More clarity for data types

[brendagutman]

What are the downstream process' expecting?
Can we give common sense estimates per field?

Example:
File.size exceeded INT32, the database default for Integer, in a previous harmonization so the model failed. Bumping the field up to INT64 for these records required edits to multiple dbt models across studies.

[RobertJCarroll]

I'm not sure the right way to handle this. From the other direction, how can the modeler know the defaults of any given downstream implementation? Or how the possibility of any compliant data provided might interact with the implementation? There's not really a structured place for that type of information in the model.

There are some places we "know" about, but it's not universal. Would it take a manual pass through the auto-generated models no matter what? Do we need an additional process there, perhaps?