Hi Mich,
yes, the structured field has very good selectivity. I would not achieve
perfectly equally sized buckets, but I don't expect any skew problems.
Of course, moving the structured field to top-level would allow bucketing. But
I would prefer to not change the schema, as many queries have a
Hi Michael,
I would be curious to know what advantage you are going to get by hashing a
structured field. Has that structured field got very high selectivity so you
end up with equally sized buckets (files) spread?
How about the following
hive> CREATE TABLE foo (id bigint, bar struct) C