I had to remove and re-import some images in Darktable.
My usual tagging scheme for names is
“People|Surname, Firstname”
After reimporting the images in question, their tags are shortened to a one-word format.
“People|Surname”
Particularly when several members of an entire family are present within a film roll, this makes it tedious to re-specify who exactly is shown on any particular photo. Note that the original long-form tags have not disappeared from the library, but they have to be re-assigned before additionally deleting the short versions.
Is this because tags are stored in a one-word format in the XMP files, or because “space” is not handled well instead of “underscore”?
Frequently using hierarchical tags with spaces without problems. I guess the “,” is the problematic character. Can’t check at the moment, no computer available.
As it happens on re-importing the images, I’d also guess the issue is the “,”. To be sure, we’d need to know in which metadata field those tags are stored (different fields can use different separator characters).
I’d still like to see as many details about this as possible: it looks strange that a tag format accepted by darktable cannot be exported/imported
1) I repeated the removal and re-import, with a random test image, keeping an eye on the XMP.
Before removal, the tag is stored fully in the XMP. After the re-import, the tag is shortened, even within the XMP.
Interestingly, in darktable’s list of proposed tags, the old “People|Surname, Firstname” is still marked as (1) after the re-import, so it is said to be used by one file.
However, if I right-click and “go to tag collection”, to show the files using this tag, no image shows up.
Also, the shortened version is not being adopted into the proposal list, while being displayed under the assigned tags when I select the respective image.
2) I changed the tag to one without the “,” [It’s a German thing…].
“People|Surname Firstname”
Now the entire tag remains unchanged, both in Darktable and within the XMP.
I guess the “,” really was the source of the error. However, it is a shame to have no separator between name parts, no matter if you use a “Last First Middle name” or a “First Middle Last name” sequence.
Do you need any extra information - such as the XMP text - to find out where precisely this is stored?