Github user MaxGekk commented on the issue:
https://github.com/apache/spark/pull/21909
> Please document it in the migration guide.
@gatorsmile Updated by the notice:
- Since Spark 2.4, text-based datasources like CSV and JSON don't parse
input lines if the required schema pushed down to the datasources in read is
empty. The schema can be empty in the case of count(), for example. To set
`true` to `spark.sql.legacy.bypassParserForEmptySchema` restores the previous
behavior when the underlying parser is always invoked even for the empty
schema. This option will be removed in Spark 3.0.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
