@JelteF Is this the type of output we expect when selecting a union datatype column?
Also Can you help with some examples of how union_extract and union_tag functions will be used with the datatypes? Some sample queries...

Yes, that looks great. I think we should at least support the following:

select union_extract(r['u'], 'str') from duckdb.query($$ select * from tbl1 $$) r;
select union_tag(r['u']) from duckdb.query($$ select * from tbl1 $$) r;

This would also be nice, but since duckdb doesn't support indexing a union this way I think that will be impossible to do, without changes to DuckDB.

select r['u']['str'] from duckdb.query($$ select * from tbl1 $$) r;

I think you could in theory implement the following syntax without changes to duckdb. But I think a better way would be to add index support for unions to duckdb itself:

select r['u']::duckdb.union['str'] from duckdb.query($$ select * from tbl1 $$) r;

Finally since postgres doesn't support overloading the . operator it's impossible to implement support for the following syntax without changes to DuckDB:

select r['u'].str from duckdb.query($$ select * from tbl1 $$) r;

Yes, that looks great. I think we should at least support the following:

select union_extract(r['u'], 'str') from duckdb.query($$ select * from tbl1 $$) r;
select union_tag(r['u']) from duckdb.query($$ select * from tbl1 $$) r;

This would also be nice, but since duckdb doesn't support indexing a union this way I think that will be impossible to do, without changes to DuckDB.

select r['u']['str'] from duckdb.query($$ select * from tbl1 $$) r;

I think you could in theory implement the following syntax without changes to duckdb. But I think a better way would be to add index support for unions to duckdb itself:

select r['u']::duckdb.union['str'] from duckdb.query($$ select * from tbl1 $$) r;

Finally since postgres doesn't support overloading the . operator it's impossible to implement support for the following syntax without changes to DuckDB:

select r['u'].str from duckdb.query($$ select * from tbl1 $$) r;

I think the first syntax is the most straigth forward , and have implemented that and updated output in PR description...