On Spark 1.6.0, calling json_tuple() with an emoji character in one of the
values returns nulls:
Input:
"""
"myJsonBody": {
"field1": "📻"
}
"""
Query:
"""
...
LATERAL VIEW JSON_TUPLE(e.myJsonBody,'field1') k AS field1,
...
"""
This looks like a platform-dependent issue; the parsing works fine on my
local computer (OSX, 1.6.3) and fails on the remote cluster(Centos7, 1.6.0)
I noticed that in 1.6.0, json_tuple was implemented this way:
https://github.com/apache/spark/pull/7946/files
So far I have:
- Checked all java system properties related to charsets on drivers and
executors
- Turned up logging to debug level and checked for relevant messages
Any more input? Should I try the dev mailing list?
