late binding for sub processes

[tom.baeyens]

"Jeff" wrote:
it
| appears to me that jBPM binds the sub-process at
| super-process deployment time, not at run-time as I would
| have expected. Late binding (aka dynamic subprocesses is a
| really nice feature of a sub-process. That way if the
| subprocess changes you get the latest version.
|
| Looking at how to implement it though is not trivial, the way
| the current classes are structured. SubProcessResolver takes
| an Element, and it is not straightforward to add an Element
| to the hibernate mapping. But in any case it is something we
| should look at doing.

makes definitely sense. it didn't fit my original picture where i thought that managing sub process bindings was something that only an operator could do explicitely. that's why it isn't there in the first place.

david, managing sub process bindings is definitely something that would make sense in the admin concole.

to implement it late binding, it only takes an extra field/column for the process name. and resolve the process at execute time, rather then at parse time. an extra attribute binding="late" should be added as well to maintain backwards compatibility.

if you have a failing test case for this, please add it to the sources on HEAD. i'll make sure the test succeeds. should only require some small, local surgery to get this working.

[tom.baeyens]

http://jira.jboss.com/jira/browse/JBPM-733

[jeffdelong]

I did a bit of "local surgery, and was able to get late binding to work. The SubProcessResolver interface requires the subProcessElement from the jpdl. Unfortunately I was not able to persist this XML Element as part of ProcessState (this may just be a limitation of my Hibernate experience having never worked with HIbernate and XML). Instead I created subProcessDefinitionName and subProcessVersion attributes on the ProcessState (both Strings), and then added a method to ProcessState to lookup the processdefinition (basically a cut and pase of DBSubProcessResolver). It would be better if either the subProcessElement was persisted, or if the SubProcessResolver interface took the two string values for arguments instead of the XML.

I do have a TestCase created. It uses process definitions from PAR files, however I can tweak it to include these in the test case source.

[spidek]

Hello, it will be useful for me and probably to other people if this attribute will be accessible from gpd.

I've opened JIRa wish for it and attached patch:
https://jira.jboss.org/jira/browse/GPD-232

[jbarrez]

I agree, this is something that should be in the GPD/documentation