[patch] Workflow tweak: handle "not-state" transition specifications

[Morris]

I found myself wanting things like this in my [ticket-workflow]:

reassign = * -> * 
reassign.operations = set_owner,leave_status 
reassign.permissions = TICKET_MODIFY 

needinfo = * -> needinfo 
needinfo.name = need info 
needinfo.operations = set_owner 
needinfo.permissions = TICKET_MODIFY

…but this would create two action options for a ticket in the "needinfo" state: "reassign and leave as needinfo", and "assign to new owner and leave as needinfo" …which are of course the same thing. I wanted to be able to do this instead:

needinfo = *,!needinfo -> needinfo 
needinfo.name = need info 
needinfo.operations = set_owner 
needinfo.permissions = TICKET_MODIFY

This would indicate that the "needinfo" state was a valid next state from every other state except "needinfo". To make this work, make the following tweak to /trac/ticket/default_workflow.py (line 161):

 if oldstates == ['*'] or status in oldstates:

…becomes…

 if ('*' in oldstates or status in oldstates) and ("!%s"%status not in oldstates):

I haven't tested this extensively, and even though it should be backwards-compatible… use with extra initial scrutiny.

[ThurnerRupert]

hmm … would it be possible to extend it more "workflow like" (​http://wfmc.org)?

i.e., currently operations exist for the arrows. the other possibility wuold be to have operations on the nodes:

• automatically executed preops (before entering the node)
• automatically executed postops (before leaving the node)
• choosable actions (possible when in the node), like "set_owner"

added some brainstorming at the bottom of TracWorkflow. see also #5865.

[Remy Blank]

Interesting idea.

[jevans]

Seems like this makes a lot of sense.

Another example

close = *,!closed -> closed

Could it be scheduled a bit sooner than 1.0?

[Christian Boos]

Milestone 1.0 deleted

[Christian Boos]

Milestone triaging deleted

[Eli Carter]

#9547 was duplicate of this ticket; adding that reporter to this ticket.

[Ryan J Ollos]

* implies all the set of all states, and !<state> is only needed to subtract from a set of all states. Therefore I think we could have !<state> imply all states except <state> (i.e. * - <state>). The following patterns would then be valid:

• * -> newstate: a transition from all states
• !oldstate1,!oldstate2 -> newstate: a transition from all states except oldstate1, oldstate2
• oldstate1,oldstate2 -> newstate: a transition from oldstate1, oldstate2