openerp/tools/html_sanitize.py

   1
   2 import lxml.html
   3 import re
   4
   5 def html_sanitize(x):
   6     if not x:
   7         return x
   8     if type(x) == str:
   9         x = unicode(x, "utf8", "replace")
  10     root = lxml.html.fromstring("<div>%s</div>" % x)
  11     result = handle_element(root)
  12     res = ""
  13     for el in children(result[0]):
  14         if type(el) == str or type(el) == unicode:
  15             res += el
  16         else:
  17             el.tail = ""
  18             res += lxml.html.tostring(el)
  19     return res
  20
  21 to_remove = set(["script", "head", "meta", "title", "link", "img"])
  22 to_unwrap = set(["html", "body"])
  23
  24 javascript_regex = re.compile("""^\s*javascript\s*\:.*$""")
  25 def handle_a(el, new):
  26     href = el.get("href", "#")
  27     if javascript_regex.search(href):
  28         href = "#"
  29     new.set("href", href)
  30 special = {
  31     "a": handle_a,
  32 }
  33
  34 def handle_element(el):
  35     if type(el) == str or type(el) == unicode:
  36         return [el]
  37     if el.tag in to_remove:
  38         return []
  39     if el.tag in to_unwrap:
  40         return reduce(lambda x,y: x+y, [handle_element(x) for x in children(el)])
  41     new = lxml.html.fromstring("<%s />" % el.tag)
  42     for i in children(el):
  43         append_to(handle_element(i), new)
  44     if el.tag in special:
  45         special[el.tag](el, new)
  46     return [new]
  47
  48 def children(el):
  49     res = []
  50     if el.text is not None:
  51         res.append(el.text)
  52     for i in el.getchildren():
  53         res.append(i)
  54         if i.tail is not None:
  55             res.append(i.tail)
  56     return res
  57
  58 def append_to(new_ones, el):
  59     for i in new_ones:
  60         if type(i) == str or type(i) == unicode:
  61             children = el.getchildren()
  62             if len(children) == 0:
  63                 el.text = i
  64             else:
  65                 children[-1].tail = i
  66         else:
  67             el.append(i)