The syntax for explicitly creating a table looks like this:

db["cats"].create({
    "id": int,
    "name": str,
    "weight": float,
}, pk="id")

So using a dictionary of column name -> Python type here would be good.

I think it becomes a new optional argument to .insert_all() and .insert() and .upsert_all() and .upsert(). Question is, what should that argument be called?

Options:

• columns
• types
• type_overrides
• extra_columns

I like the first two best. I'm leaning towards columns - it would serve two purposes: you can use it to ensure extra columns are created, and you can use it to specify the type on extra OR deteced columns.

Maybe something like this:

db["authors"].insert_all(
    [{"id": 1, "name": "Sally", "score": "2"}],
    pk="id",
    not_null={"name", "score"},
    defaults={"score": 1},
    columns={"score": int, "weight": float}
)

Reminder: the current list of arguments for those methods is:

I'm going with columns={"score": int}