Alagarr is a request-response helper library for serverless/faas functions^* invoked via HTTP events (e.g. via API Gateway). Alagarr makes your code portable: it abstracts the event-context-callback function signatures of various serverless-providers so that you can spend less time writing boring function-as-a-service-related boilerplate.

Alagarr is a higher-order function which abstracts the the programming models of various serverless-cloud providers and adds a standardized request-response model extensible through composable middleware functions. It's API is concise and will be familiar to anyone who's worked with Express.js. It comes with built-in error handling which makes it trivial to implement error-recovery strategies.

^*Currently: AWS Lambda/API Gateway. Next: GCP & Azure

Without Alagarr:

// AWS Lambda / API Gateway
module.exports.myHandler = function(event, context, callback) {
  callback(null, {
    statusCode: 200,
    body: JSON.stringify({ foo: 'bar' }),
    headers: {
      'content-type': 'application/json',
    },
  })
}

With Alagarr:

const alagarr = require('alagarr')

module.exports.myHandler = alagarr(() => ({ foo: 'bar' }))

Contents

1. Features
2. Full Example
3. Installation & Usage
4. Configuration
   1. Options
5. API
6. Error Handling
7. Logging
8. Middleware
   1. Request Middleware
   2. Response Middleware
   3. Custom Middleware
9. Contributing
10. Similar Projects
11. Related Thingies
12. License

Features

• Concise & familiar API
• Zero dependencies
• Fully tested
• Built-in error handling makes catching and throwing errors a breeze
• Kibana-ready request logging
• Middleware for common tasks included
• Request cookie parsing
• Normalized request headers
• Includes request body parsers
• Response CSP headers
• Response gzipping/deflate
• Easily respond with images/binary data
• Support for custom middleware

Full Example

Alagarr helps you cut out all the boilerplate involved with handling HTTP requests in serverless functions. Albeit somewhat contrived, here is a before-and-after example of a common pattern frequently found in AWS Lambda function's:

Without Alagarr 😭

const got = require('got')

module.exports.handler = function(event, context, callback) {
  const { queryStringParameters: { currency } } = event

  if (!currency) {
    callback(null, {
      statusCode: 400,
      body: JSON.stringify({
        message: 'Please provide the "currency" query parameter.',
      }),
      headers: {
        'content-type': 'application/json',
      },
    })
  }

  got(`https://api.coinmarketcap.com/v1/ticker/${currency}`)
    .then(response => {
      callback(null, {
        statusCode: 200,
        body: JSON.stringify(response),
        headers: {
          'content-type': 'application/json',
        },
      })
    })
    .catch(error => {
      callback(null, {
        statusCode: error.statusCode,
        body: JSON.stringify({
          error: error.response,
        }),
        headers: {
          'content-type': 'application/json',
        },
      })
    })
}

With Alagarr 😁

const { alagarr, ClientError } = require('alagarr') // @TODO: this require is wrong
const got = require('got')

module.exports.handler = alagarr((request, response) => {
  const { query: { currency } } = request

  if (!currency) {
    throw new ClientError('Please provide the "currency" query parameter.')
  }

  return got(`https://api.coinmarketcap.com/v1/ticker/${currency}`)
})

There are a few things being handled for you in the above Alagarr example:

• The programming model has been normalized. You can run this code without modification on any of the supported cloud/faas/serverless providers. Not just AWS Lambda. Alagarr makes your code portable.
• The callback() is being handled for you. Alagarr will set the status code, content-type, and body appropriately. More on this behavior here.
• Errors are caught for you and turned into something appropriate for the client based on the type of error. If you don't like the default behavior, you can provide your own error handler.

Installation & Usage

Install Alagarr with NPM or Yarn:

npm install alagarr

Then include it in your serverless function:

const alagarr = require('alagarr')

module.exports.exampleHandler = alagarr(request => {
  const { path, provider } = request

  return `You've ended up at ${path} on the ${provider} cloud.`
})

Configuration

Alagarr ships with default configuration that should work for most use-cases. But, it's possible to pass a configuration object as the second parameter to the alagar() function:

const alagarr = require('alagarr')

module.exports.handler = alagarr(() => 'Hello world!', {
  headers: {},
  logger: console.log,
})

Configuration Options

The available configuration options are outlined here:

Option	Default	Description
cspPolicies	[]	List of CSP policies to include in the response headers
errorHandler		Provide a custom error handler. See the section on Error Handling for more details
headers	{}	Headers to include in every response
logger		Logger to use to log requests. If undefined, Alagarr will use an internal logger. Logging can be disabled by setting to false. See the section on Logging for more details
requestMiddleware		Array of custom request middleware to use. See the section on Request Middleware for more details
responseMiddleware		Array of custom response middleware to use. See the section on Response Middleware for more details

API

Alagarr module

• alagarr()
• requestMiddleware
• responseMiddleware
• ClientError()
• ServerError()

Request methods

• request.body
• request.context
• request.cookies
• request.headers
• request.hostname
• request.meta
• request.method
• request.path
• request.provider
• request.query
• request.source
• request.timestamp

Response methods

• response.json()
• response.html()
• response.text()
• response.svg()
• response.png()
• response.jpeg()
• response.respondTo()
• response.raw()

---

alagarr(handlerFunction, configurationOptions?): void

@TODO

const alagarr = require('alagarr')

const configurationOptions = {
  logger: false,
}

const handlerFunction = function(request, response) {
  const { query: { name } } = request
  return response.html(`Hello ${name}.`)
}

module.exports.handler = alagar(handlerFunction, configurationOptions)

The handlerFunction has a function signature of:

export type HandlerFunction = (
  request: any,
  response: any,
) => string | object | void | Promise<string | object | void>

If your handlerFunction returns falsey, then it's your responsibility to call the appropriate response method to end the invocation (e.g. response.json()). For convenience, if the handlerFunction returns a string, the result will be passed to response.html() or response.text() for you. Alternatively, if the handler returns an object, it will be passed to response.json(). You may also return a Promise (or make your handler async).

---

requestMiddleware

@TODO

---

responseMiddleware

@TODO

---

ClientError(message, statusCode = 400)

@TODO

---

ServerError(message, statusCode = 500)

@TODO

---

request.body

The request body, if any. If using default request middleware, or another body parser, this value will contain the parsed contents of the request body.

readonly body: string | object

---

request.context

The provider context object.

On AWS Lambda this is the second parameter passed to a Lambda function's handler.

readonly context: object

---

request.cookies

An object containing the cookies included with the request.

readonly cookies: {
  readonly [name: string]: string
}

---

request.headers

An object containing all of the headers included in the request.

readonly headers: {
  readonly [name: string]: string
}

---

request.hostname

The request's hostname. Derived from the request's Host header.

readonly hostname: string

---

request.meta

An object containing some meta data about the invocation. It includes:

readonly meta: {
  readonly coldStart: boolean, // was this a cold start?
  readonly invocationCount: number, // number of times this container has been invoked
}

---

request.method

The request HTTP method. E.g GET or POST.

readonly method: enum {
  'GET',
  'POST',
  'PATCH',
  'DELETE',
}

---

request.path

The request path.

readonly path: string

---

request.provider

The name of the current request's provider. Possible values include: aws

readonly provider: enum {
  'aws'
}

---

request.query

An object of query parameters included in the request.

Given a request:

GET http://example.com?foo=1&bar=2

request.query will contain:

{
  foo: '1',
  bar: '2',
}

readonly query: {
  readonly [name: string]: string
}

---

request.source

The name of the current request's invocation source. Possible values include: api-gateway

readonly source: enum {
  'api-gateway'
}

---

request.timestamp

Timestamp at the time of the first middleware's execution.

readonly timestamp: number

---

response.respondTo(formats, statusCode = 200, options = {}): void

Respond according to request's Accept header with formats provided in formats map. Kind of like a Ruby on Rails respond_to do |format| block.

response.respondTo({
  json: {},
  html: '<html />',
})

---

response.raw(error: Error | null, result?: object | boolean | number | string): void

Exposes the underlying callback method.

response.raw(null, { something: 'custom' })

Error Handling

Throw em. Alagarr will catch them.

@TODO

Logging

Yes.

@TODO

Middleware

Alagarr uses a pipeline of middleware functions to process the incoming request and outgoing response objects. This lets you customize how your requests and responses are handled as well as provide custom middleware in addition to those provided by Alagarr.

Request Middleware

Alagarr includes the following request middleware:

Provider	Name	Default	Description
All	meta	built-in	Adds meta data about the request including whether the invocation is a coldStart, and invocation count
All	normalize-headers	built-in	Normalizes request headers.
All	normalize-programming-model	built-in	Normalizes the programming models of different providers.
All	timestamp	built-in	Adds a request-start timestamp under request.timestamp which can be used to determine the ellapsed duration of the invocation
Any	cookies	enabled	Parses cookies out of request header and makes them accessible under request.cookies
Any	hostname	enabled	Sets a convenience hostname property on the request object based on the request headers
Any	json-body	enabled	Body parser for request bodies with content-type of application/json
Any	url-encoded-body	enabled	Body parser for request bodies with content-type of application/x-www-form-urlencoded
AWS	base64-body	enabled	Decodes base64-encoded request bodies when isBase64Encoded on the API Gateway request is truthy

Response Middleware

Alagarr includes the following response middleware:

Provider	Name	Default	Description
All	enforced-headers	built-in
All	log	built-in
Any	compress	disabled	Compress response body with deflate or gzip when appropriate
Any	content-length	enabled	Adds a content-length header to the response
Any	csp	enabled	Adds Content-Security-Policy headers to the response
Any	etag	disabled	Adds an Entity Tag (ETag) header to the response

Custom Middleware

All middleware are functions. Middleware which is included in Alagarr are all pure, but this is not required for custom middleware. Middleware may return Promises which are resolved before the next middleware is called. Middleware should not mutate state, but instead return new values—but this is not required in custom middleware. However, everytime middleware mutates state, a cute cuddly koala dies somewhere in Australia. So.. Yea.

Request middleware act on a request object and must always return a new request object. Request middleware have the following function signature:

type requestMiddleware = (
  request: InterfaceRequest,
  options: InterfaceAlagarrOptions,
) => InterfaceRequest

Response middleware act on the response object and must always return a new response object. Response middleware have the following function signature:

type responseMiddleware = (
  response: InterfaceResponseData,
  request: InterfaceRequest,
  options: InterfaceAlagarrOptions,
) => InterfaceResponseData

Example

An example of custom middleware might be middleware which handles user sessions. The request middleware would restore a session from some data store, while the response middleware might ensure a session is updated and a cookie is set.

Request Middleware

module.exports.restoreSession = async function(request) {
  const { cookies: { sessionId } } = request
  const session = (await getSessionFromDatabase(sessionId)) || undefined

  return {
    ...request,
    session,
  }
}

Response Middleware

module.exports.saveSession = async function(responsePayload, request) {
  const sessionCookie = await saveSessionToDatabase(request.session)

  return {
    ...responsePayload,
    headers: {
      ...responsePayload.headers,
      'Set-Cookie': `session=${sessionCookie}`, // @TODO: refactor once #5 is closed.
    },
  }
}

This custom middleware could then be used with Alagarr in a serverless function handler with:

const handler = require('alagarr')
const { restoreSession, saveSession } = require('./custom-middleware')

const alagarrConfig = {
  requestMiddleware: ['default', restoreSession],
  responseMiddleware: ['default', saveSession],
}

module.exports.userDashboardHandler = handler((request, response) => {
    const session = request.session

    if (!session) {
      return response.redirect('/login')
    }

    return `<h1>Welcome back, ${session.username}!</h1>`
  }
  alagarrConfig,
)

Contributing

The codebase tries to follow declarative, functional(ish) programming paradigms. Many functional styles are enforced through TSLint linter utilised by the project. These include immutablity rules (no-let, no-object-mutation) and rules which prohibit imperative code (no-expression-statement, no-loop-statement). Disabling the linter for code should be avoided. Exceptions are made where satisfying a linting rule is impractical or otherwise untenable. In practice, this tends to be areas where the code touches 3rd party modules and in tests due to Jest's imperative-style.

Similar Projects

• Middy
• corgi
• node-lambda-req
• apigateway-utils
• serverless-utils
• @graphcool/lambda-helpers

Related Thingies

• aws-kms-thingy
• aws-s3-thingy

License

Alagarr © Marco Lüthy. Released under the MIT license.
Authored and maintained by Marco Lüthy with help from contributors.

github.com/adieuadieu · GitHub @adieuadieu · Twitter @adieuadieu · Medium @marco.luethy