epub.rb in softcover-1.0.beta13

- old
+ new

@@ -5,12 +5,16 @@
     # Returns the name of the cover file.
     # We support (in order) JPG/JPEG, PNG, and TIFF.
     def cover_img
       extensions = %w[jpg jpeg png tiff]
       extensions.each do |ext|
-        file = Dir[path("#{images_dir}/cover.#{ext}")].first
-        return File.basename(file) if file
+        origin = "images/cover.#{ext}"
+        target = "#{images_dir}/cover.#{ext}"
+        if File.exist?(origin)
+          FileUtils.cp(origin, target)
+          return File.basename(target)
+        end
       end
       return false
     end
 
     def cover?
@@ -37,10 +41,11 @@
         if manifest.markdown?
           opts = options.merge({ source: :polytex, origin: :markdown })
           self.manifest = Softcover::BookManifest.new(opts)
         end
         remove_html
+        remove_images
         create_directories
         write_mimetype
         write_container_xml
         write_ibooks_xml
         write_toc
@@ -62,15 +67,21 @@
       # All the HTML is generated, so this clears out any unused files.
       def remove_html
         FileUtils.rm(Dir.glob(path('epub/OEBPS/*.html')))
       end
 
+      # Removes images in case they are stale.
+      def remove_images
+        rm_r images_dir
+      end
+
       def create_directories
         mkdir('epub')
         mkdir(path('epub/OEBPS'))
         mkdir(path('epub/OEBPS/styles'))
         mkdir(path('epub/META-INF'))
+        mkdir(images_dir)
         mkdir('ebooks')
       end
 
       # Writes the mimetype file.
       # This is required by the EPUB standard.
@@ -274,14 +285,27 @@
       def clean_book_id(filename)
         File.read(filename).gsub(/#book/, '')
       end
 
       # Copies the image files from the HTML version of the document.
-      # We remove PDF images, which are valid in PDF documents but not in EPUB.
       def copy_image_files
+        # Copy over all images to guarantee the same directory structure.
         FileUtils.cp_r(File.join('html', 'images'),
                        File.join('epub', 'OEBPS'))
-        File.delete(*Dir['epub/OEBPS/images/**/*.pdf'])
+        # Parse the full HTML file with Nokogiri to get images actually used.
+        html = File.read(manifest.full_html_file)
+        html_image_filenames = Nokogiri::HTML(html).css('img').map do |node|
+                                 node.attributes['src'].value
+                               end
+        # Form the corresponding EPUB image paths.
+        used_image_filenames = html_image_filenames.map do |filename|
+                                 "epub/OEBPS/#{filename}"
+                               end.to_set
+        # Delete unused images.
+        Dir.glob("epub/OEBPS/images/**/*").each do |image|
+          next if File.directory?(image)
+          rm image unless used_image_filenames.include?(image)
+        end
       end
 
       # Make the EPUB, which is basically just a zipped HTML file.
       def make_epub(options={})
         filename = manifest.filename
\ No newline at end of file