Data Discovery: Alerts when building something that probably won't work.

A number of ideas to prevent DD users from building something that will not work anyway - or even worse: That will drain system resources.

We should validate the basic mistakes to avoid getting the users in trouble:

• Make sure relations can only be made between compatible data types
• Make sure you can't connect fact tabels directly (and in general check for many-to-one relations)
• Make sure that measures are not autocreated from tables that work as dimension tables in the data model
• Set upper limit on complexity - maybe max 2 fact tables in one model