I'm working on an app that I plan to deploy to an Ion, and I do have the presumption that it will be fast. I'm curious why you ask, do you expect it to not be fast?

I think... The Entity api is given an id/key give me the value. Ions in aws run on top of aws dynamodb which is a key value store. I assume the id lookup would be constant time. But im probably missing more then half the story. In general, The faster you make it, the more space you have to use and it also helps to know ahead of time you were going to get asked. Basically, fast answers are ones that are usually almost the question themselves. (Tiny rant over)

Ah, so the network round trip might be too slow to do something like the following?

(map #(db/pull db '[*] %) dbids)

Im not sure my understanding of the entity api is strong enough to answer dustins question. But get id, should be constant time. But the "entity api" might include a lot more.

Yeah i think (d/entity db e) is not much more than (d/datoms db {:index :eavt, :components [e]}) wrapped in datafy/nav – the problem is if the client is remote then each nav is a round trip with no way to batch them, which the ions model solves by colocating the app process/classpath with the index

Dustin, you said "ions model" but did you mean "on prem" ? I thought the ion client wasn't collocated with the index cache anymore. (It's been a while since I thought about the internals)

DynamoDB isn't pertinent to queries in Datomic Cloud. It's only used by the transactor to guarantee serialization of the transactions (enforce ACID properties of transactions).

"I thought the ion client wasn't collocated with the index cache anymore" I dont know anything about this, this is my first time using cloud

I am not using onprem – onprem has the real entity api so we'd just use that

The Ion is our code added to a Datomic Cloud server instance. It has direct access to the instance's Java Heap. An Ion doesn't make requests through the network, except of course when the Datomic Cloud server instance has no local caches and will download them to satisfy a query. This is all transparent to the Ion.

Hi @U1QJACBUM, pinging one of you cognitects, hoping you might bring some of your perspective to the OP's question! 🙏:skin-tone-3: I'm very interested in using Hyperfiddle in an Ion context, and I know many of us are too.

hyperfiddle/photon will work (assuming websocket or other transport); my q is just about a demo datomic browser app that i want entity walking for

When using hyperfiddle/photon with onprem do you access the server side entity api over a wire from the browser?

no

@U0CJ19XAM, photon does relays the browser's need for DB entity api navigation, but the DB calls are made strictly from the backend. So nothing special for you to consider about the browser. Just Clojure and Datomic Cloud. So what is sought by Photon in Datomic Cloud Ions, is "simply" some kind of entity api or datafy/nav in the context of an Ion's clojure process.

I think it's easy enough to build, and then Onprem and Ion users will get a fast db explorer app, Client (remote) users will get a slower app but it probably won't matter – and our abstraction is streaming, so slow information will just stream in late without slowing down the UX

When I said "browser" I meant in the REBL sense