Creating Pentaho Solutions, Page 15.

[fouedex]

Hi all,

I read that a visual process editor was planned, as a new feature in the eclipse plugin workbench.

Does this include the possibility to plug graphically component output into inputs of other components ?

I know it is possible to aggregate several actions into an action sequence, is it possible to plug an action sequence into one other ?

If i understand the philiosophy of this process modeling, we should be able to define atomic actions (with templating), gather them into an action sequence, and then build the process model, which the scheduler will instanciate.
But, is there a way to give existence to a process, without the scheduler, for instance run a process every time a predefined condition occurs, or when an input becomes available (from another output component). Is this what triggers are made for ?

Lbnl, page 15 of the "Creating Pentaho Solutions" guide, i read :

There are two implicit parameters instance-id and solution-id that are always available and do not need to be specified as inputs or outputs. They are the... well I'm sure you guess what they are.
Not precisely ...

[dmoran]

fouedex,

Yes, there is a plan to have a visual process editor. The purpose is to build workflows that graphically plug Action Sequences together. We intend to use a workflow engine (currently shark) to manage the flow and execution. The intent is to have a visual editor similar to JaWE, with knowledge of Action Sequences, generate the XPDL document that defines the workflow.

We did not want to recreate a workflow engine with the Action Sequence but saw the need for stringing together several actions for simple tasks. Over time we have added more capability like mapping inputs and looping but so far have resisted calling Action Sequences from other Action Sequences and having conditional logic.

One mechanism for plugging one Action Sequence into another is via the workflow engine. There is an example of this in the Bursting Samples. We do not have a component that directly calls one Action Sequence from another – so some external method is required – Scheduler, custom code, application trigger. An Action Sequences is URL addressable and can be initiated from any application that can call a URL.

The instance-id is the instance id (a unique identifier) of the current runtime and the solution-id is the id of the solution that the Action Sequence is a part of.

We have been concentrating on the reporting features and will be moving on to adding workflow features in the upcoming months – see the roadmap. We are very open to suggestions and use cases for real world requirements.

Please give us your input.

Doug

Post edited by: dmoran, at: 01/05/2006 11:31