Merge pull request #123 from alphagov/fix-body-content-parsing

Refactor parsing of body content

app/models/concerns/publishing_api/content.rb

@@ -27,7 +27,7 @@ def content
       values_from_parts = document_hash.dig(:details, :parts)&.map do
         # Add the part title as a heading to help the search model better understand the structure
         # of the content
-        ["<h1>#{_1[:title]}</h1>", ContentWithMultipleTypes.new(_1[:body]).html_content]
+        ["<h1>#{_1[:title]}</h1>", BodyContent.new(_1[:body]).html_content]
       end
 
       [*values_from_json_paths, *values_from_parts]

app/models/concerns/publishing_api/metadata.rb

@@ -117,7 +117,7 @@ def parts
           {
             slug: _1[:slug],
             title: _1[:title],
-            body: ContentWithMultipleTypes.new(_1[:body]).summarized_text_content,
+            body: BodyContent.new(_1[:body]).summarized_text_content,
           }
         end
     end

app/models/publishing_api/body_content.rb

@@ -0,0 +1,33 @@
+module PublishingApi
+  class BodyContent
+    def initialize(raw_content)
+      @content = case raw_content
+                 when String
+                   raw_content
+                 when Array
+                   raw_content.find { _1[:content_type] == "text/html" }&.dig(:content)
+                 end
+    end
+
+    def html_content
+      content
+    end
+
+    def text_content
+      return nil unless html_content
+
+      Loofah
+        .document(html_content)
+        .to_text(encode_special_chars: false)
+        .squish
+    end
+
+    def summarized_text_content(length: 75, omission: "…", separator: " ")
+      text_content&.truncate(length, omission:, separator:)
+    end
+
+  private
+
+    attr_reader :content
+  end
+end

spec/models/concerns/publishing_api/metadata_spec.rb

@@ -291,7 +291,7 @@
         it "contains the expected body with HTML stripped and truncated" do
           expect(extracted_parts.map { _1[:body] }).to eq([
             "Lorem ipsum dolor sit amet, consectetur adipiscing elit. Curabitur…",
-            "",
+            nil,
           ])
         end
       end

spec/models/publishing_api/body_content_spec.rb

@@ -0,0 +1,78 @@
+RSpec.describe PublishingApi::BodyContent do
+  subject(:body_content) { described_class.new(content) }
+
+  context "when the content is a plain string" do
+    let(:content) { "Hello, world!" }
+
+    describe "#html_content" do
+      subject(:html_content) { body_content.html_content }
+
+      it { is_expected.to eq("Hello, world!") }
+    end
+
+    describe "#text_content" do
+      subject(:text_content) { body_content.text_content }
+
+      it { is_expected.to eq("Hello, world!") }
+    end
+
+    describe "#summarized_text_content" do
+      subject(:summarized_text_content) { body_content.summarized_text_content(length: 6) }
+
+      it { is_expected.to eq("Hello…") }
+    end
+  end
+
+  context "when the content is an array of typed content that includes text/html content" do
+    let(:content) do
+      [
+        { content_type: "application/json", content: '{"foo": "bar"}' },
+        { content_type: "text/html", content: "<blink>Hello, world!</blink>" },
+      ]
+    end
+
+    describe "#html_content" do
+      subject(:html_content) { body_content.html_content }
+
+      it { is_expected.to eq("<blink>Hello, world!</blink>") }
+    end
+
+    describe "#text_content" do
+      subject(:text_content) { body_content.text_content }
+
+      it { is_expected.to eq("Hello, world!") }
+    end
+
+    describe "#summarized_text_content" do
+      subject(:summarized_text_content) { body_content.summarized_text_content(length: 6) }
+
+      it { is_expected.to eq("Hello…") }
+    end
+  end
+
+  context "when the content is an array of typed content that doesn't include text/html content" do
+    let(:content) do
+      [
+        { content_type: "application/json", content: '{"foo": "bar"}' },
+      ]
+    end
+
+    describe "#html_content" do
+      subject(:html_content) { body_content.html_content }
+
+      it { is_expected.to be_nil }
+    end
+
+    describe "#text_content" do
+      subject(:text_content) { body_content.text_content }
+
+      it { is_expected.to be_nil }
+    end
+
+    describe "#summarized_text_content" do
+      subject(:summarized_text_content) { body_content.summarized_text_content }
+
+      it { is_expected.to be_nil }
+    end
+  end
+end