Mixing Heterogeneity and Relationships

Data Services
1

I want to use heterogeneity and relationships in my application to create different views on my data.

Lets say I have the following objects:

  • Folders
  • Items

and the following rules apply:

  • A Folder can contain (sub)folders or items.
  • An Item must be in a Folder

Is there any advice on getting the best performance out of my application?
At the moment I’m modeling the data graph so that I can create the UI views that I want. I want to be able to:

  • View all Folders and Items in an Overview
  • View all Folders and Items in a Tree

This has led to the following model graph

grafik
grafik786×1160 115 KB

However, I’m concerned that when I load the Tree that the performance will suffer as:

  1. I load the Folders at root level
  2. I load the linked Folders and Items
  3. To load the linked Folders and Items I need to search the two subtypes, Folder_DM and Item_DM
  4. Item_DM is a subtype of Folder_DM so could be loaded / searched twice

Is this all taken care of for me through Heterogeneity Resolution in Data Services or do I need to change my modeling?

2

Hi @malcolm-silver-ice, from Heterogeneity perspective, if FoldersAndItems_DM is supertype of Folder_DM and Item_DM, then I think the resolution is cover this case completely. Hence the relationship here might be self-referenced relationship, in QueryAPI we provide maxDepth parameter when querying nested relationship structure like this then the performance is guaranteed. I think this case has a similar example in A12 showcase please have a look at this <INTERNAL_LINK>

3

Thanks @loc-fleet-crest,

So any performance issue is more likely linked to the unlimited depth of the self-referencing relationship. Right?

4

Hello @loc-fleet-crest,

could you please answer @malcolm-silver-ice additional question in his comment?

All the best
your Discourse Moderation Team

5

yes, it is. The deeper nested links the more impacted performance, that’s why we introduce maxDepth parameter, to prevent unlimited depth.