Currently if --ignoreempty is active all tables are still created on disk, but empty tables are subsequently deleted. A better approach (that will also help with providing html2csv as a module) would be to determine which tables will be created before creating them (possibly storing in memory, though this could become an issue for very large pages).