Clojurians Log v2
datalevin 2025-09-19
I've been looking into Datalevin, and I'm unsure of the best way to store insertion order of entries. In SQL I'd use an auto-increment column. In Datalevin (and Datomic), I've considered maintaining a counter field that I update within a transaction, and I've also thought about using a sequential ID algorithm like Flake. Is there a better, more idiomatic option that I'm missing?
Thanks for the advice! That's the conclusion I was coming to as well. I guess what I really want is a deterministic time ordering, so I could sort by timestamp then ID to resolve any potential conflicts.
Yeah that’s what I would do. Honestly I’d do the same thing in SQL too, At least with Postgres, sequences can get out of order or skip due to a variety of reasons.
I'm not sure about datalevin specifically, but in Datomic you'd use a https://docs.datomic.com/transactions/transaction-functions.html#writing that first queries the highest existing id, increments it, then stores it. It seems like Datalevin has https://cljdoc.org/d/datalevin/datalevin/0.9.22/doc/transaction#transaction-functions-in-datalog-store. That said, I would strongly recommend against data models that need to maintain insertion order. I've certainly done this to meet very specific business requirements, but it's a pain and should be avoided if at all possible.
Does datalevin come with the same caveats for long string values like datomic? (Curious about datalog and KV mode)
Per the limits https://github.com/juji-io/datalevin/blob/master/doc/limits.md individual value size can be 2 GB
I store fairly large strings in datalevin with no issues, I do not use the search index though