FEATURE: Translate categories

Author: nattsw

app/jobs/regular/translate_categories.rb

@@ -0,0 +1,45 @@
+# frozen_string_literal: true
+
+module Jobs
+  class TranslateCategories < ::Jobs::Base
+    cluster_concurrency 1
+    BATCH_SIZE = 50
+
+    def execute(args)
+      return unless SiteSetting.translator_enabled
+      return unless SiteSetting.experimental_category_translation
+
+      locales = SiteSetting.automatic_translation_target_languages.split("|")
+      return if locales.blank?
+
+      cat_id = args[:from_category_id] || Category.order(:id).first&.id
+      last_id = cat_id
+
+      # we're just gonna take all categories and keep it simple
+      # instead of checking in the db which ones are absent
+      categories = Category.where("id >= ?", cat_id).order(:id).limit(BATCH_SIZE)
+      return if categories.empty?
+
+      categories.each do |category|
+        CategoryLocalization.transaction do
+          locales.each do |locale|
+            next if CategoryLocalization.exists?(category_id: category.id, locale: locale)
+            begin
+              DiscourseTranslator::CategoryTranslator.translate(category, locale)
+            rescue => e
+              Rails.logger.error(
+                "Discourse Translator: Failed to translate category #{category.id} to #{locale}: #{e.message}",
+              )
+            end
+          end
+        end
+        last_id = category.id
+      end
+
+      # from batch if needed
+      if categories.size == BATCH_SIZE
+        Jobs.enqueue_in(10.seconds, :translate_categories, from_category_id: last_id + 1)
+      end
+    end
+  end
+end

app/jobs/scheduled/automatic_category_translation.rb

@@ -0,0 +1,18 @@
+# frozen_string_literal: true
+
+module Jobs
+  class AutomaticCategoryTranslation < ::Jobs::Scheduled
+    every 12.hours
+    cluster_concurrency 1
+
+    def execute(args)
+      return unless SiteSetting.translator_enabled
+      return unless SiteSetting.experimental_category_translation
+
+      locales = SiteSetting.automatic_translation_target_languages.split("|")
+      return if locales.blank?
+
+      Jobs.enqueue(:translate_categories)
+    end
+  end
+end

app/services/discourse_translator/category_translator.rb

@@ -0,0 +1,22 @@
+# frozen_string_literal: true
+
+module DiscourseTranslator
+  class CategoryTranslator
+    # unlike post and topics, categories do not have a detected locale
+    # and will translate two fields, name and description
+
+    def self.translate(category, target_locale = I18n.locale)
+      return if category.blank? || target_locale.blank?
+
+      # locale can come in various forms
+      # standardize it to a _ symbol
+      target_locale_sym = target_locale.to_s.sub("-", "_").to_sym
+
+      translator = DiscourseTranslator::Provider::TranslatorProvider.get
+      translated_name = translator.translate_text!(category.name, target_locale_sym)
+      translated_description = translator.translate_text!(category.description, target_locale_sym)
+
+      category.update!(name: translated_name, description: translated_description)
+    end
+  end
+end

app/services/discourse_translator/translator.rb

@@ -0,0 +1,18 @@
+# frozen_string_literal: true
+
+module DiscourseTranslator
+  # The canonical class for all your translation needs
+  class Translator
+    # this invokes the specific methods
+    def translate(translatable, target_locale = I18n.locale)
+      target_locale_sym = target_locale.to_s.sub("-", "_").to_sym
+
+      case translatable.class.name
+      when "Post", "Topic"
+        DiscourseTranslator::Provider.TranslatorProvider.get.translate(translatable, target_locale_sym)
+      when "Category"
+        CategoryTranslator.translate(translatable, target_locale)
+      end
+    end
+  end
+end

config/settings.yml

@@ -129,6 +129,9 @@ discourse_translator:
   experimental_inline_translation:
     default: false
     client: true
+  experimental_category_translation:
+    default: false
+    hidden: true
   discourse_translator_verbose_logs:
     default: false
     client: false

spec/jobs/translate_categories_spec.rb

@@ -0,0 +1,127 @@
+# frozen_string_literal: true
+
+require "rails_helper"
+
+describe Jobs::TranslateCategories do
+  let(:translator) { mock }
+
+  def localize_all_categories(*locales)
+    Category.all.each do |category|
+      locales.each { |locale| Fabricate(:category_localization, category:, locale:, name: "x") }
+    end
+  end
+
+  before do
+    SiteSetting.translator_enabled = true
+    SiteSetting.experimental_category_translation = true
+    SiteSetting.automatic_translation_backfill_rate = 100
+    SiteSetting.automatic_translation_target_languages = "pt|zh_CN"
+
+    DiscourseTranslator::Provider.stubs(:get).returns(translator)
+    Jobs.run_immediately!
+  end
+
+  it "does nothing when translator is disabled" do
+    SiteSetting.translator_enabled = false
+
+    translator.expects(:translate_text!).never
+
+    subject.execute({})
+  end
+
+  it "does nothing when experimental_category_translation is disabled" do
+    SiteSetting.experimental_category_translation = false
+
+    translator.expects(:translate_text!).never
+
+    subject.execute({})
+  end
+
+  it "does nothing when no target languages are configured" do
+    SiteSetting.automatic_translation_target_languages = ""
+
+    translator.expects(:translate_text!).never
+
+    subject.execute({})
+  end
+
+  it "does nothing when no categories exist" do
+    Category.destroy_all
+
+    translator.expects(:translate_text!).never
+
+    subject.execute({})
+  end
+
+  it "translates categories to the configured locales" do
+    number_of_categories = Category.count
+    DiscourseTranslator::CategoryTranslator
+      .expects(:translate)
+      .with(is_a(Category), "pt")
+      .times(number_of_categories)
+    DiscourseTranslator::CategoryTranslator
+      .expects(:translate)
+      .with(is_a(Category), "zh_CN")
+      .times(number_of_categories)
+
+    subject.execute({})
+  end
+
+  it "skips categories that already have localizations" do
+    localize_all_categories("pt", "zh_CN")
+
+    category1 =
+      Fabricate(:category, name: "First Category", description: "First category description")
+    Fabricate(:category_localization, category: category1, locale: "pt", name: "Primeira Categoria")
+
+    # It should only translate to Chinese, not Portuguese
+    DiscourseTranslator::CategoryTranslator.expects(:translate).with(category1, "pt").never
+    DiscourseTranslator::CategoryTranslator.expects(:translate).with(category1, "zh_CN").once
+
+    subject.execute({})
+  end
+
+  it "continues from a specified category ID" do
+    category1 = Fabricate(:category, name: "First", description: "First description")
+    category2 = Fabricate(:category, name: "Second", description: "Second description")
+
+    DiscourseTranslator::CategoryTranslator
+      .expects(:translate)
+      .with(category1, any_parameters)
+      .never
+    DiscourseTranslator::CategoryTranslator
+      .expects(:translate)
+      .with(category2, any_parameters)
+      .twice
+
+    subject.execute(from_category_id: category2.id)
+  end
+
+  it "handles translation errors gracefully" do
+    localize_all_categories("pt", "zh_CN")
+
+    category1 = Fabricate(:category, name: "First", description: "First description")
+    DiscourseTranslator::CategoryTranslator
+      .expects(:translate)
+      .with(category1, "pt")
+      .raises(StandardError.new("API error"))
+    DiscourseTranslator::CategoryTranslator.expects(:translate).with(category1, "zh_CN").once
+
+    expect { subject.execute({}) }.not_to raise_error
+  end
+
+  it "enqueues the next batch when there are more categories" do
+    Jobs::TranslateCategories.const_set(:BATCH_SIZE, 1)
+
+    Jobs
+      .expects(:enqueue_in)
+      .with(10.seconds, :translate_categories, from_category_id: any_parameters)
+      .times(Category.count)
+
+    subject.execute({})
+
+    # Reset the constant
+    Jobs::TranslateCategories.send(:remove_const, :BATCH_SIZE)
+    Jobs::TranslateCategories.const_set(:BATCH_SIZE, 50)
+  end
+end

spec/services/category_translator_spec.rb

@@ -0,0 +1,65 @@
+# frozen_string_literal: true
+
+describe DiscourseTranslator::CategoryTranslator do
+  fab!(:category) do
+    Fabricate(:category, name: "Test Category", description: "This is a test category")
+  end
+
+  describe ".translate" do
+    let(:target_locale) { :fr }
+    let(:translator) { mock }
+
+    before { DiscourseTranslator::Provider::TranslatorProvider.stubs(:get).returns(translator) }
+
+    it "translates the category name and description" do
+      translator
+        .expects(:translate_text!)
+        .with(category.name, target_locale)
+        .returns("Catégorie de Test")
+      translator
+        .expects(:translate_text!)
+        .with(category.description, target_locale)
+        .returns("C'est une catégorie de test")
+
+      DiscourseTranslator::CategoryTranslator.translate(category, target_locale)
+
+      expect(category.name).to eq("Catégorie de Test")
+      expect(category.description).to eq("C'est une catégorie de test")
+    end
+
+    it "handles locale format standardization" do
+      translator.expects(:translate_text!).with(category.name, :fr_CA).returns("Catégorie de Test")
+      translator
+        .expects(:translate_text!)
+        .with(category.description, :fr_CA)
+        .returns("C'est une catégorie de test")
+
+      DiscourseTranslator::CategoryTranslator.translate(category, "fr-CA")
+
+      expect(category.name).to eq("Catégorie de Test")
+      expect(category.description).to eq("C'est une catégorie de test")
+    end
+
+    it "returns nil if category is blank" do
+      expect(DiscourseTranslator::CategoryTranslator.translate(nil)).to be_nil
+    end
+
+    it "returns nil if target locale is blank" do
+      expect(DiscourseTranslator::CategoryTranslator.translate(category, nil)).to be_nil
+    end
+
+    it "uses I18n.locale as default when no target locale is provided" do
+      I18n.locale = :es
+      translator.expects(:translate_text!).with(category.name, :es).returns("Categoría de Prueba")
+      translator
+        .expects(:translate_text!)
+        .with(category.description, :es)
+        .returns("Esta es una categoría de prueba")
+
+      DiscourseTranslator::CategoryTranslator.translate(category)
+
+      expect(category.name).to eq("Categoría de Prueba")
+      expect(category.description).to eq("Esta es una categoría de prueba")
+    end
+  end
+end