Clone
Paul Rogers <progers@cloudera.com>
committed
on 28 Oct 18
DRILL-6809: Handle repeated map in schema inference
It turns out that the RowSet utilities build a repeated map without including the hidden… Show more
DRILL-6809: Handle repeated map in schema inference

It turns out that the RowSet utilities build a repeated map without including the hidden $offsets$ vector in the metadata for the map. But, other parts in Drill do include this vector.

The RowSet behavior might be a bug which can be addressed in another PR.

This PR:

* Adds unit tests for map accessors at the row set level. Looks like these were never added originally. They are a simplified form of the ResultSetLoader map tests.

* Verified that the schema inference can infer a schema from a repeated map (using the RowSet style.)

* Added a test to reproduce the case from the bug.

* Made a tweak to the RowSetBuilder to allow access to the RowSetWriter which is needed by the new tests.

* Could of minor clean-ups.

closes #1513

Show less

master + 2 more