Raúl created DRILL-7935:
---------------------------
Summary: Error: Hash join does not support schema changes in probe
side.
Key: DRILL-7935
URL: https://issues.apache.org/jira/browse/DRILL-7935
Project: Apache Drill
Issue Type: Bug
Affects Versions: 1.18.0
Reporter: Raúl
Let's see if someone can help us, we have a problem:
We have a collection in mongo with 58,948 documents, in which a field of the
collection is null in 58,270 and with value (not null) 714. In a select, when
we join on that collection, it returns the following error:
{code:java}
SQL Error: UNSUPPORTED_OPERATION ERROR: Hash join does not support schema
changes in probe side.
Prior schema :
BatchSchema [fields=[[`CITY_CODE` (VARCHAR:OPTIONAL)], [`COMPANY_CODE`
(INT:OPTIONAL)], [`LOCATION_CODE0` (VARCHAR(65535):OPTIONAL)]],
selectionVector=NONE]
New schema :
BatchSchema [fields=[[`CITY_CODE` (VARCHAR:OPTIONAL)], [`COMPANY_CODE`
(VARCHAR:OPTIONAL)], [`LOCATION_CODE0` (VARCHAR(65535):OPTIONAL)]],
selectionVector=NONE]
Unsupported schema change
Fragment: 0:0
{code}
The problem that I observe is that when the field has so many nulls, it
considers it INT, and when it crosses it, it sees that a value is varchar and
that is when it gives the error.
1.) How can I solve this problem? I have changed different parameters regarding
mongo and nothing makes it work
2.) Why does Drill consider null to be an int? It should not consider it as
varchar.
If I make a select only of the COMPANY_CODE field and I try to sort it, it
gives me the following error:
{code:java}
SQL Error: INTERNAL_ERROR ERROR: You tried to write a VarChar type when you are
using a ValueWriter of type NullableIntWriterImpl.
Fragment: 0:0
Please, refer to logs for more information.
{code}
Thank you!!
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
