[
https://issues.apache.org/jira/browse/NIFI-5791?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16676746#comment-16676746
]
ASF GitHub Bot commented on NIFI-5791:
--------------------------------------
Github user stevedlawrence commented on the issue:
https://github.com/apache/nifi/pull/3130
I'm not too familiar with the Reader/Writer idiom. From what I can tell, a
Reader converts data to a Record, and a Writer converts those Records back to
the data format? Is that accurate? My one concern is that although
DFDL/Daffodil can handle record oriented data, often times it is much more
complex. For example, the [daffodil
examples](https://daffodil.apache.org/examples/) page shows two examples of how
DFDL can convert data to XML. The first example is CSV data and is clearly
record oriented. But the second example is PCAP (used in the above template)
which could be seen as records, but is a complex nesting and there's a global
header that isn't really a record. Most data formats we've seen DFDL used for a
more like the latter.
The SchemaRegistry concept that looks to be used by Readers/Writers seems
like a really nice way to provide a DFDL schema (if possible without using the
Records?), but I'm not sure how well the Record concept fits in. Does it make
sense to keep the XML/JSON output as in the PR, and then if a particular
infoset does map well to a Record the user could add a JSON/XMLReader to
convert to a record?
> Add Apache Daffodil parse/unparse processor
> -------------------------------------------
>
> Key: NIFI-5791
> URL: https://issues.apache.org/jira/browse/NIFI-5791
> Project: Apache NiFi
> Issue Type: Bug
> Components: Extensions
> Reporter: Steve Lawrence
> Priority: Major
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
