doc2text 0.1

checksums.yaml

@@ -0,0 +1,7 @@
+---
+SHA1:
+  metadata.gz: 1821e833815ea821090507cea37fdec0c68dc2af
+  data.tar.gz: 7815dd2e3f7fbf1f822959e4a120ef0e0bcbef79
+SHA512:
+  metadata.gz: 5b3a0e9729eccccd888432527455336214d52ad996920593e30edab1d2e5f29bcdefdca19dfdbf9f7bf686bfabb8a6888cf706a386f058d7ee07b9bacac1e9e8
+  data.tar.gz: 6ffbb43bd9c8e4eac000b4733f3bcb149a6cee3b8cd608a22a0ca4516fbbc7317a2d30b4b0f15d8dee0c103df509ee24b029393503bd4bb40e16d22ed6c1543a

data/lib/doc2text.rb

@@ -0,0 +1,11 @@
+require 'nokogiri'
+#require 'nokogiri/xml'
+require 'fileutils'
+
+require 'doc2text/odt'
+require 'doc2text/odt_xml_node'
+require 'doc2text/namespaces'
+require 'doc2text/markdown'
+require 'doc2text/errors'
+
+require 'doc2text/content'

data/lib/doc2text/content.rb

@@ -0,0 +1,25 @@
+module Doc2Text
+  module Odt
+    module Content
+      class Document < ::Nokogiri::XML::SAX::Document
+        def initialize(markdown_document)
+          @markdown_document = markdown_document
+        end
+
+        def start_element_namespace(name ,attrs = [], prefix = nil, uri = nil, ns = [])
+          @markdown_document.new_node prefix, name, attrs
+        end
+
+        def end_element_namespace(name, prefix = nil, uri = nil)
+          @markdown_document.close_node prefix, name
+        end
+
+        def characters(string)
+          unless string.strip.empty?
+            @markdown_document << string
+          end
+        end
+      end
+    end
+  end
+end

data/lib/doc2text/errors.rb

@@ -0,0 +1,5 @@
+module Doc2Text
+  class XmlError < StandardError
+
+  end
+end

data/lib/doc2text/markdown.rb

@@ -0,0 +1,85 @@
+module Doc2Text
+  module Markdown
+    class Document
+      def initialize(output)
+        @output = output
+        @automatic_styles = {}
+      end
+
+      def new_node(prefix, name, attrs)
+        unless @xml_root
+          @xml_root = @current_node = Odt::XmlNodes::Node.create_node prefix, name, nil, attrs, self
+        else
+          new_node = Odt::XmlNodes::Node.create_node prefix, name, @current_node, attrs, self
+          @current_node.children << new_node
+          @current_node = new_node
+          self << @current_node.open
+        end
+      end
+
+      def close_node(prefix, name)
+        if Odt::XmlNodes::Node.create_node(prefix, name).eql? @current_node
+          if @current_node.delete_on_close?
+            remove_current_node!
+          else
+            remove_current_node! false
+          end
+        elsif Odt::XmlNodes::Node.create_node(prefix, name).eql? @current_node.parent
+          if @current_node.parent.delete_on_close?
+            remove_current_node!
+            remove_current_node!
+          else
+            remove_current_node! false
+            remove_current_node! false
+          end
+        else
+          # TODO remove this redundant(tree build algorithm) checks
+          raise Doc2Text::XmlError, "!Close node child #{prefix} #{name} IS NOT correct, CURRENT_ELEM #{@current_node}"
+        end
+      end
+
+      def remove_current_node!(remove = true)
+        return if !@current_node
+        self << @current_node.close
+        node_for_deletion = @current_node
+        @current_node = @current_node.parent
+        return unless @current_node
+        if remove
+          @current_node.remove_last_child! node_for_deletion
+        end
+      end
+
+      def <<(string)
+        @output << string
+      end
+
+      def close
+        @output.close
+      end
+
+      def print_tree(node)
+        puts node
+        node.children.each do |child|
+          print_tree child
+        end
+      end
+
+      # Select nodes xpath style
+      # - supports selecting from the root node
+      def xpath(string)
+        if /^(\/[\w:\-]+)+$/ =~ string
+          path = string.scan /[\w:\-]+/
+          seek_nodes = [@xml_root]
+          path.each_with_index do |xml_name, index|
+            seek_nodes.select! { |node| node.xml_name == xml_name }
+            seek_nodes = seek_nodes.map(&:children).flatten unless index == path.length - 1
+            break if seek_nodes.empty?
+          end
+          seek_nodes
+        else
+          raise Doc2Text::XmlError, 'it does not support this xpath syntax'
+        end
+      end
+    end
+  end
+end

data/lib/doc2text/namespaces.rb

@@ -0,0 +1,169 @@
+module Doc2Text
+  module Odt
+    module XmlNodes
+      class Generic
+        include Node
+      end
+
+      #
+      # These are the namespaces available in the open document format
+      # http://docs.oasis-open.org/office/v1.2/os/OpenDocument-v1.2-os.html
+      #
+      module Office
+        class AutomaticStyles
+          include Node
+
+          def visit
+            :automatic_styles
+          end
+
+          def delete_on_close?
+            false
+          end
+        end
+
+        class DocumentContent
+          include Node
+
+          def delete_on_close?
+            false # required for testing purposes. After a document has been parsed, some tests could be run against the tree built
+          end
+        end
+      end
+
+      module Animation; end
+      module Chart; end
+      module Config; end
+      module Database; end
+      module Dr3d; end
+      module Drawing; end
+      module Form; end
+      module Manifest; end
+      module Meta; end
+      module DataStyle; end
+      module Presentation; end
+      module Script; end
+      module Table; end
+      module Style
+        class Style
+          include Node
+
+          def delete_on_close?
+            false
+          end
+        end
+
+        class TextProperties
+          include Node
+
+          def delete_on_close?
+            false
+          end
+        end
+      end
+      module XslFoCompatible; end
+      module SvgCompatible; end
+      module SmilCompatible; end
+      module Of; end
+
+      module Text
+        def initialize(parent = nil, attrs = [], prefix = nil, name = nil, markdown_document = nil)
+          super parent, attrs, prefix, name, markdown_document
+          @markdown_document = markdown_document
+          style_index = attrs.index { |attr| attr.prefix == 'text' && attr.localname == 'style-name' }
+          @enclosing_style = []
+          if style_index
+            fetch_style attrs[style_index].value
+          end
+        end
+
+        def fetch_common_style(style)
+          if style
+            style.children.select { |style_property| style_property.xml_name == 'style:text-properties' }.each { |text_property|
+              text_property.attrs.each { |attr|
+                if attr.prefix == 'style'
+                  if attr.localname == 'font-style-complex' && attr.value == 'italic'
+                    @enclosing_style << '_'
+                  elsif attr.localname == 'font-weight-complex' && attr.value == 'bold'
+                    @enclosing_style << '**'
+                  end
+                end
+              }
+            }
+          end
+        end
+
+        def fetch_style(style_name)
+          styles = @markdown_document.xpath '/office:document-content/office:automatic-styles/style:style'
+          style = styles.find { |style| style.attrs.index { |attr| attr.prefix == 'style' && attr.localname == 'family' && attr.value == self.class.style_family } &&
+              style.attrs.index { |attr| attr.prefix == 'style' && attr.localname == 'name' && attr.value == style_name } }
+          fetch_common_style style
+        end
+
+        # http://docs.oasis-open.org/office/v1.2/os/OpenDocument-v1.2-os-part1.html#__RefHeading__1419256_253892949
+        class P
+          include Node
+          include Text
+
+          def initialize(parent = nil, attrs = [], prefix = nil, name = nil, markdown_document = nil)
+            super parent, attrs, prefix, name, markdown_document
+          end
+
+          def self.style_family
+            'paragraph'
+          end
+
+          def open
+            "\n#{@enclosing_style.join}"
+          end
+
+          def close
+            "#{@enclosing_style.reverse.join}\n"
+          end
+        end
+
+        class LineBreak
+          include Node
+
+          def open
+            '<br/>'
+          end
+        end
+
+        # http://docs.oasis-open.org/office/v1.2/os/OpenDocument-v1.2-os-part1.html#__RefHeading__1419264_253892949
+        class Span
+          include Node
+          include Text
+
+          def self.style_family
+            'text'
+          end
+
+          def open
+            @enclosing_style.join
+          end
+
+          def close
+            @enclosing_style.reverse.join
+          end
+        end
+
+        # http://docs.oasis-open.org/office/v1.2/os/OpenDocument-v1.2-os-part1.html#__RefHeading__1415154_253892949
+        class ListItem
+          include Node
+          include Text
+
+          not_enclosing 'p'
+
+          def open
+            '* '
+          end
+
+          def close
+            "\n"
+          end
+        end
+      end
+    end
+  end
+end

data/lib/doc2text/odt.rb

@@ -0,0 +1,73 @@
+require 'zip'
+
+module Doc2Text
+  module Odt
+    class Document
+      EXTRACT_EXTENSION = 'unpacked_odt'
+
+      def self.parse_and_save(input, output_filename)
+        odt = new input
+        begin
+          odt.unpack
+          output = File.open output_filename, 'w'
+          markdown = Markdown::Document.new output
+          begin
+            odt.parse markdown
+          ensure
+            markdown.close
+          end
+        ensure
+          odt.clean
+        end
+      end
+
+      def parse(markdown)
+        content = ::Doc2Text::Odt::Content::Document.new markdown
+        parser = Nokogiri::XML::SAX::Parser.new(content) # { |config| config.strict}
+        parser.parse open 'content.xml'
+      end
+
+      def initialize(document_path)
+        @document_path = document_path
+      end
+
+      def unpack
+        Zip::File.open(@document_path) {
+            |zip_file|
+          Dir.mkdir(extract_path)
+          zip_file.each do |entry|
+            zipped_file_extract_path = File.join extract_path, entry.name
+            FileUtils.mkdir_p File.dirname(zipped_file_extract_path)
+            zip_file.extract entry, zipped_file_extract_path
+          end
+        }
+      end
+
+      def clean
+        if [extract_path, File.join(extract_path, 'content.xml'), File.join(extract_path, 'mimetype')].all? { |file| File.exist?(file) }
+          FileUtils.rm_r extract_path
+        end
+      end
+
+      # Open file from the current odt
+      def open(filename)
+        File.open File.join(extract_path, filename), 'r'
+      end
+
+      # Parse xml file from the current odt
+      def xml_file(filename, rood_node_name)
+        Nokogiri::XML::Document.parse(open(filename)) { |config| config.strict }
+        root_node = doc.root
+        if root_node.name != rood_node_name or root_node.namespace.prefix != 'office'
+          raise XmlError, 'Document does not have correct root element'
+        else
+          open(filename)
+        end
+      end
+
+      def extract_path
+        File.join File.dirname(@document_path), ".#{File.basename(@document_path)}_#{EXTRACT_EXTENSION}"
+      end
+    end
+  end
+end

data/lib/doc2text/odt_xml_node.rb

@@ -0,0 +1,98 @@
+module Doc2Text
+  module Odt
+    module XmlNodes
+      module Node
+        attr_reader :parent, :children, :attrs, :prefix, :name
+
+        def self.create_node(prefix, name, parent = nil, attrs = [], markdown_document = nil)
+          begin
+            clazz = XmlNodes.const_get "#{titleize prefix}::#{titleize name}"
+          rescue NameError => e
+            Generic.new(parent, attrs, prefix, name, markdown_document)
+          else
+            clazz.new(parent, attrs, prefix, name, markdown_document)
+          end
+        end
+
+        def self.titleize(tag)
+          tag.split('-').map(&:capitalize).join
+        end
+
+        def initialize(parent = nil, attrs = [], prefix = nil, name = nil, markdown_document = nil)
+          @parent, @attrs, @prefix, @name = parent, attrs, prefix, name
+          @children = []
+          @has_text = false
+        end
+
+        def root?
+          !@parent
+        end
+
+        def has_text?
+          @has_text
+        end
+
+        def open
+          ''
+        end
+
+        def close
+          ''
+        end
+
+        def <<(child)
+          @children << child
+        end
+
+        def delete_on_close?
+          true
+        end
+
+        def eql?(object)
+          return false unless object.is_a? Node
+          object.xml_name == xml_name
+        end
+
+        def generic?
+          instance_of? Node
+        end
+
+        def remove_last_child!(child)
+          unless child === @children.last
+            # TODO remove this redundant(tree build algorithm) checks
+            raise Doc2Text::XmlError, "!The child #{child} IS NOT among the children of #{self}"
+          else
+            @children.pop
+          end
+        end
+
+        def xml_name
+          "#{@prefix}:#{@name}"
+        end
+
+        def to_s
+          "#{xml_name} : #{attrs}"
+        end
+
+        def not_enclosing?
+          !root? && parent.class.not_enclosing_tags && parent.class.not_enclosing_tags.find do |tag|
+            @prefix == parent.prefix && @name == tag
+          end
+        end
+
+        def self.included(base)
+          base.extend ClassMethods
+        end
+
+        module ClassMethods
+          attr_reader :not_enclosing_tags
+
+          def not_enclosing(tag)
+            @not_enclosing_tags ||= []
+            @not_enclosing_tags << tag
+          end
+        end
+      end
+    end
+  end
+end

metadata

@@ -0,0 +1,51 @@
+--- !ruby/object:Gem::Specification
+name: doc2text
+version: !ruby/object:Gem::Version
+  version: '0.1'
+platform: ruby
+authors:
+- Valentin Aitken
+autorequire: 
+bindir: bin
+cert_chain: []
+date: 2014-07-12 00:00:00.000000000 Z
+dependencies: []
+description: Parses odt to markdown
+email: bostko@gmail.com
+executables: []
+extensions: []
+extra_rdoc_files: []
+files:
+- lib/doc2text.rb
+- lib/doc2text/content.rb
+- lib/doc2text/errors.rb
+- lib/doc2text/markdown.rb
+- lib/doc2text/namespaces.rb
+- lib/doc2text/odt.rb
+- lib/doc2text/odt_xml_node.rb
+homepage: https://github.com/bostko/doc2text
+licenses:
+- GPL
+metadata: {}
+post_install_message: 
+rdoc_options: []
+require_paths:
+- lib
+required_ruby_version: !ruby/object:Gem::Requirement
+  requirements:
+  - - ">="
+    - !ruby/object:Gem::Version
+      version: '0'
+required_rubygems_version: !ruby/object:Gem::Requirement
+  requirements:
+  - - ">="
+    - !ruby/object:Gem::Version
+      version: '0'
+requirements: []
+rubyforge_project: 
+rubygems_version: 2.3.0
+signing_key: 
+specification_version: 4
+summary: Translates odt to markdown
+test_files: []
+has_rdoc: