Getting a type mismatch on String where CharSequence is expected in Scala on Avro generated java class

Question

I have created a simple Avro (1.7.7) schema and let Avro generate my Java classes which are jarred up and added to my Spark project. In the generated java class I have:

public java.util.List<java.util.Map<java.lang.CharSequence,java.lang.CharSequence>> attributes;

As part of the code created from the schema that looks like this:

{
  "namespace": "com.rbh.avro.metric",
  "type": "record",
  "name": "Metric",
  "fields": [
    { "name": "consumerId", "type": "long" },
    {
      "name": "data",
      "type": {
        "name": "Data",
        "type": "record",
        "fields": [
          { "name": "name", "type": "string" },
          { "name": "attributes",
            "type": {
              "type": "array",
              "items": {
                "type": "map",
                "values": "string"
              }
            }
          }
        ]
      }
    }
  ]
}

So, as a String is a CharSequence, some of the code I'm testing out looks like this:

val data:Data = Data.newBuilder.setName("myData").build
val x = Map("a" -> "1", "b" -> "2", "c" -> "3").asJava
val y = Map("x" -> "42", "y" -> "27", "z" -> "-1").asJava
val z = Map("g" -> "a", "h" -> "*", "i" -> "~").asJava
val xyz = List(x,y,z).asJava
data.setAttributes(xyz)

Eclipse complains about this immediately and on compile (Scala 2.10.5 & SBT 0.13.8) I see:

[error] /home/bkarels/dev/spark/event-gen/src/main/scala/com/rbh/generator/Generator.scala:60: type mismatch;
[error]  found   : java.util.List[java.util.Map[String,String]]
[error]  required: java.util.List[java.util.Map[CharSequence,CharSequence]]
[error]         data.setAttributes(xyz)
[error]                            ^
[warn] 5 warnings found
[error] one error found
[error] (compile:compileIncremental) Compilation failed
[error] Total time: 11 s, completed Jun 26, 2015 8:12:31 AM

There begins my confusion...

So, I opened a REPL session and this worked:

scala> def foo(cs:CharSequence) = {
     |   println(cs)}
foo: (cs: CharSequence)Unit

scala> def s:String = "bar"
s: String

scala> foo(s)
bar

I also, for some kind of verification, put this code in and it compiled and ran as expected:

val data:Data = Data.newBuilder.setName("myData").build
val cs0:CharSequence = "xxx"
val w = Map(cs0 -> cs0).asJava
val wl = List(w).asJava
data.setAttributes(wl)

As one might expect, this kind of clumsy verbosity is not a "solution" that has wings...

So something is amiss, but I am at a loss as to what it might be. Please let me know if you have thoughts - thank you.

Answer 1

Map[String, String] is not Map[CharSequence, CharSequence] because Map[CharSequence, CharSequence] allows the addition of key/value pairs that aren't String. See this SO question for an excellent answer.

Answer 2

I have faced this issue in Java as well while generating POJO from avroschema where the variables are created with a CharSequence for all string declarations. As mentioned in the comments above CharSequence is an interface and String implements. So a direct assignment of a String value always failed. I was using a clumsy quick solution of using .toString() to get the String of the value.

Upon googling I happened to see there was a maven-compiler tag which was to convert string as String and now I'm able to forget completely about CharSequence in my POJOs created from avro schemas. Not sure how SBT would be able to provide this feature