Merge pull request #6 from ARPC/ocr

Implement OCR

.vscode/extensions.json

@@ -0,0 +1,7 @@
+{
+  "recommendations": [
+    "rebornix.ruby",
+    "oderwat.indent-rainbow",
+    "ms-vsliveshare.vsliveshare"
+  ]
+}

.vscode/settings.json

@@ -0,0 +1,13 @@
+{
+  "editor.rulers": [
+    100,
+    120
+  ],
+  "files.exclude": {
+    "**/.git": true,
+    "**/.svn": true,
+    "**/.hg": true,
+    "**/.DS_Store": true,
+    "**/tmp": true
+  }
+}

Gemfile.lock

@@ -1,7 +1,7 @@
 PATH
   remote: .
   specs:
-    pdfservices (0.1.1)
+    pdfservices (0.1.2)
       http (~> 5.1)
       json (~> 2.6)
       jwt (>= 1.0, < 3.0)
@@ -22,7 +22,7 @@ GEM
       ffi (>= 1.0.0)
       rake
     hashdiff (1.0.1)
-    http (5.1.0)
+    http (5.1.1)
       addressable (~> 2.8)
       http-cookie (~> 1.0)
       http-form_data (~> 2.2)
@@ -31,7 +31,7 @@ GEM
       domain_name (~> 0.5)
     http-form_data (2.3.0)
     json (2.6.2)
-    jwt (2.5.0)
+    jwt (2.6.0)
     llhttp-ffi (0.4.0)
       ffi-compiler (~> 1.0)
       rake (~> 13.0)
@@ -74,9 +74,9 @@ GEM
       hashdiff (>= 0.4.0, < 2.0.0)
 
 PLATFORMS
+  arm64-darwin-21
   ruby
   x86_64-linux
-  arm64-darwin-21
 
 DEPENDENCIES
   minitest (~> 5.0)

README.md

@@ -34,8 +34,10 @@ Then you need to create a json file with your credentials:
 }
 
 ```
+### Supported API calls:
 
-Right now the only supported API is document merge. See `test/pdf_services_sdk/test_integration.rb` for an example usage.
+- Document merge. See `test/pdf_services_sdk/test_document_merge.rb` for an example usage.
+- OCR. See `test/pdf_services_sdk/test_ocr.rb` for an example usage.
 
 ## Development
 

lib/pdfservices.rb

@@ -5,6 +5,7 @@
 require "jwt"
 require "multipart_parser/reader"
 require "yaml"
+require "pdfservices/version"
 
 module PdfServices
   autoload :CredentialsBuilder, "pdfservices/credentials_builder"
@@ -15,4 +16,9 @@ module DocumentMerge
     autoload :Operation, "pdfservices/document_merge/operation"
     autoload :Result, "pdfservices/document_merge/result"
   end
+
+  module Ocr
+    autoload :Operation, "pdfservices/ocr/operation"
+    autoload :Result, "pdfservices/ocr/result"
+  end
 end

lib/pdfservices/ocr/operation.rb

@@ -0,0 +1,104 @@
+# frozen_string_literal: true
+
+require "http"
+require "pdfservices/jwt_provider"
+require "pdfservices/ocr/result"
+require "yaml"
+
+module PdfServices
+  module Ocr
+    class Operation
+      PRESIGNED_URL_ENDPOINT = "https://pdf-services.adobe.io/assets"
+      OCR_ENDPOINT = "https://pdf-services.adobe.io/operation/ocr"
+      ASSETS_ENDPOINT = "https://pdf-services.adobe.io/assets"
+
+      def initialize(credentials = nil)
+        @credentials = credentials
+      end
+
+      def get_presigned_url
+        response = api.post(PRESIGNED_URL_ENDPOINT, json: {mediaType: "application/pdf"})
+        if response.status == 200
+          JSON.parse(response.body.to_s)
+        else
+          Result.new(nil, "Unexpected response status from get presigned url: #{response.status}")
+        end
+      end
+
+      def upload_asset(source_pdf)
+        presigned_url = get_presigned_url
+        upload_uri = presigned_url["uploadUri"]
+        asset_id = presigned_url["assetID"]
+        aws = HTTP.headers({"Content-Type": "application/pdf"})
+        response = aws.put(upload_uri, body: File.open(source_pdf))
+        if response.status == 200
+          asset_id
+        else
+          Result.new(nil, "Unexpected response status from asset upload: #{response.status}")
+        end
+      end
+
+      def delete_the_asset(asset_id)
+        api.delete("#{ASSETS_ENDPOINT}/#{asset_id}")
+      end
+
+      def execute(source_pdf)
+        asset_id = upload_asset(source_pdf)
+        response = api.post(OCR_ENDPOINT, json: {assetID: asset_id})
+        if response.status == 201
+          document_url = response.headers["location"]
+          poll_document_result(document_url, asset_id)
+        else
+          Result.new(nil, "Unexpected response status from ocr endpoint: #{response.status}\nasset_id: #{asset_id}")
+        end
+      end
+
+      private
+
+      def api_headers
+        {
+          Authorization: "Bearer #{JwtProvider.get_jwt(@credentials)}",
+          "x-api-key": @credentials.client_id,
+          "Content-Type": "application/json"
+        }
+      end
+
+      def api
+        @api ||= HTTP.headers(api_headers)
+      end
+
+      def poll_document_result(url, original_asset_id)
+        sleep(1)
+        response = api.get(url)
+        if response.status == 200
+          json_response = JSON.parse(response.body.to_s)
+          ocr_asset_id = json_response&.[]("asset")&.[]("assetID")
+          case json_response["status"]
+          when "in progress"
+            poll_document_result(url, original_asset_id)
+          when "done"
+            # download_the_asset
+            response = HTTP.get(json_response["asset"]["downloadUri"])
+            # delete the assets
+            delete_the_asset(original_asset_id) if !original_asset_id.nil?
+            delete_the_asset(ocr_asset_id) if !ocr_asset_id.nil?
+            # return the result
+            Result.new(response.body, nil)
+          when "failed"
+            # delete the original asset
+            delete_the_asset(original_asset_id) if !original_asset_id.nil?
+            Result.new(nil, "OCR Failed")
+          else
+            # delete the original asset
+            delete_the_asset(original_asset_id) if original_asset_id.present?
+            Result.new(nil, "Unexpected status from polling: #{json_response["status"]}")
+          end
+        else
+          # delete the original asset
+          delete_the_asset(original_asset_id) if original_asset_id.present?
+          Result.new(nil, "Unexpected response status from polling: #{json_response["status"]}")
+        end
+      end
+    end
+  end
+end

lib/pdfservices/ocr/result.rb

@@ -0,0 +1,24 @@
+require "json"
+require "multipart_parser/reader"
+
+module PdfServices
+  module Ocr
+    class Result
+      attr_accessor :document_body, :error
+
+      def initialize(document_body, error)
+        @document_body = document_body
+        @error = error
+      end
+
+      def success?
+        @document_body != nil
+      end
+
+      def save_as_file(file_path)
+        location = File.join(Dir.pwd, file_path)
+        File.write(location, @document_body)
+      end
+    end
+  end
+end

lib/pdfservices/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module PdfServices
-  VERSION = "0.1.1"
+  VERSION = "0.1.2"
 end

test/fixtures/ocr_done.json

@@ -0,0 +1,3 @@
+{
+  "status": "in progress"
+}

test/fixtures/ocr_in_progress.json

@@ -0,0 +1,7 @@
+{
+  "status": "done",
+  "asset": {
+    "assetID": "ocr'd:asset-id",
+    "downloadUri": "https://ocr.file.url"
+  }
+}

test/fixtures/presigned_upload_url_response.json

@@ -0,0 +1,4 @@
+{
+  "assetID": "urn:a-real-long-asset-asset-id",
+  "uploadUri": "https://a.presigned.url"
+}

test/integration_spikes/ocr.rb

@@ -0,0 +1,19 @@
+# frozen_string_literal: true
+
+require "lib/pdfservices"
+
+credentials = ::PdfServices::CredentialsBuilder.new
+  .with_client_id(ENV["PDF_SERVICES_CLIENT_ID"])
+  .with_client_secret(ENV["PDF_SERVICES_CLIENT_SECRET"])
+  .with_organization_id(ENV["PDF_SERVICES_ORGANIZATION_ID"])
+  .with_account_id(ENV["PDF_SERVICES_ACCOUNT_ID"])
+  .with_private_key(ENV["PDF_SERVICES_PRIVATE_KEY"])
+  .build
+
+operation = ::PdfServices::Ocr::Operation.new(credentials)
+
+result = operation.execute("test/fixtures/files/not_yet_ocr.pdf")
+
+puts(result.error)
+
+result.save_as_file("tmp/ocr_result.pdf")

test/pdf_services_sdk/test_integration.rb → test/pdf_services_sdk/test_document_merge.rb

@@ -2,7 +2,7 @@
 
 require "test_helper"
 
-class IntegrationTest < Minitest::Test
+class DocumentMergeTest < Minitest::Test
   def test_it_works
     stub_valid_response_sequence
 

test/pdf_services_sdk/test_ocr.rb

@@ -0,0 +1,88 @@
+# frozen_string_literal: true
+
+require "test_helper"
+
+class OcrTest < Minitest::Test
+  def test_it_works
+    stub_valid_response_sequence
+
+    # Initial setup, create credentials instance.
+    credentials = valid_credentials
+
+    # source file
+    source_pdf_path = File.join(Dir.pwd, "test", "fixtures", "files", "not_yet_ocr.pdf")
+
+    operation = ::PdfServices::Ocr::Operation.new(credentials)
+
+    # Execute the operation
+    result = operation.execute(source_pdf_path)
+
+    assert result.success?
+    assert_equal "this is fake ocr'd pdf\n", result.document_body.to_s
+  end
+
+  private
+
+  def stub_valid_response_sequence
+    # get JWT for requests
+    stub_request(:post, "https://ims-na1.adobelogin.com/ims/exchange/jwt/")
+      .to_return(status: 200, body: json_fixture("valid_jwt_response"))
+
+    # get a presigned url to upload the source pdf
+    stub_request(:post, "https://pdf-services.adobe.io/assets")
+      .with(headers: secured_headers)
+      .to_return(
+        status: 200,
+        headers: json_headers,
+        body: json_fixture("presigned_upload_url_response")
+      )
+
+    # upload the source pdf
+    stub_request(:put, "https://a.presigned.url").to_return(status: 200)
+
+    # request the OCR operation
+    stub_request(:post, "https://pdf-services.adobe.io/operation/ocr")
+      .with(headers: secured_headers)
+      .to_return(
+        status: 201,
+        headers: {"location" => "https://some.polling.url"}.merge(json_headers)
+      )
+
+    # poll for the result
+    stub_request(:get, "https://some.polling.url")
+      .with(headers: secured_headers)
+      .to_return(status: 200, headers: json_headers, body: json_fixture("ocr_in_progress"))
+      .to_return(status: 200, headers: json_headers, body: json_fixture("ocr_in_progress"))
+      .to_return(status: 200, headers: json_headers, body: json_fixture("ocr_done"))
+
+    # download the ocr'd pdf
+    stub_request(:get, "https://ocr.file.url")
+      .to_return(status: 200, headers: pdf_headers, body: file_fixture("fake_ocr_done.pdf"))
+
+    # delete the original asset
+    stub_request(:delete, "https://pdf-services.adobe.io/assets/urn:a-real-long-asset-asset-id")
+      .with(headers: secured_headers)
+      .to_return(status: 200, body: "", headers: {})
+
+    # delete the ocr'd asset
+    stub_request(:delete, "https://pdf-services.adobe.io/assets/ocr'd:asset-id")
+      .with(headers: secured_headers)
+      .to_return(status: 200, body: "", headers: {})
+  end
+
+  def secured_headers
+    {
+      Authorization: "Bearer fake1.fake2.fake3",
+      "Content-Type": "application/json",
+      "X-Api-Key": "123someclientid"
+    }
+  end
+
+  def json_headers
+    {"Content-Type" => "application/json;charset=UTF-8"}
+  end
+
+  def pdf_headers
+    {"Content-Type" => "application/pdf"}
+  end
+end