Fork me on GitHub
#datomic
<
2020-05-18
>
hanDerPeder12:05:13

newb question about datamodeling. if we have a fact like john likes jane it would make sense for likes to be an attribute with valueType ref. but what if we want to quantify how strong this affection is on a scale from 1 to 10? should likes be it's own entity now?

tvaughan14:05:37

A tuple could work here too

unbalanced16:05:44

I would probably have a concept of a "relationship". Or, if it's only one way (yikes!) then perhaps make "interest" and attribute of John. And yes, I think a ref makes sense. You might also make "interests" have multiple cardinality. And then the "interest" would have the "object of interest" and perhaps "strength of interest"

calebp14:05:24

@peder.refsnes That sounds reasonable. There’s nothing stopping you from asserting another attribute for the strength on john, but if likes is cardinality many, you end up having to parse attribute names to make associations between likes and strengths. That’s usually how I divide these cases, if the cardinality is one, it’s easier not to have to look up another entity, but if the cardinality is many I usually create new entities.

calebp15:05:21

I have just upgraded my (cloud) storage and compute stacks. I have another cloudformation stack that creates the resources for exposing my Ions through API Gateway (with Lambda proxy). Once I upgraded, the API Gateway endpoints stopped working. The immediate problem was API Gateway didn’t have permission to call the Lambdas. My working theory is that the Lambdas were recreated as part of the upgrade. They have the same names and the same ARNs, but AWS seems to see them as different resources. I can remove those resources from my template, update the stack, add them back, update again and everything works, but that adds a little down time. Wondering if anybody has figured something else out for this.