Deterministic ActiveStorage blob keys

lib/parklife/rails/activestorage.rb

@@ -1,4 +1,6 @@
 # frozen_string_literal: true
+require_relative 'blob_modifications'
+
 module Parklife
   module Rails
     module ActiveStorage

lib/parklife/rails/blob_modifications.rb

@@ -0,0 +1,65 @@
+# frozen_string_literal: true
+module Parklife
+  module Rails
+    module BlobModifications
+      def self.prepended(base)
+        base.before_validation :set_parklife_key
+      end
+
+      # Generate a deterministic MD5 hash for the Blob based on its attributes.
+      #
+      # This means that for a completely fresh Parklife build the same file with
+      # the same contents will be stored with the same filename so there's no
+      # unnecessary cache-busting between freshly-built deployments.
+      #
+      # The hash is also guaranteed to be unique so it's possible that the same
+      # filename/contents creates a different value, in this case it's also
+      # produced deterministically so successive hashes will be the same - but
+      # ideally you should prevent this from happening in the first place.
+      #
+      # @return [String]
+      def generate_parklife_key
+        key = nil
+        i = 0
+
+        loop do
+          key = OpenSSL::Digest::MD5.hexdigest(
+            [
+              byte_size,
+              checksum,
+              content_type,
+              filename,
+              service_name,
+              i,
+            ].map(&:to_s).join
+          )
+
+          break if self.class.where(key: key).none?
+
+          i += 1
+        end
+
+        key
+      end
+
+      # Override the standard setter so `has_secure_token :key` does nothing.
+      def key=(_)
+        nil
+      end
+
+      # Override the standard getter which otherwise assigns a random token.
+      def key
+        self[:key]
+      end
+
+      private
+        def set_parklife_key
+          self[:key] ||= generate_parklife_key
+        end
+    end
+  end
+end
+
+ActiveSupport.on_load(:active_storage_blob) do
+  ActiveStorage::Blob.prepend Parklife::Rails::BlobModifications
+end

spec/parklife/blob_modifications_spec.rb

@@ -0,0 +1,88 @@
+require 'rails_helper'
+
+RSpec.describe Parklife::Rails::BlobModifications do
+  context 'when the same file/filename is attached' do
+    it 'the same key is assigned' do
+      post = FactoryBot.create(:post)
+      post.hero.attach(
+        filename: 'plasma.jpg',
+        io: file_fixture('plasma.jpg').open,
+      )
+
+      id1 = post.hero.blob.id
+      key1 = post.hero.blob.key
+
+      post.hero.purge
+      post.hero.attach(
+        filename: 'plasma.jpg',
+        io: file_fixture('plasma.jpg').open,
+      )
+
+      id2 = post.hero.blob.id
+      key2 = post.hero.blob.key
+
+      expect(id2).not_to eql(id1)
+      expect(key2).to eql(key1)
+    end
+  end
+
+  context 'when the same file with a different filename is attached' do
+    it 'a different key is assigned' do
+      post1 = FactoryBot.create(:post)
+      post1.hero.attach(
+        filename: 'foo.jpg',
+        io: file_fixture('plasma.jpg').open,
+      )
+
+      post2 = FactoryBot.create(:post)
+      post2.hero.attach(
+        filename: 'bar.jpg',
+        io: file_fixture('plasma.jpg').open,
+      )
+
+      key1 = post1.hero.blob.key
+      key2 = post2.hero.blob.key
+
+      expect(key1).not_to eql(key2)
+    end
+  end
+
+  context 'when a duplicate file/filename is attached' do
+    it 'a different key is assigned' do
+      post1 = FactoryBot.create(:post)
+      post1.hero.attach(
+        filename: 'plasma.jpg',
+        io: file_fixture('plasma.jpg').open,
+      )
+
+      post2 = FactoryBot.create(:post)
+      post2.hero.attach(
+        filename: 'plasma.jpg',
+        io: file_fixture('plasma.jpg').open,
+      )
+
+      key1 = post1.hero.blob.key
+      key2 = post2.hero.blob.key
+
+      expect(key1).not_to eql(key2)
+
+      post1.hero.purge
+      post2.hero.purge
+
+      post1.hero.attach(
+        filename: 'plasma.jpg',
+        io: file_fixture('plasma.jpg').open,
+      )
+      post2.hero.attach(
+        filename: 'plasma.jpg',
+        io: file_fixture('plasma.jpg').open,
+      )
+
+      key3 = post1.hero.blob.key
+      key4 = post2.hero.blob.key
+
+      expect(key3).to eql(key1)
+      expect(key4).to eql(key2)
+    end
+  end
+end