oauth-dropins

About

This is a collection of drop-in Google App Engine request handlers for the initial OAuth client flows for many popular sites, including Blogger, Disqus, Dropbox, Facebook, Flickr, Google+, IndieAuth, Instagram, Medium, Tumblr, Twitter, and WordPress.com.

Requires either the App Engine Python SDK or the Google Cloud SDK (aka gcloud) with the gcloud-appengine-python and gcloud-appengine-python-extras components. All other dependencies are handled by pip and enumerated in requirements.txt. We recommend that you install with pip in a virtualenv. App Engine details here.

If you clone the repo directly or want to contribute, see Development for setup instructions.

This software is released into the public domain. See LICENSE for details.

Quick start

Here’s a full example of using the Facebook drop-in.

  1. Make sure you have either the App Engine Python SDK version 1.9.15 or later (for vendor support) or the Google Cloud SDK (aka gcloud) installed and on your $PYTHONPATH, e.g. export PYTHONPATH=$PYTHONPATH:/usr/local/google_appengine. oauth-dropins’s setup.py file needs it during installation.

  2. Install oauth-dropins into a virtualenv somewhere your App Engine project’s directory, e.g. local/:

    shell  source local/bin/activate  pip install oauth-dropins

  3. Add this to the appengine_config.py file in your project’s root directory (background):

    py  from google.appengine.ext import vendor  vendor.add('local')  from oauth_dropins.appengine_config import *

  4. Put your Facebook application’s ID and secret in two plain text files in your app’s root directory, facebook_app_id and facebook_app_secret. (If you use git, you’ll probably also want to add them to your .gitignore.)

  5. Create a facebook_oauth.py file with these contents:

    `python from oauth_dropins import facebook import webapp2

    application = webapp2.WSGIApplication([ (‘/facebook/start_oauth’, facebook.StartHandler.to(‘/facebook/oauth_callback’)), (‘/facebook/oauth_callback’, facebook.CallbackHandler.to(‘/next’))] ```

  6. Add these lines to app.yaml:

    yaml  - url: /facebook/(start_oauth|oauth_callback)    script: facebook_oauth.application    secure: always

Voila! Send your users to /facebook/start_oauth when you want them to connect their Facebook account to your app, and when they’re done, they’ll be redirected to /next?access_token=... in your app.

All of the sites provide the same API. To use a different one, just import the site module you want and follow the same steps. The filenames for app keys and secrets also differ by site; appengine_config.py has the full list.

Usage details

There are three main parts to an OAuth drop-in: the initial redirect to the site itself, the redirect back to your app after the user approves or declines the request, and the datastore entity that stores the user’s OAuth credentials and helps you use them. These are implemented by StartHandler, CallbackHandler, and auth entities, respectively.

The request handlers are full WSGI applications and may be used in any Python web framework that supports WSGI (PEP 333). Internally, they’re implemented with webapp2.

StartHandler

This HTTP request handler class redirects you to an OAuth-enabled site so it can ask the user to grant your app permission. It has two useful methods:

  • to(callback_path, scopes=None) is a factory method that returns a request handler class you can use in a WSGI application. The argument should be the path mapped to CallbackHandler in your application. This also usually needs to match the callback URL in your app’s configuration on the destination site.

    If you want to add OAuth scopes beyond the default one(s) needed for login, you can pass them to the scopes kwarg as a string or sequence of strings, or include them in the scopes query parameter in the POST request body. This is currently supported with Facebook, Google+, Blogger, and Instagram.

    Some of the sites that use OAuth 1 support alternatives. For Twitter, StartHandler.to takes an additional access_type kwarg that may be read or write. It’s passed through to Twitter x_auth_access_type. For Flickr, the start handler accepts a perms POST query parameter that may be read, write or delete; it’s passed through to Flickr unchanged. (Flickr claims it’s optional, but sometimes breaks if it’s not provided.)

  • redirect_url(state=None) returns the URL to redirect to at the destination site to initiate the OAuth flow. StartHandler will redirect here automatically if it’s used in a WSGI application, but you can also instantiate it and call this manually if you want to control that redirect yourself:

class MyHandler(webapp2.RequestHandler):
  def get(self):
    ...
    handler_cls = facebook.StartHandler.to('/facebook/oauth_callback')
    handler = handler_cls(self.request, self.response)
    self.redirect(handler.redirect_url())

However, this is not currently supported for Google+ and Blogger. Hopefully that will be fixed in the future.

CallbackHandler

This class handles the HTTP redirect back to your app after the user has granted or declined permission. It also has two useful methods:

  • to(callback_path) is a factory method that returns a request handler class you can use in a WSGI application, similar to StartHandler. The callback path is the path in your app that users should be redirected to after the OAuth flow is complete. It will include a state query parameter with the value provided by the StartHandler. It will also include an OAuth token in its query parameters, either access_token for OAuth 2.0 or access_token_key and access_token_secret for OAuth 1.1. It will also include an auth_entity query parameter with the string key of an auth entity that has more data (and functionality) for the authenticated user. If the user declined the OAuth authorization request, the only query parameter besides state will be declined=true.

  • finish(auth_entity, state=None) is run in the initial callback request after the OAuth response has been processed. auth_entity is the newly created auth entity for this connection, or None if the user declined the OAuth authorization request.

    By default, finish redirects to the path you specified in to(), but you can subclass CallbackHandler and override it to run your own code inside the OAuth callback instead of redirecting:

class MyCallbackHandler(facebook.CallbackHandler):
  def finish(self, auth_entity, state=None):
    self.response.write('Hi %s, thanks for connecting your %s account.' %
        (auth_entity.user_display_name(), auth_entity.site_name()))

However, this is not currently supported for Google+ and Blogger. Hopefully that will be fixed in the future.

Auth entities

Each site defines an App Engine datastore ndb.Model class that stores each user’s OAuth credentials and other useful information, like their name and profile URL. The class name is of the form SiteAuth, e.g. FacebookAuth. Here are the useful methods:

  • site_name() returns the human-readable string name of the site, e.g. “Facebook”.
  • user_display_name() returns a human-readable string name for the user, e.g. “Ryan Barrett”. This is usually their first name, full name, or username.
  • access_token() returns the OAuth access token. For OAuth 2 sites, this is a single string. For OAuth 1.1 sites (currently just Twitter, Tumblr, and Flickr), this is a (string key, string secret) tuple.

The following methods are optional. Auth entity classes usually implement at least one of them, but not all.

  • api() returns a site-specific API object. This is usually a third party library dedicated to the site, e.g. Tweepy or python-instagram. See the site class’s docstring for details.
  • urlopen(data=None, timeout=None) wraps urllib2.urlopen() and adds the OAuth credentials to the request. Use this for making direct HTTP request to a site’s REST API. Some sites may provide get() instead, which wraps requests.get().
  • http() returns an httplib2.Http instance that adds the OAuth credentials to requests.

Troubleshooting/FAQ

  1. If you get this error:

    bash: ./bin/easy_install: ...bad interpreter: No such file or directory

You’ve probably hit this open virtualenv bug (fixed but not merged): virtualenv doesn’t support paths with spaces.

The easy fix is to recreate the virtualenv in a path without spaces. If you can’t do that, then after creating the virtualenv, but before activating it, edit the activate, easy_install and pip files in local/bin/ to escape any spaces in the path.

For example, in activate, VIRTUAL_ENV=".../has space/local" becomes VIRTUAL_ENV=".../has\ space/local", and in pip and easy_install the first line changes from #!".../has space/local/bin/python" to #!".../has\ space/local/bin/python".

This should get virtualenv to install in the right place. If you do this wrong at first, you’ll have installs in /usr/local/lib/python2.7/site-packages that you need to delete, since they’ll prevent virtualenv from installing into the local site-packages.

  1. If you’re using Twitter, and import requests or something similar fails with:

    ImportError: cannot import name certs

    or you see an exception like:

    File ".../site-packages/tweepy/auth.py", line 68, in _get_request_token    raise TweepError(e)  TweepError: must be _socket.socket, not socket

    …you need to configure App Engine’s SSL. Add this to your app.yaml:

    libraries:  - name: ssl    version: latest

If you use dev_appserver, you’ll also need to apply this workaround (more background). Annoying, I know.

  1. If you see errors importing or using tweepy, it may be because six.py isn’t installed. Try pip install six manually. tweepy does include six in its dependencies, so this shouldn’t be necessary. Please let us know if it happens to you so we can debug!

  2. If you get an error like this:

    File "oauth_dropins/webutil/test/__init__.py", line 5, in <module>      import dev_appserver  ImportError: No module named dev_appserver  ...  InstallationError: Command python setup.py egg_info failed with error code 1 in /home/singpolyma/src/bridgy/src/oauth-dropins-master

…you either don’t have /usr/local/google_appengine in your PYTHONPATH, or you have it as a relative directory. pip requires fully qualified directories.

  1. If you get an error like this:

    Running setup.py develop for gdata  ...  error: option --home not recognized  ...  InstallationError: Command /usr/bin/python -c "import setuptools, tokenize; __file__='/home/singpolyma/src/bridgy/src/gdata/setup.py'; exec(compile(getattr(tokenize, 'open', open)(__file__).read().replace('\r\n', '\n'), __file__, 'exec'))" develop --no-deps --home=/tmp/tmprBISz_ failed with error code 1 in .../src/gdata

…you may be hitting Pip bug 1833. Are you passing -t to pip install? Use the virtualenv instead, it’s your friend. If you really want -t, try removing the -e from the lines in requirements.freeze.txt that have it.

Changelog

1.10 - 2017-12-10

Mostly just internal changes to webutil to support granary v1.10.

1.9 - 2017-10-24

Mostly just internal changes to webutil to support granary v1.9. * Flickr: * Handle punctuation in error messages.

1.8 - 2017-08-29

  • Facebook:
    • Upgrade Graph API from v2.6 to v2.10.
  • Flickr:
    • Fix broken FlickrAuth.urlopen() method.
  • Medium:
    • Bug fix for Medium OAuth callback error handling.
  • IndieAuth:
    • Store authorization endpoint in state instead of rediscovering it from me parameter, which is going away.

1.7 - 2017-02-27

  • Updates to bundled webutil library, notably WideUnicode class.

1.6 - 2016-11-21

1.5 - 2016-08-25

1.4 - 2016-06-27

  • Upgrade Facebook API from v2.2 to v2.6.

1.3 - 2016-04-07

  • Add IndieAuth.
  • More consistent logging of HTTP requests.
  • Set up Coveralls.

1.2 - 2016-01-11

  • Flickr:
    • Add upload method.
    • Improve error handling and logging.
  • Bug fixes and cleanup for constructing scope strings.
  • Add developer setup and troubleshooting docs.
  • Set up CircleCI.

1.1 - 2015-09-06

  • Flickr: split out flickr_auth.py file.
  • Add a number of utility functions to webutil.

1.0 - 2015-06-27

  • Initial PyPi release.

Development

You’ll need the App Engine Python SDK version 1.9.15 or later (for vendor support) or the Google Cloud SDK (aka gcloud) with the gcloud-appengine-python and gcloud-appengine-python-extras components. Add them to your $PYTHONPATH, e.g. export PYTHONPATH=$PYTHONPATH:/usr/local/google_appengine, and then run:

git submodule init
git submodule update
virtualenv local
source local/bin/activate
pip install -r requirements.txt

# We install gdata in source mode, and App Engine doesn't follow .egg-link
# files, so add a symlink to it.
ln -s ../../../src/gdata/src/gdata local/lib/python2.7/site-packages/gdata
ln -s ../../../src/gdata/src/atom local/lib/python2.7/site-packages/atom

python setup.py test

Most dependencies are clean, but we’ve made patches to gdata-python-client below that we haven’t (yet) tried to push upstream. If we ever switch its submodule repo for, make sure the patches are included!

To deploy:

python -m unittest discover && git push && gcloud -q app deploy oauth-dropins *.yaml

The docs are built with Sphinx, including apidoc, autodoc, and napoleon. Configuration is in docs/conf.py To build them, first install Sphinx with pip install sphinx. (You may want to do this outside your virtualenv; if so, you’ll need to reconfigure it to see system packages with virtualenv --system-site-packages local.) Then, run docs/build.sh.

To convert README.md to README.rst for PyPI or index.rst for Sphinx:

TODO

  • Google+ and Blogger need some love:
    • handle declines
    • allow overriding CallbackHandler.finish()
    • support StartHandler.redirect_url()
    • allow more than one CallbackHandler per app
  • clean up app key/secret file handling. (standardize file names? put them in a subdir?)
  • implement CSRF protection for all sites
  • implement Blogger’s v3 API