weird text order

[femifrak]

OCRmyPDF is brilliant but sometimes i have a problem with the order of text that is underlaid. When i select the text starting from top left and go to the right end of the line and then successively down line by line, there are sometimes gaps of text which is not selected. After a few more lines these gaps are suddenly selected. Copying the selected text and pasting it into another application reveals the order, which is unfortunately wrong. I use latest stable version and have no error or warning messages.

http://www.loaditup.de/files/803343_acxm67dsue.pdf
(problem occurs in second paragraph.)

[fritz-hh]

I could not reproduce the issue:

• neither with the PDF viewer embedded in firefox v26.0
• nor with Adobe Reader XI.

It is possible, that the issue only occurs with some PDF readers? With reader do you use?

[femifrak]

I use the "Document Viewer 1.2.0" which can be launched with "atril".
When using Foxit Reader the text order is correct.
That is very good news. Thanks a lot for testing.

[femifrak]

As the output format is pdf/a i am worried why some pdf readers cannot handle the text correctly. I see three possible reasons: 1) pdf/a is not well enough defined 2) The reader is really not accaptable 3) The output is not 100% pdf/a
As i want to use OCRmyPDF for archiving files, the text output could be unreadable in future.

[fritz-hh]

1 and 3 are not the reason.
OCRmyPDF uses very mature tools (e.g. ghostscript) to generate the PDF/A file. Furthermore, OCRmyPDF validates the generated file against the PDF/A standard using jhove. Therefore the risk is very low that the file will won't be readable in the future. (Alltough I use OCRmyPDF for archiving files too, I cannot / won't provide any warranty...)
As discribed in #66, because the text layer is made of disconnected words and lines, basic PDF-reader can have difficulties to select the text accurately/reliably. Advanced PDF-readers should not have any problem. Seaching for words in the PDF should be possible even with basic PDF-readers.

If I have time, I will analyze how it is possible to improve selection accuracy on basic some PDF-readers. Therefore, I reopen the ticket.