Clojurians Log v2

you can still get a doc by id, but yes the iteration of all docs of a type might be a bad usecase

we used these single key/value maps as ids in a project, and at least it was good when inspecting the database, you always know what you are dealing with

it is a nice idea... what about using a prefix? so "<prefix>"-<rest of id>. Technically not a UUID, but u could have the <rest of id> portion be an actual uuid. This also of course assumes that there is a way to query using prefix when you're searching for "user type A", vs type B, etc.

We have used namespaced id's in XTDB v1, like

:xt/id {:user/id <uuid>}
:xt/id {:customer/id <uuid>}

v2? You can just insert all different types of users in the same users table, it doesn’t matter if they have different attributes

Sorry no, v1

in v1 there are no tables, so I don’t know how you can have “table inheritance”?

just documents

but the same applies, there is no requirement for all documents to have the same attributes

Gotcha. Thank you

This link should help explain STI: https://docs.gitlab.com/ee/development/database/single_table_inheritance.html

personally, I would mostly avoid ORMs and use a language with 1st class data

but the discussion of having some kind of “type” attribute for discrimnating between different types of docs is valid

yes. The question is not about ORMs, but about how folks would structure their documents

best is to not need an attribute just to specify a document’s “type”

you can do that by querying by some attribute that only occurs in that type, or you could even encode the type in the doc id in v1 (like {:sometype <uuid>} map as the xt id

hmm. I’ve not considered that, but that’s an interesting idea. Right now my ids are UUIDs.

I’ve been wondering about the utility of using UUIDs, and having to do a parse-uuid (give the string representation) step just to do a query

sorry, maybe I misunderstood. Are you saying to use a map as the xt id?

iirc, you can even query all docs of a “type” by doing range {:sometype <0000-…>} - {:sometype <ffff-...>}` query

yes, use the whole map as the id

on the topic of identifiers, have you looked at something like snowflake ids (not necessarily the implementation as done by twitter; more info https://en.m.wikipedia.org/wiki/Snowflake_ID) as opposed to UUIDs?

haven’t, but UUIDv7 looks nice as well

thank you for that note! re that range query, would you be able to give me an example of how to do that query?

sorry, I don’t have an example handy, but just using the <= and >= in datalog will give you range queries

no worries. I’m aware of the predicate functions; just not sure about how to query when the value in question is “nested”, instead of a “direct” EAV value

hey @jf.slack-clojurians be aware predicates over nested data won't benefit from indexing, you will essentially end up with table scans - but you can do it like: [?e :att ?some-map] [(get ?some-map :some-key) ?val] [(<= ?val 100)] you can similarly use get-in, it's just calling clojure.core/* fns

thank you for that note. In that case, I think i'll refrain from using maps