Stuff to do:
- check list/scalar context for them
- improve initial test
- add more tests