Ah yes. This is the result of a bit of a limitation in how we handle lists in value vectors. At the moment, there is no distinction in list vectors between an empty list and a null list, which is not quite right, or course.
On Sun, Jun 26, 2016 at 8:56 PM, Ted Dunning <[email protected]> wrote: > In DRILL-4754, I reported that where a field is missing, it is sometimes > replaced with an empty list. This seems very wrong. > > The data: > > > *{'a':3, 'b':[3,2], 'c':'xyz'}{'a':7, 'c':'wxy'}* > *{"a":7, "b":[]}* > > Note the missing 'b' value in the second record. > > The query: > > 0: jdbc:drill:> select * from maprfs.ted.`bug.json`; > > The result: > > 0: jdbc:drill:> select * from maprfs.ted.`bug.json`; > > > > > > > *+----+--------+-------+| a | b | c |+----+--------+-------+| 3 > | [3,2] | xyz || 7 | [] | wxy || 7 | [] | null > |+----+--------+-------+* > > Note the way that 'b' has a non-null value in the second record. Note the > way that 'c' does have a null value in the third record. >
