exiftool "ColorMode" and "ColorSpaceData" are passed through un-normalized, despite being invalid values

[LeoniePhiline]

exiftool -j may extract "ColorMode" and "ColorSpaceData" values which are not fit to be placed into sys_file_metadata.color_space unaltered.

Example:

exiftool -j "exampe-sRGB-black-and-white-file.jpg" \
    | jq '[.[] | {ColorMode, ColorSpaceData, ColorSpace}]' 
[
  {
    "ColorMode": "Grayscale",
    "ColorSpaceData": "GRAY",
    "ColorSpace": "sRGB"
  }
]

extractor defines color_space mapping as follows:

jq '.[] | select(.FAL == "color_space")' "extractor/Configuration/Services/ExifTool/default.json"
{
  "FAL": "color_space",
  "DATA": [
    "ColorMode",
    "ColorSpaceData",
    "ColorSpace->Causal\\Extractor\\Utility\\ColorSpace::normalize"
  ]
}

Since \Causal\Extractor\Service\Extraction\AbstractExtractionService::remapServiceOutput breaks upon a non-null $value, the exiftool value "ColorMode" = "Grayscale" is extracted and passed back to the TYPO3 metadata extraction service, where it is used as parameter for an INSERT INTO sys_file_metadata.

However, the sys_file_metadata.color_space field is a VARCHAR(4), and "Grayscale" does not fit, causing an error in strict mode.

Furthermore, "Grayscale" is an invalid value. According to SYSEXT:filemetadata/Configuration/TCA/Overrides/sys_file_metadata.php, the correct grayscale color_space value would be "grey".

Thus, ColorMode = "Grayscale" and ColorSpaceData = "GRAY" must be normalized to the value "grey".

To my mind, this should be handled by the ColorSpace utility, using a configuration like ...:

{
  "FAL": "color_space",
  "DATA": [
    "ColorMode->Causal\\Extractor\\Utility\\ColorSpace::normalize",
    "ColorSpaceData->Causal\\Extractor\\Utility\\ColorSpace::normalize",
    "ColorSpace->Causal\\Extractor\\Utility\\ColorSpace::normalize"
  ]
}

... and adjusting Causal\Extractor\Utility\ColorSpace::normalize to match on strings starting (lowercased) with "gray" or "grey" and replacing them with the canonical "grey" value.

[xperseguers]

Sounds legit, mind creating a PR?

[LeoniePhiline]

The premise of my initial report had been incorrect: "grey", as defined in EXT:filemetadata is in fact wrong, while "GRAY", as shown in the exiftool output above, is correct.

Generally, image file metadata color space handling is rather incorrect in TYPO3 core, and has been carried through since DAM, which was already wrong (but slightly less so, correctly omitting "sRGB").

I filed https://forge.typo3.org/issues/109630 for a full clean-up of sys_file_metadata.color_space.

Once it is fixed in TYPO3, also extractor needs some fixes, as it, too, has incorrect assumptions (stemming from TYPO3's incorrect definitions). See the forge report for details.