Social Login with Twitter using hapi.js
Twitter:
// Load modules
const Bell = require('@hapi/bell');
const Hapi = require('@hapi/hapi');
// Declare internals
const internals = {};
internals.start = async function () {
const server = Hapi.server({ port: 8000 });
// Register bell with the server
await server.register(Bell);
// Declare an authentication strategy using the bell scheme
// with the name of the provider, cookie encryption password,
// and the OAuth client credentials.
server.auth.strategy('twitter', 'bell', {
provider: 'twitter',
password: 'cookie_encryption_password_secure',
clientId: 'my_twitter_client_id',
clientSecret: 'my_twitter_client_secret',
isSecure: false // Terrible idea but required if not using HTTPS especially if developing locally
});
// Use the 'twitter' authentication strategy to protect the
// endpoint handling the incoming authentication credentials.
// This endpoint usually looks up the third party account in
// the database and sets some application state (cookie) with
// the local application account information.
server.route({
method: ['GET', 'POST'], // Must handle both GET and POST
path: '/login', // The callback endpoint registered with the provider
options: {
auth: {
mode: 'try',
strategy: 'twitter'
},
handler: function (request, h) {
if (!request.auth.isAuthenticated) {
return `Authentication failed due to: ${request.auth.error.message}`;
}
// Perform any account lookup or registration, setup local session,
// and redirect to the application. The third-party credentials are
// stored in request.auth.credentials. Any query parameters from
// the initial request are passed back via request.auth.credentials.query.
return h.redirect('/home');
}
}
});
await server.start();
};
internals.start();
Testing third-party authorization is often a painful process. They are hard to test, tidious to configure, and tend to break when running on local servers. If you are having issues running bell locally, you might want to look at the isSecure
and isSameSite
options. Since most people don't run TLS on their local test server, isSecure
must be set to false
to remove the TLS requirement. isSameSite
might need to be set to 'Lax'
in some cases.
You should also review the default scope set for each provider early in your implementation. The default scope is meant to get the minimal permissions needed to perform simple user login. In most cases, you would want to ask for more permissions as the default often does not provide access to most API calls provided by the third-party services.
The server.auth.strategy()
method requires the following strategy options:
provider
- the name of the third-party provider ('auth0'
, 'azure'
, 'bitbucket'
,
'dropbox'
, 'facebook'
, 'fitbit'
, 'foursquare'
, 'github'
, 'google'
, 'googleplus'
,
'instagram'
, 'linkedin'
, 'live'
, 'twitter'
, 'vk'
, 'arcgisonline'
, 'yahoo'
,
'nest'
, 'phabricator'
, 'pinterest'
) or an object containing a custom
provider with the following:
name
- the custom provider name. Defaults to custom
.protocol
- the authorization protocol used. Must be one of:
'oauth'
- OAuth 1.0a'oauth2'
- OAuth 2.0signatureMethod
- the OAuth signature method (OAuth 1.0a only). Must be one of:
'HMAC-SHA1'
- default'RSA-SHA1'
- in that case, the clientSecret
is your RSA private keytemporary
- the temporary credentials (request token) endpoint (OAuth 1.0a only).useParamsAuth
- boolean that determines if OAuth client id and client secret will be sent
as parameters as opposed to an Authorization header (OAuth 2.0 only). Defaults to false
.auth
- the authorization endpoint URI.token
- the access token endpoint URI.scope
- an array of scope strings (OAuth 2.0 only).scopeSeparator
- the scope separator character (OAuth 2.0 only). Only required when a
provider has a broken OAuth 2.0 implementation. Defaults to space (Facebook and GitHub
default to comma).headers
- a headers object with additional headers required by the provider (e.g. GitHub
required the 'User-Agent' header which is set by default).profileMethod
- get
or post
for obtaining user profile by profile
function. Default
is get
.profile
- a function used to obtain user profile information and normalize it. The function
signature is async function(credentials, params, get)
where:
credentials
- the credentials object. Change the object directly within the function
(profile information is typically stored under credentials.profile
).params
- the parsed information received from the provider (e.g. token, secret, and
other custom fields).get
- an OAuth helper function to make authenticated requests using the credentials
received. The get
function signature is async function(uri, params)
where:
uri
- the requested resource URI (bell will add the token or authentication
header as needed).params
- any URI query parameters (cannot include them in the URI due to signature
requirements).password
- the cookie encryption password. Used to encrypt the temporary state cookie used by
the module in between the authorization protocol steps.
clientId
- the OAuth client identifier (consumer key).
clientSecret
- the OAuth client secret (consumer secret). This is usually a client password
formatted as a string, but to allow OAuth2 custom client authentication
(e.g. client certificate-based authentication), this option can be passed as an object. This
object will be merged with the Wreck request object used to call the token endpoint. Such an
object can contain custom HTTP headers or TLS options (e.g.
{ agent: new Https.Agent({ cert: myClientCert, key: myClientKey}) }
).
To allow dynamically updating secret, this option can be passed as a function returning string
to be used as clientSecret
.
forceHttps
- A boolean indicating whether or not you want the redirect_uri to be forced to
https. Useful if your hapi application runs as http, but is accessed through https.
location
- Set the base redirect_uri manually if it cannot be inferred properly from server
settings. Useful to override port, protocol, and host if proxied or forwarded. It may be passed
either as a string (in which case request.path is appended for you), or a function which takes
the client's request
and returns a non-empty string, which is used as provided. In both cases,
an empty string will result in default processing just as if the location
option had not been
specified.
Each strategy accepts the following optional settings:
cookie
- the name of the cookie used to manage the temporary state. Defaults to
'bell-provider'
where 'provider' is the provider name. For
example, the Twitter cookie name defaults to 'bell-twitter'
.isSameSite
- sets the cookie same site option. Defaults to Strict
.isSecure
- sets the cookie secure flag. Defaults to true
.isHttpOnly
- sets the cookie HTTP only flag. Defaults to true
.ttl
- cookie time-to-live in milliseconds. Defaults to null
(session time-life - cookies are
deleted when the browser is closed).domain
- the domain scope. Defaults to null
(no domain).providerParams
- provider-specific query parameters for the authentication endpoint. It may be
passed either as an object to merge into the query string, or a function which takes the client's
request
and returns an object. Each provider supports its own set of parameters which customize
the user's login experience. For example:
display
('page', 'popup', or 'touch'), auth_type
, auth_nonce
.access_type
, approval_prompt
, prompt
, login_hint
, user_id
, hd
.force_login
, screen_name
.fields
.allowRuntimeProviderParams
- allows passing query parameters from a bell protected endpoint
to the auth request. It will merge the query params you pass along with the providerParams and
any other predefined ones. Be aware that this will override predefined query parameters! Default
to false
.scope
- Each built-in vendor comes with the required scope for basic profile information. Use
scope
to specify a different scope as required by your application. It may be passed either as
an object to merge into the query string, or a function which takes the client's request
and
returns an object. Consult the provider for their specific supported scopes.skipProfile
- skips obtaining a user profile from the provider. Useful if you need specific
scope
s, but not the user profile. Defaults to false
.config
- a configuration object used to customize the provider settings:
extendedProfile
getMethod
uri
- allows pointing to a private enterprise installation (e.g.
'https://vpn.example.com'
). See Providers documentation for more
information.profileParams
- an object of key-value pairs that specify additional URL query parameters to
send with the profile request to the provider. The built-in facebook
provider, for example,
could have fields
specified to determine the fields returned from the user's graph, which would
then be available to you in the auth.credentials.profile.raw
object.runtimeStateCallback
- allows passing additional OAuth state from initial request. This must be
a function returning a string, which will be appended to the bell internal state
parameter
for OAuth code flow.On route handlers, the authentication object may contain one or more of the following properties depending on your route configuration and whether the authentication process was completed successfully or not:
auth.credentials = {
provider: String, // provider name
};
auth.credentials = {
token: String,
secret: String,
query: Object // sign-in request query params
};
auth.credentials = {
token: String,
refreshToken: String,
expiresIn: Number,
query: Object // sign-in request query params
};
auth.artifacts = Object; // OAuth token payload response
auth.credentials = {
query: Object // sign-in request query params
};
The server.auth.strategy()
method supports string representations of its boolean and number typed
options. For example, forceHttps
can be set to 'true', 'false', 'yes', or 'no'. In effect, this
allows you to configure any strategy option using environment variables.
By default, bell will reply back with an internal error (500) on most authentication errors due to the nature of the OAuth protocol. There is little that can be done to recover from errors as almost all of them are caused by implementation or deployment issues.
If you would like to display a useful error page instead of the default JSON response, use the
hapi onPreResponse
extension point to transform
the error into a useful page or to redirect the user to another destination.
Another way to handle authentication errors is within the route handler. By default, an
authentication error will cause the handler to be skipped. However, if the authentication mode is
set to 'try'
instead of 'required'
, the handler will still be called. In this case, the
request.auth.isAuthenticated
must be checked to test if authentication failed. In that case,
request.auth.error
will contain the error.
To keep track of the token expiry time, request.auth.credentials.expiresIn
provides you the
duration (in seconds) after which you could send a refresh token request using the
request.auth.credentials.refreshToken
to get a new token.
Testing applications using third-party login can be challenging given the lack of user interaction
to perform the third-party login flow as well as the multiple steps required. To assist in testing
such application without having to modify the application with custom code, Bell provides an
override method Bell.simulate()
which puts the module into simulation mode and any strategies
created while it is in this mode will return the simulated credentials.
The Bell.simulate(credentialsFunc)
takes a single argument:
credentialsFunc
- a function called for each incoming request to the protected resource that
should return the credentials object to be set in request.auth.credentials
. Note that bell
will set the default keys automatically if not present except for the provider-specific values.
has the signature function(request)
where:
request
- the hapi request object.Note that you must call Bell.simulate()
before the module is registered by your application and
need to call Bell.simulate(false)
to stop it from simulating authentication.
Sometimes, you want to use bell without using specifying a Hapi strategy. This can be the case when combining the auth logic together with another module.
bell exposes an oauth object in its plugin. Therefore, server.plugins.bell.oauth
now has all
that's needed. For example, calling the v2
method with all the settings documented above, will
handle the oauth2 flow.
You can pass a function, rather than an object, into the providerParams
and scope
config
options to allow you to customize the scope or parameters based on the user's request. For example,
this may be used you want people to be able to log in with a provider (and only need some basic
user information) but also want to let users authorize your application to post messages or status
updates on their behalf.
server.auth.strategy('twitter', 'bell', {
provider: 'twitter',
password: 'some cookie password',
location: 'http://example.com/oauth',
scope(request) {
const scopes = ['public_profile', 'email'];
if (request.query.wantsSharePermission) {
scopes.push('publish_actions');
}
return scopes;
}
});