We have QlikSense connecting to a MapR hadoop cluster via ODBC. The query that is written in Qlik is different to the query that gets passed through to the ODBC server. eg this
SELECT col1, col2, col3
FROM tblA
JOIN tblB on tblA.ID = tblB.ID
JOIN tblC on tblB.ID = tblC.ID
comes through something like
SELECT col1 AS `C_43578923459873`, col2 AS `C_345698723459873`, col3 AS `C_378456234589`
FROM tblA AS `C_4567`
LEFT OUTER JOIN tblB AS `C_3435` ON `C_4567`.ID = `C_3435`.ID
JOIN (SELECT .... FROM tblA AS `C_6789`) ON ....
etc
I've truncated the second query for brevity. Basically it mangles all the column and table names and then makes every join a subquery. If the table had 8 joins then we end up with 7 subqueries. We get subqueries within subqueries. Note I haven't written the second query 100% accurately, take it as just an example of what it is doing. The mangling of names is OK but the subqueries makes the query horrendously inefficient, turning a 21 second query into 2 hours and causing it to use massive database resources.
First question, can anyone confirm QlikSense is doing this? It is not impossible something else is doing it (but seems unlikely)
Second question, how do we stop it from doing this?
I have tested from DBeaver, Python and Hue and none of them show this behaviour. For these clients the query gets sent though exactly the same.
Thanks for any replies.
