The coverage over the entire codebase is quite low. Most of the handlers are tested for very basic things, but the main logic of images scanning/merging/etc. is not.
- History When (XMP/Lightroom)
- Metadata Date (XMP/Lightroom)
- Move exif to fully generic? Track all fields, re-introduce Object but binary serialization.
- from lightroom: region type: face but without name, means unknown person.
- sub-image view (copies, raw file, etc.)
- skip panoramas unless explicitly requested
Camera/lens database:
- pics/week average and median
Stats:
- focal length
- camera+focal length
- camera/lens
- aperture
- shutter speed
Maybe generic (1 val, 2 val, 3 val, N val) auto-compute?
- with result field: picture count, etc.
- for consistency, should try to find a landscape-oriented image first
- and with a person, if possible
-
The current copy system is really bad. A copy is declared as such even when no master is present (for its deduced original name), leading to bad behaviour when multiple copies are possible; e.g. with '-' as separator, pic.xxx, pic-a.xxx, pic-a-b.xxx will lead to bad outcomes.
-
The copy regex handling is across the full "file path", i.e. including any subdirs under its parent folder. If the regex doesn't explicitly exclude slashes and if it also can catch dirs (e.g. '-' as separator, and subdirs that include dashes), it will lead to so broken behaviour that it is funny.