[ 
https://issues.apache.org/jira/browse/NIFI-361?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15247682#comment-15247682
 ] 

ASF GitHub Bot commented on NIFI-361:
-------------------------------------

Github user davetorok commented on the pull request:

    https://github.com/apache/nifi/pull/354#issuecomment-211909411
  
    We've been working with JOLT for awhile within NiFi and appreciate having 
it supported as a standard processor within NiFi! I have some suggestions.
    
    1.  I would change the PR to support all the built-in transformations 
including 'sort' and 'cardinality' which are not in the pull request.  
    
    2. 90% of what we've seen uses ChainR as that is a composition of 1 or more 
individual transforms, almost always a "ShiftR + DefaultR".  In fact it is a 
little confusing to support options other than just "Chainr" since ChainR takes 
an array of individual transform Specs whereas the others are individual specs. 
 At minimum I would make CHAINR the default.
    
    3.  Jolt supports additional custom transforms via fully-qualified Java 
classnames.  We use a couple.  I would recommend adding documentation on how to 
support it via a drop-in jar similar to how ExecuteScript supports additional 
resources.  Example custom transform is here 
https://github.com/zacheusz/jolt-date  
    



> Create Processors to mutate JSON data
> -------------------------------------
>
>                 Key: NIFI-361
>                 URL: https://issues.apache.org/jira/browse/NIFI-361
>             Project: Apache NiFi
>          Issue Type: New Feature
>            Reporter: Aldrin Piri
>            Assignee: Oleg Zhurakousky
>            Priority: Minor
>             Fix For: 1.0.0, 0.7.0
>
>
> Creating a separate issue to track these as a pull request has been submitted 
> for related issue NIFI-356.
> Also backed by JsonPath, processors should facilitate through specification 
> of user-defined properties:
> * Add - identify path and add key/value pair
> ** Handle if the path is an array, this would ignore the name specified and 
> just add the value to the collection
> * Remove - delete the element at the specified path
> * Update - change the value for the given path to a provided value
> Need to determine if objects/arrays make sense for values or if they are 
> needed.
> While it would be nice to be able to execute several operations per processor 
> instance, it may be hard to capture all the relevant information needed for 
> multiple operations in one processor configuration in a user friendly context.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to