yeah, with the same schema, but each scenario would realistically assert subsets of the schema

think about the github domain. There are many many features. Each one of those features becomes it's own

type

. with its own authorization logic

and we have hundreds of engineering teams, each owning one or various verticals

yeah so to have a test suite with a theoretical

spicedb validate

you'd have to basically write a bash script that'd take the latest schema, append it to YAML files for each test case (containing relationships and assertions) and then execute

spicedb validate

on that output

doesn't need to be bash, but a script nonetheless

right. coming up with our own structure and then just mashalling it into a single YAML may work - unless there are conflicting fixtures

if you have conflicting fixtures they can just be separate tests with the same schema

oh that's a good thought, so run "validate" with each one

yep that's what I was suggestion earlier with appending to the YAML -- your test cases would just be the asserts/relationship sections of the YAML

and then when the tests run, it'd run each test in parallel using the same schema

but not the same instance of spicedb

got it, thought you mean "take all the files, turn it into a single blob, and run one single

validate

" 👍🏻

getting late here y'all, will ruminate on all this. Thanks a ton Jimmy and Joey!

our pleasure

thanks for the feedback! i might make an issue for adding a validate command

we actually used to have one in our POC implementation

awesome, may give it a go and implement it. You can see I'm just looking for excuses to start tinkering with it 😂

haha absolutely if you want to claim it you can

is this an appropriate channel to ask about authzed?

yep

i haven't dove into spicedb but I'm assuming the relational db is being utilized like a graph db. if so, have you considered an adapter for dgraph?

it is essentially being used to store triples

and then spicedb is responsible for turning that into a graph

so even with a data store, it'll still hold the data in mem?

er

there are things that spicedb does, such as reinterpreting relationships based on schema, that would be difficult to push down through the datastore layer

makes sense.

i'm not saying any of this would be impossible on top of a graph db

far from it