Re^2: Extracting text from a PDF (using PDF::API2)

Thanks for the suggestion desemondo!
I've been leaning very heavily on the Adobe document you referred to (especially the section you referenced) and I had actually seen that example too. Unfortunately it doesn't really help with my problem because my issue is PDF::API2-specific.

I guess it might be helpful to rephrase my question: how do I use PDF::API2 to "import" the fonts that are embedded in a PDF so that I can map the character IDs used in the content stream to unicode code points (which can then be assembled into words and sentences...etc.)?

To put the question in perspective, here is a simple example of what I'm dealing with. In the example you supplied, the characters "ABC" are in a known encoding. They would be encoded in the PDF exactly as the example states. However I am dealing with PDFs that do something like:

BT
/F13 12 Tf
288 720 Td
[<01>4<02>-1<03>-1<04>-1<05>2<06>] TJ
ET
[download]

I need a way to use PDF::API2 to map the character IDs (the things between the angle brackets) to unicode code points...

Comment on Re^2: Extracting text from a PDF (using PDF::API2) Download Code

Replies are listed 'Best First'.
Re^3: Extracting text from a PDF (using PDF::API2) by tmaly (Monk) on Dec 03, 2009 at 03:59 UTC
as a side note, there is a PDF::API3 available on CPAN. I have used CAM::PDF mainly for the tasks of extracting text. However, I have had little luck with embedded html in pdfs. You may be able to walk the root dictionary of the pdf using CAM::PDF and store information you need. There is also a module CAM::PDF::Renderer::Text that may be of some help	[reply]
Re^4: Extracting text from a PDF (using PDF::API2) by music_man1352000 (Novice) on Dec 03, 2009 at 04:17 UTC
Thanks for the suggestions tmaly. As I mentioned in the OP, I did look at CAM::PDF. The module you mentioned is kinda useful for rough output. However it doesn't handle layout at all (it is almost useless for my tabular data) and it can't handle the text encoding that I mentioned in my reply to desemondo. If I used it I'd still have to work at the stream level, but since it doesn't have any support for fonts I thought that PDF::API2 (which appears to support fonts) might be better. In other words, I was hoping I could avoid traversing the root dictionary (which if done properly would be a potentially huge job - to many object references!) by using PDF::API2. Just to clarify: the original documents for these PDFs are most likely to be Excel spreadsheets. They don't contain embedded HTML. "<\d+>" is just the way character \d+ (in the example, from font "F13") is encoded.	[reply]