can return a completely new representation of the data. Aconverter
runs beforeproperties
resolvers.
const userResolver = resolve<User, MyContext>(
{
isDrinkingAge: async (value, user, context) => {
const drinkingAge = await context.getDrinkingAge(user.country)
return user.age >= drinkingAge
},
fullName: async (value, user, context) => {
return `${user.firstName} ${user.lastName}`
}
},
{
// Convert the raw data into a new structure before running property resolvers
converter: async (rawData, context) => {
return {
firstName: rawData.data.first_name,
lastName: rawData.data.last_name
}
}
}
)
Hooks
In a Feathers application, resolvers are used through hooks to convert service query
, data
and response
. The context for these resolvers is always the hook context.
resolveData
Data resolvers use the hooks.resolveData(...resolvers)
hook and convert the data
from a create
, update
or patch
service method or a custom method. This can be used to validate against the schema and e.g. hash a password before storing it in the database or to remove properties the user is not allowed to write. It is possible to pass multiple objects containing resolvers which will run in the order they are passed. Subsequent resolver objects will receive the output from previous resolvers. schemaHooks.resolveData
can be used as an around
and before
hook.
import { hooks as schemaHooks, resolve } from '@feathersjs/schema'
import { Type } from '@feathersjs/typebox'
import type { Static } from '@feathersjs/typebox'
import type { HookContext } from '../declarations'
const messageSchema = Type.Object(
{
id: Type.Number(),
text: Type.String(),
createdAt: Type.Number(),
updatedAt: Type.Number(),
userId: Type.Number()
},
{ $id: 'Message', additionalProperties: false }
)
type Message = Static<typeof messageSchema>
// Pick the data for creating a new message
const messageDataSchema = Type.Pick(messageSchema, ['text'])
type MessageData = Static<typeof messageDataSchema>
// Resolver that automatically set `userId` and `createdAt`
const messageDataResolver = resolve<Message, HookContext>({
userId: async (value, message, context) => {
// Associate the currently authenticated user
return context.params?.user.id
},
createdAt: async () => {
// Return the current date
return Date.now()
}
})
// Resolver that automatically sets `updatedAt`
const messagePatchResolver = resolve<Message, HookContext>({
updatedAt: async () => {
// Return the current date
return Date.now()
}
})
app.service('users').hooks({
before: {
create: [schemaHooks.resolveData(messageDataResolver)],
patch: [schemaHooks.resolveData(messagePatchResolver)]
}
})
Note that as an all
hook resolveData
will run for any method that has data
, including custom methods. If you want to validate custom methods differently the hook should be registered on each service method it is used:
app.service('users').hooks({
before: {
create: [schemaHooks.resolveData(messageDataResolver)],
update: [schemaHooks.resolveData(messageDataResolver)],
patch: [schemaHooks.resolveData(messageDataResolver)],
customMethod: [schemaHooks.resolveData(customMethodDataResolver)]
}
})
resolveResult
Result resolvers use the hooks.resolveResult(...resolvers)
hook and resolve the data that is returned by a service call (context.result in a hook). This can be used to populate associations or add other computed properties etc. It is possible to pass multiple resolvers which will run in the order they are passed, using the previous data.
Important
schemaHooks.resolveResult
must be used as an around
hook. This is to ensure that the database adapters will be able to handle $select queries properly when using virtual properties.
import { hooks as schemaHooks, resolve, virtual } from '@feathersjs/schema'
import { Type } from '@feathersjs/typebox'
import type { Static } from '@feathersjs/typebox'
import type { HookContext } from '../declarations'
const userSchema = Type.Object(
{
id: Type.Number(),
email: Type.String(),
password: Type.String(),
avatar: Type.Optional(Type.String())
},
{ $id: 'User', additionalProperties: false }
)
type User = Static<typeof userSchema>
const messageSchema = Type.Object(
{
id: Type.Number(),
text: Type.String(),
createdAt: Type.Number(),
userId: Type.Number(),
user: Type.Ref(userSchema)
},
{ $id: 'Message', additionalProperties: false }
)
type Message = Static<typeof messageSchema>
export const messageResolver = resolve<Message, HookContext>({
user: virtual(async (message, context) => {
// Populate the user associated via `userId`
return context.app.service('users').get(message.userId)
})
})
app.service('messages').hooks({
around: {
all: [schemaHooks.resolveResult(messageResolver)]
}
})
resolveExternal
External (or dispatch) resolver use the hooks.resolveDispatch(...resolvers)
hook to return a safe version of the data that will be sent to external clients. It is possible to pass multiple resolvers which will run in the order they are passed, using the previous data. Returning undefined
for a property resolver will exclude the property which can be used to hide sensitive data like the user password. This includes nested associations and real-time events. schemaHooks.resolveExternal
can be used as an around
or after
hook.
import { hooks as schemaHooks, resolve } from '@feathersjs/schema'
import { Type } from '@feathersjs/typebox'
import type { Static } from '@feathersjs/typebox'
import type { HookContext } from '../declarations'
const userSchema = Type.Object(
{
id: Type.Number(),
email: Type.String(),
password: Type.String(),
avatar: Type.Optional(Type.String())
},
{ $id: 'User', additionalProperties: false }
)
type User = Static<typeof userSchema>
export const userExternalResolver = resolve<User, HookContext>({
// Always hide the password for external responses
password: async () => undefined
})
// Dispatch should be resolved on every method
app.service('users').hooks({
around: {
all: [schemaHooks.resolveExternal(userExternalResolver)]
}
})
important
In order to get the safe data from resolved associations all services involved need the schemaHooks.resolveExternal
hook registered even if it does not need a resolver (schemaHooks.resolveExternal()
).
schemaHooks.resolveExternal
should be registered first when used as an around
hook or last when used as an after
hook so that it gets the final result data.
resolveQuery
Query resolvers use the hooks.resolveQuery(...resolvers)
hook to modify params.query
. This is often used to set default values or limit the query so a user can only request data they are allowed to see. It is possible to pass multiple resolvers which will run in the order they are passed, using the previous data. schemaHooks.resolveQuery
can be used as an around
or before
hook.
In this example for a User
schema we are first checking if a user is available in our request. In the case a user is available we are ruturning the user's ID. Otherwise we return whatever value was provided for id
.
context.params.user
would only be set if the request contains a user. This is usually the case when an external request is made. In the case of an internal request we may not have a specific user we are dealing with, and we will just return value
.
If we were to receive an internal request, such as app.service('users').get(123)
, context.params.user
would be undefined
and we would just return the value
which is 123
.
import { hooks as schemaHooks, resolve } from '@feathersjs/schema'
import { Type } from '@feathersjs/typebox'
import type { Static } from '@feathersjs/typebox'
import type { HookContext } from '../declarations'
const userSchema = Type.Object(
{
id: Type.Number(),
email: Type.String(),
password: Type.String(),
avatar: Type.Optional(Type.String())
},
{ $id: 'User', additionalProperties: false }
)
type User = Static<typeof userSchema>
export const userQueryProperties = Type.Pick(userSchema, ['id', 'email'])
export const userQuerySchema = querySyntax(userQueryProperties)
export type UserQuery = Static<typeof userQuerySchema>
export const userQueryResolver = resolve<UserQuery, HookContext>({
// If there is an authenticated user, they can only see their own data
id: async (value, query, context) => {
if (context.params.user) {
return context.params.user.id
}
return value
}
})
// The query can be resolved on every method
app.service('users').hooks({
before: {
all: [resolveQuery(userQueryResolver)]
}
})
For a more complicated example. We will make a separate queryResolver
, called companyFilterQueryResolver
, that will act as a ownership filter. We will have a Company
service that is owned by a User
. We will assume our app has two registered users and two companies. Each user owning one company. For simplicity, User1
owns Company1
, and User2
owns Company2
We want to make sure only the user that owns the company can make any requests related to it. Our schema contains a ownerUser
field, this is the owner of the company. When a request is made to the company schema, we are effectivly filtering our search for companies to be only those whose ownerUser
matches the requesting user's id.
So if a GET /company
request is made by User1
, our resolver will convert our query to GET /company?name=Company1&ownerUser={User1.id}
. The result will only return an array of 1 company to User1
Similarily, if a patch request was made by User1
to modify Company2
. A 404
would occur, as resulting query would search the database for a Company2
that is owned by User1
which does not exist.
// Main data model schema
export const companySchema = Type.Object(
{
id: Type.String({ format: 'uuid' }),
name: Type.String(),
ownerUser: Type.Ref(userSchema)
},
{ $id: 'Company', additionalProperties: false }
)
// Schema for allowed query properties
export const companyQueryProperties = Type.Pick(companySchema, ['id'])
export const companyQuerySchema = Type.Intersect(
[
querySyntax(companyQueryProperties),
// Add additional query properties here
Type.Object({}, { additionalProperties: false })
],
{ additionalProperties: false }
)
export type CompanyQuery = Static<typeof companyQuerySchema>
export const companyQueryValidator = getValidator(companyQuerySchema, queryValidator)
export const companyQueryResolver = resolve<CompanyQuery, HookContext>({})
export const companyFilterQueryResolver = resolve<Company, HookContext>({
ownerUser: async (value, obj, context) => {
if (context.params.user) {
return context.params.user.id
}
return value
}
})