Forwarding files to another service.

[bhargavcp]

Hello,

First of all, thank you for this awesome library :)

I have a question: How do you suggest forwarding the files that we receive in a resolver to another service? Currently, I write the file locally with stream.pipe(createWriteStream(filename)) and then forward that file to the other service that I want. But I think there should be a better way to achieve this.

Is it BusBoy that is stopping me from doing that? Because, I remember looking at formidable and it gives me the entire file object, but BusBoy seems to only give few of the fields from the File object and does some 'magic' under the hood?

[jaydenseric]

I suppose the context is the API changes from apollo-upload-server v3 -> v4+?

It seems in the past (when formidable was used) that you were saving an upload to the filesystem, then creating a stream from it to upload it to another service. That looks like the sync scenario of the diagram below. With the new API, you get a stream you direct anywhere from the start, looking more like the async scenario.

See the docs. You can do some shallow validation based on the filename and mimetype, but importantly you want to const stream = createReadStream() and pass stream into the API for whatever cloud service you are using.

[mike-marcacci]

Hi @bhargavcp,

You shouldn't need to create an intermediary file, but should simply be able to pipe the upload's contents to your external service. The fields that are available (name, type, etc) are everything that is available before the entire content has streamed in. Anything that is calculated from the actual contents (a hash, size, etc) require that the stream be completely consumed. If you need any of these, you can use the createReadStream() API from v7.0.0 to compute these while the upload's contents stream to a permanent location:

const { filename, mimetype, createReadStream } = await upload

// pipe contents to a hash function
createReadStream()
  .on('error', handleError)
  .pipe(hash)

// pipe contents to S3, Google Storage, SAN, etc
createReadStream()
  .on('error', handleError)
  .pipe(destination)

[bhargavcp]

Thank you for your reply @jaydenseric and @mike-marcacci. I have another maybe related question. How can I server images from my GraphQL server? Example, I want the user to be able to query like this:
query { userProfile(id: .....) { image name phone } }

[jaydenseric]

When resolving the upload, I suggest:

1. Create a unique file ID.

2. Store the image under that ID in the cloud (e.g. S3, Digital Ocean Spaces, etc.).

3. Use a thumbnailing/CDN service like imgix. Use it to extract image metadata.

4. If the metadata is not valid, delete the S3 file and resolve validation errors.

5. In your database, record the image under the file ID along with key bits of metadata (user ID, image dimensions, etc.).

6. Make an Image type. Mine looks like this:

   

   The url resolver simply constructs an imgix url for the file ID. If you're lazy you could just make it return the base URL, and document that the user needs to append their own query string for size, etc.

[bhargavcp]

@jaydenseric What is the type ImageFormat? Thanks for your reply.

[jaydenseric]

@bhargavcp sorry for forgetting to get back to you all those years ago! The answer to your question can be seen here: #251 (comment) .