Github user bdesert commented on the issue: https://github.com/apache/nifi/pull/2711 @joetrite , @MikeThomsen , So, I tested the processor. It works OK. But I've got a question. In JSON it's easy - json structure will include both attribute name and attribute value. in CSV - there will be only values on the output. Don't you think it would be useful to add a header, or attribute with avro-like generated schema? I agree that specified attributes will be in provided order, but if regex is used - then user won't be able to relate a value to a name. The same is about core attributes - since we add only existing not empty. If we decide to add avro-like schema, we gonna have problems - attribute names could be non Avro-safe.
---