Do Floats and Doubles not compare as equal in Datomic?

I've been using the repl that comes with datomic (bin/repl) for all of my datomic repl needs, but I'm finally feeling the need to use a repl from emacs. However, I'm using the rest api, and I don't have a lein or boot project that cider can hook into. Should I make a dummy project just to start it with cider, or is there some way to accomplish this that I'm missing?

To answer my own previous question, Datomic does not do float/double coercion. 😞

fyi, we're working on a slack replacement and might use datomic https://hackpad.com/collection/wnikaeBENEE

How choose storage for datomic? Is there any Guiding principles? I read the documentation about how set up for various storage db, but not about choice.

@alwaysbcoding: this probably isn't terribly helpful, but your example looks a lot like what I do from python. I don't do anything special with the ?vars

the error suggests to me that your query datastructure isn't being coerced to a string before being sent?

@isaac: the canonical advice is to use whatever is operationally easiest for you. for example, if you already have postgres in production, use that. I have also seen people say that if you're on AWS, consider using dynamo.

@isaac: @bhagany the last thing is that cassandra's probably the easiest/best if you don't have any other particular choice any particular way and want distributed/perf etc and cannot use dynamo

datomic just support version of cassandra at 2.x?

yes, only cassandra 2.0 and 2.1

Is it will support cassandra higher version?

Probably, at some point.

Hey, I’m new to datomic and could use some guidance solving a problem. Let’s say I have some book entities that have a :book/color attribute which could be :red, :green, or :blue. Let’s say also that I have a map of colors to ordinal values {:red 1 :green 2 :blue 3} which I don’t have stored in datomic. I’d like to have a query that, given an eid and such a map, returns the ordinal value of the book’s color. (I realize this is trivially done after the fact; I’m actually working on a transaction fn for which I think a solution to this simple problem will help.)

My naive attempt fails:

(defn get-book-color-ordinality
  [db eid color-ordinals]
  (d/q '[:find ?ord
         :in $ ?eid ?ords
         :where
         [?eid :book/color ?color]
         [(get ?ords ?color) ?ord]]
       db eid color-ordinals))

Nevermind, this does seem to work, I had just mixed up the type of my color attrs

But it doesn’t work as a txn fn; datomic complains with:

java.lang.RuntimeException: Unable to resolve symbol: ?ords in this context

Any clue why this might be true? I even get this when I inline the color-ordinals map.

Does anyone know why I would be able to successfully use d/with but transacting the same tx-data to an unchanged connection would fail with a java.lang.ExceptionInInitializerError?

Even weirder, transacting this particular data does work against an in memory connection but it does not work against a free dev connection. I'm stumped (and blocked). Any ideas? CC @bkamphaus @stuartsierra

Sorry, I got nothing @domkm

@stuartsierra: Thanks for responding. Would you pass this on to your colleagues?

@domkm: can you share your tx-data ?

I’ve written up a simple gist illustrating the txn-fn problem I’m having: https://gist.github.com/dball/f4cd5a52dddc7b812b86

The former form is a macro which generates a txn fn suitable for installation in a schema; the macro is well tested and I have no reason not to trust it. It fails when used in a txn. The latter form works exactly as intended locally. I’m not sure why; are clojure map literals not allowed within clojure txn fns?

@marshall: I pasted a snippet above.

@marshall: It's sort of a contrived example because I don't intend to use that exact code in production (which essentially prevents any concurrent modification), but I intended to do something related and, now that I know the above works in a mem transactor but not in a dev transactor, I am concerned about the approach.

@domkm is it the invocation of the txn function that fails or the installation of it?

@marshall: Installation works fine in both mem and dev. Let me get back to you on dev invocation...

@marshall: Dev invocation works as well. I used this tx-data: '[[:fn/query {:query [:find ?e :where [?e :person/email "foo"]]}]]

@domkm: The snippet is invoking one query to make an exception, and a second query using that exception as an arg, to throw an exception if the db has been changed and no longer has the basis-t 1031?

@bkamphaus: Almost exactly. It doesn't invoke the query which makes and throws an exception. It simple returns that query in a list and the transactor invokes it after splicing it in.

@domkm: I guess I’m not clear whether this is the intended use case of opening query in a transaction function, or a contrived example meant to illustrate a failure. That said, a suspect for the deserialize exception might be something that does a sequence manipulation in a transaction function that requires a clojure map or vec specifically rather than targeting a java.util.Hashmap as the precision of type preservation for collections is only guaranteed at the interface level (well, I think, either way it’s not preserved for the exact type) on the wire. That would produce something like the symptom you describe, where something would work on mem, but not dev/free. Maybe the assoc or cons in the code for :fn/query would be the points where the failure is encountered?

(speculation, not tested at this point)

@bkamphaus: Okay, thanks, I'll test that.

@bkamphaus: Hmm, so you're correct that they are not Clojure data structures but assoc and cons are working as I expected on this: '[[:fn/query {:query [:find ?e :where [?e :person/email "foo"]]}]] I'm confused

@domkm: I’ll use your transaction function code and investigate a bit more, see if I can identify what the exact issue is. Will update after digging some.

@bkamphaus: Thank you!