Add fallback for undefined UserComment strings

[lemmi]

darktable and potentially any program that uses exiv2 will set the type for the exif UserComment tag to undefined for utf-8 encoded strings.

exiftool will happily decode such a string.

This patch adds a fallback that tries to decode this type of strings as utf-8, otherwise return the empty string.

[lemmi]

Here's a test image with UserComment set to äöüß:

Here's a test program that fails to return the string without the patch:

package main

import (
	"fmt"
	"log"
	"os"

	"github.com/bep/imagemeta"
)

func run() error {
	f, err := os.Open(os.Args[1])
	if err != nil {
		return err
	}

	defer f.Close()

	opts := imagemeta.Options{
		R:               f,
		ImageFormat:     imagemeta.JPEG,
		ShouldHandleTag: func(imagemeta.TagInfo) bool { return true },
		HandleTag: func(info imagemeta.TagInfo) error {
			fmt.Printf("%10s: %v\n", info.Tag, info.Value)
			return nil
		},
		Warnf: func(s string, a ...any) {
			fmt.Printf("Warning: "+s+"\n", a...)
		},
	}

	return imagemeta.Decode(opts)
}
func main() {
	if err := run(); err != nil {
		log.Fatal(err)
	}
}

go run . undefined_usercomment.jpg

XResolution: 72
YResolution: 72
ResolutionUnit: 2
YCbCrPositioning: 1
ExifVersion: 0232
ComponentsConfiguration: 1 2 3 0
UserComment:
FlashpixVersion: 0100
ColorSpace: 65535

exiftool:

exiftool undefined_usercomment.jpg

ExifTool Version Number         : 13.03
File Name                       : undefined_usercomment.jpg
Directory                       : .
File Size                       : 683 bytes
File Modification Date/Time     : 2025:02:03 12:03:04+01:00
File Access Date/Time           : 2025:02:03 12:03:06+01:00
File Inode Change Date/Time     : 2025:02:03 12:03:04+01:00
File Permissions                : -rw-r--r--
File Type                       : JPEG
File Type Extension             : jpg
MIME Type                       : image/jpeg
Exif Byte Order                 : Big-endian (Motorola, MM)
X Resolution                    : 72
Y Resolution                    : 72
Resolution Unit                 : inches
Y Cb Cr Positioning             : Centered
Exif Version                    : 0232
Components Configuration        : Y, Cb, Cr, -
User Comment                    : äöüß
Flashpix Version                : 0100
Color Space                     : Uncalibrated
Image Width                     : 16
Image Height                    : 16
Encoding Process                : Progressive DCT, Huffman coding
Bits Per Sample                 : 8
Color Components                : 3
Y Cb Cr Sub Sampling            : YCbCr4:2:0 (2 2)
Image Size                      : 16x16
Megapixels                      : 0.000256

exiv2:

exiv2 -pe undefined_usercomment.jpg

Exif.Image.XResolution                       Rational    1  72/1
Exif.Image.YResolution                       Rational    1  72/1
Exif.Image.ResolutionUnit                    Short       1  2
Exif.Image.YCbCrPositioning                  Short       1  1
Exif.Image.ExifTag                           Long        1  90
Exif.Photo.ExifVersion                       Undefined   4  48 50 51 50
Exif.Photo.ComponentsConfiguration           Undefined   4  1 2 3 0
Exif.Photo.UserComment                       Undefined  16  äöüß
Exif.Photo.FlashpixVersion                   Undefined   4  48 49 48 48
Exif.Photo.ColorSpace                        Short       1  65535

[bep]

Thanks for this. It seems that the new switch produces invalid strings in some cases, which I must admit I don't understand how could happen.

I have added your failing image to the test setup.

[lemmi]

It actually makes sense if you look at the dump of the failed test image:

The UserComment starts with 00 00 00 00 00 00 00 00 and is therefore caught in the new case, so it's not immediately returned as an empty string. I suspected that exiftool might just trim whitespaces. Although I know next to nothing about perl, I think this is the line:
https://github.com/exiftool/exiftool/blob/5a772cefb038d709fd20b61cd7291517a2fde08f/lib/Image/ExifTool/Exif.pm#L5490
I updated the code to do the same and remove all trailing whitespace.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 50.00000% with 3 lines in your changes missing coverage. Please review.

Project coverage is 76.95%. Comparing base (0deaa44) to head (c2fbacd).
Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
helpers.go	50.00%	2 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #37      +/-   ##
==========================================
- Coverage   77.05%   76.95%   -0.10%     
==========================================
  Files          14       14              
  Lines        1691     1697       +6     
==========================================
+ Hits         1303     1306       +3     
- Misses        303      305       +2     
- Partials       85       86       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.