I'm curious what makes it better than the approach in the blog?
If it's this:
> also maintains a schema.sql file that can be loaded as a one-off
That's exactly what the declarative files provide