Suggestion: functionality to count contained cuds objects

[yoavnash]

It could be useful (e.g., for statistics) to have a functionality that counts contained cuds objects inside of another cuds object or wrapper.

Implementation idea: maintain a counter in each cuds object that specifies how many nested objects are there.

[ahashibon]

good point @yoavnash for keeping the system clean, and compact in API, we can simply add a property (relation) managed by the system and accessed by get and one can also then iterate on it (e.g., on all the counts of all elements in a Data Space).

#Note:

• by system manged it is meant a "read only" (for users and admins) property. We had something like this in SimPhoNy 1.0
• ontologically, this relation would inherit (also, i.e., in multi inheritance sense) from the "ReadOnlyProperty" class
• this way we preserve the semantic meaning fully, as opposed to adding an ad hoc syntactic property outside the ontology.

[urbanmatthias]

Just for clarification:

• Let A has_part B and B has_part C. Then the number of contained objects of A as you propose it should be two, correct?
• Should this be available for every cuds object?
• With your proposed solution (@yoavnash) one must update the counter in potentially many cuds object in a deep tree structure, correct?
• I would rather prefer something like len(cuds_object) over a system managed relationship that can be queried by get, because:
  • it is leaner
  • you cannot encode "this relationship counts the contained cuds objects" ontologically (at least not with the current formats owl and yml)
  • there is no ReadOnlyProperty so far
  • I like to disagree with adham ;)

[urbanmatthias]

How would you handle the case where another user adds cuds objects to a database. How would the counter of your objects be updated?

[urbanmatthias]

How would you handle the case where another user adds cuds objects to a database. How would the counter of your objects be updated?

[yoavnash]

Maybe len(cuds_object) should simply be then a call for a recursive function?

[urbanmatthias]

The user wouldn't expect that a recursive function (that potentially takes a long time) is called with len()

[yoavnash]

It's either recursion when calling len() or when calling add() (to update the counters). Maybe we shouldn't support it then? or create a utility function that makes it clear that it uses recursion.

[urbanmatthias]

Yeah, maybe a utility function is the way to go

[yoavnash]

Maybe we should make a module called rec or recursive to mark all the recursive functionalities explicitly as recursive. As an illustration:

from osp.core.utils.rec import pretty_print, count