Expose IPTC metadata for PNG, JPG and Tiff #2611
Conversation
This change adds an `iptc_metadata` method (similar to `xmp_metadata` and `exif_metadata`) to the `ImageDecoder` and implements it for PNG and JPG.
This comment was marked as outdated.
This comment was marked as outdated.
1c3t3a
changed the title
| /// This method converts a `Value` to a vector of bytes. A `Value` in Tiff can have different | ||
| /// types, e.g. a byte, a short or a float. This method intents to convert all these types to | ||
| /// a vector of bytes (e.g. a u32 can be represented as a [u8; 4]). However, since this is only | ||
| /// intended to parse values stored in XMP, IPTC and EXIF metadata sections, we ignore / return | ||
| /// an error on a few Values that don't make sense in this context (e.g. Value::Ifd). | ||
| fn value_to_bytes(value: Value, bytes: &mut Vec<u8>) -> ImageResult<()> { |
There was a problem hiding this comment.
There was a problem hiding this comment.
The one thing I'd worry about here is architectural difference. Since this is doing little-endian conversion it makes me suspicious it may only work on little endian targets? Then again, to quote Linus Torvalds, "Oh Christ. Is somebody seriously working on BE support in 2025?" so I think this may be fine if we first check target_endianess and always error for non-little targets.
There was a problem hiding this comment.
Actually I came up with this piece of code during the CI run which I think was running this on BE? I might misunderstood sth there :) Alternatively we could also just handle endianess correctly? I am also fine with erroring out, though!
|
The CI failure seems to be due to a spurious network failure in a setup step. Is there a way to retrigger the run? |
|
Friendly ping on this PR, would anyone have a moment to take a look? :) |
| /// This method converts a `Value` to a vector of bytes. A `Value` in Tiff can have different | ||
| /// types, e.g. a byte, a short or a float. This method intents to convert all these types to | ||
| /// a vector of bytes (e.g. a u32 can be represented as a [u8; 4]). However, since this is only | ||
| /// intended to parse values stored in XMP, IPTC and EXIF metadata sections, we ignore / return | ||
| /// an error on a few Values that don't make sense in this context (e.g. Value::Ifd). | ||
| fn value_to_bytes(value: Value, bytes: &mut Vec<u8>) -> ImageResult<()> { |
There was a problem hiding this comment.
The one thing I'd worry about here is architectural difference. Since this is doing little-endian conversion it makes me suspicious it may only work on little endian targets? Then again, to quote Linus Torvalds, "Oh Christ. Is somebody seriously working on BE support in 2025?" so I think this may be fine if we first check target_endianess and always error for non-little targets.
2 participants
This change adds an
iptc_metadatamethod (similar toxmp_metadataandexif_metadata) to theImageDecoderand implements it for PNG and JPG.This is related to #2610. If my research was correct and IPTC metadata only exists in PNG and JPG, then this should fix the issue.