Batch multiple Feathers service calls into one
feathers-batch
allows to batch multiple service requests into one. This is useful for minimizing client side requests to any Feathers API and can additionally speed up batched requests by only performing authentication once.
It also comes with a client side module that automatically collects API requests from a Feathers client into a batch.
feathers-batch
consists of two parts:
- The server side batch service to execute batch calls
- The client side batch client to collect parallel requests from a Feathers client into a batch service request
npm install feathers-batch --save
The batch service is a normal Feathers service that executes the batch calls.
It can be registered by adding the following to your src/services/index.js|ts
:
const { BatchService } = require('feathers-batch');
module.exports = function (app) {
// ...
app.use('batch', new BatchService(app));
}
Now multiple service calls can be made by sending a create
(POST
) call to /batch
with a { calls: [] }
property. calls
is an array in the same format as the Socket.io direct connection events:
{
"calls": [
[ "method", "serviceName", /* list of parameters */ ],
...
]
}
Note: When using a Feathers client with the batch client this will be done automatically.
For example, the following will execute a batch call to app.service('users').get(1, { query: { active: true } })
and app.service('messages').find({ query: { userId } })
:
{
"calls": [
[ "get", "users", 1, { active: true } ],
[ "find", "messages", { userId } ]
]
}
The return value will be the information as returned by Promise.allSettled:
[
{
"status": : "fulfilled",
"value": { /* user object returned by app.service('users').get(1) */ }
}, {
"status": : "fulfilled",
"value": { /* page returned by app.service('messages').find({ query: { userId } }) */ }
}
]
If an error happened:
[
{
"status": : "fulfilled",
"value": { /* user object returned by app.service('users').get(1) */ }
}, {
"status": : "rejected",
"reason": { /* error JSON or object with error message */ }
}
]
If you are batching authenticated requests, it is possible to perform the authentication step only once (instead of on every service call) in a batch by adding the authenticate hook to the batch service create
method:
app.service('batch').hooks({
before: {
create: [ authenticate('jwt') ]
}
});
feathers-batch
also exports a client side module that can be used with Feathers on the client that automatically collects multiple requests that are made at the same time into a single batch call. This works for any transport mechanism (REST, Socket.io etc.).
Batching on the client can be enabled like this:
// If your module loader supports the `browser` package.json field
import { batchClient } from 'feathers-batch';
// Alternatively
import { batchClient } from 'feathers-batch/client';
const client = feathers();
// configure Feathers client here
// `batchClient` should be configured *after*
// any other application level hooks
client.configure(batchClient({
batchService: 'batch'
}));
Now you can continue to make normal service calls and whenever possible they will be automatically combined into a batch (see parallelizing requests for more information).
The following options are available for the batchClient
:
batchService
(required) - The name of the batch serviceexclude
(optional) - An array of service names that should be excluded from batchingtimeout
(optional) (default:50
) - The number of milliseconds to wait when collecting parallel requests.
At the same time means e.g. multiple components making requests to the API in parallel. The following example will NOT be collected into a batch since the calls run sequentially using await
:
const user = await client.service('users').get(userId);
const messages = await client.service('messages').find({
query: { userId }
});
If the requests are not dependent on each other and you want to batch them, Promise.all needs to be used:
const [ user, messages ] = await Promise.all([
client.service('users').get(userId),
client.service('messages').find({
query: { userId }
})
]);
Copyright (c) 2020 Feathers contributors
Licensed under the MIT license.