Hi,
I'm capturing data of the form A (1:n) B, which is a fairly standard
item-subitem pattern. In a standard DB, I'd have A and B tables with a foreign
key from B to A.
But since Hive is different -- there's no natural primary key in my data and
joins seem much more expensive -- I'm considering using an Array of Structs.
So -- some questions:
Does this make sense? How's performance? Say B has an attribute
'num', and I want to find the average of nums or something [which a B table
would lend itself to]
Is there an example of how to format the files?
Thanks,
Ranjan
