This tool is mainly intended for testing environments where PDF files are generated programmatically, and should be exactly the same from one test run to the next. It uses a bitmap comparison, so even the smallest differences in appearance will not be tolerated (some amount of fuzziness can be introduced by decreasing render resolution). At the same time, the internal structure of the file and any non-visible features are not compared.
We can imagine some more interesting ways to compare PDFs, for example:
- Semantic or textual equivalence
- Indifference to image rotation / centering / brightness (useful for scanned images)
- Optical text recognition
But pdf-comp does none of those. It just generates images for each page in the files using pdftoppm and checks if they are the same. Optionally, it generates a side-by-side comparison image of differing pages, with the differences highlighted.
There is a command line program detailed below, but the real use of this package is to includimport it in an automated testing suite.
This code is inspired by SerHack's pdf-diff. So thanks! It relies on Horst Rutter's amazing golang PDF processor, pdfcpu. Also, Google's Gemini has made a significant contribution.
Some nice near-term improvements would be: perform image difference highlight file creation in memory when the user has requested pdf but not images; execute pdftoppm asynchronously.
It's important to use a recent version of XpdfReader - specifically not the Ubuntu package poppler-utils. As of this writing, version 0.85 of pdftoppm included in poppler-utils is broken for output to stdout and will cause the program to fail. Try doing something like
$ wget https://dl.xpdfreader.com/xpdf-tools-linux-4.05.tar.gz
$ tar -xzvf xpdf-tools-linux-4.05.tar.gz
[move appropriate binaries into your PATH]
$ go install github.com/pdfcpu/pdfcpu/cmd/pdfcpu@latest
$ go install github.com/mdmcconnell/pdfcomp@latest
Have a look at cli.go for an example of how to use EqualPDFs
same, err := EqualPDFs(file1, file2, images, w, resolution, ratio)
if err != nil {
fmt.Fprintf(os.Stderr, "%s", err.Error())
} else if same {
// do something
}
if same {
// do something else
}
Usage: pdfcomp [options] file1.pdf file2.pdf
-images if set, create images for each page that is different, highlighting the differences. Names will be of the form file1.pdf-n-diff.png (with n being the page number)
-pdf compile page-by-page images into a single pdf file of differences
-resolution= integer dpi resolution for creating bitmaps, default 300dpi. May impact performance.
-ratio= integer divide dpi by this number to determine the radius of difference highlight circles to output, default 30. Only meaninfgul if images is set
0 The two PDFs are visually equivalent
1 The two PDFs are not visually equivalent
2 The program has encountered some error before completing (and normally printed an error message)
This works especially well for fixed-width fonts (with variable-width, an entire line of text after the first different character will be highlighted).
