I often have to work with pdfs. Yes, I know the advice, just avoid pdf. But I don’t have that choice.
Sometimes scanned pdfs are missing text.
Sometimes pdf-born-pdfs have corrupt text, or lose their text after processing in Ghostscript.
So far, my only option is to check each pdf in turn, and check the text. I’ve tried variations on wc -w in the command line, but they often report a word count for pdfs which have lost all words. For English-language pdfs, a quick search for pdfs which don’t contain “the” should find any which lack text, or have severely corrupted text. So can Devonsphere search for files which which don’t contain a given term?
Technically, it’s possible but I don’t suggest it since it will return tons of results, not just PDFs.
Where are you searcing and what do you intend to do with the files?
To begin with, I’d like to search my Calibre libraries, and then some other folders with files I’m working with.
I want to ensure these pdfs are searchable, and find and fix the ones which aren’t.
I have an old Kindle so it can’t read newer pdfs. I reprocess scanned pdfs with Willus’s k2pdfopt, and pdf-born-pdfs with Ghostscript with compatibility level 1.4 and sometimes compression options, to create old-Kindle-compatible versions. I can’t use tablets or other touch devices.
I hope to find the pdfs which are missing text so I can either (a) ocr them or (b) if the originals have text but Ghostscript reprocessed ones lose text, run the originals through the Quartz gray filter and then Ghostscript reprocess them again.
I have reported the bug on Ghostscript’s side, but they don’t consider it a bug there. For whatever reason Quartz + Ghostscript avoids that bug in just Ghostscript.
Unfortunately, Quartz alone can’t convert newer pdfs to old-Kindle-compatible versions.
cd into the directory of your choice and run this command:
grep -ai -L font "$PWD"/*.pdf
This should only report PDFs with no text layer. You can obviously modify this to suit your needs.
Thank you, but that doesn’t find files with corrupt text layers, e.g. losing everything except punctuation.
You can modify it, e.g., looking for the as you mentioned previously.
Thank you.
grep -aid recurse -L the *.pdf
turns up one affected file and some false positives.
grep -aid recurse -L and *.pdf
one affected file and fewer false positives
but oddly, trying a nor search…
grep -aid recurse -L “and|the” *.pdf
or
grep -aid recurse -L and *.pdf | grep -ai -L the *.pdf
adds more false positives.
P.S. using -E allows the | for OR, and
-aiE -d recurse -L -e “and|the” *.pdf
Gets this down to 2 good results and 1 false positive in my test folder.
P.S. Better results with pdfgrep -ir -L “the|and” (folder)
Good to hear but remember pdfgrep is not part of the base installation of macOS.
Yes, I use Homebrew for that.
For a shell script in Automator:
Application, Shell, Bash. Pass input as arguments.
for f in “$@”
do
suffix="-list.txt"
base=basename "$f"
outputfile=$base$suffix
/usr/local/bin/pdfgrep -ir -L “the|and” “$f” > “$f”-list.txt
done
Drop the appropriate folder onto the app icon, and it should output a list of pdfs with either unreadable text, or non-English text. Add common works from other languages you use.
It probably includes some extra cruft since I’ve adapted it from other scripts. If you use another package manager, you may need another path.