Before this commit, there where two different locks when updating the restricted
data cache. A future lock, what is faster but only works in the same thread. The
other lock is in redis, it is not so fast, but also works in many threads.
The future lock was buggy, because on a second call of update_restricted_data
the same future was reused. So on the second run, the future was already done.
I don't see any way to delete. The last client would have to delete it, but there
is no way to find out which client the last one is.
* geis does not work with channels2 and never will be (it has to be python now)
* pytest
* rewrote cache system
* use username instead of pk for admin user in tests
* Second websocket channel for the projector
* Removed use of projector requirements for REST API requests.
Refactored data serializing for projector websocket connection.
* Refactor the way that the projector autoupdate get its data.
* Fixed missing assignment slide title for hidden items.
* Release all items for item list slide and list of speakers slide. Fixed error with motion workflow.
* Created CollectionElement class which helps to handle autoupdate.
- Used raw SQL for createing default projector during inital migration.
- Removed default_password and hidden agenda items from autoupdate data for some users.
- Removed old get_collection_and_id_from_url() function.
Refactored user creation and update via REST API. Used new serializer.
Cleaned up management commands, signals and imports. Moved code from 'api.py' to 'models.py'.
Changed usage of group 'Registered'. Now the users don't have to be members to gain its permissions. Used customized auth backend for this.
Added and changed some tests.
Added api for groups.
Refactored serializers now using 'id' instead of 'url'.
Rework of tornado autoupdate functionality.
Implemented extra data in SockJS messages.
