Hello, welcome to the Drill community! The rand function is actually being pulled in from Hive, we need to fix how we are evaluating it. As it doesn't take parameters, we are likely trying to optimize by evaluating it just once and copying the result for the entire current batch of records. We need to probe the Hive metadata to see when this optimization is not appropriate (when the function is non-deterministic).
Drill comes with it's own native random() function that does what you want. On Mon, Dec 14, 2015 at 11:10 PM, <masahide.mi...@brother.co.jp> wrote: > Hi, > > I found that rand() function does not reply random value. > When I used it, all rows are same values through all rows. > > You can see this result with bellow query: > select *, rand() from sys.boot limit 10 > > > ver: 1.3.0 > client: Web( drillhost:8047 ) > environment: distributed on EMR > > Thanks. > > -- > Miura, Masahide > >
