We noticed that apparently Synapse performs data type inference when creating a view of views. When joining two openrowset based tables together (some using the WITH clause to hard code data types) to create a new view, some of the data types change between the new view and the views it is based on.
I would expect the data types to remain the same and am a little worried that there could be data loss. Even if that is not a risk, I would hope for the data types to remain consistent.
Is there anyway for the view of views to not perform data type inference or is there any another work around?