BPM- ADF binding files architecture for ADF human taskflow.

Question

Hi,
I wast to know what is the architecture behind the hundreds of the file generated for the BPM adf integration for the human taskflow.
Is there any method of debugging why data coming to BPM human taskflow is not coming properly to ADF.
I appeciate any debug pointers, although these files are clearly xml files and we have no control over generation of these files but still we want to know how data been populated and how if any error occured of data corruption, how we will resolve.
REF we are facing major issues with the data corruption issues with the ADF BPM components.
Thanks
VIpin

Answer 1

Hi Vipin
This post has an interesting discussion related to your question. I am glad BPM is infact generating all these complex xml files behind the screens and taking care of so many things. Thanks to ADF framework. Yes, I totally agree, debugging may be nightmare to begin with. But with divide and rule, you can get over it. End of the Day its just pure Workflow and a complex ADF. ADF can be stripped out carefully and run all the actonis from a normal web application. If we design properly as pure SOA, we create back end services like EJBs that can be consumed from WebServices or ADF Layer. Once we have this we can integrate these Services into any BPM Process.
Re: Integrating BPM taskflows into an ADF application
Now, coming back to your questions:
1. You have a Complex BPM Process with complex BO (business objects), Arrays of BOs. I hope you have XSD for your BOs and using them only you created BusinesObjects and ProcessDataObjects. Now lets assume you have a BPM Process that uses this complex data structure. Lets take a very first Human Task initiaor who may see a Task Details page with your compelx objects. Lets say, in Task Details page, he says a header section with 10 fields. Another section with a Table of Multiple Rows and columns (arrays of Data), etc etc.
2. NOW the main question is, when this User Saves or Submits the Task, you should be doing something with that Data ? Am I right. Are you Saving it to any Database or everything is in the Payload itself. If so, I hope you have complex. If its payload itself, then I hope you have a complex strucutre to pass arrays of data through out the process.
3. Assuming all this gets Stored into backend Database. Lets not worry about retrieving it back when task goes to second user. Always Payload will store the key IDs lke dataId like OrderId, CustomerId etc and using that it can pull any data from backend and show on task details page.
4. This is where the ADF comes into picture. So even before you integrate your complex backend data with Process, first create a simple WebAppilcation. If you have EJB methods to handle this data, you create Data controls, bindings. Bindings will help generate the Screens. So you can first test your application completely here. Once all this is working, thats when you call these Services from your Process.
And beauty is with new release of BPM Suite (should be in BPM 11.5 + FP (Feature Pack)), they have a concept of Simulation. This should help you to simulate the Process without any real data and JUST focus and test on the Flow of the Process. This comes real handy for testing.
And for the data used in your process, you have ADF Tools to create stand alone ADF Application.
We have some applications like yours with very complex data structure, arrays that gets stored like in 25 tables with references going everywhere. First we built a simple data layer using EJBs and JPA layer (JPA Entities). These Entities have annotated methods to take care of most of the things. In the EJB we exposed some generic methods instead of adf generating them, due to complex data structre. Then we created data controls and bindings. Create Custom JSPX Files. This is where we ended up having our own Layout for the Data. I hope you have already seen by default, task details layout page generated is NOT usabe at all. It generates all the data in one Single Long column. We do not use that. Since we know data, we created our own Task details page. The only thing we add is the Actions bar and comments and attachment section. When you auto generate a taskdetails page for any .TASK file, you will see DataBindins created for Payload, Actions, and also separately for coments and attachmetns. Just drag these comments and attachments on to your page and define your own layout.
Re: deploying a large Oracle BPM Application with multiple UI projects (Single taskdetails for multiple UIs and Tasks)
Re: What exactly 'PAYLOAD' Means??? (Simple definition for Payload word referred in any Process)
If you see one of the old posts, there is a way to reuse just one Single TaskDetals page for Multiple different Human .TASK files. All that matters iis the top Actions Menu that are differenet for each human task like someone can Submit, someone can only Approve or Reject etc. We can club all these Actions into single actionsMenu.jsff and put this on TaskDetails page and assing to all the .task files. Menus are automatically disabled/enabled using workflow APIs. You do not need any special coding for them. Look at code snipet for any actions button that has like isAllowedAction and pass the name of the action. Workflow APIs see if that .TASK file has a Action named with that internally.
In conclusion, there is a decent clear gap between Process and the ADF Generated Task Details page.
Thanks
Ravi Jegga