Apollo Graphql Query An Uploaded File

Apollo Server 2.0 has the ability to receive file uploads as described in this blog post.

However, all the tutorials and blog posts I found only showed how to upload a file. Nobody demonstrated how to actually retrieve the file back to display it onscreen.

Does anybody know how to properly query the file contents for display onscreen?
Also, there's the possibility that maybe there is no way of querying a file and you have to build a separate rest endpoint to retrieve the contents?

Some thoughts:
I imagine the query to be something like

query {
  fetchImage(id: 'someid')

with the respective server-side definition

type Query {
  fetchImage(id : ID!): Upload //maybe also a custom type, but how do I include the actual file contents?

Hint:Upload is a scalar type that apollo-server automatically adds to your type definition. It is used for the upload so I imaging it also being usable for the download/query. Please read the blog post mentioned above for more information.



The response from a GraphQL service is always serialized as a JSON object. Technically, a format other than JSON could be used in serialization but in practice only JSON is used because it meets the serialization requirements in the spec. So, the only way to send a file through GraphQL would be to convert the file into some format that's JSON-compatible. For example, you could convert a Buffer to a byte array and send that as an array of integers. You would also have to send the appropriate mime type. It would be up to the client to convert the byte array back into a usable format on receiving the response.

If you go this route, you'd have to use your own scalar or object type -- the Upload scalar does not support serialization so it will throw if you try to use it as an output type (and it's not really suitable for this sort of thing anyway).

However, while doing this is technically possible, it's also inadvisable. Serializing a larger file could cause you to run out of memory since there's no way to stream data through GraphQL (the entire response has to be in memory before it can be sent). It's much better to serve the file statically (ideally using nginx instead of Node). If your API needs to refer to the file, it can then just return the file's path.