Merge json and build csv with duplicates

Here's an example of some geojson that I have:

var json = {
  features: [
    {
      properties: {
        osm_key: "amenity",
        extent: [
          151.214672,
          -33.8562966,
          151.2158814,
          -33.8574149
        ],
        street: "Lower Concourse",
        name: "Sydney Opera House",
        state: "New South Wales",
        osm_id: 4960757,
        osm_type: "W",
        postcode: "2061",
        osm_value: "theatre",
        country: "Australia"
      },
      type: "Feature",
      geometry: {
        type: "Point",
        coordinates: [
          151.2152582491399,
          -33.85685575
        ]
      }
    },
    {
      properties: {
        osm_key: "tourism",
        extent: [
          151.214672,
          -33.8562966,
          151.2158814,
          -33.8574149
        ],
        street: "Lower Concourse",
        name: "Sydney Opera House",
        state: "New South Wales",
        osm_id: 4960757,
        osm_type: "W",
        postcode: "2061",
        osm_value: "attraction",
        country: "Australia"
      },
      type: "Feature",
      geometry: {
        type: "Point",
        coordinates: [
          151.2152582491399,
          -33.85685575
        ]
      }
    },
    {
      properties: {
        osm_key: "highway",
        extent: [
          -95.6987584,
          29.9960185,
          -95.6984449,
          29.9907477
        ],
        name: "Opera House Row Drive",
        state: "Texas",
        osm_id: 261793234,
        osm_type: "W",
        postcode: "77433",
        osm_value: "residential",
        city: "Cypress",
        country: "United States of America"
      },
      type: "Feature",
      geometry: {
        type: "Point",
        coordinates: [
          -95.698749,
          29.993634
        ]
      }
    },
    {
      properties: {
        osm_key: "tourism",
        street: "葆台路",
        name: "Sydney Opera House",
        state: "Beijing",
        osm_id: 3184358225,
        osm_type: "N",
        postcode: "100070",
        osm_value: "attraction",
        city: "Beijing",
        country: "China"
      },
      type: "Feature",
      geometry: {
        type: "Point",
        coordinates: [
          116.2844542,
          39.8078321
        ]
      }
    },
    {
      properties: {
        osm_key: "amenity",
        street: "Macquarie Street",
        name: "Opera House Car Park",
        state: "New South Wales",
        osm_id: 2877110066,
        osm_type: "N",
        postcode: "2000",
        osm_value: "parking",
        country: "Australia"
      },
      type: "Feature",
      geometry: {
        type: "Point",
        coordinates: [
          151.2135144,
          -33.8593646
        ]
      }
    }
  ],
  type: "FeatureCollection"
}

As you can see there are 2 entries for Sydney Opera House - one tagged with osm_value "theatre", the other tagged with "attraction".

I don't have control over the data being returned so I need a javascript function that I can pass the json to and it return a geoJson object in the same format but with the duplicate entries removed and the osm_values of the duplicates merged in csv format like:

var json = {
  features: [
    {
      properties: {
        osm_key: "amenity",
        extent: [
          151.214672,
          -33.8562966,
          151.2158814,
          -33.8574149
        ],
        street: "Lower Concourse",
        name: "Sydney Opera House",
        state: "New South Wales",
        osm_id: 4960757,
        osm_type: "W",
        postcode: "2061",
        osm_value: "theatre, attraction",
        country: "Australia"
      },
      type: "Feature",
      geometry: {
        type: "Point",
        coordinates: [
          151.2152582491399,
          -33.85685575
        ]
      }
    },
    {
      properties: {
        osm_key: "highway",
        extent: [
          -95.6987584,
          29.9960185,
          -95.6984449,
          29.9907477
        ],
        name: "Opera House Row Drive",
        state: "Texas",
        osm_id: 261793234,
        osm_type: "W",
        postcode: "77433",
        osm_value: "residential",
        city: "Cypress",
        country: "United States of America"
      },
      type: "Feature",
      geometry: {
        type: "Point",
        coordinates: [
          -95.698749,
          29.993634
        ]
      }
    },
    {
      properties: {
        osm_key: "tourism",
        street: "葆台路",
        name: "Sydney Opera House",
        state: "Beijing",
        osm_id: 3184358225,
        osm_type: "N",
        postcode: "100070",
        osm_value: "attraction",
        city: "Beijing",
        country: "China"
      },
      type: "Feature",
      geometry: {
        type: "Point",
        coordinates: [
          116.2844542,
          39.8078321
        ]
      }
    },
    {
      properties: {
        osm_key: "amenity",
        street: "Macquarie Street",
        name: "Opera House Car Park",
        state: "New South Wales",
        osm_id: 2877110066,
        osm_type: "N",
        postcode: "2000",
        osm_value: "parking",
        country: "Australia"
      },
      type: "Feature",
      geometry: {
        type: "Point",
        coordinates: [
          151.2135144,
          -33.8593646
        ]
      }
    }
  ],
  type: "FeatureCollection"
}

The following needs to be bourne in mind:

• there may or may not be any duplicates to filter out


• the returned geoJson must stay in the same order but with the duplicates removed


• duplicates may not be adjacent items in the features array


• the structure of the json should stay more or less the same but the keys in the geoJson should not be hard-coded because additional child keys (eg coordinates, street) could be added at any point


• duplicates are recognised by having both the same properties.osm_id and properties.osm_type


• jquery is being used on the page

I've seen the following page that mentions something similar but the answer by adeneo looks to be limited to hard-coded keys: http://ift.tt/1Q0QAyk whereas I need it to be dynamic.

Hopefully someone can help.