Clojurians Log v2

I think this is fairly close to optimal, the main remaining issue is to write full tree snapshots less often by using a tx-log in the db root. This is what I am looking into right now.

Hello, Another question about something that seems to be a strange behavior with :attribute-refs? activated. I declared an ident with the following schema:

{:db/ident :foo/bar
 :db/valueType :db.type/keyword
 :db/cardinality :db.cardinality/many}

(d/transact conn [{:db/id id
                   :foo/bar [:plop]}])

(d/transact conn [[:db/add id :foo/bar :plop]])

Bad entity attribute :foo/bar at [:db/add id :foo/bar :plop], expected reference number

For anybody who would like to reduce the storage size, these are the most compact settings for datahike at the moment:

{:store  {:backend ...
          :config {:compressor {:type :lz4}}}
 :keep-history? false
 :attribute-refs? true
 :schema-flexibility :write}

@dam @alekcz360 I did some REPL explorations with benchmarks. After gc! a Datom in this setting consisting of 4 longs (8 bytes) = 16 bytes costs ~20.1 bytes per index in a database with 10k stored datoms and this scales linearly in size from then on to 100k. This is also the case without lz4 compression since the value is a number in my benchmark. In case you store string data lz4 should greatly reduce size.