spark scala cannot resolve column with using agg-CodePudding

I have the following code:

def proccess(spark: SparkSession, df : DataFrame): DataFrame= {
  val mydf = df.withColumn("next_intent_temp", when (col("next_intent")=== "val", 1).otherwise(0)
  
  val mynewdf = mydf.groupBy(col("a"), col("b"), col("c")
                .agg(sum(col("next_intent_temp")).as("next_intent"))
                .select(col("market"), struct(col("next_intent"),...).alias("data")
    )
  mynewdf 

}

Running it gives:

cannot resolve 'next_intent_temp' given input columns [...]; Aggregate [sum('next_intent_temp')...

I don't really understand this error because when I do mydf.show() I do see the column next_intent_temp with the correct values. so why the sum fails with this error?

Noting: I simplified the code for the question but preserved the structure that I really have.

CodePudding user response：

You have .as("next_intent"). Use that column name.

I think you have some syntax and missing ) aspects.

This is a simpler example, but if you follow it, you should be ok.

import spark.implicits._
import org.apache.spark.sql.functions._

val df = spark.sparkContext.parallelize(Seq( (1,7,"ST"), (1,8,"XX"), (1,9,"RW"), (3,10,"ST"), (3,11,"AA"), (3,12,"RW"), (2,3,"ST"), (2,4,"TT"))).toDF("i", "c", "t")
val df2 = df.withColumn("next_intent_temp", lit(1))

val df3 = df2.groupBy(col("i"), col("c"), col("t"))
             .agg(sum(col("next_intent_temp")).as("next_intent"))
             .select(col("i"), struct(col("next_intent")).alias("data"))
                        
df3.show(false)