Linked art 'representations' structure

[f27wood]

In the linked-art schema the attached ‘representations’ are not consistent in structure. For example when trying to retrieve all ‘toy train’ objects ( http://nma.conaltuohy.com/xproc-z/object?type=toy+train&media=*&format=json-ld ) you can see three different types of representation structures:

• An array of representations which contains another array of media representations
• A single object representation (media id MA23361721) which contains an array of media representations
• An array of representations which contains a single object media representation (media id 85655)

This is likely a bug, as I would expect a single standard structure for this data. Otherwise a developer needs to account for 3x different methods to retrieve images using linked art schema.

[Conal-Tuohy]

I think part of this may be an issue for the JSON-LD "Compaction" algorithm. By default, in JSON-LD, if a property has a single value then it should be expressed as a JSON object, or if it has multiple values it should be an array of JSON-objects, and this means that the result can be irregular.

However, it's possible to specify an alternative to that default behaviour, and I think the Linked Art context is supposed to control this by specifying that certain properties are always expressed as arrays, even when they have only one value (so that the hierarchy is consistent). This would mean we'd always produce an array of "visual items", containing another of "visual items".

I will check this.

[f27wood]

To discuss

[Conal-Tuohy]

This issue is still present:

• https://data.nma.gov.au/object/168543?format=json-ld has a single representation with a single representation
• https://data.nma.gov.au/object/168025?format=json-ld has a single representation with an array of representations
• https://data.nma.gov.au/object/58698?format=json-ld has an array of representations, each with an array of representations

The JSON-LD compaction stylesheet needs revision