Sha256: 39be3254082d7adbfcf8845cfa5cd2fb5fd269099f101a4aa3434e3d94b742ca
Contents?: true
Size: 1.21 KB
Versions: 6
Compression:
Stored size: 1.21 KB
Contents
require 'nokogiri'
class HtmlToTextConverter
def initialize(html_string)
@html_string = html_string
end
def parse
return '' unless @html_string
html = pre_parse
@fragment = Nokogiri::HTML.fragment(html)
add_newline_after_block_elements
parse_whitespace
parse_lists
@fragment.content
end
private
def pre_parse
html = @html_string.gsub(/>\s+</, '><') # remove whitespace between tags
html.gsub(' ', ' ') # replace nbsp with regular space
end
def add_newline_after_block_elements
@fragment.css('br, p, div, ul, ol').each { |node| node.after("\n") }
end
def parse_whitespace
@fragment.xpath('.//text()').each do |node|
if node.content =~ /\S/ # has non-whitespace characters
node.content = node.content.squeeze(' ') # consolidate whitespace
node.content = node.content.lstrip
else
# remove nodes that are only whitespace
node.remove unless node.content.include?("\n")
end
end
end
def parse_lists
# Make lists pretty:
# * list item 1
# * list item 2
# * list item 3
@fragment.css('li').each do |node|
node.content = " * #{node.content}\n"
end
end
end
Version data entries
6 entries across 6 versions & 1 rubygems