Skip to content

Latest commit

 

History

History
53 lines (42 loc) · 3.01 KB

DiffingPDFs.md

File metadata and controls

53 lines (42 loc) · 3.01 KB
Raw

Goal: find more subtle typos like https://github.com/stakx/ecma-335/pull/17/files

I've been thinking about how I'd find diffs of big chunks of text, and decided today is the day to sit down and figure it out!

  1. Downloaded https://www.ecma-international.org/wp-content/uploads/ECMA-335_6th_edition_june_2012.pdf
  2. Printed my WIP HTML render of stakx/ecma-335 at https://carlwa.com/ecma-335/ to PDF.
  3. Came up with these commands after reading https://superuser.com/a/651406/282374
pdftotext -raw ECMA-335.pdf - | grep -v 'Ecma International 2012' > ecma.txt

pdftotext -layout stakx-ecma-335.pdf - | tr -d '\f' > stakx.txt

norm() {
  sed -e 's/[“”]/"/g' -e "s/[‘’]/'/g" -e "s//→/g" -e "s///g"
}

dwdiff -c <(<ecma.txt norm) <(<stakx.txt norm) | less -R

Worked pretty well, except some things I had to work around:

  • needed to normalize smart quotes and some other unicode
  • some wide tables got cut off in print-to-pdf
  • word wrapping inside table cells rendered on different lines
  • if print-to-pdf split a table across pages, the headers were duplicated to the next page
  • for some reason, the print-to-pdf step doubled hyperlinks
  • for every diff that found a typo in the markdown, there were 2 showing typos fixed from the original PDF!

In hindsight, probably should have used pandoc to convert MD to TXT, instead of MD to HTML to PDF to TXT.

🎉 Created PR stakx/ecma-335#22

Also, some notes on other tools I tried:

  • Tried pdfminer and pdfplumber
    • Might be interesting to come back to, if they are able to output intra-document links
    • funny discovery: The ECMA-335 PDF has a dozen links to the author's local file e.g. II.4.2 file:///C:/Users/Joel%20Marcey/Documents/My%20Dropbox/TwinRoots/CLI/Partitions/Partition%20V%20Annexes.doc%23_Sample
    • Extracting plain text from HTML seems like a solved problem, but how to do the diff?
  • Tried https://soft.rubypdf.com/software/diffpdf
  • Tried using pdfjam to merge the 500 pages to 1 huge page
    • in [[manjoro]] WSL, installed extra/texlive-binextra
    • ran script https://gist.github.com/timabell/9616807b2fe3fa60f234
    • but multiple problems trying to use this in diffpdf GUI:
    • as I scroll down the two locked views become more and more out of alignment
    • can't select-copy text from the selection to easy find markdown to change
    • different (fixedwidth?) formatting reported as a difference
    • workable with 10-page sample, but scaling to all 500 pages is far too zoomed out
  • Tried using vscode to text diff the .txt above. It timed out, and wasn't able to do the line-unaware smart diff the other tools could.
  • Tried wdiff tool first which seems more widely supported
    • Didn't have color, have to work around like: wdiff -n -w $'\033[30;41m' -x $'\033[0m' -y $'\033[30;42m' -z $'\033[0m'