Fork me on GitHub

hi. is it possible please to find information on data representation in SQL layer in datomic: which tables, relations, logic of storing the data really. Can not find it anywhere


@andmed: I don't think you would have that since it's a part of datomic internal design and implementation and it's not open source.


@nxqd got it. have now a weird task to build on top of RDBMS some schemaless data design, everything what i have seen says it is a bad idea, was interested to know about datomic


@andmed: If you are only talking about something schemaless (and not full blown datomic which is somewhat-schemaless but also maintains history) AND you don’t expect it to scale in any significant way, you should be able to get away with a dead simple table with entity, attribute, value columns. Anything more than that and your intuition about it being hard (to do right, fast, reliable) is correct. Having said that I think there are likely other OSS dbs out there that do what you want. See also: Good luck with your project!


@kahunamoore: yes, the hard case seems to be my case exactly 🙂 The guy wants to manage relations in a separate table, and have more crazy ones with "relatives", "relatees" and such like, full awfulness of which I am yet to discover. Thank you, these are good links to start


Datomic uses all backing stores the same way — as a key/value store for opaque binary blobs. You can't query them with the backing store's native query language, e.g. SQL.


A while ago I stumbled upon a tutorial that showed how to implement basic fact oriented database much like datomic/datascript. It was on github, written as a md document. The repo contained a bunch of other articles, presumably some kind of sources for a book by various authors. I can’t find it anymore. Maybe someone seen that and remembers where it was?


I’m going to implement that in haskell as a learning project