I am working on NIFI-4456 which will allow the JSON reader/writer to support the "one JSON per line" format as well as the "JSON array" format for input and output, so you will be able to read in one JSON per line and output a JSON array, using ConvertRecord (or any other record-aware processor). In the meantime, you can use the following crude script in an ExecuteGroovyScript processor to process your entire file (avoiding the Split/Merge pattern), it should get you what you want: def flowFile = session.get()
if(!flowFile) return
flowFile = session.write(flowFile, {inStream, outStream ->
outStream.write('['.bytes)
inStream.eachLine { line, i ->
if(i > 1) outStream.write(','.bytes)
outStream.write(line.bytes)
}
outStream.write(']'.bytes)
} as StreamCallback)
session.transfer(flowFile, REL_SUCCESS) The script just adds array brackets around the whole doc, and separates the lines by a comma. I did the crude version because it doesn't need to load the entire input content into memory. If you need more control over the JSON objects, you could iterate over the lines (still with eachLine), use JsonSlurper to deserialize each string into a JSON object, then add each object to an array, then use JsonOutput to serialize the whole thing back to a string. However that involves having the entire content in memory and could get unwieldy for large input flow files.
... View more
