you can use strings as tempids even with the on-prem version of datomic. is there any specific reason for using the d/tempid function?

Err no - I somehow must have just come across it and thought that was the way to generate 'the next tempid'. I can just use gensym I guess. How do people normally generate tempid strings?

i have the feeling that u might not even need to generate if you don't care about what the tempids are.

it's not necessary to explicitly specify a tempid anymore, UNLESS you want to reference a newly transacted (or modified) entity in another fact/entity within the same txn

also, tempid strings only have to be uniq within a transaction, so you can simply number them with range

maybe if u can share more specifics about your use-case, then we can help easier. im working on some tsv import code now and trying to write tests for it. it looks something like this:

(defn mk-rule [n]
  (let [rule-expr (str "rule-" (if (keyword? n) (name n) n))]
    {:db/ident  (keyword rule-expr)
     :rule/algo :regex
     :rule/expr rule-expr}))

(deftest re-import-rules
  (testing "remove rule"
    (tx [(mk-rule :to-be-deleted)
         (mk-rule :unchanged)])
    (is (= #{:rule-to-be-deleted
             :rule-unchanged}
           (q '[:find (set ?r-ident) .
                :where
                [?r :rule/algo+expr]
                [(datomic.api/ident $ ?r) ?r-ident]])))))

note how i just made up a convention of identifying my temporary rule entities with a rule- prefix in another test, i just made up a bunch of rules and simply numbered, like this:

(tx (map mk-rule (range 10)))

(tx [{:txn/id            1
        :txn/matching-rule [:rule-2 :rule-3]}])

also, if u use nested entity maps in tx-data, then the assignment of the nested entity ids to the containing entity's ref attribute is done automatically by the d/transact logic

This is a Fulcro application. The idea is that there are fulcro-tempids on client. They get sent to the server. The idea is to generate datomic tempids to go with them as pairs in a map. (key will be Fulcro, val will be datomic tempid). After transact! get two maps. Can use them to get a map of fulcro-tempid -> real-eid. The client can then use that map to do the remapping of the client state.

i've also noticed that u were talking about transact!. that's an old function, if i understood correctly. the current https://docs.datomic.com/on-prem/clojure/index.html documentation doesn't even mention it anymore. it just simply uses transact

I'm using an old version of Datomic.

and upgrading is not an option?

0.9.5703

because writing more code which is not necessary when using newer datomic feels like unnecessary pain

Well the upgrading ability ran out.

Only lasts for a year.

that seems like a recent enough version though to support string tempids and transact without a bang

Yes I'll start using transact now I know.

i.e. today.

(i also just noticed this change a few days ago, when coming back to datomic after 2-3 years ;)

So I should just generate negative number and str them?

Yeah I noticed it but ignored it!

so it sounds like you dont need a fulcro-tempid -> datomic-tempid because the fulcro-tempid can be just a string and u can use that directly in your tx-data

Yes that's what I thought too, as long as it is a string, which I can convert it to if its not.

So I might just use the random-uuid from in there.

isn't something like (->> tx-data (d/transact conn) :tempids vals) enough?

That gives me the real-eids.

what else is associated to the "fulcro-tempids" on the client side?

Well back on the client, in client state, there are client tempids (yes "fulcro-tempids").

aren't those already some uniq strings?

because it sounds like you can just use those directly as the datomic tempids

Fulcro can change them to real ids, but needs the map that can do that.

They are from that function above.

So for each one of them (a TempId) there needs to be a val which is a real-eid.

and what is that TempId? which namespace is it from for example? but i guess i can't add more to this topic now. i have to get back to work too.

The problem is already solved in my mind, doing as you say, using 'fulcro-tempid' as the tempids to datomic. String conversion not really an issue.

Thank you very much.

https://github.com/fulcrologic/fulcro/blob/develop/src/main/com/fulcrologic/fulcro/algorithms/tempid.cljc

Accessing Datomic from a separate VPC in versions older than 388 only can be achieved with VPC Peering Well I'm on the latest version so this seems out of the question too. What am I supposed to do?

Just tried adding my EKS VPC to the private datomic route 53 hosted zone, no dice on that either

You could run the SOCKS proxy in your EKS vpc

Thanks! I think I'm just going to go with Production