birdy

birdy is a super awesome Twitter API client for Python in just a little under 400 LOC.

TL;DR

Features

Future proof dynamic API with full REST and Streaming API coverage
OAuth1 (user) and OAuth2 (app) authentication workflows
Automatic JSON decoding, JSONObject
ApiResponse, StreamResponse objects
Informative exceptions
Easily customizable through subclassing
Built on top of the excellent requests and requests-ouathlib libraries

Usage

Import client and initialize it:

from birdy.twitter import UserClient
client = UserClient(CONSUMER_KEY,
                    CONSUMER_SECRET,
                    ACCESS_TOKEN,
                    ACCESS_TOKEN_SECRET)

GET example (GET users/show):

response = client.api.users.show.get(screen_name='twitter')
response.data

POST example (POST statuses/update):

response = client.api.statuses.update.post(status='Hello @pybirdy!')

Dynamic URL example (POST statuses/destroy/:id):

response = client.api.statuses.destroy['240854986559455234'].post()

Streaming API example (Public Stream POST statuses/filter):

response = client.stream.statuses.filter.post(track='twitter')

for data in response.stream():
    print data

Why another Python Twitter API client? Aren't there enough?

The concept behind birdy is so simple and awesome that it just had to be done, and the result is a super light weight and easy to use API client, that covers the whole Twitter REST API in just a little under 400 lines of code.

To achieve this, birdy relies on established, battle tested python libraries like requests and requests-ouathlib to do the heavy lifting, but more importantly it relies on Python's dynamic nature to automatically construct API calls (no individual wrapper functions for API resources needed). This allows birdy to cover all existing Twitter API resources and any future additions, without the need to update birdy itself.

Includes full support for both OAuth1 (user) and OAuth2 (application) authentication workflows.

Finally, birdy is simple and explicit by design, besides error handling and JSON decoding it doesn't process the returned data in any way, that is left for you to handle (who'd know better what to do with it).

OK, I'm sold, but how do I use it? How does this dynamic API construction work?

The easiest way to show you is by example. Lets say you want to query Twitter for @twitter user information. The Twitter API resource for this is GET users/show (Twitter docs).

First you will need to import a client, here we import UserClient (OAuth1) and than initialize it.

from birdy.twitter import UserClient
client = UserClient(CONSUMER_KEY,
                    CONSUMER_SECRET,
                    ACCESS_TOKEN,
                    ACCESS_TOKEN_SECRET)

To query the GET /users/show API resource and pass in the parameter screen_name='twitter' you do this.

resource = client.api.users.show
response = resource.get(screen_name='twitter')

What happens here is very simple, birdy translates the users.show part after client.api into the appropriate API resource path ('users/show'). Then when you call get() on the resource, birdy constructs a full resource URL, appends any parameters passed to get() to it and makes a GET request to that URL and returns the result.

Usually the above example would be shortened to just one line like this.

response = client.api.users.show.get(screen_name='twitter')

Making a post request is similar, if for example, you would like to post a status update, this is how to do it. The API resource is POST statuses/update (Twitter docs).

response = client.api.statuses.update.post(status='Hello @pybirdy!')

Like before the part after client.api gets converted to the correct path, only this time post() is called instead of get(), so birdy makes a POST request and pass parameters (and files) as part of the request body.

For cases when dynamic values are part of the API resource URL, like when deleting a tweet at POST statuses/destroy/:id (Twitter docs), birdy supports an alternative, dictionary lookup like, syntax. For example, deleting a tweet with id '240854986559455234' looks like this.

response = client.api.statuses.destroy['240854986559455234'].post()

By now it should be clear what happens above, birdy builds the API resource path and than makes a POST request, the only difference is that part of the API path is provided like a dictionary key lookup.

Actually any call can be written in this alternative syntax, use whichever you prefer. Both syntax forms can be freely combined as in the example above. Some more examples:

response = client.api['users/show'].get(screen_name='twitter')

response = client.api['users']['show'].get(screen_name='twitter')

response = client.api['statuses/destroy']['240854986559455234'].post()

Is Streaming API supported as well?

Sure, since version 0.2, birdy comes with full support for Streaming API out of the box. Access to the Streaming API is provided by a special StreamClient.

StreamClient can't be used to obtain access tokens, but you can use UserClient to get them.

To work with the Streaming API, first import the client and initialize it.

from birdy.twitter import StreamClient
client = StreamClient(CONSUMER_KEY,
                    CONSUMER_SECRET,
                    ACCESS_TOKEN,
                    ACCESS_TOKEN_SECRET)

To access resources on the Public stream, like POST statuses/filter (Twitter docs)

resource = client.stream.statuses.filter.post(track='twitter')

For User stream resource GET user (Twitter docs)

resource = client.userstream.user.get()

And for Site stream resource GET site (Twitter docs)

resource = client.sitestream.site.get()

To access the data in the stream you iterate over resource.stream() like this

for data in resource.stream():
   print data

Great, what about authorization? How do I get my access tokens?

birdy supports both OAuth1 and OAuth2 authentication workflows by providing two different clients, a UserClient and AppClient respectively. While requests to API resources, like in above examples are the same in both clients, the workflow for obtaining access tokens is slightly different.

Before you get started, you will need to register your application with Twitter, to obtain your application's CONSUMER_KEY and CONSUMER_SECRET.

OAuth1 workflow for user authenticated requests (UserClient)

Step 1: Creating a client instance

First you need to import the UserClient and create an instance with your apps CONSUMER_KEY and CONSUMER_SECRET.

from birdy.twitter import UserClient

CONSUMER_KEY = 'YOUR_APPS_CONSUMER_KEY'
CONSUMER_SECRET = 'YOUR_APPS_CONSUMER_SECRET'
CALLBACK_URL = 'https://127.0.0.1:8000/callback'

client = UserClient(CONSUMER_KEY, CONSUMER_SECRET)

Step 2: Get request token and authorization URL

Pass callback_url only if you have a Web app, Desktop and Mobile apps do not require it.

Next you need to fetch request token from Twitter. If you are building a "Sign-in with Twitter" type application it's done like this.

token = client.get_signin_token(CALLBACK_URL)

Otherwise like this.

token = client.get_authorize_token(CALLBACK_URL)

Save token.oauth_token and token.oauth_token_secret for later user, as this are not the final token and secret.

ACCESS_TOKEN = token.oauth_token
ACCESS_TOKEN_SECRET = token.oauth_token_secret

Direct the user to Twitter authorization url obtained from token.auth_url.

Step 3: OAuth verification

If you have a Desktop or Mobile app, OAUTH_VERIFIER is the PIN code, you can skip the part about extraction.

After authorizing your application on Twitter, the user will be redirected back to the callback_url provided during client initialization in Step 1.

You will need to extract the OAUTH_VERIFIER from the URL. Most web frameworks provide an easy way of doing this or you can parse the URL yourself using urlparse module (if that is your thing).

Django and Flask examples:

#Django
OAUTH_VERIFIER = request.GET['oauth_verifier']

#Flash
OAUTH_VERIFIER = request.args.get('oauth_verifier')

Once you have the OAUTH_VERIFIER you can use it to obtain the final access token and secret. To do that you will need to create a new instance of UserClient, this time also passing in ACCESS_TOKEN and ACCESS_TOKEN_SECRET obtained in Step 2 and then fetch the tokens.

client = UserClient(CONSUMER_KEY, CONSUMER_SECRET,
                    ACCESS_TOKEN, ACCESS_TOKEN_SECRET)

token = client.get_access_token(OAUTH_VERIFIER)

Now that you have the final access token and secret you can save token.oauth_token and token.oauth_token_secret to the database for later use, also you can use the client to start making API request immediately. For example, you can retrieve the users home timeline like this.

response = client.api.statuses.home_timeline.get()
response.data

That's it you have successfully authorized the user, retrieved the tokens and can now make API calls on their behalf.

OAuth2 workflow for app authenticated requests (AppClient)

Step 1: Creating a client instance

For OAuth2 you will be using the AppClient, so first you need to import it and create an instance with your apps CONSUMER_KEY and CONSUMER_SECRET.

from birdy.twitter import AppClient

CONSUMER_KEY = 'YOUR_APPS_CONSUMER_KEY'
CONSUMER_SECRET = 'YOUR_APPS_CONSUMER_SECRET'

client = AppClient(CONSUMER_KEY, CONSUMER_SECRET)

Step 2: Getting the access token

OAuth2 workflow is much simpler compared to OAuth1, to obtain the access token you simply do this.

access_token = client.get_access_token()

That's it, you can start using the client immediately to make API request on behalf of the app. It's recommended you save the access_token for later use. You initialize the client with a saved token like this.

client = AppClient(CONSUMER_KEY, CONSUMER_SECRET, SAVED_ACCESS_TOKEN)

Keep in mind that OAuth2 authenticated requests are read-only and not all API resources are available. Check Twitter docs for more information.

Any other useful features I should know about?

Of course, birdy comes with some handy features, to ease your development, right out of the box. Lets take a look at some of the goodies.

Automatic JSON decoding

JSON data returned by the REST and Streaming API is automatically decoded to native Python objects, no extra coding necessary, start using the data right away.

JSONObject

When decoding JSON data, objects are, instead of a regular Python dictionary, converted to a JSONObject, which is a read-only dictionary subclass with attribute style access in addition to regular dictionary lookup style, for convenience. The following code produces the same result

followers_count = response.data['followers_count']

followers_count = response.data.followers_count

Don't want to use JSONObject? No problem, this behavior can be changed by means of subclassing.

ApiResponse

Calls to REST API resources return a ApiResponse, which in addition to returned data, also gives you access to response headers (useful for checking rate limits) and resource URL.

response.data           # decoded JSON data
response.resource_url   # resource URL
response.headers        # dictionary containing response HTTP headers

StreamResponse

StreamResponse is returned when calling Streaming API resources and provides the stream() method which returns an iterator used to receive JSON decoded streaming data. Like ApiResponse it also gives you access to response headers and resource URL.

response.stream()       # a generator method used to iterate over the stream

for data in response.stream():
    print data

Informative exceptions

There are 4 types of exceptions in birdy all subclasses of base BirdyException (which is never directly raised).

TwitterClientError raised for connection and access token retrieval errors
TwitterApiError raised when Twitter returns an error
TwitterAuthError raised when authentication fails, TwitterApiError subclass
TwitterRateLimitError raised when rate limit for resource is reached, TwitterApiError subclass

TwitterApiError and TwitterClientError instances (exepct for access token retrieval errors) provide a informative error description which includes the resource URL and request method used (very handy when tracking errors in logs), also available is the following:

exception.request_method    # HTTP method used to make the request (GET or POST)
exception.resource_url      # URL of the API resource called
exception.status_code       # HTTP status code returned by Twitter
exception.error_code        # error code returned by Twitter
exception.headers           # dictionary containing response HTTP headers

Customize and extend through subclassing

birdy was built with subclassing in mind, if you wish to change the way it works, all you have to do is subclass one of the clients and override some methods and you are good to go.

Subclassing a client and then using the subclass instance in your code is actually the recommended way of using birdy.

For example, if you don't wish to use JSONObject you have to override get_json_object_hook() method.

from birdy.twitter import UserClient

class MyClient(UserClient):
    @staticmethod
    def get_json_object_hook(data):
        return data

client = MyClient(...)
response = client.api.users.show.get(screen_name='twitter')

Or maybe, if you want global error handling for common errors, just override handle_response() method.

class MyClient(UserClient):
    def handle_response(self, method, response):
        try:
            response = super(MyClient, self).handle_response(method, response)
        except TwitterApiError, e:
            ...
            # Your error handling code
            ...
        return response

Another use of subclassing is configuration of requests.Session instance (docs) used to make HTTP requests, to configure it, you override the configure_oauth_session() method.

class MyClient(UserClient):
    def configure_oauth_session(self, session):
        session = super(MyClient, self).configure_oauth_session(session)
        session.proxies = {'http': 'foo.bar:3128'}
    return session

Do you accept contributions and feature requests?

Yes, both contributions (including feedback) and feature requests are welcome, the proper way in both cases is to first open an issue on GitHub and we will take if from there.

Keep in mind that I work on this project on my free time, so I might not be able to respond right way.

What does the future hold? Will there be a 1.0 release?

Next release (0.3) will focus on Python 3 support, the version after that should be 1.0 which will focus on unit tests.

After 1.0 some of the possible features are:

Cursors for REST API
Automatic reconnecting for Streaming API

Credits

birdy would not exists if not for the excellent requests and requests-oauthlib libraries and the wonderful Python programing language.

Also thanks to Twython for inspiration and python-twitter for motivation.

Question, comments, ...

If you need to contact me, you can follow me on Twitter (@sect2k) or drop me an email at mitja@inueni.com.

bhanuvrat/birdy

birdy

TL;DR

Features

Usage

Why another Python Twitter API client? Aren't there enough?

OK, I'm sold, but how do I use it? How does this dynamic API construction work?

Is Streaming API supported as well?

Great, what about authorization? How do I get my access tokens?

OAuth1 workflow for user authenticated requests (UserClient)

Step 1: Creating a client instance

Step 2: Get request token and authorization URL

Step 3: OAuth verification

OAuth2 workflow for app authenticated requests (AppClient)

Step 1: Creating a client instance

Step 2: Getting the access token

Any other useful features I should know about?

Automatic JSON decoding

JSONObject

ApiResponse

StreamResponse

Informative exceptions

Customize and extend through subclassing

Do you accept contributions and feature requests?

What does the future hold? Will there be a 1.0 release?

Credits

Question, comments, ...