objective c - PDF search on the iPhone

Question

After two days trying to read annotations from a PDF using Quartz, I've managed to do it and posted my code.

Now I'd like to do the same for another frequently asked question: searching PDF documents with Quartz. Same situation as before, this question has been asked many times with almost no practical answers. So I need some pointers first, as I still haven't implemented this myself.

What I tried:

I tried using CGPDFScannerScan handling the TJ and Tj operators - returns the right text on some PDF, whereas on other documents it returns mostly random letters. Maybe it's related to text encoding? Someone pointed out that text blocks (marked by BT/ET operators) should be handled instead, but I still haven't managed to do so. Anyone managed to extract text from any PDF?

After that, searching should be easy by storing all the text in a NSMutableString and using rangeOfString (if there's a better way please let me know).

But then how to highlight the result? I know there are a few operators to find the glyph sizes, so I could calculate the resulting rect based on those values, but I've been reading the spec for hours... it's a bloated mess and I'm going insane. Anyone with a practical explanation?

Update

User Naveen Thunga found PDFKitten, "a framework for extracting data from PDFs in iOS". I just tried the demo and it seems to work as advertised. I will test it with more PDFs and will post the results soon. As a side note, the code seems very good to me -- if you are interested in how this stuff works it's pretty awesome.

Answer 1

This isn't a simple problem to implement, but it is straightforward.

For any given page you need to scan the page using the CGPDF scanner API. You need to register callbacks for PDF operators that affect text in the page - not just TJ/Tj, but also those that set font, affect the text drawing matrix, etc. You need to build a state machine that updates with each encountered tag+parameters. You need to examine text accounting for the current font's encoding. When you find text that you want to highlight, you'll need to examine the current text drawing matrix you've been updating to determine the drawing coordinates. Read the PDF specification (version 1.7 is downloadable from Adobe) to understand which operators you need to pay attention to.

Font encoding is perhaps the most difficult part since there are a handful of ways encoding can be specified, and some of them are proprietary to the font. Mostly you can cheat and fall back on a subset of ANSI encoding - but this WILL break on certain PDFs having strange fonts.

Essentially you are processing the page as if you were to render it.