This is a simple example program to show how to build a web application that access the Twitter API in Clojure.
It provides a single page web app, using Javascript (well, actually coffeescript) to call JSON apis on the server - this keeps the server quite simple, a great fit for clojure.
Note that I'm a relative clojure newbie - I'm learning as I go, but don't assume this is the best idiomatic clojure code!
The application uses the following libraries and programs:
- Clojure bits:
- Compojure - a great simple Clojure web framework
- Ring - the web layer underlying Compojure
- Leiningen - a popular Clojure build tool
- lein-ring - to load the app dynamically in development mode
- clojure-twitter - for twitter API stuff
- clj-oauth - for the underlying authentication used by twitter
- midje - for testing
- a few others as needed - for other clojure libraries, look in the main 'project.clj' file
- Front end bits:
- Sass for nicer CSS syntax
- coffeescript for a nicer JavaScript syntax
- JQuery to talk to browsers
- handlebars.js to build client-side views
- Modernizr, Underscore, and a few others to do various bits and pieces
Some of this is based on code and ideas from my loosely coupled web app skeleton
$ lein deps
$ vim stackato.yml
# Set your own TWITTER_KEY and TWITTER_SECRET
# (read the section 'Usage' below first)
# Also, set the CALLBACK_HOST if you are deploying to different URL
$ stackato push -n
- 6 Jun 2011 - added initial midje tests - also load config from a (memoized) function so tests don't need to have config set up
- 4 Jun 2011 - considerably simplified startup - as Heroku may restart the app on demand, there's little value in using atoms to store config. Instead, config now stored in environment variables - needs more setup, but probably a simpler example anyway.
Currently this can't be run standalone - it can only be run via leiningen.
- install [Leiningen]
- run "lein deps" to install dependencies - this will download a pile of libraries, and may take some time
This app needs a valid Twitter application to run. This is pretty easy to configure yourself:
- go to https://dev.twitter.com/apps
- click "register a new app"
- enter any details you want
- make sure the application website, and the callback url, match the domain you plan to use.
- for development, this should be "127.0.0.1" - you can't use "localhost" for some reason, Twitter block it
- so for example, your callback url and application website should be "http://127.0.0.1/" unless you plan to host on a real host name. The port doesn't matter, as far as I can tell.
- Note the "key" and "secret" values on the twitter app page
- Set up your environment variables as follows: export TWITTER_KEY="xxxxxx" export TWITTER_SECRET="xxxxxxxx" export CALLBACK_HOST="127.0.0.1" export CALLBACK_PORT=3000
- If you are deploying to Heroku, you can set these via
heroku config:add
- see http://devcenter.heroku.com/articles/config-vars
Note it's important to specify 127.0.0.1
not localhost
- Twitter don't let you set up an app on localhost, and the application gets the host name from the request
- run "lein ring server" - this will build the app, deploy it to a local server on (probably) port 3000, and then display the home page in a browser.
- alternateively, run "lein run -m twitter-example.core" which will run in a local jetty server on port 8080, a somewhat more production-like environment
The Twitter oauth dance is confusing if you haven't met it before. Twitter have a good description at [https://dev.twitter.com/pages/auth]
It might help however to describe all the steps in this app that happen before a Twitter API is called:
- At startup, the server (core.clj) reads configuration from the environment into a map called 'config'
- Then the application starts, and users can load the front page - from here on, most activity is driven from the client side
- When a user first loads the "/" page, the index.html page is loaded
- this includes view templates - they aren't visible to the user, but loaded by the javascript for rendering page snippets
- The code in twitter-example.coffee is then loaded (as javascript) - this contains the client-side application logic
- On page load, the TwitterExample constructor is called, which sets up view logic and calls
check_status()
check_status()
makes an ajax call to "/auth/check_status.json" on the server- The server URLs are usually handled by the
defroutes
matchers half way down core.clj - but in this case, the user has not yet authorized, so the middleware functionwith-oauth-check
kicks in
- The server URLs are usually handled by the
with-oauth-check
determines there is no auth information in the session, so it starts the oauth interaction with twitter- It makes a call to Twitter to get a request token, based on the application credentials and our local URLs. The request token is stored in the user's session
- It also determines the appropriate URL on twitter to which the user should be redirected
- It then returns a HTTP 401 error, with a JSON body containing a message, and an authURL value with the redirect URL
- The client catches the 401 error and calls
auth-error()
which renders a HTML view "noauth", showing the user the message "Please authorize in Twitter", and a link to the authURL returned from the server - When the user clicks on the "authorize" link, they are taken to Twitter, where they can authorize - in which case they are redirected to the app, with the URL "/twitter-oauth-response" with url parameters containing extra data to confirm the user is authorized
- The server gets the "/twitter-oauth-response" request (it's actually stored in 'oauth-response-path', matching the line
(GET oauth-response-path [oauth_token oauth_verifier]
in core.clj - It fetches the request token from the session, and makes another API call to Twitter to convert the request token and verifier into an access token
- The access token is stored in the session, so future calls to "/auth" URLs won't result in a 401 error. (note the request token is no longer needed, and should probably be thrown away at this stage)
- The server then redirects the user back to "/", the starting page of the application
- a more friendly application could try to store what the user originally requested, and after geting authorization, complete that request, but it's more than this little app can do
- The client reloads the page again, as before, which calls
check_status()
which calls "/auth/check_status.json" - This time the user has an access token in the session, so the
with-oauth-check
middleware does nothing - The
defroutes
matcher for "/auth/check_status.json" matches the request, and returns a simple payload consisting of the user's twitter name - The client (in the
check_status()
function) renders the user's name, then callsfetch_tweets()
- This results in another ajax call, this time to "/auth/tweets.json"
- The server matches "/auth/tweets.json" and makes a Twitter API call to retrieve the user's home timeline
- The JSON response from Twitter is filtered to remove unneeded information, and then returned to the client
- The client renders the resulting twitter messages.
Phew! This may look complex, but it's a pretty common workflow for applications trying to do OAuth. It's actually somewhat simpler in clojure/coffeescript than other languages I've done this in!
Testing is somewhat rudimentary - there are a few tests included, but they only test a few simple things.
Testing uses the wonderful midje library. To run tests, run
lein midje
Note that midje provides quite powerful mocking, using the 'provided' command - none of these tests need to actually call Twitter, the twitter API is mocked out.
I'm still a bit unsure how far to TDD clojure - it's obviously possible, but there's a lot of functionality that probably doesn't gain much from (unit) tests. Unless I define the whole namespace as a unit, and only mock external interfaces. Hmm...
CoffeeScript and Sass are precompilers - they take .coffee and .scss files, and build javascript and CSS respectively.
You can work with the resulting javascript and CSS files if you want - they are in resources/public/javascript and resources/public/stylesheets. But if you want to make non-trivial changes, you'll want to be able to compile the original sources.
You need to install coffeescript - this requires node.js, and explaining how is beyond the scope of this readme. See [http://jashkenas.github.com/coffee-script/] for more.
You also need to install sass, which is a ruby library - again, beyond the scope of this readme! See [http://sass-lang.com/] for more.
You can manually compile the scss/coffee files with: ./precompile.sh Or you can run a background script that compiles files whenever they change with: ./watch_all.sh These only work on Unix/Mac systems - Windows users will have to work out a Windows equivalent.
See separate TODO.markdown file
Heroku deployment was a dream - the only hiccup I had (apart from a bunch of work to remove file-based configuration) was some extra dependencies.
For some reason I had to add [ring/ring-core "0.3.8"] [ring/ring-jetty-adapter "0.3.8"] to the dependencies in project.clj - leiningen standalone seems to handle these automagically, but on heroku I had to be explicit.
Other than that I followed the example at https://gist.github.com/1001206
gem install heroku
heroku keys:add
heroku create <my app name> --stack cedar
git push heroku master
It's that simple! heroku ps
lists running processes, heroku logs
shows the log. You can even run heroku run lein repl
and get a REPL!
Note I've confirmed with Heroku, despite some slight lack of clarity on their site, this is free for the basic level of service (750 dynamo-hours per month).
Copyright (c) 2011 Kornelis Sietsma
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.