Workflow variables data types

Question

Is there any documentation regarding how the workflow variables are being stored?

e.g. two workflows use similar form property types (term), but one is stores as an ArrayList, while the other is stored in a String
Class of impactedDST is class java.lang.String
Class of dataElement is class java.util.ArrayList

arthurburkhardt · Answer

There are two things you can do:

print log lines related to the script
Introspect task form data

In any case, you need to programmatically start the workflow:

payload = {
  "workflowDefinitionId": "3fa85f64-5717-4562-b3fc-2c963f66afa6",
  "businessItemIds": [
    "3fa85f64-5717-4562-b3fc-2c963f66afa6"
  ],
  "businessItemType": "ASSET",
  "formProperties": {
    "additionalProp1": "string",
    "additionalProp2": "string",
    "additionalProp3": "string"
  },
  "guestUserId": "3fa85f64-5717-4562-b3fc-2c963f66afa6",
  "sendNotification": true
}

r = c.post('workflowInstances', json=payload)

Print log lines related to the script

You can poll for new log lines continuously. There are some issues with the API though:

It is a private API => No documentation
There are no checkpoints to guarantee you read every line. You can approximate by checking timestamps and downloading the whole file when you see you missed lines.
There’s a ton of garbage because of the stacktraces whenever an error occurs. This is absolutely useless for workflow debugging, so you might consider just disregarding lines starting with a tab.

from collibra import Console
import time

co = Console()

try:
    max_timestamp = ''
    while true:
        lines = co.get(f'log/service/85a195a8-b519-483e-9ba0-823d6f896d30/content?filename=dgc.log&lines=200&_=1667911772639').json()
        for line in lines:
            if line[:23] > max_timestamp:
                print(line)
        max_timestamp = line[:23]
        time.sleep(3) #poll every 3 seconds
        
except KeyboardInterrupt:
    print("Stopped polling logs")

Introspect workflow task form data

You can query tasks and update them with API calls:

Using APIs, it would even be possible to build test scripts and automatically test workflows whenever new deployment occur.

rostislav · Answer

Following this discussion, we’ve added a Class of stored value column to the Form property types documentation .

arthurburkhardt · Answer

Yay, I completed the remaining steps in 30min this morning. That’s a nice way to start the week
I was afraid I’d have to compile an extension for Eclipse, but it turns out there’s a simple Run External tools command, that allows to easily configure with project variables.

So now, indeed, I can just left click to package + upload.
Last step: add the execute (with optional passing messages if there’s a user task) + logging.

arthurburkhardt · Answer

BTW, @alvin.useree.lloydsbanking.com , after our discussion, I realized I needed to accelerate the feedback loop beetween IDE and workflow engine, I’m halfway through it, shouldn’t be too long to finish it. In the IDE, my project structure is like this wfProcessId/ ├── wfProcessId.bpmn ├── wfProcessId-scriptTaskId1.groovy └── wfProcessId-scriptTaskId2.groovy wfProcessId2/ ... Python scripts to either download workflow definitions from a Collibra instance or extract from a .cma archive and store them with that convention. I absolutely HATE copy pasting code from groovy editor to bpmn workflow (too error prone), so I instead created a python script to package the .bpmn workflow definition and upload it to a collibra instance. It just opens the .bpmn file, looks up for script tasks and replaces the script element in the XML with the content of the corresponding groovy script files. Added bonus: I can add custom logic, for example to inject supporting functions, or remove parts of the scripts (I declare beans and classes to inspect the objects, which is useful while developing the script, but should not be deployed). e.g. Now for the grand finale: I’ll look into how to add an entry into Eclipse context menu, so that I can right click the workflow folder, package and upload to Collibra. Right click to execute the workflow and see log results from the console directly from Eclipse So, one click to package + deploy, then one click to execute and get results (without useless stacktrace) directly in Eclipse! Are you interested in any of that? Do you have also good practices to recommend?

Former User · Answer

Hey mate,

For that specific case, I’m going to go out on a limb and say that it depends on whether you’ve set the multiValue form value as true, false or not set it at all

If set true, then the returned class is an Array of strings for each UUID if set to false or not set then it returns the UUID in string format.

Outside of that example I 100% agree with this point, I haven’t been able to find documentation on the returned classes so I generally run a getClass() and log it to the console to remind myself of the class which is a seriously inefficient way to this!

Also there a some hidden gotchas that could easily be mitigated with this documentation. For example, if you use a textarea, you can’t use the outputs of the textarea for anything unless you cast it as a string first, when I train people on workflows, these are the types of things I package as “just remember this as a rule” which frankly isn’t the best look!

If you find the documentation, give me a nudge mate!

Cheers!

Alvin Useree

1	Grant Rollerson SUNCORP CORPORATE SERVICES PTY LTD
2	Tom Friesen Algonquin College
3	SB S Benson Plus3 IT
4	S stefan_busch
5	Brian Lloyd USA TODAY Co.
6	D daverichard