Adding new data results in recursive process / infinite loop

[atomicdata-dev]

Adding new data results in recursive process / infinite loop

[joepio]

Fixed in 0.20

[joepio]

This problem might come back.

I've fixed it for the default store by adding the populate_base_models function, which adds some core models (some Properties, like shortname and description. This is necessary to prevent a loop where Property X (like the shortname Property) cannot be added, because it's Property Y (like description) has to be fetched before it can be added, which in turn has property Property X (shortname) which needs to be fetched before.

I'm afraid this could happen again, for example when a user defined two Properties that have each other as properties. Not sure when this would happen, but it seems possible. A malicious actor might freeze a server by doing this.

Some approaches to preventing / limiting this problem:

Throw error instead of looping forever

What currently happens, is that the calls stack becomes too large, and the app crashes. But is there a way to know if one of the last fetches was.. itself?

Perhaps... We could have a get_property_stack vector in the store struct, which contains the URLs of the last two Subjects that were retrieved. When get_property is called and in turn needs to fetch the resource, we add one URL to it.

Limit the properties on Property instances

If we simply constrain which kinds of properties are allowed on Properties, we could achieve the same effect.
But this would get ugly pretty fast. get_property might need a custom fetch function, which parses only the required props and ignores the rest... etc. Hmm.