The formula syntax is the very least of your problems. You can write your own, use Java's embeddable script engine capability and plug in GraalVM's JavaScript, shell out to Perl, etc.

It seemed like the most challenging part to me. If “user-defined calculated fields” are the least of my problems, what is the most challenging aspect in your mind? 😅

Mostly in jest. But someone who is mildly dissatisfied with a complex monster of over-clocked Excel might be difficult to please. If at all possible, you could have them hire and fire another consultant before you heroically step in with a solution. As to that solution, the foundation is data which you will need to model in space and time. The offhand reference to versioning (free with Datomic) made me think this modeling might not yet be complete. 🙂

Looks like the as-unqualified-maps result set builder might be what you're looking for? https://github.com/seancorfield/next-jdbc/blob/e21a7ab59011a005b6609b9b1abb1f4b53fb1f58/doc/result-set-builders.md?plain=1#L10

wow thanks. I was looking for a solution at the clojure level, didn't realise I can configure the driver

One Clojure-level solution would be something like

(update-keys #:apppage{:page_id 2, :page_name "testing", :order_key 5, :page_type "task"}
             #(keyword nil (name %)))

;; => {:order_key 5, :page_id 2, :page_name "testing", :page_type "task"}

ah, nice!

Why do you want to strip the namespace? The default is that way for a reason: qualified keys are good - they help avoid collisions between, say, :person/id and :address/id. You can use aliases to make them shorter to type if you really can't bear typing a few extra characters for the table name:grin:

I always avoid calling any columns "id" for that exact reason. I use <tablename>_id instead.

I find that too repetitive in joins: join foo on foo.foo_id = bar.foo_id but I've seen both styles in dba recommendations over the years...

... And mapping :foo/id to foo.id is what I'm used to from HoneySQL, rather than :foo/foo-id

if you name your columns like that you can use both USING and NATURAL, so it doesn’t need to be repetitive :) I don’t do this though, just offering it as trivia instead of recommending the approach

In the Clojure-realm, this typically means the library is done and safe for use. You can use random-uuid when you say snowflake id maybe cuid would fit the bill or do you mean 0.% probability collision rate? (check table, generate unique or generate again)