One way (in NiFi 0.5.0+) is to use the ExecuteScript processor, which gives
you full control over the session and flowfile(s). For example if you had
JSON in your "kafka.key" attribute such as "{"data": {"myKey": "myValue"}}"
, you could use the following Groovy script to parse out the value of the
'data.myKey' field:
def flowfile = session.get()
if(!flowfile) return
def json = new
groovy.json.JsonSlurper().parseText(flowfile.getAttribute('kafka.key'))
flowfile = session.putAttribute(flowfile, 'myKey', json.data.myKey)
session.transfer(flowfile, REL_SUCCESS)
I put an example of this up as a Gist (
https://gist.github.com/mattyb149/478864017ec70d76f74f)
A possible improvement could be to add a "jsonPath" function to Expression
Language, which could take any value (including an attribute) along with a
JSONPath expression to evaluate against it...
Regards,
Matt
On Mon, Mar 21, 2016 at 1:48 PM, McDermott, Chris Kevin (MSDU -
STaTS/StorefrontRemote) <chris.mcderm...@hpe.com> wrote:
> Joe,
>
> Thanks for the reply. I think I was not clear.
>
> The JSON I need to evaluate is in a FlowFile attribute (kafka.key) which I
> need to be able to evaluate without modifying the original FlowFile content
> (which was read from the Kafka topic). What I can’t figure out is how to
> squirrel away the flowfile content so that I can write the value of the
> kafka.key attribute to the FlowFile content, so that I can process it with
> EvaluateJsonPath, and then read content I squirreled away back into the
> FlowFile content. I considered using the the DistributedMapCache, but there
> would be no guarantee what I added to the cache would still be there when I
> needed it back.
>
>
>
>
> On 3/21/16, 1:37 PM, "Joe Witt" <joe.w...@gmail.com> wrote:
>
> >Chris,
> >
> >Sounds like you have the right flow in mind already. EvaluateJSONPath
> >does not write content. It merely evaluates the given jsonpath
> >expression against the content of the flowfile and if appropriate
> >creates a flowfile attribute of what it finds.
> >
> >For example if you have JSON from Twitter you can use EvaluateJsonPath
> >and add a property with a name
> >'twitter.user' and a value of '$.user.name'
> >
> >Once you run the tweets through each flow file will have an attribute
> >called 'twitter.user' with the name found in the message. No
> >manipulation of content at all. Just promotes things it finds to flow
> >file attributes.
> >
> >Thanks
> >Joe
> >
> >On Mon, Mar 21, 2016 at 1:34 PM, McDermott, Chris Kevin (MSDU -
> >STaTS/StorefrontRemote) <chris.mcderm...@hpe.com> wrote:
> >> What I need to do is read a file from Kafka. The Kafka key contains a
> JSON string which I need to turn in FlowFile attributes while preserving
> the original FlowFile content. Obviously I can use EvaluteJsonPath but
> that necessitates replacing the FlowFile content with the kaka.key
> attribute, thus loosing the original FlowFile content. I feel like I’m
> missing something fundamental.
>
