Build model failed if any of the source table schema had been changed, saying deleted unrelated columns.

How the error occurred?

During building a model, KE will construct a new SELECT sql from KE's table metadata, to read data from source tables. But if the user had deleted some unused columns from any source table, KE's metadata will be different with Hive's metastore, as a consequence, `SparkSession::sql` will throws validation exceptions because of the mismatched columns found in the SELECT sql.

How to solve it?

Using the intersection between KE's columns and Hive metastore's.