[
https://issues.apache.org/jira/browse/NIFI-9206?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17653323#comment-17653323
]
Chris Sampson commented on NIFI-9206:
-------------------------------------
Rebased the original PR(s) against latest {{main}} and addressed previous
comments.
> Create a processor that is capable of removing fields from records
> ------------------------------------------------------------------
>
> Key: NIFI-9206
> URL: https://issues.apache.org/jira/browse/NIFI-9206
> Project: Apache NiFi
> Issue Type: New Feature
> Components: Extensions
> Reporter: Peter Gyori
> Assignee: Chris Sampson
> Priority: Major
> Time Spent: 1.5h
> Remaining Estimate: 0h
>
> A processor should be created that is capable of removing fields from records
> (RemoveRecordField might be a name for it).
> The processor should have 3 properties:
> * Record Reader (a Reader controller service could be specified)
> * Record Writer (a Writer controller service could be specified)
> * Field To Remove (expects a RecordPath that points to the field to be
> removed)
> The processor should be able to accept additional dynamic properties that
> specify further fields (by RecordPath) to be removed from the record.
> +*Example*+
> +input:+
> {code:java}
> {
> "id": 1,
> "name": "John",
> "address": {
> "zip": 1111,
> "street": "Main",
> "building": 11
> }
> }
> {code}
> +Field to remove:+ /address/building
> +output:+
> {code:java}
> {
> "id": 1,
> "name": "John",
> "address": {
> "zip": 1111,
> "street": "Main"
> }
> }
> {code}
> The record's schema should be modified accordingly (removing the
> /address/building field from the schema). Field removal should be permitted
> regardless of the field being nullable or not.
> Generally, the removal of a field should include the field's removal from the
> schema AND the data. The exception is if the removal is data-dependent (the
> field should be removed if its value equals "xyz"). In this case no schema
> modification should occur.
> The processor should be able to remove one or more elements from arrays
> (e.g.: /addresses[ 1 ] shoud remove the element from the addresses array from
> the 1st position). When removing a field from elements of an array, the
> array's schema should only be modified if the removal is applied to all
> elements of the array (i.e.: /addresses[ * ]/building should modify the
> schema of the array, but /addresses[ 1 ]/building should not).
> The same rule should be applied when handling Map datatype.
> If the record does not contain the field that is expected to be removed, the
> record should still be transferred to the 'success' relationship, with no
> modification. The expectation is that if /x/y is expected to be removed from
> the record, then the record leaving the processor should not contain /x/y
> field.
> If a certain field can be of different types (e.g. the address field can be
> "string" as well as "record" and possibly another "record" with a schema
> different from the former record type) then if /address/building is expected
> to be removed from the record, the processor is expected to remove the
> building field from the schema of all the possible types of the address field
> regardless of the address field being whatever concrete type in the
> particular record that is being processed.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
