I submitted a pull request to fix this error, as it is blocking HIVE-8988. Julian Hyde, what do you think?

I added a review comment. Can you also add a test that ResultSetMetaData.getColumnName(int) works for columns in a GROUPING SETS query. I believe that info in the JDBC driver comes from RelMdColumnOrigins. TableInRootSchemaTest.testAddingTableInRootSchema does something similar to this.

Thanks! I just changed the line according to your review. I also added a test (RelMdColumnOriginsTest) based on the query that was producing the problem in Hive: the test does not pass, but I think the problem is not related to RelMdColumnOrigins (is it a problem related to the data type of the grouping(col) column?).

Just to clarify, the problem in the test query:

SELECT TABLE1.ID, TABLE2.ID FROM
 (SELECT GROUPING(A) AS ID FROM T1 
  GROUP BY ROLLUP(A,B)) TABLE1 
JOIN 
 (SELECT GROUPING(A) AS ID FROM T1 
  GROUP BY ROLLUP(A,B)) TABLE2 
ON TABLE1.ID = TABLE2.ID

is in the results, not in the metadata. In particular:

Expected: <[ID=1; ID=1, ID=0; ID=0]>
     but: was <[ID=1; ID=1, ID=0; ID=0 x 25]>

If you agree, I can skip the results check in this test, and open a new JIRA for this.

The actual results look OK. Isn't it right that you get (0, 0) repeating 25 times and (1, 1) once. Your expected result has (0, 0) only once, and that looks wrong.

Note that the 'x 25' comes from how ImmutableMultiSet.toString works when there are repeats in the set. It's not data!

OK, now I get it, thanks! I'm uploading a new patch.

Fixed in http://git-wip-us.apache.org/repos/asf/incubator-calcite/commit/4e95a8ca.

Closing now that 1.0.0-incubating has been released.