metanorma-tools 0.1.0

data/spec/metanorma/tools/cli_spec.rb

@@ -0,0 +1,102 @@
+# frozen_string_literal: true
+
+require 'spec_helper'
+require 'tempfile'
+
+RSpec.describe Metanorma::Tools::Cli do
+  let(:fixture_path) { File.join(__dir__, '../../fixtures/document-en.dis.presentation.xml') }
+  let(:temp_dir) { Dir.mktmpdir }
+  let(:output_path) { File.join(temp_dir, 'test_output.zip') }
+
+  after { FileUtils.rm_rf(temp_dir) }
+
+  describe '#extract_images' do
+    it 'extracts images from fixture document' do
+      cli = described_class.new
+      cli.options = { output_dir: temp_dir, zip: false }
+      cli.extract_images(fixture_path)
+
+      # Verify files were created
+      png_files = Dir.glob(File.join(temp_dir, '*.png'))
+      expect(png_files.length).to eq(6)
+
+      # Check for expected filenames
+      filenames = png_files.map { |f| File.basename(f) }
+      expect(filenames.any? { |f| f.include?('figA1') }).to be true
+      expect(filenames.any? { |f| f.include?('figC1') }).to be true
+      expect(filenames.any? { |f| f.include?('figC2') }).to be true
+    end
+
+    it 'shows help when no arguments provided' do
+      expect { described_class.start([]) }.to output(/Commands:/).to_stdout
+    end
+
+    it 'shows help for extract-images command' do
+      expect { described_class.start(['help', 'extract-images']) }.to output(/Usage:/).to_stdout
+    end
+
+    it 'handles invalid input file' do
+      cli = described_class.new
+      cli.options = { output_dir: temp_dir }
+      expect { cli.extract_images('non_existent.xml') }.to raise_error(SystemExit)
+    end
+
+    it 'creates output directory if it does not exist' do
+      nested_output = File.join(temp_dir, 'nested', 'dir')
+
+      cli = described_class.new
+      cli.options = { output_dir: nested_output }
+      cli.extract_images(fixture_path)
+
+      expect(Dir.exist?(nested_output)).to be true
+      png_files = Dir.glob(File.join(nested_output, '*.png'))
+      expect(png_files.length).to eq(6)
+    end
+
+    it 'retains original filenames when option is enabled for ISO documents' do
+      cli = described_class.new
+      cli.options = { output_dir: temp_dir, retain_original_filenames: true }
+      cli.extract_images(fixture_path)
+
+      png_files = Dir.glob(File.join(temp_dir, '*.png'))
+      filenames = png_files.map { |f| File.basename(f) }
+
+      # Should include original filename parts
+      expect(filenames.any? { |f| f.include?('_a1.png') }).to be true
+      expect(filenames.any? { |f| f.include?('_b1.png') }).to be true
+      expect(filenames.any? { |f| f.include?('_c2-a.png') }).to be true
+      expect(filenames.any? { |f| f.include?('_c2-b.png') }).to be true
+      expect(filenames.any? { |f| f.include?('_c2-c.png') }).to be true
+    end
+
+    it 'does not retain original filenames when option is disabled' do
+      cli = described_class.new
+      cli.options = { output_dir: temp_dir, retain_original_filenames: false }
+      cli.extract_images(fixture_path)
+
+      png_files = Dir.glob(File.join(temp_dir, '*.png'))
+      filenames = png_files.map { |f| File.basename(f) }
+
+      # Should NOT include original filename parts
+      expect(filenames.any? { |f| f.include?('_a1.png') }).to be false
+      expect(filenames.any? { |f| f.include?('_b1.png') }).to be false
+      expect(filenames.any? { |f| f.include?('_c2-a.png') }).to be false
+    end
+  end
+
+  describe 'command line integration' do
+    it 'can be executed via command line' do
+      # Test the actual CLI execution using Open3 for cross-platform compatibility
+      require 'open3'
+
+      stdout, stderr, status = Open3.capture3("bundle exec metanorma-tools extract-images #{fixture_path} --output-dir #{temp_dir}")
+
+      expect(status.success?).to be true
+      expect(stdout).to include('Successfully extracted')
+    end
+
+    it 'shows version information' do
+      expect { described_class.start(['--version']) }.to output(/#{Metanorma::Tools::VERSION}/).to_stdout
+    end
+  end
+end

data/spec/metanorma/tools/document_metadata_spec.rb

@@ -0,0 +1,308 @@
+# frozen_string_literal: true
+
+require 'spec_helper'
+
+RSpec.describe Metanorma::Tools::DocumentMetadata do
+  describe 'initialization and attributes' do
+    it 'creates metadata with all attributes' do
+      metadata = described_class.new(
+        title: 'Test Document Title',
+        docnumber: '17301',
+        stage: 'DRAFT International Standard',
+        substage: '00',
+        docidentifier: 'ISO/DIS 17301-1:2023',
+        standard_number: '17301',
+        part_number: '1',
+        edition: '1',
+        stage_code: 'dis',
+        stage_abbreviation: 'DIS',
+        flavor: 'iso'
+      )
+
+      expect(metadata.title).to eq('Test Document Title')
+      expect(metadata.docnumber).to eq('17301')
+      expect(metadata.stage).to eq('DRAFT International Standard')
+      expect(metadata.substage).to eq('00')
+      expect(metadata.docidentifier).to eq('ISO/DIS 17301-1:2023')
+      expect(metadata.standard_number).to eq('17301')
+      expect(metadata.part_number).to eq('1')
+      expect(metadata.edition).to eq('1')
+      expect(metadata.stage_code).to eq('dis')
+      expect(metadata.stage_abbreviation).to eq('DIS')
+      expect(metadata.flavor).to eq('iso')
+    end
+
+    it 'sets default flavor to iso' do
+      metadata = described_class.new(
+        title: 'Test Document',
+        docnumber: '17301'
+      )
+
+      expect(metadata.flavor).to eq('iso')
+    end
+
+    it 'allows custom flavor' do
+      metadata = described_class.new(
+        title: 'Test Document',
+        docnumber: '17301',
+        flavor: 'iec'
+      )
+
+      expect(metadata.flavor).to eq('iec')
+    end
+  end
+
+  describe 'auto_prefix generation' do
+    context 'with ISO flavor' do
+      it 'generates correct prefix for DIS stage' do
+        metadata = described_class.new(
+          standard_number: '17301',
+          edition: '1',
+          stage_abbreviation: 'DIS',
+          flavor: 'iso'
+        )
+
+        expect(metadata.auto_prefix).to eq('17301_dis_ed1')
+      end
+
+      it 'generates correct prefix for PWI stage' do
+        metadata = described_class.new(
+          standard_number: '17301',
+          edition: '1',
+          stage_abbreviation: 'PWI',
+          flavor: 'iso'
+        )
+
+        expect(metadata.auto_prefix).to eq('17301_pwi_ed1')
+      end
+
+      it 'generates correct prefix for FDIS stage' do
+        metadata = described_class.new(
+          standard_number: '17301',
+          edition: '2',
+          stage_abbreviation: 'FDIS',
+          flavor: 'iso'
+        )
+
+        expect(metadata.auto_prefix).to eq('17301_fdis_ed2')
+      end
+
+      it 'handles nil stage_abbreviation gracefully' do
+        metadata = described_class.new(
+          standard_number: '17301',
+          edition: '1',
+          stage_abbreviation: nil,
+          flavor: 'iso'
+        )
+
+        expect(metadata.auto_prefix).to eq('17301__ed1')
+      end
+    end
+
+    context 'with other flavors' do
+      it 'generates correct prefix for IEC flavor' do
+        metadata = described_class.new(
+          standard_number: '62304',
+          edition: '1',
+          stage_abbreviation: 'DIS',
+          flavor: 'iec'
+        )
+
+        expect(metadata.auto_prefix).to eq('iec_62304_dis_ed1')
+      end
+
+      it 'generates correct prefix for custom flavor' do
+        metadata = described_class.new(
+          standard_number: '12345',
+          edition: '1',
+          stage_abbreviation: 'DRAFT',
+          flavor: 'custom'
+        )
+
+        expect(metadata.auto_prefix).to eq('custom_12345_draft_ed1')
+      end
+    end
+  end
+
+  describe 'string representation' do
+    context 'with docidentifier and title' do
+      it 'formats with docidentifier and title' do
+        metadata = described_class.new(
+          title: 'Céréales et légumineuses — Spécification et méthodes d\'essai — Riz (DIS)',
+          docidentifier: 'ISO/DIS 17301-1:2023'
+        )
+
+        expected = 'ISO/DIS 17301-1:2023 - Céréales et légumineuses — Spécification et méthodes d\'essai — Riz (DIS)'
+        expect(metadata.to_s).to eq(expected)
+      end
+    end
+
+    context 'without docidentifier' do
+      it 'formats with standard information' do
+        metadata = described_class.new(
+          standard_number: '17301',
+          edition: '1',
+          stage_code: 'dis',
+          stage_abbreviation: 'DIS'
+        )
+
+        expect(metadata.to_s).to eq('ISO 17301 Edition 1 Stage dis (DIS)')
+      end
+    end
+
+    context 'with minimal information' do
+      it 'handles missing fields gracefully' do
+        metadata = described_class.new(
+          title: 'Test Document'
+        )
+
+        expect(metadata.to_s).to eq('ISO  Edition  Stage  ()')
+      end
+    end
+  end
+
+  describe 'serialization' do
+    it 'can be serialized to hash' do
+      metadata = described_class.new(
+        title: 'Test Document',
+        docnumber: '17301',
+        stage: 'DRAFT International Standard',
+        substage: '00',
+        docidentifier: 'ISO/DIS 17301-1:2023',
+        standard_number: '17301',
+        part_number: '1',
+        edition: '1',
+        stage_code: 'dis',
+        stage_abbreviation: 'DIS',
+        flavor: 'iso'
+      )
+
+      hash = metadata.to_hash
+      expect(hash['title']).to eq('Test Document')
+      expect(hash['docnumber']).to eq('17301')
+      expect(hash['stage']).to eq('DRAFT International Standard')
+      expect(hash['substage']).to eq('00')
+      expect(hash['docidentifier']).to eq('ISO/DIS 17301-1:2023')
+      expect(hash['standard_number']).to eq('17301')
+      expect(hash['part_number']).to eq('1')
+      expect(hash['edition']).to eq('1')
+      expect(hash['stage_code']).to eq('dis')
+      expect(hash['stage_abbreviation']).to eq('DIS')
+      expect(hash['flavor']).to eq('iso')
+    end
+
+    it 'can be created from hash' do
+      hash = {
+        'title' => 'Test Document',
+        'docnumber' => '17301',
+        'stage' => 'DRAFT International Standard',
+        'substage' => '00',
+        'docidentifier' => 'ISO/DIS 17301-1:2023',
+        'standard_number' => '17301',
+        'part_number' => '1',
+        'edition' => '1',
+        'stage_code' => 'dis',
+        'stage_abbreviation' => 'DIS',
+        'flavor' => 'iso'
+      }
+
+      metadata = described_class.from_hash(hash)
+      expect(metadata.title).to eq('Test Document')
+      expect(metadata.docnumber).to eq('17301')
+      expect(metadata.auto_prefix).to eq('17301_dis_ed1')
+    end
+  end
+
+  describe 'real-world examples' do
+    it 'handles fixture document metadata correctly' do
+      metadata = described_class.new(
+        title: 'Céréales et légumineuses — Spécification et méthodes d\'essai — Riz (DIS)',
+        docnumber: '17301',
+        stage: 'DRAFT International Standard',
+        substage: '00',
+        docidentifier: 'ISO/DIS 17301-1:2023',
+        standard_number: '17301',
+        part_number: '1',
+        edition: '1',
+        stage_code: 'dis',
+        stage_abbreviation: 'DIS'
+      )
+
+      expect(metadata.auto_prefix).to eq('17301_dis_ed1')
+      expect(metadata.to_s).to include('ISO/DIS 17301-1:2023')
+      expect(metadata.to_s).to include('Céréales et légumineuses')
+    end
+
+    it 'handles multi-part standards' do
+      metadata = described_class.new(
+        title: 'Information technology — Security techniques — Part 3: Guidelines',
+        docnumber: '27001',
+        standard_number: '27001',
+        part_number: '3',
+        edition: '2',
+        stage_code: 'fdis',
+        stage_abbreviation: 'FDIS'
+      )
+
+      expect(metadata.auto_prefix).to eq('27001_fdis_ed2')
+      expect(metadata.part_number).to eq('3')
+    end
+
+    it 'handles different document stages' do
+      stages = [
+        { code: 'pwi', abbr: 'PWI' },
+        { code: 'nwip', abbr: 'NWIP' },
+        { code: 'wd', abbr: 'WD' },
+        { code: 'cd', abbr: 'CD' },
+        { code: 'dis', abbr: 'DIS' },
+        { code: 'fdis', abbr: 'FDIS' },
+        { code: 'is', abbr: 'IS' }
+      ]
+
+      stages.each do |stage|
+        metadata = described_class.new(
+          standard_number: '12345',
+          edition: '1',
+          stage_code: stage[:code],
+          stage_abbreviation: stage[:abbr]
+        )
+
+        expect(metadata.auto_prefix).to eq("12345_#{stage[:abbr].downcase}_ed1")
+      end
+    end
+  end
+
+  describe 'edge cases' do
+    it 'handles empty strings' do
+      metadata = described_class.new(
+        title: '',
+        docnumber: '',
+        standard_number: '',
+        edition: '',
+        stage_abbreviation: ''
+      )
+
+      expect(metadata.auto_prefix).to eq('__ed')
+      expect(metadata.to_s).to be_a(String)
+    end
+
+    it 'handles special characters in title' do
+      metadata = described_class.new(
+        title: 'Test — Document with "special" characters & symbols',
+        docidentifier: 'ISO/DIS 12345:2023'
+      )
+
+      expect(metadata.to_s).to include('Test — Document with "special" characters & symbols')
+    end
+
+    it 'handles very long titles' do
+      long_title = 'A' * 500
+      metadata = described_class.new(
+        title: long_title,
+        docidentifier: 'ISO/DIS 12345:2023'
+      )
+
+      expect(metadata.to_s).to include(long_title)
+    end
+  end
+end

data/spec/metanorma/tools/figure_extractor_spec.rb

@@ -0,0 +1,265 @@
+# frozen_string_literal: true
+
+require 'spec_helper'
+require 'tempfile'
+require 'zip'
+
+RSpec.describe Metanorma::Tools::FigureExtractor do
+  let(:fixture_path) { File.join(__dir__, '../../fixtures/document-en.dis.presentation.xml') }
+  let(:extractor) { described_class.new }
+  let(:temp_dir) { Dir.mktmpdir }
+
+  after { FileUtils.rm_rf(temp_dir) if Dir.exist?(temp_dir) }
+
+  describe '#extract' do
+    context 'with the fixture document containing figures in annexes' do
+      it 'extracts all figures from the document' do
+        # Capture output to verify extraction
+        output = capture_stdout do
+          extractor.extract(fixture_path, temp_dir)
+        end
+
+        expect(output).to include('Found 6 figures')
+        expect(output).to include('Successfully extracted 6 figures')
+
+        # Verify files were created
+        png_files = Dir.glob(File.join(temp_dir, '*.png'))
+        expect(png_files.length).to eq(6)
+      end
+
+      it 'extracts document metadata correctly' do
+        output = capture_stdout do
+          extractor.extract(fixture_path, temp_dir)
+        end
+
+        expect(output).to include('ISO/DIS 17301-1:2023')
+        expect(output).to include('Auto-generated prefix: 17301_dis_ed3')
+        expect(output).to include('Document metadata extraction: Yes')
+      end
+
+      describe 'figure extraction details' do
+        before do
+          capture_stdout { extractor.extract(fixture_path, temp_dir) }
+        end
+
+        it 'extracts all 6 figures with correct data' do
+          png_files = Dir.glob(File.join(temp_dir, '*.png'))
+          expect(png_files.length).to eq(6)
+
+          # Verify all files have content
+          png_files.each do |file|
+            expect(File.size(file)).to be > 1000 # PNG files should be reasonably sized
+          end
+        end
+
+        it 'generates correct ISO graphic filenames for all figures' do
+          png_files = Dir.glob(File.join(temp_dir, '*.png')).map { |f| File.basename(f) }
+
+          # All files should follow the ISO DRG pattern
+          png_files.each do |filename|
+            expect(filename).to match(/^17301_dis_ed3fig.+\.png$/)
+          end
+
+          # Check for specific expected patterns
+          expect(png_files.any? { |f| f.match(/figA1/) }).to be true
+          expect(png_files.any? { |f| f.match(/figC1/) }).to be true
+          expect(png_files.any? { |f| f.match(/figC2/) }).to be true
+        end
+      end
+
+      describe 'image format and data validation' do
+        before do
+          capture_stdout { extractor.extract(fixture_path, temp_dir) }
+        end
+
+        it 'correctly processes all images as PNG with valid data' do
+          png_files = Dir.glob(File.join(temp_dir, '*.png'))
+
+          png_files.each do |file|
+            # Verify PNG file signature
+            File.open(file, 'rb') do |f|
+              signature = f.read(8)
+              expected_signature = "\x89PNG\r\n\x1A\n".dup.force_encoding('ASCII-8BIT')
+              expect(signature).to eq(expected_signature)
+            end
+          end
+        end
+      end
+
+      describe 'archive creation' do
+        let(:zip_extractor) { described_class.new(zip: true) }
+
+        it 'creates a zip archive with all figures' do
+          output = capture_stdout do
+            zip_extractor.extract(fixture_path, temp_dir)
+          end
+
+          expect(output).to include('ZIP archive created')
+
+          zip_files = Dir.glob(File.join(temp_dir, '*.zip'))
+          expect(zip_files.length).to eq(1)
+
+          Zip::File.open(zip_files.first) do |zip_file|
+            expect(zip_file.entries.length).to eq(6)
+
+            zip_file.entries.each do |entry|
+              expect(entry.name).to match(/^17301_dis_ed3fig.+\.png$/)
+              expect(entry.size).to be > 0
+            end
+          end
+        end
+
+        it 'creates correctly named files in archive' do
+          capture_stdout do
+            zip_extractor.extract(fixture_path, temp_dir)
+          end
+
+          zip_files = Dir.glob(File.join(temp_dir, '*.zip'))
+
+          Zip::File.open(zip_files.first) do |zip_file|
+            filenames = zip_file.entries.map(&:name).sort
+
+            # Verify we have the expected number of files
+            expect(filenames.length).to eq(6)
+
+            # All should be PNG files with proper naming
+            filenames.each do |filename|
+              expect(filename).to match(/^17301_dis_ed3fig.+\.png$/)
+            end
+          end
+        end
+      end
+    end
+
+    context 'with error handling' do
+      it 'exits for non-existent file' do
+        expect do
+          capture_stdout { extractor.extract('non_existent.xml') }
+        end.to raise_error(SystemExit)
+      end
+
+      it 'handles malformed XML gracefully' do
+        malformed_xml = Tempfile.new(['malformed', '.xml'])
+        malformed_xml.write('not valid xml content')
+        malformed_xml.close
+
+        expect do
+          capture_stdout { extractor.extract(malformed_xml.path) }
+        end.to raise_error(SystemExit)
+      ensure
+        malformed_xml&.unlink
+      end
+    end
+
+    context 'with CLI integration' do
+      it 'works with the CLI extract-images command' do
+        # Test that the CLI can process the fixture file
+        output = `bundle exec metanorma-tools extract-images #{fixture_path} --output-dir #{temp_dir} 2>&1`
+        expect($?.success?).to be true
+
+        # Verify files were created
+        png_files = Dir.glob(File.join(temp_dir, '*.png'))
+        expect(png_files.length).to eq(6)
+      end
+    end
+  end
+
+  describe 'individual component testing' do
+    describe Metanorma::Tools::Figure do
+      it 'creates figure with valid data' do
+        figure = described_class.new(
+          'A.1',
+          'iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mP8/5+hHgAHggJ/PchI7wAAAABJRU5ErkJggg==',
+          :datauri_png,
+          'test-figure.png'
+        )
+
+        expect(figure.autonum).to eq('A.1')
+        expect(figure.format).to eq(:datauri_png)
+        expect(figure.content).not_to be_empty
+        expect(figure.original_filename).to eq('test-figure.png')
+      end
+    end
+
+    describe Metanorma::Tools::DocumentMetadata do
+      it 'creates metadata with valid data' do
+        metadata = described_class.new(
+          title: 'Test Document',
+          docnumber: '12345',
+          stage: 'DRAFT',
+          substage: '00',
+          docidentifier: 'ISO/DIS 12345:2023',
+          standard_number: '12345',
+          part_number: '1',
+          edition: '1',
+          stage_code: 'dis',
+          stage_abbreviation: 'DIS'
+        )
+
+        expect(metadata.title).to eq('Test Document')
+        expect(metadata.docnumber).to eq('12345')
+        expect(metadata.stage).to eq('DRAFT')
+        expect(metadata.auto_prefix).to eq('12345_dis_ed1')
+      end
+    end
+
+    describe Metanorma::Tools::IsoGraphicFilename do
+      it 'generates correct filename format' do
+        filename = described_class.new(
+          standard_number: 17301,
+          part_number: 1,
+          stage_code: 'dis',
+          edition_number: 3,
+          content_type: 'figure',
+          figure_number: 'A1',
+          original_filename: 'figureA-1.png',
+          file_extension: 'png'
+        )
+
+        expect(filename.to_s).to eq('17301-1_dis_ed3figA1_figureA-1.png')
+      end
+
+      it 'handles UUID-based figure IDs' do
+        filename = described_class.new(
+          standard_number: 17301,
+          part_number: 1,
+          stage_code: 'dis',
+          edition_number: 3,
+          content_type: 'text',
+          text_number: 1,
+          original_filename: '_85f711f6-478d-a680-b5b9-3bc85332dfd1.png',
+          file_extension: 'png'
+        )
+
+        expect(filename.to_s).to eq('17301-1_dis_ed3figText1__85f711f6-478d-a680-b5b9-3bc85332dfd1.png')
+      end
+
+      it 'handles subfigure naming correctly' do
+        filename = described_class.new(
+          standard_number: 17301,
+          part_number: 1,
+          stage_code: 'dis',
+          edition_number: 3,
+          content_type: 'figure',
+          figure_number: 'C2',
+          subfigure: 'a',
+          original_filename: 'figureC-2-a.png',
+          file_extension: 'png'
+        )
+
+        expect(filename.to_s).to eq('17301-1_dis_ed3figC2a_figureC-2-a.png')
+      end
+    end
+  end
+
+  private
+
+  def capture_stdout
+    original_stdout = $stdout
+    $stdout = StringIO.new
+    yield
+    $stdout.string
+  ensure
+    $stdout = original_stdout
+  end
+end