2 * X queries vs. 1

currently a single transaction with a batch X of documents does 2 * X SQL queries.
it looks whether the document exists first and then inserts or updates it:

(defrecord JdbcDocumentStore [ds dbtype]
  db/DocumentStore
  (submit-docs [this id-and-docs]            ;;    batch of documents
    ;; ...
    (doseq [[id doc] id-and-docs
      (if-not (do

Right now in insertElementRecursive(), DocLayer does NOT try to store arrays with null values as sparse arrays, i.e. for null value it still insert a k-v for it, which is kind of against what it attempts to achieve later when doing update (skipping the null values and effectively store those arrays as spars arrays). We should make the behavior consistent and do the same in insertion.

Currently, there are some examples in README.md with Elasticsearch queries and corresponding uptasticsearch code. That code is effectively pseudocode right now, as it references a fictional Elasticsearch cluster.

I think this could lead to a bad experience with the docs and lead people to walk away from the project and not come back.

I would love if someone changed those examples to be r