Thread by @angealbertini

generic advice to make PDF manipulations by hand (fill forms, modify contents...):
1 run `qpdf -qdf` to normalize the PDF [optionally with --stream-data=uncompress].
2 do your changes.
3 clean up the PDF with `mutool clean`.

remember that some fields like author can be stored as Unicode:
so 'ange' = 00 61 00 6E 00 67 00 65

inline kerning is often used (by LaTeX, etc...) in the pages contents:
it's an array of a short string (w/ parenthesis), a number, a short string...

so 'ange' can be stored like:
[(an) 3.0 (ge)]

if you want to modify graphical elements (watermarks...), don't miss @GEndignoux ' cheat sheet of PDF operators:
github.com/gendx/pdf-chea…

@GEndignoux Some viewers (evince, sumatra) display the changes on the fly, which is very neat to poke around.
It saves a lot of time when experimenting.
Mutool should get you a compatible PDF in the end anyway.

@GEndignoux PDFs - especially after `qpdf -qdf` - may look ASCII, but they are fully binary, so don't use text editors that remove non-ASCII or modifies whitespace.
You can turn a PDF to pure ASCII with "ascii hex encoding" via `mutool clean -a`
=> thgs like \xff\xd8 will be encoded FFD8.

If you want to learn the syntax and logic by yourself:
speakerdeck.com/ange/lets-writ…

and if you're curious about removing watermarks or checking for potential leaks:
speakerdeck.com/ange/an-overvi…