In the personinfo example (https://github.com/linkml/linkml/blob/main/examples/PersonSchema/personinfo/personinfo.py) we end up with

@dataclass
class Person(NamedThing):
    ...
    has_employment_history: Optional[Union[Union[dict, "EmploymentEvent"], List[Union[dict, "EmploymentEvent"]]]] = empty_list()

This is nice as it allows us to be very loose when instantiating objects. However, it is hard to read. And it also doesn't inform either us or our IDE of what the 'normalized' form of the object is

E,g,

if I do

p = Person('x', has_employment_history={'is_current': 1})

it will normalize this from a single dict to a list of objects

>>> type(p.has_employment_history)
<class 'list'>

I am not sure if we need this level of looseness in creation - it is handy to convert from a dict to an object but auto-converting objects to lists seems to encourage bad use. TBD: check if this is required for jsonld strategy

Should we:

• add an option to have stricter constructors in pythongen
• leave as is, and focus on pydanticgen (@kevinschaper ) as the preferred method for making an OM?