Hi @Fredi , Its hard to say what is happening without looking at the data where optionalDict is not empty. You only provided data when its empty. Keep in mind that this is not true Python its actually flavor of it called Jython so its not apple to apple when comparing to python. If I can suggest two alternatives: 1 - Since Jython script is going to  be deprecated starting from version 2.0 , then I would recommend using groovy instead . Actually parsing json in groovy is much simpler than Jython. Im not sure what version you are using but there is a dedicated processor for executing groovy script called ExecuteGroovyScript that is probably faster than running traditional ExecuteScritp which you can still use it. The script looks like this based on your input :       import org.apache.commons.io.IOUtils import java.nio.charset.StandardCharsets import groovy.json.JsonSlurper import groovy.json.JsonOutput flowFile = session.get() if(!flowFile) return def text = '' // Cast a closure with an inputStream parameter to InputStreamCallback session.read(flowFile, {inputStream -> text = IOUtils.toString(inputStream, StandardCharsets.UTF_8) } as InputStreamCallback) def jsonSlurper = new JsonSlurper() def jsonData = jsonSlurper.parseText(text) if(jsonData.directories[0]) { session.remove(flowFile) jsonData.directories.each { d -> newflowfile = session.create() newflowfile = session.write(newflowfile, {inputStream, outputStream -> outputStream.write(JsonOutput.toJson(d).getBytes(StandardCharsets.UTF_8)) } as StreamCallback) newflowfile = session.putAttribute(newflowfile, "setId", jsonData.setId.toString()) newflowfile = session.putAttribute(newflowfile, "setName", jsonData.setName) newflowfile = session.putAttribute(newflowfile, "absolute.path", d.path) if(jsonData.optionalDict) { newflowfile = session.putAttribute(newflowfile, "value1", jsonData.optionalDict.set_entity_relation.intValue.toString()) newflowfile = session.putAttribute(newflowfile, "value2", jsonData.optionalDict.set_entity_relation.stringValue) } session.transfer(newflowfile, REL_SUCCESS) } } else session.transfer(flowfile, REL_FAILURE)       I have tried the script for both scenarios and it worked as expected. 2- The other alternative is to use other processors (nifi way) to achieve what you want without executing script (not the nifi way) . The execute processor should be left as last option incase the out of the box processors dont suffice or you looking to improve performance in case the flow gets very complicated and inefficient. For this I would use the following processors: 1- JsonEvaluatePath to extract common attribues: setId, SetName, optionalDict.value 1 & 2..etc. 2-Do JsonSplit or QueryRecords on the directories object: this will produce different flowfiles and each flowfile will have the common attribute. 3- JsonEvaluatePath to extract each directory attributes even though its already part of the flowfile content. Hopefully that helps. If it does please accept the solution. Thanks ... View more

Just for everyones information; There was nothing wrong with the code. It was simply that the docker image I used to run Nifi did not build properly and therefore my modifications weren't taken into account. ... View more

Hey @Fredi , I believe the answer for your problem is the processor UpdateRecord.   Update record allows you to directly manipulate the fields in your file content. You add dynamic properties to the processor where the key of the property is /<field> (so in your case, '/api_value'), and in the value of this dynamic property you can write down some logic to determine what value to insert into api_value.   In the processor, there is a field called "Replacement Value Strategy", which defines how the value of the property will be read. If you set this to "Record Path Value", it means you can now give a path to a different field in your file (url_value!) - I can't test this right now because I'm not at my office, but I'm not entirely sure whether you can manipulate the result after giving a record path (to extract the api_value from the evaluated url_value).   Regardless, I'm just about 100% sure this can be done with two processors - One EvaluateJsonPath to extract the url_value into an attribute, then UpdateRecord that uses the 'Literal Value' replacement strategy - with this strategy, you can just add a property with key '/api_value' and value '${url_value}' (or whatever attribute name you gave to the extracted url_value) and once you can access url_value with the expression language (via  ${url_value}) you can use all the available functions to manipulate expression language variables.   Here's an article with a couple of examples on UpdateRecord: https://community.cloudera.com/t5/Community-Articles/Update-the-Contents-of-FlowFile-by-using-UpdateRecord/ta-p/248267  (I noticed in the article they used some recordPath related functions like "replaceRegex", so I believe there might be a way to use these and then limit the entire issue to just one UpdateRecord processor! Sadly I'm not too familiar with these myself and this was the first time I've seen them)   And here's the expression language documentation: https://nifi.apache.org/docs/nifi-docs/html/expression-language-guide.html  You can see there are lots of useful functions to extract your api_value once you have ${url_value} as an attribute variable, for example "substring", "find"/"replace", "ifElse", etc. all of which you can try and use to ensure only the api_value is left in the end.   Hope this helps! I'm sure using ReplaceText and possibly JoltTransform could provide alternate solutions to the issue, however I believe UpdateRecord is the cleanest solution for this and truly makes use of the processor's abilities. If you struggle to use it correctly, you can reply with an example json and expected output and I'll try to write down the flow when I have time. ... View more