in reply to Re^2: Mechanize Firefox text Method
in thread Mechanize Firefox text Method

PDF does not always contain text. I've seen lots of PDF files that were composed of images (scanned texts, no OCR involved). So getting no text or much less text than expected is not always a problem in your code.

PDF is a "postscript print job on steroides". PDF is basically postscript, with lots of addons that aren't really relevant for your problem. Postscript describes how to print a page. Most times, it works roughly in reading order, but neither postscript nor PDF have a problem with a print job that first emits all "A"s, then all "B"s, then all "C"s, and so on. It inflates the print job, and it makes it really hard to extract the original text, and there seems to be software written for exactly this purpose.

I think a much cleaner way is to determinate the URL of the PDF file (using Mechanize), download the PDF file (using LWP or Mechanize), and process the PDF file using tools like pdftotext.

Note that you still need some OCR software for scanned images, pdftotext just extracts text from the PDF file.

Update: There are several commercial OCR programs that can take PDF files (including those composed of scanned images) as input and deliver text or Word documents.

Alexander

--
Today I will gladly share my knowledge and experience, for there are no sweeter words than "I told you so". ;-)

Replies are listed 'Best First'.
Re^4: Mechanize Firefox text Method
by fionbarr (Friar) on May 06, 2013 at 18:10 UTC
    have a look at CAM::PDF. I have used to do text searchs in PDFs with hundreds of pages.
[reply]
Re^4: Mechanize Firefox text Method
by halweitz (Novice) on May 05, 2013 at 16:28 UTC

    Thanks again for your reply. Let me clarify a bit. Since I can read the documents in the browser I know they contain only text so OCR is not an issue. All the documents follow a similar set of templates but the content changes for each. I have viewed hundreds of these and any document that does not conform will be skipped.

    Your comments on downloading and then using a pdftotext tool on the local file is inline with my current thinking as long as it can be scripted and run without intervention. Are there any other suggestions I should examine?

[reply]
      Let me clarify a bit. Since I can read the documents in the browser I know they contain only text so OCR is not an issue.

      I think we have a little communication problem: Sure you can read text displayed in Firefox, because it was rendered from something like <html><body><h1>Hello</h1>. But you can also read text displayed in Firefox that was rendered from something like <html><body><img src="http://www.example.com/pics/hello.gif" alt="">. Your computer can't, at least not as easy as you. To extract the text from the latter, you need OCR.

      All the documents follow a similar set of templates but the content changes for each.

      Any chance to get access to the data before the template engine creates the PDF? Perhaps as XML, JSON, CSV or even HTML?

      Alexander

      --
      Today I will gladly share my knowledge and experience, for there are no sweeter words than "I told you so". ;-)
[reply]
[d/l]
[select]

        Hi Alexander. Thanks for the reply. Instead of using the content method of Mechanize::Firefox I used CAM::PDF to extract the text from each PDF page and then search the text for my filter phrases. The only wrinkle is that I have to save the URL in a file and then load the file into CAM::PDF. When the script ends I get multiple clean up errors about removing a call back on an undefined value. This does not affect the script in any way.

[reply]

In Section Seekers of Perl Wisdom