Filterable behavior for Image Block raw transformations

[danielbachhuber]

One of the plugins I work on, Tasty Pins, offers the ability to add "Pinterest Text" to an image (related #8472). When the image ends up in the Classic Editor, its markup looks like this:

Notice the inclusion of data-pin-description="This cashew coffee is delicious, creamy, and easy. Perfect for a chilly morning! | pinchof yum.com".

When this image is transformed to an Image Block, the data-pin-description attribute is lost, even though I've registered it as a custom attribute to the block (related #8470, #6878).

What's the best way to filter the Image Block transform behavior? It doesn't appear plugins have much control over the behavior:

gutenberg/core-blocks/image/index.js

Lines 112 to 125 in 804abec

	transform: ( node ) => {
	// Search both figure and image classes. Alignment could be
	// set on either. ID is set on the image.
	const className = node.className + ' ' + node.querySelector( 'img' ).className;
	const alignMatches = /(?:^|\s)align(left|center|right)(?:$|\s)/.exec( className );
	const align = alignMatches ? alignMatches[ 1 ] : undefined;
	const idMatches = /(?:^|\s)wp-image-(\d+)(?:$|\s)/.exec( className );
	const id = idMatches ? idMatches[ 1 ] : undefined;
	const anchorElement = node.querySelector( 'a' );
	const linkDestination = anchorElement && anchorElement.href ? 'custom' : undefined;
	const href = anchorElement && anchorElement.href ? anchorElement.href : undefined;
	const blockType = getBlockType( 'core/image' );
	const attributes = getBlockAttributes( blockType, node.outerHTML, { align, id, linkDestination, href } );
	return createBlock( 'core/image', attributes );

[designsimply]

Noting I deleted a comment which was saying the same thing over and over many times per minute in various locations. Spamming and unacceptable behavior towards anyone, or this community, will not be allowed. If you want to give feedback, please do it without spamming or personal attacks.

My apologies for those who were effected during this. Please know we have a zero tolerance policy over this type of behavior.

[danielbachhuber]

I made some progress on an implementation:

if ( 'core/image' !== name ) {
    return settings;
}

const newTransforms = [];
settings.transforms.from.forEach(( transformObj ) => {
    if ( 'raw' === transformObj.type ) {
        const origTransform = transformObj.transform;
        const transformWrap = ( node ) => {

            // Parse the Pinterest attributes and add to our central store.
            const img = node.querySelector( 'img' );
            const className = node.className + ' ' + img.className;
            const idMatches = /(?:^|\s)wp-image-(\d+)(?:$|\s)/.exec( className );
            const id = idMatches ? Number( idMatches[ 1 ] ) : undefined;
            if ( id ) {
                const valueKey = 'image_' + id;
                // img.dataset['pin-description'] is undefined because it's already been sanitized
                setStoreValue( 'descriptions', valueKey, img.dataset['pin-description'] );
            }

            return origTransform( node );
        };
        transformObj = Object.assign( transformObj, {
            transform: transformWrap,
        });
    } else if ( 'shortcode' === transformObj.type ) {
        // todo handle this too
    }
    newTransforms.push( transformObj );
});
settings.transforms.from = newTransforms;

However, the node HTML is unexpectedly sanitized before the transformation is called, which means my data is already missing.

[mcsf]

HTML is unexpectedly sanitized before the transformation is called

@danielbachhuber, is this helpful?

#8648 (comment)