Programmatically change font color of text in PDF

Question

I'm not familiar with the PDF specification at all. I was wondering if it's possible to directly manipulate a PDF file so that certain blocks of text that I've identified as important are highlighted in colors of my choice. Language of choice would be python.

Answer 1

It's possible, but not necessarily easy, because the PDF format is so rich. You can find a document describing it in detail here. The first elementary example it gives about how PDFs display text is:

BT
/F13 12 Tf
288 720 Td
(ABC) Tj
ET

BT and ET are commands to begin and end a text object; Tf is a command to use external font resource F13 (which happens to be Helvetica) at size 12; Td is a command to position the cursor at the given coordinates; Tj is a command to write the glyphs for the previous string. The flavor is somewhat "reverse-polish notation"-oid, and indeed quite close to the flavor of Postscript, one of Adobe's other great contributions to typesetting.

The problem is, there is nothing in the PDF specs that says that text that "looks" like it belongs together on the page as displayed must actually "be" together; since precise coordinates can always be given, if the PDF is generated by a sophisticated typography layout system, it might position text precisely, character by character, by coordinates. Reconstructing text in form of words and sentences is therefore not necessarily easy -- it's almost as hard as optical text recognition, except that you are given the characters precisely (well -- almost... some alleged "images" might actually display as characters...;-).

pyPdf is a very simple pure-Python library that's a good starting point for playing around with PDF files. Its "text extraction" function is quite elementary and does nothing but concatenate the arguments of a few text-drawing commands; you'll see that suffices on some docs, and is quite unusable on others, but at least it's a start. As distributed, pyPdf does just about nothing with colors, but with some hacking that could be remedied.

reportlab's powerful Python library is entirely focused on generating new PDFs, not on interpreting or modifying existing ones. At the other extreme, pure Python library pdfminer entirely focusing on parsing PDF files; it does do some clustering to try and reconstruct text in cases in which simpler libraries would be stumped.

I don't know of an existing library that performs the transformational tasks you desire, but it should be feasible to mix and match some of these existing ones to get most of it done... good luck!

Answer 2

Highlight is possible in pdf file using PDF annotations but doing it natively is not that easy job. If any of the mentioned library provide such facility is something that you may look for.