Reading metadata from PDF (XMP)

[canpolat]

Hi,

It looks like there is a problem with reading XMP metadata from PDF files in my setup (2.63).

Here is how to reproduce it:

1. Select a book and modify its metadata (add a tag, for example)
2. Embed the metadata to file using Calibre's built-in utility
3. Open the file location and copy the PDF file to a temporary place
4. Add the copied file to Calibre.
5. Observe that the modified metadata is not present (Calibre uses the file name to populate some of the fields).

I first thought the problem is related to embedding XMP into the PDF, but then I used http://regex.info/exif.cgi to validate that the XMP data is actually in the file and Calibre does not use it.

In "The Add Process" in my Calibre preferences, "Read metadata from file contents rather than the file name" is selected.

Any pointers would be much appreciated.

[BetterRed]

Roundtrip as you describe seems to works for me

Imports DC elements like title, tags etc, AND custom columns from http://calibre-ebook.com/xmp-namespace.xml-->calibre:custom_metadata (bag) items.

My add process settings are same as yours

I re-added the book both by drag-drop the file into book list and via Windows Send to:calibre - no difference.

BR

[canpolat]

Hi BetterRed,

Thanks for taking the time to confirm that it works on your setup. Since this is such a widely used function I was almost sure that the problem is on my setup, but hoping that I was missing an obvious setting. Now, I just need to find what the problem is

[BetterRed]

Originally Posted by canpolat

Hi,

5. Observe that the modified metadata is not present (Calibre uses the file name to populate some of the fields).

@canpolat - when I re-added the book I deliberately changed the name of the file, there is no evidence of that name in the added book, so everything had to come from embedded metadata.

Essentially I copied a PDF (Romeo a Julia - William Shakespeare.pdf) from my test library to the desktop, renamed it to junk.pdf, and added that file to an empty clone of my test library. The following image demonstrates that.



The following is from Jeffrey Friedl's Metadata Viewer



BR

[kovidgoyal]

There are two settings that can affect this:

1) Preferences->Add books->Read metadata from file content has to be enabled

2) Preferences->Plugins the PDF metadata reader/writer plugins need to be enabled.

[canpolat]

@BetterRed, my "Adding books" preferences are identical to yours (the ones in the red rectangles).

@kovidgoyal, Thank you for your reply. Both of those settings are set in my installation.

In the meantime, I reinstalled Calibre just to see if it was an installation glitch. I installed the same version (2.63.0) since that is the one that comes from the repositories of my distribution.

Funny thing is, I started getting an error that says "'QResizeEvent' object has no attribute 'tip'" (this was not there before). Apparently that is a bug fixed in a later version (and I'm not worried about not). But the fact that I started getting this error even though versions are the same should mean something.

After the reinstall all my settings were intact. I really don't want to install Calibre from scratch (removing all my settings), but I suppose that is the only thing left.

[BetterRed]

If you didn't install from links on the calibre downloads page then I suggest you do so.

Another thing to try is to use one of the alternate locations, they're https links. ISPs have been known to play silly buggers with their caching mechanisms, https prevents that.

BR