How to write a query simulating a very primitive recommendation system?

Users Dgraph
2

Try this.

{

  U1 as var(func: eq(name, "Product 1"))

  var(func: type(Order)) @cascade {
    products @filter(uid(U1))
      also_bought: products @filter(NOT uid(U1)) {
        G as uid
      }
    }

      also_bought(func: uid(G)) {
        name
      }
}

Modified dataset

{
   "set": [
      {
         "uid": "_:0x21",
         "dgraph.type": "Product",
         "name": "Product 1"
      },
      {
         "uid": "_:0x22",
         "dgraph.type": "Product",
         "name": "Product 2"
      },
      {
         "uid": "_:0x23",
         "dgraph.type": "Product",
         "name": "Product 3"
      },
      {
         "uid": "_:0x24",
         "dgraph.type": "Product",
         "name": "Product 4"
      },
      {
         "uid": "_:0x25",
         "dgraph.type": "Product",
         "name": "Product 5"
      },
      {
         "dgraph.type": "Order",
         "products": [
            {
               "uid": "_:0x21"
            },
            {
               "uid": "_:0x22"
            }
         ]
      },
      {
         "dgraph.type": "Order",
         "products": [
            {
               "uid": "_:0x21"
            },
            {
               "uid": "_:0x23"
            }
         ]
      },
      {
         "dgraph.type": "Order",
         "products": [
            {
               "uid": "_:0x23"
            },
            {
               "uid": "_:0x24"
            }
         ]
      },
      {
         "dgraph.type": "Order",
         "products": [
            {
               "uid": "_:0x21"
            },
            {
               "uid": "_:0x22"
            },
            {
               "uid": "_:0x23"
            }
         ]
      }
   ]
}