DTPO changing/reencoding PDFs on save?

DEVONthink Tips
1

I was provided with a number of scanned documents that the provider OCR’d prior to handing over. In the original file handed to me, the text is selectable and the OCR accuracy seems high. Copy and pasting is almost 100% accurate, searching works, and seeking definitions works almost 100% perfectly.

The file metadata states:
PDF Producer: Adobe Acrobat 9.0 Paper Capture Plug-in with ClearScan
Content Producer: Adobe Acrobat 9

Oddly, when I make a few highlights then save that file, the metadata changes:
PDF Producer: Mac OS X 10.10.2 Quartz PDFContext
Content Producer: Adobe Acrobat 9
After this, the text is still selectable, but the OCR accuracy is now 0%. Copy and pasting copies and pastes nothing. No definitions can be sought for words. Search does not function.

I notice that this happens in both Preview and in DTPO’s built-in viewer, and it only occurs with some files, not all. Mostly scans of text that were subjected to some type of OCR.

The file size also increases. For example an 11mb original PDF changes to about 16mb after saving after viewing in DTPO or Preview.

Your thoughts are appreciated!

2

After doing some additional testing I can reproduce this in Preview, DTP, and DTPO on 10.10.2, 10.10.1, and on three different devices (rMBP, MBP and Mac Pro). So since I can reproduce this in Preview I suppose this isn’t explicitly a DT issue… but perhaps someone else here has had this experience?

3

When you change a document in DTPO, it uses OS X’s pdf routines to write the output, so it does change the structure of the PDF. It should be mostly the same, but PDF internals like object streams, types of compression, etc might be different.

That said, it should be equivalent! This process shouldn’t be losing information the way you describe. That suggests there is some sort of bug. Since you see the same bug in Preview, it’s an OS-level bug.

4

Yeah I’m beginning to think that. I no longer have a 10.9 machine to test it on to see if it was introduced with Yosemite. I’ll have to do some more poking around.