Iterate through a column in Dataset which have array of key value pairs and find out a pair with max value

问题

I have data in a dataframe , which was obtained from azure eventhub. Then I convert this data to json object and stored the required data into a dataset as shown below.

Code for obtaining data from eventhub and store it into a dataframe.

val connectionString = ConnectionStringBuilder(<ENDPOINT URL>)
    .setEventHubName(<EVENTHUB NAME>).build

val currTime = Instant.now
val ehConf = EventHubsConf(connectionString)
    .setConsumerGroup("<CONSUMER GRP>")
    .setStartingPosition(EventPosition
             .fromEnqueuedTime(currTime.minus(Duration.ofMinutes(30))))
    .setEndingPosition(EventPosition.fromEnqueuedTime(currTime))

val reader =  spark.read.format("eventhubs").options(ehConf.toMap).load()

var SIGNALS =  reader
    .select(get_json_object(($"body").cast("string"),"$.NUM").alias("NUM"),
            get_json_object(($"body").cast("string"),"$.SIG1").alias("SIG1"),
            get_json_object(($"body").cast("string"),"$.SIG2").alias("SIG2"),
            get_json_object(($"body").cast("string"),"$.SIG3").alias("SIG3"),
            get_json_object(($"body").cast("string"),"$.SIG4").alias("SIG4")
     )

val SIGNALSFiltered = SIGNALS.filter(col("SIG1").isNotNull &&
    col("SIG2").isNotNull && col("SIG3").isNotNull && col("SIG4").isNotNull)

The data obtained at SIGNALSFiltered is shown below.

+-----------------+--------------------+--------------------+--------------------+--------------------+
|              NUM|                SIG1|                SIG2|                SIG3|                SIG4|
+-----------------+--------------------+--------------------+--------------------+--------------------+
|XXXXX01|[{"TIME":15695605310...|[{"TIME":15695605310...|[{"TIME":15695605310...|[{"TIME":15695605310...|
|XXXXX02|[{"TIME":15695604780...|[{"TIME":15695604780...|[{"TIME":15695604780...|[{"TIME":15695604780...|
|XXXXX03|[{"TIME":15695605310...|[{"TIME":15695605310...|[{"TIME":15695605310...|[{"TIME":15695605310...|
|XXXXX04|[{"TIME":15695605310...|[{"TIME":15695605310...|[{"TIME":15695605310...|[{"TIME":15695605310...|
|XXXXX05|[{"TIME":15695605310...|[{"TIME":15695605310...|[{"TIME":15695605310...|[{"TIME":15695605310...|
|XXXXX06|[{"TIME":15695605340...|[{"TIME":15695605340...|[{"TIME":15695605340...|[{"TIME":15695605340...|
|XXXXX07|[{"TIME":15695605310...|[{"TIME":15695605310...|[{"TIME":15695605310...|[{"TIME":15695605310...|
|XXXXX08|[{"TIME":15695605310...|[{"TIME":15695605310...|[{"TIME":15695605310...|[{"TIME":15695605310...|

If we check entire data for a single row it will be as below.

|XXXXX01|[{"TIME":1569560531000,"VALUE":3.7825},{"TIME":1569560475000,"VALUE":3.7812},{"TIME":1569560483000,"VALUE":3.7812},{"TIME":1569560491000,"VALUE":34.7875}]|
    [{"TIME":1569560537000,"VALUE":3.7825},{"TIME":1569560481000,"VALUE":34.7825},{"TIME":1569560489000,"VALUE":34.7825},{"TIME":1569560497000,"VALUE":34.7825}]|
    [{"TIME":1569560505000,"VALUE":34.7825},{"TIME":1569560513000,"VALUE":34.7825},{"TIME":1569560521000,"VALUE":34.7825},{"TIME":1569560527000,"VALUE":34.7825}]|
    [{"TIME":1569560535000,"VALUE":34.7825},{"TIME":1569560479000,"VALUE":34.7825},{"TIME":1569560487000,"VALUE":34.7825}]

I want only the highest TIME pair from each column, not the entire TIME VALUE pairs. Output should be as shown below.

+-----------------+-----------------------------+---------------------------------------+---------------------------------------+----------------------------------------+
|              NUM|                         SIG1|                                   SIG2|                                   SIG3|                                    SIG4|
+-----------------+-----------------------------+---------------------------------------+---------------------------------------+----------------------------------------+
|XXXXX01|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":4.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":5.7825}]|
|XXXXX02|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":6.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":7.7825}]|
|XXXXX03|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":9.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":8.7825}]|

How to Iterate through each column in each row and get the highest TIME-VALUE pair?
After getting highest in each columns (SIG1,....SIG4) have to update only the value of TIME in all columns with highest among them.
Is there Any way to convert the base dataset as below?. Each elements in a column should be converted to a new row.

+-----------------+-----------------------------+---------------------------------------+---------------------------------------+----------------------------------------+
|    NUM|                                   SIG1|                                   SIG2|                                   SIG3|                                    SIG4|
+-----------------+-----------------------------+---------------------------------------+---------------------------------------+----------------------------------------+
|XXXXX01|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|
|XXXXX01|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|
|XXXXX01|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|        null                           |[{"TIME":1569560531000,"VALUE":3.7825}]|
|XXXXX01|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|
|XXXXX02|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|
|XXXXX02|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|
|XXXXX02|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|
|XXXXX02|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|[{"TIME":1569560531000,"VALUE":3.7825}]|```

Any leads or help is appreciated! Thanks in Advance.

回答1:

You have to write one user defined function like below. which will loop your data and get Max Time Value. Note: UDF is just for reference, you can change it as per requirement

How to Iterate through each column in each row and get the highest TIME-VALUE pair?

scala> import org.apache.spark.sql.expressions.{UserDefinedFunction} 

scala> def MaxTime:UserDefinedFunction = udf((json:String) => {
   val pars = JSON.parseFull(json)
   var output=""
   pars.foreach{ x => val y = x.asInstanceOf[List[Any]]
     var i = 1
     var TimeMap = scala.collection.mutable.Map[String, Long]()
     var ValueMap = scala.collection.mutable.Map[String, Double]()
     y.foreach{ zz => val z =  zz.asInstanceOf[Map[String,Double]]
       TimeMap(i.toString) =  z("TIME").toLong
       ValueMap(i.toString) =  z("VALUE")
       i = i + 1
     }
   output = """[{"TIME" : """ + TimeMap.maxBy(_._2)._2.toString + """ ,"VALUE": """ + ValueMap(TimeMap.maxBy(_._2)._1) + """}]"""
  } 
output})

scala> SIGNALSFiltered.withColumn("SIG1", MaxTime(col("SIG1")).withColumn("SIG2", MaxTime(col("SIG2")))).withColumn("SIG3", MaxTime(col("SIG3"))).withColumn("SIG4", MaxTime(col("SIG4"))).show(false)

After getting highest in each columns (SIG1,....SIG4) have to update only the value of TIME in all columns with highest among them.

Write same UDF like above and pass complete row as a parameter. Then parse each column value into Map and get Maximum among all columns.

来源：https://stackoverflow.com/questions/58128746/iterate-through-a-column-in-dataset-which-have-array-of-key-value-pairs-and-find

标签

dataframe

apache-spark

apache-spark-sql

apache-spark-dataset