Newbie: Defining Metrics Programmatically Not Working

[prime-minister-of-fun]

Hi,
I'm attempting to build a scraper that returns both values and the metric definition dynamically.

If I hardcode two metrics at the global level of the script:
DEMO_EVENTS = Gauge('demo_notification', 'A value')
OTHER_EVENTS = Gauge('other_notification', 'A value', ,['server', 'region'])

def foo (a_dictionary):
 for key in a_dictionary:
  ..parse dictionary into commands..
    exec(DEMO_EVENTS.set(2))
    exec(OTHER_EVENTS.labels( server=tomato, region=us-west).set(3)
    exec(OTHER_EVENTS.labels( server=orange, region=us-east).set(3)
  return()
 foo(a_dictionary)

Everything works as expected.
If I add the below, and remove the global declaration, then python returns an error.

def bar (a_dictionary)
 for key in a_dictionary:
  .... parse dictionary into commands....
  exec("global DEMO_EVENTS")
  exec("OTHER_EVENTS = Gauge('other_notification', 'A value', ,['server', 'region'])")
  
  bar(a_dictionary)
  foo(a_dictionary)

This example returns an error on attempting .set(). The error:"name 'OTHER_EVENTS' is not defined". Python does not register the global object. You can delete the global command and the error is the same.

My gut feeling is, I'm missing something. Is dynamically generating the metric not supported?

[csmarchbanks]

Is there a reason that you are using exec rather than just creating and calling OTHER_EVENTS directly from within the loop?

I think you need to add global OTHER_EVENTS not DEMO_EVENTS in your second example.