request tracker kanban extension

12 mar 2016


lately we extended the famous request tracker software with a 看板-interface:


  • jQuery UI

  • drag’n’drop operations:

    • drag ticket to other ‘state’
    • drag ticket to lane ‘user’ to have it assigned to a certain user+status
    • using the javascript callback you can basically do many cool & creative things
  • WebSocket are used to update all kanban-views in realtime after a ticket has changed

  • by overriding perl functions we added an introspection layer on sub _Set(), sub Create() and sub MergeInto() function calls

  • using a reverse proxy (apache) we made look like one webpage even though it is internally served using two webservers

  • mojolicious reuses the session cookies form RT to indirectly verify user credentials

  • using the REST 1.0 interface of RT tickets are queried from the kanban-view


we first started with twiggy + AnyEvent but it turned out that it was so much easier implementing it using mojolicious due to its async nature, feature richness and wonderful documentation.

redis & normalization

redis was used as IPC between the two webservers. one problem was that RT would often send 4 ticket updates for the same ticket within 200ms, thus we had to add a normalizer to reduce the amount of events passed to the UI.


since the reverse proxy forwards all the cookies from the browser, not only to RT but also to the mojolicious webserver, we just used that to check if a user is allowed to subscribe to the ticket-life update feed.
this was implemented by brute-forcing the REST 1.0 interface with a bogus query and the given cookies.

UI consistency

the UI requires to catch all events but a user might have used ‘suspend’ or was otherwise disconnect for a short time-period. we implemented a sequence number and thus every sent updates increases it. if the client notices that his internal sequence number differs from the supplied one it will just ‘rest’ the view by reloading all tickets.


the merge/link buttons have a auto-completion feature using jQuery’s autocomplete which is very cool. just play with it and see how easy it is to integrate into your webpages!

we also integrated a full-screen mode so that you can put the kanban on a big screen.


if you like to play with the implementation or are interested in the source code:


this was a very cool project and we think that the mojolicious implementation can be adapted/used for other rt-extensions or even other perl based projects.

we would like to thank LiHAS (sponsor) and Best Practical Solutions and #perl ppl for their help.

best wishes,
joachim & paul

article source