en masse inline editing in an uncompressed PDF

Question

I have a large PDF (~20mb, 160 mb. uncompressed). I need to do a find and replace in the text in it, about 1000 times. Here is what I tried.

1. Via SVG

   • Tranform to SVG (inkscape)
   • Read SVG line by line and do the replace in the file
   • Transform back to PDF

=> bad output, probably due to some geometric transform matrix in the SVG, the text is not well rendered

2. Creating ~1000 sed command

   • Uncompress PDF
   • Perform each replace with a sed command
   • Recompress PDF

=> way too long. each sed command takes about 20 sec, leading to several hours of process

3. Read line-by-line and replace

   • Uncompress PDF
   • Read line by line the PDF
     • find text to be replaced
     • replace using perl
     • write line to a new file
   • Compress the new file

=> due to left data-stream in the uncompressed PDF, the new file is apparently damaged (writing binary as lines of text)

I wonder if it would be possible to read line-by-line the uncompressed PDF, but do the editing directly in it. How could I do this?

I have searched for perl inline editing, but it performs the changes in the whole file at once, while I'd like to edit a single line.

Other ideas are more than welcome ;)

Following advise, I used CAM::PDF, this was the most efficient and simple solution

Answer 1

There is no difference between 2. and 3. Sed reads the input file line by line and writes changed lines into the output file. If you fed -i switch to it, sed just opens the input file and then unlinks (it's what rm do) then opens the output file with the same name and writes into. That's it. No magic involved. So if you damaged content by Perl, but not by sed you do something different than by sed. The main difference is, you can make Perl script way faster for replacing many strings. See Using sed on text files with a csv

The main trick is you can compile regexp for search nad replace which works in linear time.

my %replace = ( foo => 'bar' );
my $re = join '|', map quotemeta, keys %replace;
$re = qr/($re)/;

while (<>) {
    s/$re/$replace{$1}/g;
}

You can use it with your original approach, but I would recommend to make it in Perl script which allows you to keep the regexp and replace hash between pdf files. You can also try it to combine with CAM::PDF. There is the example script changepagestring.pl in it. You can also look at PDF::API2 which would require more work but may provide better result. But remember, PDF format is not intended for modification.

Answer 2

You can follow the pdftk steps as described in How to find and replace text in a existing PDF file with PDFTK (or other command line application)

You can first split the PDF into smaller documents with a few pages each, replace the text and again merge them together - all using pdftk.

There is also the PDFEdit software (http://pdfedit.cz/en/index.html). It is a GUI app with a scripting interface. You can process individual pages and then do a find replace using scripting commands. See if it loads your PDF.