open_office.rb

# frozen_string_literal: true

require 'date'
require 'nokogiri'
require 'cgi'
require 'zip/filesystem'
require 'roo/font'
require 'roo/tempdir'
require 'base64'
require 'openssl'

module Roo
  class OpenOffice < Roo::Base
    extend Roo::Tempdir

    ERROR_MISSING_CONTENT_XML = 'file missing required content.xml'
    XPATH_FIND_TABLE_STYLES   = "//*[local-name()='automatic-styles']"
    XPATH_LOCAL_NAME_TABLE    = "//*[local-name()='table']"

    # initialization and opening of a spreadsheet file
    # values for packed: :zip
    def initialize(filename, options = {})
      packed       = options[:packed]
      file_warning = options[:file_warning] || :error

      @only_visible_sheets = options[:only_visible_sheets]
      file_type_check(filename, '.ods', 'an Roo::OpenOffice', file_warning, packed)
      # NOTE: Create temp directory and allow Ruby to cleanup the temp directory
      #       when the object is garbage collected. Initially, the finalizer was
      #       created in the Roo::Tempdir module, but that led to a segfault
      #       when testing in Ruby 2.4.0.
      @tmpdir = self.class.make_tempdir(self, find_basename(filename), options[:tmpdir_root])
      ObjectSpace.define_finalizer(self, self.class.finalize(object_id))
      @filename = local_filename(filename, @tmpdir, packed)
      # TODO: @cells_read[:default] = false
      open_oo_file(options)
      super(filename, options)
      initialize_default_variables

      unless @table_display.any?
        doc.xpath(XPATH_FIND_TABLE_STYLES).each do |style|
          read_table_styles(style)
        end
      end

      @sheet_names = doc.xpath(XPATH_LOCAL_NAME_TABLE).map do |sheet|
        if !@only_visible_sheets || @table_display[attribute(sheet, 'style-name')]
          sheet.attributes['name'].value
        end
      end.compact
    rescue
      self.class.finalize_tempdirs(object_id)
      raise
    end

    def open_oo_file(options)
      Zip::File.open(@filename) do |zip_file|
        content_entry = zip_file.glob('content.xml').first
        fail ArgumentError, ERROR_MISSING_CONTENT_XML unless content_entry

        roo_content_xml_path = ::File.join(@tmpdir, 'roo_content.xml')
        content_entry.extract(roo_content_xml_path)
        decrypt_if_necessary(zip_file, content_entry, roo_content_xml_path, options)
      end
    end

    def initialize_default_variables
      @formula                = {}
      @style                  = {}
      @style_defaults         = Hash.new { |h, k| h[k] = [] }
      @table_display          = Hash.new { |h, k| h[k] = true }
      @font_style_definitions = {}
      @comment                = {}
      @comments_read          = {}
    end

    def method_missing(m, *args)
      read_labels
      # is method name a label name
      if @label.key?(m.to_s)
        row, col = label(m.to_s)
        cell(row, col)
      else
        # call super for methods like #a1
        super
      end
    end

    # Returns the content of a spreadsheet-cell.
    # (1,1) is the upper left corner.
    # (1,1), (1,'A'), ('A',1), ('a',1) all refers to the
    # cell at the first line and first row.
    def cell(row, col, sheet = nil)
      sheet ||= default_sheet
      read_cells(sheet)
      row, col = normalize(row, col)
      if celltype(row, col, sheet) == :date
        yyyy, mm, dd = @cell[sheet][[row, col]].to_s.split('-')
        return Date.new(yyyy.to_i, mm.to_i, dd.to_i)
      end

      @cell[sheet][[row, col]]
    end

    # Returns the formula at (row,col).
    # Returns nil if there is no formula.
    # The method #formula? checks if there is a formula.
    def formula(row, col, sheet = nil)
      sheet ||= default_sheet
      read_cells(sheet)
      row, col = normalize(row, col)
      @formula[sheet][[row, col]]
    end

    # Predicate methods really should return a boolean
    # value. Hopefully no one was relying on the fact that this
    # previously returned either nil/formula
    def formula?(*args)
      !!formula(*args)
    end

    # returns each formula in the selected sheet as an array of elements
    # [row, col, formula]
    def formulas(sheet = nil)
      sheet ||= default_sheet
      read_cells(sheet)
      return [] unless @formula[sheet]
      @formula[sheet].each.collect do |elem|
        [elem[0][0], elem[0][1], elem[1]]
      end
    end

    # Given a cell, return the cell's style
    def font(row, col, sheet = nil)
      sheet ||= default_sheet
      read_cells(sheet)
      row, col   = normalize(row, col)
      style_name = @style[sheet][[row, col]] || @style_defaults[sheet][col - 1] || 'Default'
      @font_style_definitions[style_name]
    end

    # returns the type of a cell:
    # * :float
    # * :string
    # * :date
    # * :percentage
    # * :formula
    # * :time
    # * :datetime
    def celltype(row, col, sheet = nil)
      sheet ||= default_sheet
      read_cells(sheet)
      row, col = normalize(row, col)
      @formula[sheet][[row, col]] ? :formula : @cell_type[sheet][[row, col]]
    end

    def sheets
      @sheet_names
    end

    # version of the Roo::OpenOffice document
    # at 2007 this is always "1.0"
    def officeversion
      oo_version
      @officeversion
    end

    # shows the internal representation of all cells
    # mainly for debugging purposes
    def to_s(sheet = nil)
      sheet ||= default_sheet
      read_cells(sheet)
      @cell[sheet].inspect
    end

    # returns the row,col values of the labelled cell
    # (nil,nil) if label is not defined
    def label(labelname)
      read_labels
      return [nil, nil, nil] if @label.size < 1 || !@label.key?(labelname)
      [
        @label[labelname][1].to_i,
        ::Roo::Utils.letter_to_number(@label[labelname][2]),
        @label[labelname][0]
      ]
    end

    # Returns an array which all labels. Each element is an array with
    # [labelname, [row,col,sheetname]]
    def labels(_sheet = nil)
      read_labels
      @label.map do |label|
        [label[0], # name
         [label[1][1].to_i, # row
          ::Roo::Utils.letter_to_number(label[1][2]), # column
          label[1][0], # sheet
         ]]
      end
    end

    # returns the comment at (row/col)
    # nil if there is no comment
    def comment(row, col, sheet = nil)
      sheet ||= default_sheet
      read_cells(sheet)
      row, col = normalize(row, col)
      return nil unless @comment[sheet]
      @comment[sheet][[row, col]]
    end

    # returns each comment in the selected sheet as an array of elements
    # [row, col, comment]
    def comments(sheet = nil)
      sheet ||= default_sheet
      read_comments(sheet) unless @comments_read[sheet]
      return [] unless @comment[sheet]
      @comment[sheet].each.collect do |elem|
        [elem[0][0], elem[0][1], elem[1]]
      end
    end

    private

    # If the ODS file has an encryption-data element, then try to decrypt.
    # If successful, the temporary content.xml will be overwritten with
    # decrypted contents.
    def decrypt_if_necessary(
      zip_file,
      content_entry,
      roo_content_xml_path, options
    )
      # Check if content.xml is encrypted by extracting manifest.xml
      # and searching for a manifest:encryption-data element

      if (manifest_entry = zip_file.glob('META-INF/manifest.xml').first)
        roo_manifest_xml_path = File.join(@tmpdir, 'roo_manifest.xml')
        manifest_entry.extract(roo_manifest_xml_path)
        manifest        = ::Roo::Utils.load_xml(roo_manifest_xml_path)

        # XPath search for manifest:encryption-data only for the content.xml
        # file

        encryption_data = manifest.xpath(
          "//manifest:file-entry[@manifest:full-path='content.xml']"\
        "/manifest:encryption-data"
        ).first

        # If XPath returns a node, then we know content.xml is encrypted

        unless encryption_data.nil?

          # Since we know it's encrypted, we check for the password option
          # and if it doesn't exist, raise an argument error

          password = options[:password]
          if !password.nil?
            perform_decryption(
              encryption_data,
              password,
              content_entry,
              roo_content_xml_path
            )
          else
            fail ArgumentError, 'file is encrypted but password was not supplied'
          end
        end
      else
        fail ArgumentError, 'file missing required META-INF/manifest.xml'
      end
    end

    # Process the ODS encryption manifest and perform the decryption
    def perform_decryption(
      encryption_data,
      password,
      content_entry,
      roo_content_xml_path
    )
      # Extract various expected attributes from the manifest that
      # describe the encryption

      algorithm_node            = encryption_data.xpath('manifest:algorithm').first
      key_derivation_node       =
        encryption_data.xpath('manifest:key-derivation').first
      start_key_generation_node =
        encryption_data.xpath('manifest:start-key-generation').first

      # If we have all the expected elements, then we can perform
      # the decryption.

      if !algorithm_node.nil? && !key_derivation_node.nil? &&
         !start_key_generation_node.nil?

        # The algorithm is a URI describing the algorithm used
        algorithm           = algorithm_node['manifest:algorithm-name']

        # The initialization vector is base-64 encoded
        iv                  = Base64.decode64(
          algorithm_node['manifest:initialisation-vector']
        )
        key_derivation_name = key_derivation_node['manifest:key-derivation-name']
        iteration_count     = key_derivation_node['manifest:iteration-count'].to_i
        salt                = Base64.decode64(key_derivation_node['manifest:salt'])

        # The key is hashed with an algorithm represented by this URI
        key_generation_name =
          start_key_generation_node[
            'manifest:start-key-generation-name'
          ]

        hashed_password = password

        if key_generation_name == 'http://www.w3.org/2000/09/xmldsig#sha256'

          hashed_password = Digest::SHA256.digest(password)
        else
          fail ArgumentError, "Unknown key generation algorithm #{key_generation_name}"
        end

        cipher = find_cipher(
          algorithm,
          key_derivation_name,
          hashed_password,
          salt,
          iteration_count,
          iv
        )

        begin
          decrypted = decrypt(content_entry, cipher)

          # Finally, inflate the decrypted stream and overwrite
          # content.xml
          IO.binwrite(
            roo_content_xml_path,
            Zlib::Inflate.new(-Zlib::MAX_WBITS).inflate(decrypted)
          )
        rescue StandardError => error
          raise ArgumentError, "Invalid password or other data error: #{error}"
        end
      else
        fail ArgumentError, 'manifest.xml missing encryption-data elements'
      end
    end

    # Create a cipher based on an ODS algorithm URI from manifest.xml
    # params: algorithm, key_derivation_name, hashed_password, salt, iteration_count, iv
    def find_cipher(*args)
      fail ArgumentError, 'Unknown algorithm ' + algorithm unless args[0] == 'http://www.w3.org/2001/04/xmlenc#aes256-cbc'

      cipher = ::OpenSSL::Cipher.new('AES-256-CBC')
      cipher.decrypt
      cipher.padding = 0
      cipher.key     = find_cipher_key(cipher, *args[1..4])
      cipher.iv      = args[5]

      cipher
    end

    # Create a cipher key based on an ODS algorithm string from manifest.xml
    def find_cipher_key(*args)
      fail ArgumentError, 'Unknown key derivation name ', args[1] unless args[1] == 'PBKDF2'

      ::OpenSSL::PKCS5.pbkdf2_hmac_sha1(args[2], args[3], args[4], args[0].key_len)
    end

    # Block decrypt raw bytes from the zip file based on the cipher
    def decrypt(content_entry, cipher)
      # Zip::Entry.extract writes a 0-length file when trying
      # to extract an encrypted stream, so we read the
      # raw bytes based on the offset and lengths
      decrypted = ''
      File.open(@filename, 'rb') do |zipfile|
        zipfile.seek(
          content_entry.local_header_offset +
            content_entry.calculate_local_header_size
        )
        total_to_read = content_entry.compressed_size

        block_size = 4096
        block_size = total_to_read if block_size > total_to_read

        while (buffer = zipfile.read(block_size))
          decrypted     += cipher.update(buffer)
          total_to_read -= buffer.length

          break if total_to_read == 0

          block_size = total_to_read if block_size > total_to_read
        end
      end

      decrypted + cipher.final
    end

    def doc
      @doc ||= ::Roo::Utils.load_xml(File.join(@tmpdir, 'roo_content.xml'))
    end

    # read the version of the OO-Version
    def oo_version
      doc.xpath("//*[local-name()='document-content']").each do |office|
        @officeversion = attribute(office, 'version')
      end
    end

    # helper function to set the internal representation of cells
    def set_cell_values(sheet, x, y, i, v, value_type, formula, table_cell, str_v, style_name)
      key = [y, x + i]
      @cell_type[sheet] ||= {}
      @cell_type[sheet][key] = value_type.to_sym if value_type
      @formula[sheet] ||= {}
      if formula
        ['of:', 'oooc:'].each do |prefix|
          if formula[0, prefix.length] == prefix
            formula = formula[prefix.length..-1]
          end
        end
        @formula[sheet][key] = formula
      end
      @cell[sheet] ||= {}
      @style[sheet] ||= {}
      @style[sheet][key] = style_name
      case @cell_type[sheet][key]
      when :float
        value = (table_cell.attributes['value'].to_s.include?(".") || table_cell.children.first.text.include?(".")) ? v.to_f : v.to_i
        value = 'true' if formula == '=TRUE()'
        value = 'false' if formula == '=FALSE()'
        @cell[sheet][key] = value
      when :percentage
        @cell[sheet][key] = v.to_f
      when :string
        @cell[sheet][key] = str_v
      when :date
        # TODO: if table_cell.attributes['date-value'].size != "XXXX-XX-XX".size
        if attribute(table_cell, 'date-value').size != 'XXXX-XX-XX'.size
          #-- dann ist noch eine Uhrzeit vorhanden
          #-- "1961-11-21T12:17:18"
          @cell[sheet][key]      = DateTime.parse(attribute(table_cell, 'date-value').to_s)
          @cell_type[sheet][key] = :datetime
        else
          @cell[sheet][key] = table_cell.attributes['date-value']
        end
      when :time
        hms = v.split(':')
        @cell[sheet][key] = hms[0].to_i * 3600 + hms[1].to_i * 60 + hms[2].to_i
      else
        @cell[sheet][key] = v
      end
    end

    # read all cells in the selected sheet
    #--
    # the following construct means '4 blanks'
    # some content <text:s text:c="3"/>
    #++
    def read_cells(sheet = default_sheet)
      validate_sheet!(sheet)
      return if @cells_read[sheet]

      sheet_found = false
      doc.xpath("//*[local-name()='table']").each do |ws|
        next unless sheet == attribute(ws, 'name')

        sheet_found = true
        col         = 1
        row         = 1
        ws.children.each do |table_element|
          case table_element.name
          when 'table-column'
            @style_defaults[sheet] << table_element.attributes['default-cell-style-name']
          when 'table-row'
            if table_element.attributes['number-rows-repeated']
              skip_row = attribute(table_element, 'number-rows-repeated').to_s.to_i
              row      = row + skip_row - 1
            end
            table_element.children.each do |cell|
              skip_col   = attribute(cell, 'number-columns-repeated')
              formula    = attribute(cell, 'formula')
              value_type = attribute(cell, 'value-type')
              v          = attribute(cell, 'value')
              style_name = attribute(cell, 'style-name')
              case value_type
              when 'string'
                str_v      = ''
                # insert \n if there is more than one paragraph
                para_count = 0
                cell.children.each do |str|
                  # begin comments
                  #=begin
                  #- <table:table-cell office:value-type="string">
                  #  - <office:annotation office:display="true" draw:style-name="gr1" draw:text-style-name="P1" svg:width="1.1413in" svg:height="0.3902in" svg:x="2.0142in" svg:y="0in" draw:caption-point-x="-0.2402in" draw:caption-point-y="0.5661in">
                  #      <dc:date>2011-09-20T00:00:00</dc:date>
                  #      <text:p text:style-name="P1">Kommentar fuer B4</text:p>
                  #    </office:annotation>
                  #    <text:p>B4 (mit Kommentar)</text:p>
                  #  </table:table-cell>
                  #=end
                  if str.name == 'annotation'
                    str.children.each do |annotation|
                      next unless annotation.name == 'p'
                      # @comment ist ein Hash mit Sheet als Key (wie bei @cell)
                      # innerhalb eines Elements besteht ein Eintrag aus einem
                      # weiteren Hash mit Key [row,col] und dem eigentlichen
                      # Kommentartext als Inhalt
                      @comment[sheet]      = Hash.new unless @comment[sheet]
                      key                  = [row, col]
                      @comment[sheet][key] = annotation.text
                    end
                  end
                  # end comments
                  if str.name == 'p'
                    v          = str.content
                    str_v      += "\n" if para_count > 0
                    para_count += 1
                    if str.children.size > 1
                      str_v += children_to_string(str.children)
                    else
                      str.children.each do |child|
                        str_v += child.content #.text
                      end
                    end
                    str_v = str_v.gsub(/&apos;/, "'") # special case not supported by unescapeHTML
                    str_v = CGI.unescapeHTML(str_v)
                  end # == 'p'
                end
              when 'time'
                cell.children.each do |str|
                  v = str.content if str.name == 'p'
                end
              when '', nil, 'date', 'percentage', 'float'
                #
              when 'boolean'
                v = attribute(cell, 'boolean-value').to_s
              end
              if skip_col
                if !v.nil? || cell.attributes['date-value']
                  0.upto(skip_col.to_i - 1) do |i|
                    set_cell_values(sheet, col, row, i, v, value_type, formula, cell, str_v, style_name)
                  end
                end
                col += (skip_col.to_i - 1)
              end # if skip
              set_cell_values(sheet, col, row, 0, v, value_type, formula, cell, str_v, style_name)
              col += 1
            end
            row += 1
            col = 1
          end
        end
      end
      doc.xpath("//*[local-name()='automatic-styles']").each do |style|
        read_styles(style)
      end

      fail RangeError unless sheet_found

      @cells_read[sheet]    = true
      @comments_read[sheet] = true
    end

    # Only calls read_cells because Roo::Base calls read_comments
    # whereas the reading of comments is done in read_cells for Roo::OpenOffice-objects
    def read_comments(sheet = nil)
      read_cells(sheet)
    end

    def read_labels
      @label ||= doc.xpath('//table:named-range').each_with_object({}) do |ne, hash|
        #-
        # $Sheet1.$C$5
        #+
        name              = attribute(ne, 'name').to_s
        sheetname, coords = attribute(ne, 'cell-range-address').to_s.split('.$')
        col, row          = coords.split('$')
        sheetname         = sheetname[1..-1] if sheetname[0, 1] == '$'
        hash[name] = [sheetname, row, col]
      end
    end

    def read_styles(style_elements)
      @font_style_definitions['Default'] = Roo::Font.new
      style_elements.each do |style|
        next unless style.name == 'style'
        style_name = attribute(style, 'name')
        style.each do |properties|
          font                                = Roo::OpenOffice::Font.new
          font.bold                           = attribute(properties, 'font-weight')
          font.italic                         = attribute(properties, 'font-style')
          font.underline                      = attribute(properties, 'text-underline-style')
          @font_style_definitions[style_name] = font
        end
      end
    end

    def read_table_styles(styles)
      styles.children.each do |style|
        next unless style.name == 'style'
        style_name = attribute(style, 'name')
        style.children.each do |properties|
          display = attribute(properties, 'display')
          next unless display
          @table_display[style_name] = (display == 'true')
        end
      end
    end

    # helper method to convert compressed spaces and other elements within
    # an text into a string
    # FIXME: add a test for compressed_spaces == 0. It's not currently tested.
    def children_to_string(children)
      children.map do |child|
        if child.text?
          child.content
        else
          if child.name == 's'
            compressed_spaces = child.attributes['c'].to_s.to_i
            # no explicit number means a count of 1:
            compressed_spaces == 0 ? ' ' : ' ' * compressed_spaces
          else
            child.content
          end
        end
      end.join
    end

    def attribute(node, attr_name)
      node.attributes[attr_name].value if node.attributes[attr_name]
    end
  end
end