And if the answer is to build the query dynamically, I'm banging my head on why these two queries are different: A single result with three values (the goal):

(d/q {:query {:find ['(sample 3 ?tid)]
              :in '[$]
              :where '[[(q '{:find [(min 10 ?exp)]
                             :where [[_ :st.oauth.token/expires-at ?exp]]} $)
                        [[?oldests]]]
                       [?t :st.oauth.token/id ?tid]
                       [(identity ?oldests) [?expires-at ...]]
                       [?t :st.oauth.token/status ?s]
                       [?t :st.oauth.token/expires-at ?expires-at]]}
      :args [db]})

(d/q {:query {:find [('sample 3 '?tid)]
              :in '[$]
              :where '[[(q '{:find [(min 10 ?exp)]
                             :where [[_ :st.oauth.token/expires-at ?exp]]} $)
                        [[?oldests]]]
                       [?t :st.oauth.token/id ?tid]
                       [(identity ?oldests) [?expires-at ...]]
                       [?t :st.oauth.token/status ?s]
                       [?t :st.oauth.token/expires-at ?expires-at]]}
      :args [db]})

In your second example

('sample 3 '?tid) => ?tid

you are invoking the symbol sample

on 3

with a failed-lookup value of ?tid

You want (list 'sample 3 '?tid)

Thank you -I'm embarrassed I didn't catch that myself. I'm struggling with understanding the evaluation context of queries -and doubly so with nested subqueries - getting the inner 10 on the min collection aggregate to be parameterized is proving even harder than the outer 3 on the sample. But regardless, invoking a symbol as a fn at that outer layer should have been an obvious fail.

This may help https://github.com/brandonbloom/backtick

That is awesome. I've been looking for something like that without realizing it -and not just for this issue but also when building rules and such in Datalog.

Lets you write (let [sample-n 3] (template [:find [(sample ~sample-n ?tid)] …)) without a giant mess of quote-unquote to keep the datalog symbols unqualified

Yep. That's a struggle. @U053S2W0V is probably going to keep me from pulling my hair out. Thanks to both of you.

Ok, it seems the issue is actually what the message says, the documentation is not clear on this, and you need to use [:] instead of '[*]

Still seems wrong that the *gets namespace-qualified somehow

pulling using * as a symbol should work

Yeah, I find it odd. If I use * directly, it just complains about clojue_STAR something

this is the only way I found that actually worked

I think you need to change pattern into ?pattern ?

No, that wouldn't work, the '?'convention is only for query arguments, patterns don't use that. If I replace the * with a different pattern, it works

Ok, i thought it became a query argument because of the usage of :in. But i can be wrong of course

What I mean by query argument, is this '[?page :page/id ?id]'

inside the (pull ….) expression, you can't use '?''

Ah i see

or at least, that's what I have read

but I seem to remember trying that out in another situation, and I got an error because of that

You are doing this correctly. https://docs.datomic.com/cloud/query/query-data-reference.html#separation-of-concerns

No ?. There is magic support for pull pattern arguments (that I think this example shows may be slightly broken)