Split PDF files by bookmarks

Posted on 2016-02-05 Edit on GitHub

Sejda is a Java library and command line utility for manipulating PDF files. It's the best solution I've been able to find for a free, CLI utility that can split PDF documents by bookmarks.

There are many websites and desktop applications, mostly non-free, for manipulating PDF files. Not many, however, offer the ability to split PDF files by bookmark. Adobe Acrobat does offer this capability, but it's much too heavy-weight for the simple tasks I'd like to perform.

PDF files are nice because they're portable and (relatively) tool-friendly (e.g. you can search PDF documents with pdfgrep), although not as Emacs-friendly as Info files. Emacs, however, cannot display large PDF files efficiently, because the default DocView mode has to render each page of the PDF document as an image.

The Common Lisp Recipes book weighs in at 755 pages, which is much too for DocView to handle. What to do? Well, we'll split the book by chapter.

First, give yourself execution permission on the file: 

chmod +x bin/sejda-console

Next, check out the tutorial for available commands. We want splitbybookmarks, which

Splits a given pdf document before each page where exists a GoTo action in the document outline (bookmarks) at the specified level (optionally matching a provided regular expression).

The command we want is 

sejda-console splitbybookmarks -f <file-name>.pdf -o <output-dir> -l 2 -e "(.*)([1-9I])(.+)" -p "[BOOKMARK_NAME_STRICT]"

Explanations of the command options:

  • -l <level> means split by bookmarks at <level>, in this case 2, which is the level for Common Lisp Recipes' chapters.
  • -e <regexp> means split only on bookmarks matching the <regexp>. ([1-9I]) means "bookmarks with names including a digit or the letter I", the letter "I" being the first letter of the "Index" chapter. The way that we have to specify regexp is a bit strange (having to use (.*) and (.+)).
  • -p <output-prefix-option> means to use an output prefix to name files. [BOOKMARK_NAME_STRICT] uses the bookmark (i.e. chapter) name, but strips out non-alphanumeric characters.