If you're initializing something (i.e. the amount of data are small), you could prep the schema in a script by connecting to a ":memory:" database and resolve all of the column challenges, then write the fully polished file out to disk.