endzeit created NIFI-12386:
------------------------------
Summary: Add a FilterAttributes processor
Key: NIFI-12386
URL: https://issues.apache.org/jira/browse/NIFI-12386
Project: Apache NiFi
Issue Type: New Feature
Components: Core Framework
Reporter: endzeit
Assignee: endzeit
Flows in Apache NiFi can get quite sophisticated, consisting of a long chains
of both {{ProcessGroup}} and {{Processor}} components.
Oftentimes {{Processor}} components, including those in the NiFi standard
bundle, enrich an incoming {{FlowFile}} with additional FlowFile attributes.
This can lead to a fair amount of different FlowFile attributes accumulating
over the FlowFile's lifecycle.
In order to prevent subsequent {{ProcessGroup}} / {{Processor}} components to
accidentally rely on implementation details of preceding components, a good
practice is to:
# define which FlowFile attributes should exist at selected points in the
{{Flow}}
# reduce the attributes of the FlowFile at the selected point to those defined
This can be achieved by using the
[UpdateAttribute|https://nifi.apache.org/docs/nifi-docs/components/org.apache.nifi/nifi-update-attribute-nar/1.23.2/org.apache.nifi.processors.attributes.UpdateAttribute/index.html]
processor of the standard processor bundle.
However, the {{UpdateAttribute}} processor allows only for a regular expression
to define a set of attributes to remove. Instead, the outlined practice above
desires to explicitly state a set of attributes to keep. One can do so with a
regular expression as well, but writing the reverse lookup to achieve this is
not the easiest endeavor to put it mildly.
This issue proposes a new processor {{FilterAttributes}} to be added to the
library of {{{}nifi-standard-processors{}}}, which can be configured with a set
of attributes and removes all attributes of an incoming FlowFile other than the
ones configured.
The processor should
* have a required, non-blank property "Attributes to keep", which takes a list
of attribute names separated by delimiter, e.g. comma (,).
** trailing whitespace around attribute names should be ignored
** leading or trailing delimiters should be ignored
* have a required, non-blank property "Delimiter", which is used to delimit
the individual attribute names, with a default value of "," (comma).
* have a single relationship "success" to which all FlowFiles are routed,
similar to {{UpdateAttribute}}
* have an
[InputRequirement|https://www.javadoc.io/doc/org.apache.nifi/nifi-api/latest/org/apache/nifi/annotation/behavior/InputRequirement.html]
of
[INPUT_REQUIRED|https://www.javadoc.io/doc/org.apache.nifi/nifi-api/latest/org/apache/nifi/annotation/behavior/InputRequirement.Requirement.html]
*
[@SupportsBatching|https://www.javadoc.io/doc/org.apache.nifi/nifi-api/latest/org/apache/nifi/annotation/behavior/SupportsBatching.html]
* be
[@SideEffectFree|https://www.javadoc.io/doc/org.apache.nifi/nifi-api/latest/org/apache/nifi/annotation/behavior/SideEffectFree.html]
Some possible extension might be:
* have a required property "Core attributes", with allowable values of "Keep
UUID only", "Keep all", with a default of "Keep UUID only"
** an additional allowable value e.g. "Specify behaviour" may be added, which
allows for more customization
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
