[WIP] remove expensive api from InternalRow

[davies]

This is a follow up PR for #6792, to use narrow down the APIs in InternalRow (remove those expensive ones).

[davies]

cc @liancheng @rxin @JoshRosen

[cloud-fan]

hi @davies , I met a bug at #6825 that we didn't handle BinaryType(byte[]) correctly at equals and hashCode. We should use java.util.Arrays.equals and java.util.Arrays.hashCode. Could you fix it here?

[SparkQA]

Test build #35092 has finished for PR 6869 at commit 444382e.

• This patch fails Spark unit tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
  • trait Row extends InternalRow
  • abstract class InternalRow extends Serializable
  • final class SpecificMutableRow(val values: Array[MutableValue]) extends Row with MutableRow
  • class GenericRow(protected[sql] val values: Array[Any]) extends Row
  • abstract class InternalOutputWriter extends OutputWriter

[rxin]

do we need size? maybe just keep one of length or size.

[cloud-fan]

agree. It's internal and we don't need to follow scala convention to provide size.

[rxin]

@davies can you update the pr description to include what you actually removed?

[davies]

@cloud-fan it will be fixed by #6876

[cloud-fan]

As it's internal row, should we just return UTF8String?

[cloud-fan]

Maybe we can only define getUTF8String at InternalRow and define getString at user-facing Row?

[davies]

Close this, in favor of #7003