You have to manually create the table definition.
Unfortunately there are a few incompatibilities between Impala and MR's
Parquet implementations at the moment, so you might run into trouble
querying your Parquet table. These will be fixed in the next Impala
release, but for now you may want to create a table in Hive over the
original Avro files (which you can then query with Impala). You can also
use Impala to convert an Avro table to a Parquet table.
On Wed, May 22, 2013 at 3:07 AM, Christopher Ng wrote:
we have a bunch of structured data in avro that i would like to play with
in impala. i've managed to get parquet-avro to write out the parquet
files, but i'm wondering if there's an easy way to create the table
definition (ie the columns) to make the files visible in impala. or do i
have to look at the parquet format and manually convert it into a table def?
