tree 1f562b0c513d5dc67c92e5dd9b7d4034c476d81b
parent b7965d8240f3a5cd6972351b7fe1116e4a0835c1
author Adam Tamas <tadam@cloudera.com> 1596096625 +0200
committer Gabor Kaszab <gaborkaszab@cloudera.com> 1599298966 +0000

IMPALA-10012: ds_hll_sketch() results ascii codec decoding error fix

While the ds_hll_sketch() generates a string value as output the data
is not an ascii encoded text but a bitsketch, because of this, when
the shell get this data it disconnect while it tries to decode it.

The issue can be reproduced with a simple method like using unhex
with a wrong input.
Example: SELECT unhex("aa");

This patch contains a solution, where we replace any not UTF-8
decodable characters if we run into an UnicodeDecodeError after
fetching it.

This solution is working with the Thrift 0.9.3 autogenerated gen-py
but still fails with Thrift 0.11.0.

For Thrift 0.11.0 the error is catched and an error message is sent
(not working with beeswax protocol, because it generates a different
error (TypeError) which can come for other reasons too).

Testing:
-manual testing with these protocols: 'hs2-http', 'hs2', 'beeswax'

Change-Id: I0c5f1290356e21aed8ca7f896f953541942aed05
Reviewed-on: http://gerrit.cloudera.org:8080/16418
Tested-by: Impala Public Jenkins <impala-public-jenkins@cloudera.com>
Reviewed-by: Gabor Kaszab <gaborkaszab@cloudera.com>