this works on drill .06 on the mapr drill sandbox : create or replace view tview as select cast(convert_from(row_key, 'UTF8') as bigint) as cust_id, cast(t.loyalty.membership as varchar(20)) as membership from dfs.`/tables/customers` t;
select sum(orders.order_total) as sales, tview.membership from hive.orders, tview where orders.cust_id=tview.cust_id group by tview.membership; ______________________________________ HOWEVER this gives an error: create or replace view tview as select cast(convert_from(row_key, 'UTF8') as bigint) as cust_id, cast(t.loyalty.membership as varchar(20)) as membership from maprdb.customers t; select sum(orders.order_total) as sales, tview.membership from hive.orders, tview where orders.cust_id=tview.cust_id group by tview.membership; ______________________________________ this works: create or replace view tview as select cast(row_key as bigint) as cust_id, cast(t.loyalty.membership as varchar(20)) as membership from maprdb.customers t; select sum(orders.order_total) as sales, tview.membership from hive.orders, tview where orders.cust_id=tview.cust_id group by tview.membership; ______________________________________ the data in this HBase (maprdb) table is from a importtsv bulkimport of a csv file so it is all strings converted to bytearrays. Is it recommended to just use cast in this case, which always works ? It is still not clear to me when to use convert_from for an hbase table.
