Maxim Gekk created SPARK-26303:
----------------------------------
Summary: Return partial results for bad JSON records
Key: SPARK-26303
URL: https://issues.apache.org/jira/browse/SPARK-26303
Project: Spark
Issue Type: Improvement
Components: SQL
Affects Versions: 2.4.0
Reporter: Maxim Gekk
Currently, JSON datasource and JSON functions return row with all null for a
malformed JSON string in the PERMISSIVE mode when specified schema has the
struct type. All nulls are returned even some of fields were parsed and
converted to desired types successfully. The ticket aims to solve the problem
by returning already parsed fields. The corrupted column specified via JSON
option `columnNameOfCorruptRecord` or SQL config should contain whole original
JSON string.
For example, if the input has one JSON string:
{code:json}
{"a":0.1,"b":{},"c":"def"}
{code}
and specified schema is:
{code:sql}
a DOUBLE, b ARRAY<INT>, c STRING, _corrupt_record STRIN
{code}
expected output of `from_json` in the PERMISSIVE mode:
{code}
+---+----+---+--------------------------+
|a |b |c |_corrupt_record |
+---+----+---+--------------------------+
|0.1|null|def|{"a":0.1,"b":{},"c":"def"}|
+---+----+---+--------------------------+
{code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
