Basic diffing is complete.

Author: hansonw

Diff for: .gitignore

@@ -3,3 +3,4 @@ pkg
 *.png
 *.wmf
 *.emf
+~*

Diff for: Gemfile

@@ -6,3 +6,4 @@ gem 'htmlentities'
 gem 'roman-numerals'
 gem 'rmagick'
 gem 'ruby-prof'
+gem 'diff-lcs'

Diff for: examples/image2.docx

@@ -8,10 +8,9 @@
 module YDocx
   class Builder
     include MarkupMethod
-    attr_accessor :contents, :files, :style, :title
+    attr_accessor :contents, :style, :title
     def initialize(contents)
       @contents = contents
-      @files = Pathname.new('.')
       @style = false
       @title = ''
       init
@@ -99,11 +98,7 @@ def build_tag(tag, content, attributes, mode=:html)
       unless attributes.empty?
         attributes.each_pair do |key, value|
           next if mode == :xml and key.to_s =~ /(id|style|colspan)/u
-          if tag == :img and key == :src
-            _attributes << " src=\"#{resolve_path(value.to_s)}\""
-          else
-            _attributes << " #{key.to_s}=\"#{value.to_s}\""
-          end
+          _attributes << " #{key.to_s}=\"#{value.to_s}\""
         end
       end
       if mode == :xml
@@ -134,12 +129,18 @@ def style
 }
 td {
   padding: 5px 10px;
+}
+td.add {
+  background: rgba(124,252,0,0.3);
+}
+td.delete {
+  background: rgba(255,0,0,0.1);
+}
+td.modify {
+  background: rgba(0,0,255,0.1);
 }
       CSS
       style.gsub(/\s\s+|\n/, ' ')
     end
-    def resolve_path(path)
-      @files.join path
-    end
   end
 end

Diff for: examples/test1.docx

@@ -2,6 +2,7 @@
 # encoding: utf-8
 
 require 'ydocx'
+require 'ydocx/differ'
 
 module YDocx
   class Command
@@ -72,9 +73,8 @@ def run_diff
               files << path
             end
           end
-          docs = files.map do |file|
-            puts YDocx::Document.open(file).contents.hash
-          end
+          docs = files.map { |f| YDocx::Document.open(f) }
+          puts YDocx::Differ.diff(*docs)
         end
       end
       def version

Diff for: examples/test2.docx

@@ -0,0 +1,151 @@
+#!/usr/bin/env ruby
+# encoding: utf-8
+
+require 'ydocx'
+require 'diff/lcs'
+
+module YDocx
+  class Differ
+    def self.chunk_similarity(chunk1, chunk2)
+      text1 = chunk1.map { |c| c.join('') }
+      text2 = chunk2.map { |c| c.join('') }
+      # Find the LCS between the runs
+      lcs = Diff::LCS.LCS(text1, text2)
+      lcs_len = 0
+      lcs.each do |run|
+        lcs_len += run.length
+      end
+      tlen = 0
+      (text1 + text2).each do |run|
+        tlen += run.length
+      end
+      if tlen == 0
+        return 1
+      else
+        return 2.0 * lcs_len / tlen
+      end
+    end
+    # Return the % similarity between two blocks (paragraphs, tables)
+    def self.block_similarity(block1, block2)
+      if block1.class != block2.class
+        return 0
+      end
+      if block1.is_a? Paragraph
+        chunk_similarity(block1.get_chunks, block2.get_chunks)
+      elsif block1.is_a? Table
+        return 1
+      else
+        return 0
+      end
+    end
+    def self.get_chunks(paragraphs)
+      chunks = []
+      paragraphs.each_with_index do |p, i|
+        if i > 0
+          chunks << [Run.new("\r", Style.new)]
+        end
+        chunks += p.get_chunks
+      end
+      chunks
+    end
+    def self.diff(doc1, doc2)
+      blocks1 = doc1.contents.blocks
+      blocks2 = doc2.contents.blocks
+      # Do an n^2 LCS diff on the blocks.
+      n = blocks1.length
+      m = blocks2.length
+      lcs = Array.new(n+1) { Array.new(m+1, 0) }
+      action = Array.new(n+1) { Array.new(m+1, -1) }
+      blocks1.reverse.each_with_index do |a, ii|
+        blocks2.reverse.each_with_index do |b, jj|
+          if n*m > 1000
+            sim = (a == b ? 1 : 0)
+          else
+            sim = block_similarity(a, b)
+          end
+          i = n-1-ii
+          j = m-1-jj
+          lcs[i][j] = lcs[i+1][j]
+          action[i][j] = 0
+          if lcs[i][j+1] > lcs[i][j]
+            lcs[i][j] = lcs[i][j+1]
+            action[i][j] = 1
+          end
+          if sim > 0.5 && lcs[i+1][j+1] + sim > lcs[i][j]
+            lcs[i][j] = lcs[i+1][j+1] + sim
+            action[i][j] = 2
+          end
+        end
+      end
+      i = 0
+      j = 0
+      lblocks = []
+      rblocks = []
+      diff_blocks = []
+      while i < n || j < m        
+        if j == m || action[i][j] == 0
+          lblocks << blocks1[i]
+          i += 1
+        elsif i == n || action[i][j] == 1
+          rblocks << blocks2[j]
+          j += 1
+        else
+          unless lblocks.empty? && rblocks.empty?
+            if lblocks.empty? || rblocks.empty? ||
+               chunk_similarity(get_chunks(lblocks), get_chunks(rblocks)) > 0.5
+              diff_blocks << [lblocks.dup, rblocks.dup]
+            else
+              puts lblocks.to_s
+              puts rblocks.to_s
+              diff_blocks << [lblocks.dup, []]
+              diff_blocks << [[], rblocks.dup]
+            end
+            lblocks = []
+            rblocks = []
+          end
+          diff_blocks << [[blocks1[i]], [blocks2[j]]]
+          i += 1
+          j += 1
+        end
+      end
+      unless lblocks.empty? && rblocks.empty?
+        if lblocks.empty? || rblocks.empty? ||
+           chunk_similarity(get_chunks(lblocks), get_chunks(rblocks)) > 0.5
+          diff_blocks << [lblocks.dup, rblocks.dup]
+        else
+          diff_blocks << [lblocks.dup, []]
+          diff_blocks << [[], rblocks.dup]
+        end
+      end
+      
+      table = Table.new
+      diff_blocks.each do |block|
+        c1 = Cell.new
+        c1.blocks = block[0]
+        c2 = Cell.new
+        c2.blocks = block[1]
+        if block[0].empty?
+          c2.class = 'add'
+        elsif block[1].empty?
+          c1.class = 'delete'
+        elsif block[0] != block[1]
+          c1.class = c2.class = 'modify'
+        end
+        table.cells << [c1, c2]
+      end
+      
+      [doc1, doc2].each do |doc|
+        if !doc.images.empty?
+          doc.create_files
+        end
+      end
+      
+      html_doc = ParsedDocument.new
+      html_doc.blocks << table
+      builder = Builder.new(html_doc)
+      builder.title = 'Diff Results'
+      builder.style = true
+      builder.build_html
+    end
+  end
+end

Diff for: examples/test3.docx

@@ -37,7 +37,6 @@ def to_html(output=false)
       files = output_directory
       @builder = Builder.new(@contents) do |builder|
         builder.title = @path.basename
-        builder.files = files.relative_path_from(files.dirname)
         builder.style = true
         html = builder.build_html
       end
@@ -64,7 +63,6 @@ def to_xml(output=false)
       end
       xml
     end
-    private
     def create_files
       files_dir = output_directory
       mkdir Pathname.new(files_dir) unless files_dir.exist?
@@ -76,6 +74,7 @@ def create_files
         organize_image(origin_path, source_path, image[:data])
       end
     end
+   private
     def organize_image(origin_path, source_path, data)
       if source_path.extname != origin_path.extname # convert
         output_file = output_directory.join(source_path)
@@ -118,7 +117,7 @@ def read(file)
         end
       end
       rel = @zip.find_entry('word/_rels/document.xml.rels').get_input_stream
-      @parser = Parser.new(doc, rel, rel_files) do |parser|
+      @parser = Parser.new(doc, rel, rel_files, output_directory) do |parser|
         @contents = parser.parse
         @images = parser.images
       end