History information

[heiko.braun]

I started working on the reporting for jbpm4 and realized several problems so far:


instance history doesn't populate the end date & duration.
And i am not sure how canceled instances should appear here.

activity history doesn't populate the task assignee.

process definition / deployment history doesn't exists at all

[tom.baeyens]

i'll check the first 2 asap.

process definition / deployment history doesn't need history info. for those, just use the repository service/schema directly. or am i missing something?

[tom.baeyens]

i've created https://jira.jboss.org/jira/browse/JBPM-2261 so that i don't forget

[heiko.braun]

How to we deal with canceled process instances or tasks? IMO they should show up in the history as well, with a proper state, i.e. 'canceled'.
If we only keep successfully ended executions in the history then we loose one important correlation aspect: success/failure ratio.

[camunda]

+1!

And they should influence statistics of the already finished tasks in this instance as well in my opinion

[tom.baeyens]

i agree cancelled process instances / tasks should also remain in the history.

there is a related aspect that I wanted to bring up: I want to split the execution state from the from the user provided end state.

currently users can provide a custom end state value. i believe the state of a (history)process instance should always be put to STATE_ENDED. and that we should have a second field indicating the user defined end state. it's the user defined end state that could be cancelled, aborted, error, failure and so on.

then i believe users can distinct better between ended processes and active processes in their queries.

[tom.baeyens]

Heiko, as for your original request. There is a ProcessInstanceHistoryTest that shows that the start, end dates and duration are filled. If that still fails for you, can you give an indication of how exactly you see this problem ?

For the HistoryTaskInstance, i see the problem. The API doesn't allow for accessing that info. Hibernate does know about the actual classes. So it is already possible to write reports for it if you use hibernate queries. Please comment on the jira issue if you need API access.

for deployment, i don't see the need for history information. do you ? Also please comment on jira issue

https://jira.jboss.org/jira/browse/JBPM-2261

[camunda]

Tom wrote:

I want to split the execution state from the from the user provided end state.

currently users can provide a custom end state value. i believe the state of a (history)process instance should always be put to STATE_ENDED. and that we should have a second field indicating the user defined end state. it's the user defined end state that could be cancelled, aborted, error, failure and so on.

I agree and spontaneously I don't see reasons against it.

[kukeltje]

I agree as well and even non-spontaneously (meaning I gave it some thought :-P) see no reasons against it.

[tom.baeyens]

guys, i realized that we have the end dates in the history as well. so having the end state set to 'ended' would be duplicating information that is already easy queryable.

select ..
from ...
where histProcInst.state = 'ended'

==

select ..
from ...
where histProcInst.end is not null

so i'll keep a single state in the history process instance entity.

[heiko.braun]

right, user defined end states are implicitly denoted through the end activity already. so we won't needed that either.