Github user markap14 commented on the issue:
https://github.com/apache/nifi/pull/2587
@JohannesDaniel thanks for the update! I commented above re: the use of
Expression Language in the property descriptor.
I do still feel like the check for 'record tag names' is unnecessary, as
the reader should not be responsible for filtering the data but rather just for
reading it. There already exist mechanisms for filtering the data (You could
use PartitionRecord + RouteOnAttribute, ValidateRecord, or QueryRecord just off
the top of my head to achieve this). Additionally, we have the Schema for the
Record Reader. So if the element name matches the top-level Schema name (or one
of them, if the top-level field is a UNION/CHOICE element), then we could use
that. So, with your example above, if you only want to read the
`<record/><record/>` part, your schema should indicate that the top-level field
name is `record`. In that case, it should filter out the `other` record. Does
that make sense?
---
