serverless-functions
v0.5.4
Published
A toolkit for writing, testing, running and deploying serverless functions (e.g. AWS Lambda).
Downloads
28
Maintainers
Readme
A toolkit for writing, testing, running and deploying serverless functions (e.g. AWS Lambda).
Concepts
"Serverless" means splitting an API into a set of independent functions (API endpoints). Each function is a directory having a function.json
description file and an index.js
main code file (which can import
any other code) which exports default async function
which takes a parameters
object (HTTP GET query parameters, HTTP POST body, etc) and returns a response
which could be, for example, a JSON object.
Use
Create a new project.
Create a .babelrc
file in it:
{
"presets": [
["@babel/preset-env", {
"targets": {
"node": "8.10"
}
}]
],
"plugins": [
"@babel/plugin-proposal-object-rest-spread",
"@babel/plugin-proposal-class-properties"
]
}
(The above example is for Babel 7. For Babel 6 see version 0.2.x.)
(Support for babel.config.js
hasn't been tested; .babelrc
seems to work)
Create a serverless.json
file in it:
{
"name": "project-name"
}
To create a function create a directory anywhere inside the project directory and put index.js
and function.json
files in that directory.
function.json
is the description of the function, i.e. its name, which URL does it respond on, to which HTTP method, etc.
{
"name": "function-name",
"path": "/example-function/{parameterName}",
"method": "GET"
}
index.js
is the function code:
export default async function({ params, body, query }) {
return {
parameter: params.parameterName
}
}
Install serverless-functions
package:
npm install serverless-functions --save
Add a new script
to package.json
:
{
"scripts": {
"run-locally": "serverless run dev 8888"
}
}
Run the functions locally:
npm run run-locally
Go to http://localhost:8888/example-function/123
. It should respond with { parameter: 123 }
.
Input
The function receives the following parameters:
params
— URL path parameters (e.g./users/{id}
).query
— URL query parameters.body
— HTTP request body.headers
— HTTP request headers.
Output
The function can return anything. If it doesn't return anything then an empty object is assumed (in order for the HTTP response to be a valid JSON).
Errors
To return an error from a function one can throw an instance of Error
with an optional httpStatusCode
property. The HTTP Response status code is gonna be the value of the httpStatusCode
property (or 500
) and the HTTP Response JSON object is gonna contain error?: object
and statusCode: number
. The error
object contains message: string
and all properties from error.data
, if it's present.
export default async function() {
throw new Error('Test')
}
There are some pre-defined errors available:
Unauthenticated
Unauthorized
NotFound
Conflict
InputRejected
import { Unauthorized } from 'serverless-functions/errors'
export default async function() {
// Throws a 403 Forbidden error.
throw new Unauthorized('You are not authorized to perform this action')
}
Streaming
Currently serverless functions seem to not support streaming HTTP request/response data. Use the old-school Node.js stuff for that (e.g. express
).
AWS Setup
See the AWS Lambda guide.
Extending
One can customize the code template used for generating functions. The template is:
$initialize()
export async function handler(event, context, callback) {
try {
await $onCall(event, context)
const parameters = $createFunctionParameters(event, context)
let result = await $handler(parameters)
result = (await $onReturn(result)) || result
callback(null, $createResponse(result))
} catch (error) {
await $onError(error)
callback(null, $createErrorResponse(error))
} finally {
await $finally()
}
}
Each of the $
parameters (except $handler
) can be customized by adding a respective code
entry in serverless.json
:
{
"name": "project-name",
"code": {
"initialize": "./custom/initialize.js"
}
}
./custom/initialize.js
file path is resolved against the root project directory (process.cwd()
).
./custom/initialize.js
import Database from './database'
import config from './config'
function $initialize() {
const database = new Database(config[STAGE].database)
database.connect()
// Make the `database` accessible from functions.
global.database = database
}
Relative import
s inside "custom" code pieces are resolved against the root project directory too.
The global database
variable can then be used inside functions:
export default async function() {
return {
items: await database.items.findAll({ limit: 10 })
}
}
Globals
The execution envirnoment provides the following global constants:
STAGE : string
— the current "stage" (e.g.dev
,prod
,test
).FUNCTION : object
— the contents offunction.json
.LOCAL : boolean
— whether the functions are being run locally (serverless run
) rather than being deployed in a cloud.
Testing
For each function one should also create an index.test.js
file with unit tests. Example using mocha
/chai
:
import func from '.'
describe('function-name', () => {
it('should return some result for some input', async () => {
expect(await func(input)).to.deep.equal(output)
})
})
API
import { run } from 'serverless-functions'
import config from './serverless.json'
await run(stage, port, config, { cwd: process.cwd() })