Every catalog application supports an image or set of images bound to this item. For that purpose, in the next chapter, we will see how to work with binary objects in MongoDB. However, now is the time to decide how to semantically link the binary data to an item document. Extending the model schema in such a way that it contains binary base64 representation of binary data in the document is anything but a good idea, as mixing literally encoded and binary data in one format is never a good idea. It increases the complexity of the application and makes it error-prone:

{
  "_id": "5a4c004b0eed73835833cc9a",
  "itemId": "1",
  "itemName": "Sports Watch",
  "price": 100,
  "currency": "EUR",
  "categories": [
    "Watches",
    "Sports Watches"
  ],
  "image":" 
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 
"} 

Imagine how large a result for a non-filtered query can become just for a few hundred items, if all of them had an image binary representation as a value of a JSON attribute. To avoid that, we will return the image for each item at a URL that is logically linked to the resource's URL—/catalog/v2/item/{itemId}/image.

That way, if there is an image assigned to an item, it will be served at a known location. This approach, however, does not semantically link the binary item to its corresponding resource, as when accessing the item at /catalog/v2/item/{itemId}, there would be no indication of whether it has an image assigned or not. To solve this, let's use a custom HTTP header in response of the item route:

GET http://localhost:3000/catalog/v2/item/1 HTTP/1.1 
Host: localhost:3000 
Connection: Keep-Alive 
User-Agent: Apache-HttpClient/4.1.1 (java 1.5) 

HTTP/1.1 200 OK 
X-Powered-By: Express 
Content-Type: application/json; charset=utf-8 
Content-Length: 152 
Image-Url: http://localhost:3000/catalog/v2/item/1/image
ETag: W/"98-2nJj2mZdLV2YDME3WYCyEwIXfuA" 
Date: Thu, 01 Feb 2018 13:50:43 GMT 
Connection: keep-alive 

{
  "_id": "5a4c004b0eed73835833cc9a",
  "itemId": "1",
  "itemName": "Sports Watch",
  "price": 100,
  "currency": "EUR",
  "__v": 0,
  "categories": [
    "Watches",
    "Sports Watches"
  ]
}

When present in the response, the Image-Url header indicates that the item has an additional resource bound to it, and the header value provides the address where it is available. Using this approach, we linked a binary resource semantically to our document.

In the next chapter, we will implement the routes that will handle the manipulation of arbitrary items bound to the items in the catalog.