The use of PDF to present Arabic text can follow at least a few different strategies, none of which bode well for the extraction of Unicode Arabic text from a PDF file. Some or all of the text may actually be stored as image data rather than as character data, and to the extent that there are portions of text comprised of discrete characters, those characters use numeric assignments that bear no discernible relation to Unicode Arabic code points.
in reply to Parsing Arabic PDF using in perl
I remember spending a few hours one time (a couple years back) trying to find web references that would explain the PDF character encoding scheme for Arabic, but I never succeeded. Of course, I'm ignorant enough about PDF details in general that I can't even assess how inadequate that attempt was.