As title, consult a great god told,

CodePudding user response:

Parse the document structure, and then according to the pages of output