That is the value that was changed?

This appears to be just one empty transaction. Try different start and end parameters to get others

What does 50 mean in that response?

That’s an attribute

Likely :db/txInstant

Ah, hmm

Consider using d/tx-range also

If all you really want is the full log it will be faster and offers laziness

Oh snap, that looks like exactly what I’m after, thanks a ton!

Yeah, I can see everything I’m looking for now, thanks again!

In datomic, [?r ?k ?v] would bind a number (the attribute entity id) to ?k

that would be ?r in my testing with datascript

but we can see the list of ?k's returned in the result list here and they are keywords

e.g., the following has attribute keys in the "middle" position: https://github.com/tonsky/datascript/blob/cd129b7a9ef2fc31fa95873c376691f272752832/test/datascript/test/query_fns.cljc#L131

Could it be some edge case with qualified symbols and predicate clauses?

try this:

(d/q '[:find [?k ...] :where [?e :root/user ?r][?r ?k ?v][(not= ?k :user/feature-flags)]] @connection)

note the quoting changed

[?k]` would expand to [current-ns/?k]

maybe that ?k isn’t unifying with the one inside the (not= ?k ...) for whatever reason

Yea I tried ?k# too to remove the ns

Ok I'll try outer single quote

oooh, the combination of outer single quote and no quote on the pred fn worked

weird

core=> (cljs.pprint/pprint (d/q '[:find [?k ...] :where [?e :root/user ?r][?r ?k ?v][(not= ?k :user/feature-flags)]] @connection))
[:user/id
 :user/disable-global-role?
 :user/has-to-accept-license?
 :user/first-name
 :user/last-name
 :user/role
 :user/company-id 
 :user/avatar-url 
 :user/app-roles 
 :user/email]
nil
core=> (cljs.pprint/pprint (d/q '[:find [?k ...] :where [?e :root/user ?r][?r ?k ?v][(= ?k :user/feature-flags)]] @connection))
[:user/feature-flags]
nil

thanks!

interestingly i have the following elsewhere in my codebase for query generation and it works (notice the backtick):

(defn query-root-scalar [key]
  `[:find ?v# . :where [~root-id ~(keyword "root" (name key)) ?v#]])

oh, it’s because 'not= expands to the quoted form

you needed ~'not=

doh! yep. good call

so will it hurt anything that all of my ?s will be "namespace-bound"

I was trying to get fancy by using # suffix to make them unique

so it was invoking ('not= x :something) which is symbol-lookup of not= in collection x, or :something as not found, so it was always returning :something thus everything matched.

yea but here's the weird part: https://github.com/tonsky/datascript/blob/0.11.6/src/datascript/query.cljc#L140

I was getting "Unknown predicate error ..." when the predicate wasn't found in bulit-ins linked to above

re “weird part”, yeah but it’s actually (quote not=) which isn’t in that list

right which should produce an "Unknown predicate error ..." 🙂

or, it will just eval to itself?

mm, not sure. my brain is getting twisted at these meta levels. something weird was going on for sure.

>  so will it hurt anything that all of my ?s will be “namespace-bound” It’s just weird, you may shake out edge cases

It defies the usual assumptions

yea but that seems to be sort of the SOP in these datascripts

perhaps going with the flow is called for 🙂

https://github.com/brandonbloom/backtick may make it easier to mix quasi-quoting and unqualified symbols

no

how about the ability to apply a "result transform" to the result of a :find ?

such as the: into {} portion of the following:

core=> (cljs.pprint/pprint (into {} (d/q '[:find ?k ?v :where [?e :root/user ?r][?r ?k ?v][(not= ?k :user/feature-flags)]] @connection)))

the library i'm using won't allow the "outside wrapper" transform such as above, so i'd need a way to get it inside the query itself

(map #(update % 0 dissoc the-field)) ?

where would that fit into the overall [:find ... } ?

actually i don't think that would work. update and dissoc require a map which in my example doesn't exist until into {} is executed

look at the reg- functions below to understand why any "outer wrapper" solution won't work: https://github.com/denistakeda/re-posh/blob/master/src/re_posh/subs.cljc of course, it could be done with two subs, one taking the input from the first and then transforming it. but imagine that we want to do this for 100s of subscriptions and suddently 250 subs become 500 subs (which is overly verbose)

Update doesn’t require a map

Is there a reason you don’t enumerate the fields you need?

yea, verbosity. it's N-1 fields that I need

and N is on the larger side

but i think i will just restructure the facts around this data, break it out to the top level

upon insertion

which will be fine for this case. but it's unfortunate that there isn't a better solution for this

in the general case