@zhangliang to accomplish that i would use UpdateRecord Since your data is csv and structured we can use record manipulation to accomplish this. First I would treat all your values as string and build an avro schema to use: {
"type":"record",
"name":"nifiRecord",
"namespace":"org.apache.nifi",
"fields":[
{"name":"test_a","type":["null","string"]},
{"name":"test_b","type":["null","string"]},
{"name":"test_c","type":["null","string"]},
{"name":"test_d","type":["null","string"]},
{"name":"test_e","type":["null","string"]}
]
} Then I would configure my UpdateRecord to use a CSV Reader and a CSV Writer I would configure the CSV Reader like this: Use schema text property Schema Text = Put your avro schema there Value Separator = | And the CSV Writer leave everything default except: Value Separator = | Finally the UpdateRecord processor will need 2 user fields. In this case we want to update the fields "test_c" and "test_d" And then we can use Record path manipulation and in particular for this use case the substringBefore function to only give us everything before the DOT "." Here is what you should configure: This will then take an input like this: test_a|test_b|test_c|test_d|test_e
a|b|3.0|4.0|5.0
a|b|3.0|4.0|5.0
a|b|3.0|4.0|5.0 and produce an output like this: test_a|test_b|test_c|test_d|test_e
a|b|3|4|5.0
a|b|3|4|5.0
a|b|3|4|5.0
... View more
