Details
-
Improvement
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
Description
There is a Spark unit test that includes reading JSON floating point values that are NaN's (validity bit is set). The Jackson parser in Arrow version 0.4 allowed for these by default, but looks like the updated version requires the conf ALLOW_NON_NUMERIC_NUMBERS to allow this.
Attachments
Issue Links
- links to