mogge
Morning! It's quiet around here. I feel like an impostor as I'm writing hardly any Clojure these days. Trying to persuade one of the companies I work for that Datomic Cloud is the solution to all their problems, but I don't think they believe me 🙂
You are always welcome here, regardless of how much Clojure you write! And we feel ya... We've all are/been there.
I will report back if I make any headway, but I'm not optimistic...
👍
Yes, please let us know. And maybe there is someone here who can give you tips and/or tricks regarding Datomic Cloud.
The problem is that they are a Python shop and might be a bit too wedded to that way of doing things. Currently they have a very confused data model spread across 8 DynamoDB tables that they make multiple queries against to construct a graph. Consequently, some of their API calls take upwards of 12 seconds!!! They will probably move to PostgreSQL so they can stick with Python, this will be an improvement but IMHO still not the best solution. Maybe I will get some time to put together a small demonstrator project for them.
have you had a look at Rama?
Yes, I took a couple of training days last year to evaluate it and worked through the tutorial. I like it, but it feels like overkill unless you need the write throughput.
Are you using it?
I'm not using it, but have tried out a few things. and even if write throughput isn't your problem it might still help. just my 2 cents of course
Datomic can be a hard sell, especially into non-Clojure envs. Part of why I "prefer" XTDB is that it can look "just like PostgreSQL" (pgwire protocol) and you can continue using SQL for CRUD etc. I think the XTDB folks have customers using Python, so there would be precedent there. And, of course, many of the XTDB folks are right there in the UK...
Oh yes, I have met Malcolm et al. For me, one of the selling points of Datomic Cloud is the ions, and they now have HTTP direct support for API Gateway so no longer any need for a lambda shim. Your API handlers are performant because of data locality and the deployment story is a nice one. The bigger difficulty, I think, is the language barrier.
...difficulty in selling it...
could you make a prototype? (with AI of course 😉 ). that might help.
Yes, I was thinking about making a prototype, it's just a question of finding the time. I would not use a LLM, though, for practical and ethical reasons. Some of my colleagues find such tools useful and can turn a blind eye to the ethical issues, but my plan is to retire before I am forced to use them 🙂
fair enough
Mornin'!!