Clojurians Log v2

Forms service gotcha: if you try to redirect the client on a succesful form service operation, you need to wrap rfe/push-state in e/drain, otherwise the unserializable routing result tries to transfer to server and causes a Reactor Failure:

(e/defn EditAccountOp [!errors account-uuid account-data]
  (e/server
    (let [user-ent    (d/entity *db [:id *user-id])
          account-ent (d/entity *db [:id account-uuid])
          rx (e/Offload
               (e/Snapshot
                 #(forms/try-ok
                    (throw (Exception. "test errors"))
                    (let [rx (api.account/update-account!
                               {:acl *acl :conn *conn}
                               user-ent account-ent
                               account-data)]
                      (prn 'rx rx)
                      rx))))
          success? (= [::forms/ok] rx)]
      (e/client
        (if success?
          (e/drain ; if you don't drain here, Electric will complain about unserializable routing result from rfe/push-state.
            (reset! !errors nil)
            (rfe/push-state ::layout/accounts {}))
          (reset! !errors "error!")))
      rx)))

2nd Update: I restarted my REPL & Electric, and now it's fast again 😕. dev-time memory leak? We are seeing very slow rendering times when navigating away from an edit form with 15 fields (and their associated layout elements) to a list of DB entries due to the dismount time of form fields. I debugged this with @xifi and confirmed that if we hide the form fields and navigate, the render is instant. We're displaying a table of accounts. We checked that the Datomic d/pull-many takes 0.5ms itself. The table thead & tfoot render instantly, but the e/for over accounts collection, blocks for several seconds while the form fields from the previous route are unmounted. We cannot see why the e/for would block, other than for dismount. Is there any fix on horizon to speed up dismount, or avoid new route blocking on previous route's unmount? We could do a workaround where form is always present but hidden, but we'd have to combine our listing vs editing routes, which muddies concerns. Update: weirdly, the delay only seems to occur when we navigate the user via rfe/push-href, not when viewing the form and clicking on a list route nav link directly. This suggests something is blocking when it shouldn't.