Search code examples
scalaapache-sparkapache-spark-sql

How to compare two StructType sharing same contents?


It seems like StructType preserves order, so two StructType containing same StructFields are not considered equivalent.

For example:

val st1 = StructType(
StructField("ii",StringType,true) :: 
StructField("i",StringType,true) :: Nil)

val st2 = StructType(
StructField("i",StringType,true) :: 
StructField("ii",StringType,true) :: Nil)

println(st1 == st2)

returns false even though they both have StructField("i",StringType,true) and StructField("ii",StringType,true), just in different order.

I need a test that can say that these two are equivalent because for my purpose, these two are not different.

val schema1 = StructType(StructField("A",ArrayType(st1,true),true) :: Nil)

val schema2 = StructType(StructField("A",ArrayType(st2,true),true) :: Nil)

val final_schema = StructType((schema1 ++ schema2).distinct)

The result of final_schmea should only have one StructType of A instead of two, but distinct considers these two StructType as different, so I end up getting two different StructField named A. So my question is, is there a way to compare two StructTypes based on their contents, not on orders?

EDIT:

After further investigation, since StructType is basically Seq<StructField>, I can do content comparison for that works for Seq, but I am trying to think of a way I can do comparison for embedded StructType most efficiently.


Solution

  • This can probably be cleaned up, but it works and handles nested StructType:

    def isEqual(struct1: StructType, struct2: StructType): Boolean = {
      struct1.headOption match {
        case Some(field) => {
          if(field.dataType.typeName != "struct") {
            struct2.find(_ == field) match {
             case Some(matchedField) => isEqual(StructType(struct1.filterNot(_ == field)), StructType(struct2.filterNot(_ == field)))
             case None => false
            }
          } else {
            val isEqualContents = struct2.find(x => x.name == field.name && x.nullable == field.nullable && x.dataType.typeName == "struct") match {
              case Some(matchedField) => isEqual(field.dataType.asInstanceOf[StructType], matchedField.dataType.asInstanceOf[StructType])
              case None => false
            }
            if(isEqualContents) isEqual(StructType(struct1.filterNot(_ == field)), StructType(struct2.filterNot(_ == field))) else false
          }
        }
        case None => struct2.size == 0
      }
    }
    
    val st1 = StructType(
    StructField("ii",StringType,true) :: 
    StructField("i",StringType,true) :: 
    StructField("iii", StructType(StructField("iv", StringType, true) :: Nil), true) :: Nil)
    
    val st2 = StructType(
    StructField("i",StringType,true) :: 
    StructField("ii",StringType,true) :: 
    StructField("iii", StructType(StructField("v", StringType, true) :: Nil), true) :: Nil)
    
    isEqual(st1, st2)
    

    It could also use a little more love to become tail recursive, too.