Shrine Gem Ruby on Rails Serverless Image Handler, Background Job

Question

I'm using the Shrine Gem to process files in my Ruby on Rails project using S3 multipart upload. I want to use the Derivatives plugin to point to a AWS Lamdba serverless image handler as my derivative remote_url, and process these derivatives in a background job. I've just about got this working, with one issue:

Directly after uploading a file, that record is saved with the Shrine cache storage url. Once the background job finishes processing and the derivatives are created, the promote job updates the attachment_url with the store endpoint.

So, what I want to do is promote the original URL to the store bucket first (so that default_url will point to the store bucket, not cache), then process the derivatives after this. However, I can't quite figure out how to do this. My Shrine initializer and uploader are below.

# config/initializers/shrine.rb

require 'shrine'
require 'shrine/storage/s3'
# require "shrine/storage/file_system"

s3_options = Rails.application.credentials.s3

Shrine.storages = {
  cache: Shrine::Storage::S3.new(prefix: 'cache', public: true, **s3_options),
  store: Shrine::Storage::S3.new(prefix: 'media', public: true, **s3_options),
}

Shrine.plugin :activerecord           # Load Active Record integration
Shrine.plugin :cached_attachment_data # For retaining cached file on form redisplays
Shrine.plugin :determine_mime_type
Shrine.plugin :infer_extension
Shrine.plugin :instrumentation
Shrine.plugin :pretty_location
Shrine.plugin :remote_url, max_size: 40*1024*1024 # ~40mb
Shrine.plugin :restore_cached_data    # Refresh metadata for cached files
Shrine.plugin :type_predicates
Shrine.plugin :uppy_s3_multipart      # Enable S3 multipart upload for Uppy https://github.com/janko/uppy-s3_multipart
Shrine.plugin :url_options, store: { host: Rails.application.credentials.asset_host } if Rails.application.credentials.asset_host.present?

# app/uploaders/file_uploader.rb

class FileUploader < Shrine
  plugin :derivatives
  plugin :default_url
  plugin :backgrounding

  # TODO: images returned by Shrine.remote_url have file extension set as .jpeg, not .jpg, which is annoying
  # TODO: set up URL fallbacks for backgrounded derivatives? https://shrinerb.com/docs/processing#url-fallbacks

  # The Cloudfront URL generated by the serverless image handler
  SERVERLESS_IMAGE_HOST = Rails.application.credentials.image_handler_host
  DEFAULT_EDITS = {
    rotate: 'auto',
    quality: 60,
    progressive: true,
    chromaSubsampling: '4:4:4',
    withoutEnlargement: true,
    sharpen: true
  }

  # Fall back to the original file URL when the derivative
  # https://shrinerb.com/docs/processing#url-fallbacks
  Attacher.default_url do |derivative: nil, **|
    file&.url if derivative
  end

  # Perform derivative transformations inside a background job
  Attacher.promote_block do |**options|
    if file.image?
      # TODO: the initially promoted file URL is saved to the record as the cache URL. We need to
      # promote the original image first, then perform the derivative job.
      # promote
      AttachmentDerivativeJob.perform_later(self.class.name, record.class.name, record.id, name, file_data)
    else
      promote
    end
  end

  # The derivatives plugin allows storing processed files ("derivatives") alongside the main attached file
  # https://shrinerb.com/docs/plugins/derivatives
  Attacher.derivatives do |original|
    def serverless_image_request(edits = {})
      request_path = Base64.strict_encode64({
        bucket: Shrine.storages[:cache].bucket.name,
        key: [Shrine.storages[:cache].prefix, record.attachment.id].reject(&:blank?).join('/'), # The aws object key of the original image in the `store` S3 bucket
        edits: edits
      }.to_json).chomp
      "#{SERVERLESS_IMAGE_HOST}/#{request_path}"
    end

    if file.image?
      process_derivatives(:image, original)
    else
      process_derivatives(:file, original)
    end
  end

  Attacher.derivatives :image do |original|
    {
      thumb: Shrine.remote_url( serverless_image_request({
        resize: {
          width: 200,
          height: 200,
          fit: 'cover'
        }.reverse_merge(DEFAULT_EDITS)
      })),
      small: Shrine.remote_url( serverless_image_request({
        resize: {
          width: 600,
          height: 600,
          fit: 'inside'
        }.reverse_merge(DEFAULT_EDITS)
      })),
      medium: Shrine.remote_url( serverless_image_request({
        resize: {
          width: 1200,
          height: 1200,
          fit: 'inside'
        }.reverse_merge(DEFAULT_EDITS)
      })),
      large: Shrine.remote_url( serverless_image_request({
        resize: {
          width: 2200,
          height: 2200,
          fit: 'inside'
        }.reverse_merge(DEFAULT_EDITS)
      }))
    }
  end

  Attacher.derivatives :file do |original|
    {}
  end
end

Answer 1

An update - I was able to configure my Uploader class in such a way that it performs how I wanted by overriding the activerecord_after_save method in the Attacher class. Not sure if this is the best solution, but it seems to behave how I expect.

Now, when I upload a series of images they are uploaded directly to S3 and the original file is promoted to permanent Shrine store. In the after save callback, I trigger the AttachmentDerivativeJob, which handles processing of derivatives using AWS Lambda's default serverless image handler.

class FileUploader < Shrine
  plugin :derivatives
  plugin :default_url

  # The Cloudfront URL generated by the serverless image handler
  SERVERLESS_IMAGE_HOST = Rails.application.credentials.image_handler_host
  DEFAULT_EDITS = {
    rotate: 'auto',
    quality: 60,
    progressive: true,
    chromaSubsampling: '4:4:4',
    withoutEnlargement: true,
    sharpen: true
  }

  # Active Record - Overriding callbacks
  # https://shrinerb.com/docs/plugins/activerecord#overriding-callbacks
  class Attacher
    private

    def activerecord_after_save
      super

      if file.image? && derivatives.blank?
        AttachmentDerivativeJob.perform_later(self.class.name, self.record.class.name, self.record.id, self.name, self.file_data)
      end
    end
  end

  # Fall back to the original file URL when the derivative
  # https://shrinerb.com/docs/processing#url-fallbacks
  Attacher.default_url do |derivative: nil, **|
    file&.url if derivative
  end

  # The derivatives plugin allows storing processed files ("derivatives") alongside the main attached file
  # https://shrinerb.com/docs/plugins/derivatives
  Attacher.derivatives do |original|
    def serverless_image_request(edits = {})
      request_path = Base64.strict_encode64({
        bucket: Shrine.storages[:store].bucket.name,
        key: [Shrine.storages[:store].prefix, record.attachment.id].reject(&:blank?).join('/'), # The aws object key of the original image in the `store` S3 bucket
        edits: edits
      }.to_json).chomp
      "#{SERVERLESS_IMAGE_HOST}/#{request_path}"
    end

    if file.image?
      process_derivatives(:image, original)
    else
      process_derivatives(:file, original)
    end
  end

  Attacher.derivatives :image do |original|
    {
      thumb: Shrine.remote_url( serverless_image_request({
        resize: {
          width: 200,
          height: 200,
          fit: 'cover'
        }.reverse_merge(DEFAULT_EDITS)
      })),
      small: Shrine.remote_url( serverless_image_request({
        resize: {
          width: 600,
          height: 600,
          fit: 'inside'
        }.reverse_merge(DEFAULT_EDITS)
      })),
      medium: Shrine.remote_url( serverless_image_request({
        resize: {
          width: 1200,
          height: 1200,
          fit: 'inside'
        }.reverse_merge(DEFAULT_EDITS)
      })),
      large: Shrine.remote_url( serverless_image_request({
        resize: {
          width: 2200,
          height: 2200,
          fit: 'inside'
        }.reverse_merge(DEFAULT_EDITS)
      }))
    }
  end

  Attacher.derivatives :file do |original|
    {}
  end
end