PARTITION BY support?


#1

Hi,
great job, ultra fast. We are testing mapd with customer data (oracle bi) on 822lc with 4 tesla. Problem is that sql (generated by BI tool) has PARTITON BY in sql which is not working, do you have any sugestion, roadmap … ?

sql:
SELECT D3.c43 AS c1,

D3.c9 AS c54
FROM (SELECT D1.c1 AS c1,

D1.c54 AS c54
FROM (SELECT SUM (D1.c1) OVER (PARTITION BY D1.c9) AS c1,

SUM (D1.c7) OVER (PARTITION BY D1.c44) AS c41,
D1.c42 AS c42,

D1.c45 AS c45,
SUM (
D1.c1)
OVER (
PARTITION BY D1.c44,

                   ...
                   ROW_NUMBER ()
                   OVER (
                      PARTITION BY D1.c9, D1.c26, D1.c34, D1.c44, D1.c54
                      ORDER BY
                         D1.c9 ASC,
                      ...
                      AS c55
              FROM (  SELECT SUM (
                                CASE
                                   WHEN xxxxDATE (
                                                        ...
                                   THEN
                                      xxxx
                                END)
                                AS c1,
                             SUM (
                                CASE
                                   WHEN xxxDATUM BETWEEN TO_DATE (
                                                ...
                                   THEN
                                      ...
                                END)
                                AS c2,
                             ...
                                     END)
                    GROUP BY ...
					....
					...
					...) D1) D1
     WHERE (D1.c55 = 1)) D3

#2

Hi

Thanks for trying MapD.

There is no PARTITION support in the product currently. It is on the roadmap.

I would suggest as a workaround to look at using the Immerse Dashboard tool. This potentially alleviates the requirements for the complex partition queries used to generate what we would think of as traditional reporting. With Immerse and MapD Core the end user can drill into their data in an interactive way and look at exactly the data relevant to their problem at hand.

regards


#3

Hi,
thanks on quick answer. Ok, we’ll try immerse way.