Updating raw vector and map values fails about datahike

[kordano]

Thanks, I'll have a look at it.

[whilo]

You are peeking into the internals, which is ok, but you have to use :eavt-durable, we still have to change that to :eavt with the next cleanup (probably rebase on upstream datascript). This should also address the pull issue. The actual bug you are describing is visible in

(d/q '[:find ?e ?v
       :where [?e :my/vec ?v]]
  @conn)

My suspicion is that something is wrong with schema handling the memory backend.

[grischoun]

I started having look at the problem.
When I try the failing transaction:
@(d/transact conn [[:db/add [:my/id "x"] :my/vec ["x" "y" "z"]] [:db/add [:my/id "x"] :my/map {:x "x" :y "y" :z "z"}]])

I get the following NullPointerException:

@pm4r, do you see the same thing?

[pm4r]

@grischoun Actually I do have same stack trace at this transaction. I didn't realize it earlier because It turns out if I evaluated expressions one by one in my editor (Atom + ProtoRepl) it didn't evaluate derefing future so I didn't see exception, but executing larger block at once indeed throws it. That explains why db had nothing related to that transaction.

@whilo I didn't know about :eavt-durable, will remember that in future. I used :eavt almost exclusively to check what's inside when I see something unexpected,

[whilo]

We will get rid of the :eavt-durable and :eavt distinction. Right now we are framing datahike as a joint database for ClojureScript and Clojure and will probably rebase it on a current version of datascript to make it cleaner. This should be transparent to users through the import/export upgrade procedure.

[whilo]

Uh that went quick at closing the issue, I will reopen it for now until we have the fixed hitchhiker tree tested with datahike.

[whilo]

This fixes the issue, I have tested it.

[whilo]

There are still the schema problems open though. I think datahike should throw if you try to assign the same attribute multiple times, which it doesn't here.