Twitter OAuth wizard.
Twiz does authentication and/or authorization to Twitter with OAuth 1.0a, has built in REST
api support and also supports third party STREAM
and REST
libs.
twiz-client | twiz-server |
---|---|
twiz-client
is a js script.
twiz-server
is Express middleware.
Many Twitter apis require user authentication (access token
) before usage. OAuth 1.0a
is (essencially a digital signature) process of letting know who (which app) wants to use an api and on which user's behalf. In other words you tell Twitter who you are, if twitter is ok with you it lets you to ask an user of your website (with twitter account), on authorization page, if he agrees that you act on his behalf (like post tweets on user's profile ect ...)
It happens to be a 3-leg (step) dance, it's implementation could look like this:
As you can see there are 3 actors. Your web app/site, your server and twitter.
Twitter apis for authorization and authentication do not use CORS
, that is, they do not emit CORS
headers. So we cant send request to twitter directly from browser, but rather proxy them trough a Server, since request sent from server do not have any CORS
restrictions applied. Your app is identified with CONSUMER_KEY
and CONSUMER_SECRET
. Both of which you get from twitter when you create new app.
On Twitter, user is internally associated with the
access token
. Like user'saccess token
, your app'skey
andsecret
are extreamly sensitive information. If anyone whould to get your app's key/secret then it can send requests to twitter just like they were sent from your app. It can try to mock you. Usually that's not what you want. Also, javascript is in plain text form in browser so we should never have ourCONSUMER_KEY/SECRET
there but on server. This is another good reason why we need a server to proxy our requests. Likewise if anyone is to get user'saccess token
, then it may be able to send requests on user's behalf (who may actually never visited an app) and/or authorized such actions for that app. Altough, my guess is not in easy and straightforward way.
-
You ask your server to get you a
request token
. Server prepares and signs your request and sends it to twitter. Twitter checks yourCONSUMER_KEY
and signature (by reproducing it). If all is ok, it grants (returns) you an unauthorizedrequest token
. It's function is to be authorized (approved) by user in step 2, so it can be used to get user'saccess token
in step 3. -
Uppon receiving
request token
data, user is redirected to twitter to obtain user authorization :
-
if you are using
/oauth/authorize
enpoint for redirection, then every time user is redirected there it lands on authorization (interstitials) page. Even if user previously authorized your app. -
if you are using
/oauth/authenticate
enpoint for redirection, then only first time user is redirected it lands on authorization page. On any subsequent redirection twitter remembers first one and user is directed back again to the app. No authorization page is showed, the user is not involved directly. Historicaly it didn't work like it should, for a time. Two were actually the same where/oauth/authenticate
acted like/oauth/authorize
.Even there is no authorization page showed, there still can be a redirection page flash.
- Since user approved your
request token
, now it is used to get user'saccess token
. Server signs your request and sends it. Twitter does things similarly like in first step and grants you anaccess token
which belongs to the user who authorized it in second step.
After we get user's access token
the 3-leg OAuth is finished and we can use it to send request to twitter on user's behalf (like posts tweets on user's profile). In OAuth
parlance this process of sending requests with the access token is called accessing protected resources
, but it is not part of the OAuth
.
We can see that in 3-rd leg access token
is send back to web app. Which usually is not good idea, because of security implications it could have, like we mentioned earlier.
Let's see what twiz is doing with OAuth:
Three differences are:
-
Optimized load
Preparing an
OAuth
leg (step) mostly refers to the assembling ofSignature Base String
andAuthorization Header string
. It is all done in browser (Consumer
) in an effort to ease the server load for actions that bear no security concerns and thus do not need to be executed in server. Already prepared requests come to the server who acts mostly like a signing/stamping authority by merely inserting sensitive data and signing requests. -
Security
Another important point is that the user's access token is never send back to the browser by any means.
-
Haste
On the server we have a decision point where if we already have user's
access token
(stored from a previous authorization ie. in database) we can usehaste
(on diagram the yes branch ).Haste is a process where you verify
access token
freshness (verify credentials) with twitter and if it's fresh you can immediately go for a twitter api requests user actually wants. Checking token freshness is checking that user didn't revoke right to your app of doing things on it's behalf and such.This is usefull for scenarios where you save user's
access token
after first authorization and then just check for it's freshness before you go for an api request. User does not need to be bothered every time with 3-legOAuth
, there is no interstitials page. With haste you are the one who remembers user authorization instead of letting twitter to do it (like it does on the/oauth/authenticate
). All in order to have smooth user experience, like for instance in case/oauth/authenticate
stops working as expected. Or when redirection page flashes for the moment before user is returned to app and you would like to remove that behaviour.If this is the first time a user is making a request (and we dont have the
access token
) then we just continue the wholeOAuth flow
(on diagram the no branch). One of twiz's features is very easy switching between any of yourOAuth
workflows while having a redundant mechanism for smooth user experience (haste
) as an option.
In order to efficiently and safely use twiz make sure you:
- Provide HTTPS all the way (client ---> server --> twitter)
- In browser install
twiz-client
, on server installtwiz-server
- Create app account on twitter
- Users (of your app) must have twitter accounts
in browser:
CDN: <script src="https://cdn.jsdelivr.net/npm/twiz-client/src/twiz-client_bundle.min.js"></script>
yarn: yarn add twiz-client
local:
npm install twiz-client && npm run build
Then drop it in script tag: <script src="src/twiz-client_bundle.js"></script>
on server:
npm install twiz-server
browser:
// Let's say this code is in your page -> https://myApp.com
let twizlent = twizClient();
btn.addListener('onClick', function(){ // lets say we initiate oauth on click event
let args = {
server_url: 'https://myServer.com/route', // address of your node server
redirection_url: 'https://myApp.com', // address of your web app/site (where twitter will direct
// user after authorization)
options:{ // twitter request options
method: 'POST',
path: 'statuses/update.json'
params: {
status: "Hooray, new tweet!"
}
}
}
twizlent.OAuth(args)
.then(function fulfilled(o){
if(o.error) // not 200OK responses (has o.error.statusCode, o.error.statusText, o.error.data)
if(o.data) // (200OK) will have data on succesfull twiz.haste(accessToken) call on server
if(o.redirection) // Will have an o.redirection set to *true* when twiz.continueOAuth() is called on // server and user is redirected. Serves as a notifier for redirections.
o.xhr // Always present in case you need to pull some data from response
// (like custom server headers you might be sending)
}, function rejected(err){ // twiz errors
// err is instance of Error()
// has err.name, err.message, err.stack ...
})
})
// finishOAuth() Can be called asap in page
// Makes 3-rd step from diagram
// We dont need the redirection url for this step, but it will be ignored so we can pass same args
// It will fire after twitter (re)directs back to app, only on valid redirection (authorization) urls from twitter.
twizlent.finishOAuth(args);
.then(function fulfilled(o){
if(o.error) // not 200OK responses
if(o.data) // (200OK) will have data on succesfull twiz.continueOAuth() call on server
o.xhr // Always present in case you need to pull some data from response
// (like custom server headers you might be sending)
}, function rejected(err){ // twiz errors
// err is instance of Error()
// has err.name, err.message, err.stack ...
})
Notice that our redirection_url
is same as url of the page from which we are making a request. Making this a SPA use case.
The only presumtions about a succesfull request is one with 200OK
status code, so anything that does not have that status code will still be in fulfilled handler but in o.error
, left to your code logic.
twizlent.OAuth(..)
will bring api data (o.data
) if twiz.haste(accessToken)
was called on the server and had 200OK
response. If not and the twiz.continueOAuth()
is called it will receive request token
and redirect user to twitter.
Then o.redirection
is set to true
in fullfuled handler. Also note that here everything (redirection to twitter, twitter's (re)direction back to app) happens in same window/tab in browser. Check web site workflow for popUps.
Server is writen as express middleware.
node.js:
var twizServer = require('twiz-server');
var express = require('express');
var app = express();
var twizer = twizServer({
consumer_secret: process.env.CONSUMER_SECRET,
consumer_key: process.env.CONSUMER_KEY,
key: fs.readFileSync('yourServerPrivateKey.pem'),
cert: fs.readFileSync('yourServerCert.pem') // can be self signed certificate
})
app.use(twizer); // use the twiz-server
app.on('hasteOrOAuth', function(twiz, verifyCredentials){ // event where we pick haste or oauth
// When you don't have access token (or just don't want to use haste) you continue the oauth flow
twiz.continueOAuth();
// 1. user gets request token
// 2. is redirected for authorization (or authentication), twizlent.OAuth(..) has
// o.redirection set to *true*
// 3. with twiz.finishOAuth() in browser users gets api data in o.data
/ * . . . */
// Note that here in *hasteOrOAuth* handler is where you should go for user's access token
// since 'hasteOrOAuth' event will only be emitted for certain requests. Otherwise you'll hog your server
// cpu/io unnecessary. Verifyng credentials and using haste is completely optional step.
verifyCredentials(accessToken,{ skip_status: true}) // When you have accessToken
.then(function fullfilled(credentials){ // You can inspect returned credentials object
twiz.haste(accessToken) // Gets api data and sends back to browser
// (to twiz.OAuth(..) fullfiled handler)
}, function rejected(err){ // non 200OK responses from verifyCredentials
twiz.continueOAuth() // likely you would want to send it to reauthorization of access token
})
.catch(function(err){ // errors that might happen in fullfiled handler
})
})
app.on('tokenFound', function(found){ // when whole oauth process is finished you will get the user's
// access token
found // promise
.then(function(accessToken){
// user's access token received from twitter which you can put in database
}, function rejected(err){ // twiz errors
})
})
]
By default twizlent.OAuth(..)
will use the /oauth/authorize
endpoint , but you can use the /oauth/authenticate
like this:
browser:
let args = {
...
endpoints:{
authorize: 'authenticate' // sets authenticate instead of authorize (notice no forward slash)
}
}
This is the so called Sign in with Twitter flow, the one that uses /oauth/authenticate
endpoint. By default, twiz gets you an access token and gets your api data immediately afterwards. After these actions you can specify your own end with twiz.onEnd(). This is handy if you are using template rendering to show the signed-in UI to user, wheater on front-end or back-end.
Since it gets you an access token and api data in one swoop, twiz makes the sign-in-with-twitter button unecessary, the whole proccess can happen in one click of a button.
As opposed to:
1. user clicks the sign-in-with-twitter button to authorize your app (you display the signed-in user UI)
2. user clicks another button to get data it wants from an twitter api
Test drive the Sign in with twitter + api data with single click. (SPA with /oauth/authenticate
). Running on heroku free plan. May appear slow when dyno is waking up.
Currently the minimum of what twiz see as valid access token
is an object that has properties oauth_token
and oauth_token_secret
set. But it can have other parameters, like screen_name
.
The twiz-server
(here twizer) is by default an ending middleware, that is it will end the request. So call it before your error handling middlewares, if any. There are cases when twiz does not end the request, check Stream usage. Errors will be sent to the next error handling midleware with next(err)
calls and same errors will also be piped back to the browser.
If your app is not on same domain your browser will preflight request because of CORS
. So you need to use some preflight middleware before twiz-server
:
node.js:
...
app.use(yourPreflight);
app.use(twizer);
Currently you only have to set Access-Control-Allow-Origin
to your app's fqdn address.
The credentials
object in fulfileld handler can contain a lot of information. In order to ease the memory
footprint you can use parameters object (like one with skip_status
) to leave out information you don't need. Here is the list of params you can use.
Web Site workflow is very similar to that of a SPA
. You just need to put the new_window
object to args to specifiy your new popUp / window characteristics and call twizlent.finishOAuth(..)
from code in that popUp / window . Note that browser doesn't differentiate much between a popUp and a new window (new tab). Main difference is in dimentions.
browser:
// Let's say this code is in your page -> https://myApp.com
let twizlent = twizClient();
btn.addListener('onClick', function(){ // lets say we initiate oauth on click event
let args = {
server_url: 'https://myServer.com/route', // address of your node server
redirection_url: 'https://myApp.com/popUpWindow', // address of your popUp/window page
new_window:{
name: 'myPopUpWindow',
features: 'resizable=yes,height=613,width=400,left=400,top=300'
},
options:{ // twitter request options
method: 'POST',
path: 'statuses/update.json'
params: {
status: "Hooray, new tweet!"
}
}
}
twizlent.OAuth(args)
.then(function fulfilled(o){
if(o.error) // not 200OK responses (has o.error.statusCode, o.error.statusText, o.error.data)
if(o.data) // (200OK) will have data on succesfull twiz.haste(accessToken) call on server
if(o.window) // When redirection happens instead of o.redirection notification you'le have
// reference to the popUp/window and the redirection will happen from that window. // Like you would expect.
o.xhr // Always present in case you need to pull some data from response
// (like custom server headers you might be sending)
}, function rejected(err){ // Twiz errors
// err is instance of Error()
// has err.name, err.message, err.stack ...
})
})
The redirection_url
is now different then the page url from which we are making the request. Also we have new_window
where we specify the window/popUp features where redirection_url
will land . Making this more of a website use case.
The new_window
object contains two properties, name
and features
, they act the same as windowName
and windowFeatures
in window.open(). Note o.window
reference to newly opened window / popUp instead of o.redirection
.
browser(different page):
// code in https://myApp.com/popUpWindow
twizlent.finishOAuth(args); // Also can be called asap in page
.then(function fulfilled(o){
if(o.error) // not 200OK responses
if(o.data) // (200OK) will have data on succesfull twiz.continueOAuth() call on server
o.xhr // always present in case you need to pull some data from response
// (like custom server headers you might be sending)
}, function rejected(err){ // twiz errors
// err is instance of Error()
// has err.name, err.message, err.stack ...
})
What this enables is to have completely custom popUp pages but same familiar popUp, like for instance when you whould like to share something on twitter by pressing a twitter share button. Currently the downside is that users of the web site use case will get a popUp warning by browser which they have to allow before popUp apears.
Test drive the web site (custom pop-up) workflow. Running on heroku free plan. May appear slow when dyno is waking up.
node.js:
// Same code as in SPA use case;
error.name | error.message |
---|---|
redirectionUrlNotSet | You must provide a redirection_url to which users will be redirected. |
serverUrlNotSet | You must proivide server_url to which request will be sent. |
optionNotSet | Check that method and path are set. |
noCallbackFunc | You must specify a callback function. |
callbackURLnotConfirmed | Redirection(callback) url you specified wasn't confirmed by Twitter. |
noContentType | Failed to get content-type header from response. Possible CORS restrictions or header is missing. |
chunkedResponseWarning | Stream is consumed chunk by chunk in xhr.onprogress(..) callback. |
error.name | error.message |
---|---|
verifierNotFound | "oauth_verifier" string was not found in redirection(callback) url . |
tokenNotFound | "oauth_token" string was not found in redirection(callback) url . |
tokenMissmatch | Request token and token from redirection(callback) url do not match. |
requestTokenNotSet | Request token was not set. |
requestTokenNotSaved | Request token was not saved. Check that page url from which you make request match your redirection_url. |
chunkedResponseWarning | Stream is consumed chunk by chunk in xhr.onprogress(..) callback. |
noRepeat | Cannot make another request with same redirection(callback) url. |
spaWarning | Twitter authorization data not found in url. |
spaWarning
and noRepeat
are errors that have informative character and usually you dont have to pay attention to them. They happen when user loads/relods page where twizlent.finishOAuth(..)
is called on every load, imediately (which is valid). They are indications that twizlent.finishOAuth(..)
will not run. For example, spaWarning
means twizlent.finishOAuth(..)
won't run on url that doesn't contain valid twitter authorization data. noRepeat
means that you cannot make two requests with same twitter authorization data (like same request token
). Check the Stream for explanation of chunkedResponseWarning
.
Errors are ones that can happen on request
or response
streams (lower level) and they are hanled by calling next(..)
. There are no twiz errors currently for this function. Not 200OK
responses are only piped back to client and are not considered as errors.
Errors work same as twizlent.continueOAuth(..)
Any not 200OK
response are considered as an accessTokenNotVerified
error. Express' next(..)
is called and promise is rejected with the same error.
error.name | error.message |
---|---|
accessTokenNotVerified | json string |
Note that the error.message
will be a json string
taken from response
payload so you can have exact twitter error description, error code etc ...
Before making changes:
npm run start // runs `build` and `watch` scripts
When you're done:
npm run lint
Build:
npm run build
Test:
npm test
Make changes.
Lint:
npm run lint
Test:
npm run test