Some code that parses a PDF file into objects. There's also some very preliminary WIP code to display the file structure on a web page.

At the lowest level, a PDF file is a sequence of 8-bit bytes. Use hexdump or xxd or something to see it.

At the next lowest level, a PDF file is a set of objects, wrapped in some file structure (header and trailer). This is still too low-level to be practically useful for anything, but the code here only concerns itself with this level.

Specifically, there are 8 types of objects:

• Boolean Objects
• Numeric Objects
• String Objects
• Name Objects
• Array Objects
• Dictionary Objects
• Stream Objects
• Null Object

To read a PDF file, we need to be able to parse all these kinds of objects, and also, additionally:

• Indirect Object definitions (like 12 0 obj … endobj)
• Indirect object references (like 12 0 R)
• File structure: Header, body, cross-reference table, trailer.

This project parses (most) PDF files, into the above structures. It does nothing further in terms of interpreting the objects in any useful way.

It is also (sort of, WIP) a "literate program" to some extent: read here.

Some notes:

• It can now round-trip via JSON. That is, if you dump to JSON and read back, you will get the exact same bytes.

  • This is not as impressive as it sounds, because we could in principle just dump the sequence of bytes into JSON as an array of numbers. However, here we're doing slightly more than that.

• It assumes the input is valid, e.g. does not check in dict for unique keys, does not check for stream length, etc. In fact, parses the file "forwards", rather than starting with the trailer first.

• Using from Rust code: src/lib.rs has a library that parses a PDF file into PDF objects.

• Web interface (WIP, calls the parser but does not display anything much yet):

  • Visit https://shreevatsa.net/pdf-explorer/ (last working version: https://638396b5cb23920d58f8adf4--fastidious-ganache-d72698.netlify.app/) or
  • Run build.sh and python3 -m http.server (or equivalent), then access http://[::]:8000/

I haven't yet tried either of these, but they seem to be further along (IIUC they're written in Python and generate HTML):

• https://github.com/desgeeko/pdfsyntax
  • git clone https://github.com/desgeeko/pdfsyntax && cd pdfsyntax and then python3 -m pdfsyntax inspect foo.pdf > output.html
• https://github.com/trailofbits/polyfile
  • pip3 install polyfile and then polyfile --html output.html foo.pdf

There's also a Java app:

• https://github.com/itext/i7j-rups (https://itextpdf.com/products/rups)
  • Download the release and then java -jar ~/Downloads/itext-rups-7.2.5.jar etc.

Others:

• https://www.reportmill.com/snaptea/PDFViewer/ = https://www.reportmill.com/snaptea/PDFViewer/pviewer.html (drag PDF onto it)
• https://pdf.hyzyla.dev/
• https://sourceforge.net/projects/pdfinspector/ (an "example" of https://superficial.sourceforge.net/)
• https://www.o2sol.com/pdfxplorer/overview.htm

Status currently:

• As of 2022-05-01 14:20 (e54b45e): Works for 19426 out of 19492 PDF files on my laptop. So "fails" for 66 files. Looked at each of them. They are all malformed in some way or the other.