Framing and @reverse

[azaroth42]

Further to #311, should the predicate that triggered the use of an @reverse property in a frame be removed from its resource to avoid a necessarily circular reference?

For example, if the data shape is:

X includes Y
Z is Y
Z in X

And there's a @reverse of "in" called "proxies" which is then @embed always via a Frame, should the Z resources also have in back to X?

I would have expected:

{
  "id": "X",
  "includes":  ["Y"],
  "uses": [{"id": "Z", "is": "Y"}]
}

But instead the in property is still present in Z.

Use Case: http://openarchives.org/ore/0.9/jsonld#proxies

[gkellogg]

@azaroth42 could you come up with a specific, minimal example which illustrates the issue? You might use my distiller as a way of expressing the actual behavior, at least until the last bits of framing support are in the playground.

[azaroth42]

Input JSON-LD:

{
  "@context": {
    "eg": "https://example.org/ns/",
    "reversed": {"@reverse": "eg:forwards", "@type": "@id"},
    "forwards": {"@id": "eg:forwards", "@type": "@id"},
    "Thing": {"@id": "eg:Thing"}
  },
  "@graph": [
    {
      "@id": "http://example.com/1",
      "@type": "Thing"
     },
    {
      "@id": "http://example.com/2",
      "@type": "Thing",
      "forwards": "http://example.com/1"
    }
  ]
}

Input Frame:

{
  "@context": {
    "eg": "https://example.org/ns/",
    "reversed": {"@reverse": "eg:forwards", "@type": "@id"},
    "forwards": {"@id": "eg:forwards", "@type": "@id"},
    "Thing": {"@id": "eg:Thing"}
  },
  "@id": "http://example.com/1",
  "reversed": {"@embed": "@always"}
}

Output JSON-LD:

{
  "@context": {
    "eg": "https://example.org/ns/",
    "reversed": {
      "@reverse": "eg:forwards",
      "@type": "@id"
    },
    "forwards": {
      "@id": "eg:forwards",
      "@type": "@id"
    },
    "Thing": {
      "@id": "eg:Thing"
    }
  },
  "@graph": [
    {
      "@id": "http://example.com/1",
      "@type": "Thing",
      "reversed": {
        "@id": "http://example.com/2",
        "@type": "Thing",
        "forwards": "http://example.com/1"
      }
    }
  ]
}

Note the presence of both reversed and forwards, which is encoding the same triple (2 forwards 1) twice.

[gkellogg]

@azaroth42 doesn't look like this was ever resolved or put into the framing spec, do you want to replicate it over there?

[tanuja3088]

Yes I am also facing an issue regarding the same, About
attached sample input and frame
sample.zip

[gkellogg]

It's not an error to include this triple, but you can construct a frame which will eliminate it (playground):

{
  "@context": {
    "eg": "https://example.org/ns/",
    "reversed": {"@reverse": "eg:forwards", "@type": "@id"},
    "forwards": {"@id": "eg:forwards", "@type": "@id"},
    "Thing": {"@id": "eg:Thing"}
  },
  "@id": "http://example.com/1",
  "reversed": {
    "@embed": "@always",
    "@explicit": true
  }
}

[tanuja3088]

Hi, I quite a dint understand what changes I need to do in the frame as per my sample example?

[gkellogg]

I added "@explicit": true to the "reversed" property, so that it only emits those properties explicitly listed.

[tanuja3088]

òk..I tried doing it but still get the same response there is no part info test:hasPart in the output

[gkellogg]

I can't parse your example, as there may be encoding issues. Can you send a (shortened) link to the playground with your example loaded?

[tanuja3088]

Here is the URL, the output I get of Playground is not that I get from the jsonld-java

https://tinyurl.com/yd55h8d3

[gkellogg]

jsonld-java may not support @reverse in frames yet. It's a new feature in 1.1.

[tanuja3088]

ahh ok