Contributor

This PR fixes 3 small bugs:

1. When you call asin(2.0) or acos(5.0) (values outside the valid range), PySpark is supposed to return NaN. DuckDB
   was returning NULL instead, which breaks code that expects PySpark behavior. Fixed by using a SQL workaround since
   the internal C++ layer converts NaN to NULL by design.

2. The spark.read.json() method had all the code written, but a raise NotImplementedError at the end blocked it
   from ever returning. One line removal fixed it.

3. Instead of hitting a generic NotImplementedError when encountering DuckDB union types (which PySpark doesn't
   support), it now throws a clear ContributionsAcceptedError explaining the limitation.

[minor] add suport for types and print schema

b32e901

fix typo

748031d

add treeString support for printSchema

84b3708

fix formatting

1db145c

fix NaN response

c42e3f8

minor fixes

320deed

fix ruff check

bfa84c1

fix test

e3e4b96

fix precommit

0bfeccb

revert unncessary change

e410e7a

remove unncessary comment

b8c499c

Collaborator

Thanks! I rebased onto v1.4-andium. When the tests pass I'll do a review.