Why are there so many #Unsupported for supported Expressions?

[manticore-projects]

I would like to understand better, why so many productions return #Unsupported.
Example:

SELECT NEXT VALUE FOR a from b;

This valid statement is parsed perfectly fine, but instead of a NEXT VALUE FOR Node it would return an Unsupported Node.

SQL Text
 └─DirectSqlStatement [1:1-1:31]
    └─Select [1:1-1:30]
       ├─SelectList [1:8-1:23]
       │  └─SelectItem [1:8-1:23]
       │     └─Unsupported [1:8-1:23]
       │        └─SchemaQualifiedName [1:23-1:23]
       │           └─Identifier: a [1:23-1:23]
       └─TableExpression [1:25-1:30]
          └─FromClause [1:25-1:30]
             └─TableName [1:30-1:30]
                └─Identifier: b [1:30-1:30]

Please care to elaborate the idea behind, I thank you a lot in advance.

[manticore-projects]

Being very curios, I just changed the production:

void next_value_expression() #NextValueFor:
{}
{
    "NEXT" "VALUE" "FOR" schema_qualified_name()
}

And as expected, I get an useful AST immediately:

SELECT NEXT VALUE FOR a from b;

SQL Text
 └─DirectSqlStatement [1:1-1:31]
    └─Select [1:1-1:30]
       ├─SelectList [1:8-1:23]
       │  └─SelectItem [1:8-1:23]
       │     └─NextValueFor [1:8-1:23]
       │        └─SchemaQualifiedName [1:23-1:23]
       │           └─Identifier: a [1:23-1:23]
       └─TableExpression [1:25-1:30]
          └─FromClause [1:25-1:30]
             └─TableName [1:30-1:30]
                └─Identifier: b [1:30-1:30]

[kaikalur]

We started with what Presto currently supports (for execution) only as real nodes and others as unsupported. In fact, I started with everything as unsupported and changed only those features supported by Presto as real nodes. That way when someone wants to implement a features, they can start with the grammar and go from there.

[manticore-projects]

Thank you, this makes it clear.
So would I be allowed to transform the #Unsuppported into real Nodes as long as a) they work for particular test cases and b) the work with the complete test suite?

[kaikalur]

Yeah, we can probably relax that now. See if you can make it conditional so that we give error for unsupported features if/when we use it in Presto.

Another option would be to add a validator class that has a hard-coded list of nodes that we currently have and any other node kind will be error out (and use that in the Presto version)