def get_body(doc):
[ elem.drop_tree() for elem in doc.xpath('.//script | .//link | .//style') ]
raw_html = unicode(tostring(doc.body or doc))
return clean_attributes(raw_html)
def sanitize(self, node, candidates):
MIN_LEN = self.options.get('min_text_length', self.TEXT_LENGTH_THRESHOLD)
for header in self.tags(node, "h1", "h2", "h3", "h4", "h5", "h6"):
if self.class_weight(header) < 0 or self.get_link_density(header) > 0.33:
header.drop_tree()
for elem in self.tags(node, "form", "iframe", "textarea"):
elem.drop_tree()
allowed = {}
# Conditionally clean <table>s, <ul>s, and <div>s
for el in self.reverse_tags(node, "table", "ul", "div"):
if el in allowed or el in self.keep_elements:
continue
weight = self.class_weight(el)
if el in candidates:
content_score = candidates[el]['content_score']
# print '!',el, '-> %6.3f' % content_score
else:
content_score = 0
tag = el.tag
if weight + content_score < 0:
self.debug("Cleaned %s with score %6.3f and weight %-3s" %
(describe(el), content_score, weight, ))
el.drop_tree()
elif el.text_content().count(",") < 10:
counts = {}
for kind in ['p', 'img', 'li', 'a', 'embed', 'input']:
counts[kind] = len(el.findall('.//%s' %kind))
counts["li"] -= 100
content_length = text_length(el) # Count the text length excluding any surrounding whitespace
link_density = self.get_link_density(el)
parent_node = el.getparent()
if parent_node is not None:
if parent_node in candidates:
content_score = candidates[parent_node]['content_score']
else:
content_score = 0
# if parent_node is not None:
# pweight = self.class_weight(parent_node) + content_score
# pname = describe(parent_node)
# else:
# pweight = 0
# pname = "no parent"
to_remove = False
reason = ""
# if el.tag == 'div' and counts["img"] >= 1:
# continue
if counts["p"] and counts["img"] > counts["p"]:
reason = "too many images (%s)" % counts["img"]
to_remove = True
elif counts["li"] > counts["p"] and tag != "ul" and tag != "ol":
reason = "more <li>s than <p>s"
to_remove = True
elif counts["input"] > (counts["p"] / 3):
reason = "less than 3x <p>s than <input>s"
to_remove = True
elif content_length < (MIN_LEN) and (counts["img"] == 0 or counts["img"] > 2):
reason = "too short content length %s without a single image" % content_length
to_remove = True
elif weight < 25 and link_density > 0.2:
reason = "too many links %.3f for its weight %s" % (link_density, weight)
to_remove = True
elif weight >= 25 and link_density > 0.5:
reason = "too many links %.3f for its weight %s" % (link_density, weight)
to_remove = True
elif (counts["embed"] == 1 and content_length < 75) or counts["embed"] > 1:
reason = "<embed>s with too short content length, or too many <embed>s"
to_remove = True
# if el.tag == 'div' and counts['img'] >= 1 and to_remove:
# imgs = el.findall('.//img')
# valid_img = False
# self.debug(tounicode(el))
# for img in imgs:
#
# height = img.get('height')
# text_length = img.get('text_length')
# self.debug ("height %s text_length %s" %(repr(height), repr(text_length)))
# if to_int(height) >= 100 or to_int(text_length) >= 100:
# valid_img = True
# self.debug("valid image" + tounicode(img))
# break
# if valid_img:
# to_remove = False
# self.debug("Allowing %s" %el.text_content())
# for desnode in self.tags(el, "table", "ul", "div"):
# allowed[desnode] = True
# find x non empty preceding and succeeding siblings
i, j = 0, 0
x = 1
siblings = []
for sib in el.itersiblings():
# self.debug(sib.text_content())
sib_content_length = text_length(sib)
if sib_content_length:
i =+ 1
siblings.append(sib_content_length)
if i == x:
break
for sib in el.itersiblings(preceding=True):
# self.debug(sib.text_content())
sib_content_length = text_length(sib)
if sib_content_length:
j =+ 1
siblings.append(sib_content_length)
if j == x:
break
# self.debug(str(siblings))
if siblings and sum(siblings) > 1000 :
to_remove = False
self.debug("Allowing %s" % describe(el))
for desnode in self.tags(el, "table", "ul", "div"):
allowed[desnode] = True
if to_remove:
self.debug("Cleaned %6.3f %s with weight %s cause it has %s." %
(content_score, describe(el), weight, reason))
# print tounicode(el)
# self.debug("pname %s pweight %.3f" %(pname, pweight))
el.drop_tree()
return clean_attributes(tounicode(node))
def sanitize(self, node, candidates):
MIN_LEN = self.options.get('min_text_length', self.TEXT_LENGTH_THRESHOLD)
for header in self.tags(node, "h1", "h2", "h3", "h4", "h5", "h6"):
if self.class_weight(header) < 0 or self.get_link_density(header) > 0.33:
header.drop_tree()
for elem in self.tags(node, "form", "iframe", "textarea"):
elem.drop_tree()
allowed = {}
# Conditionally clean <table>s, <ul>s, and <div>s
for el in self.reverse_tags(node, "table", "ul", "div"):
if el in allowed or el in self.keep_elements:
continue
weight = self.class_weight(el)
if el in candidates:
content_score = candidates[el]['content_score']
#print '!',el, '-> %6.3f' % content_score
else:
content_score = 0
tag = el.tag
if weight + content_score < 0:
self.debug("Cleaned %s with score %6.3f and weight %-3s" %
(describe(el), content_score, weight, ))
el.drop_tree()
elif el.text_content().count(",") < 10:
counts = {}
for kind in ['p', 'img', 'li', 'a', 'embed', 'input']:
counts[kind] = len(el.findall('.//%s' %kind))
counts["li"] -= 100
content_length = text_length(el) # Count the text length excluding any surrounding whitespace
link_density = self.get_link_density(el)
parent_node = el.getparent()
if parent_node is not None:
if parent_node in candidates:
content_score = candidates[parent_node]['content_score']
else:
content_score = 0
#if parent_node is not None:
#pweight = self.class_weight(parent_node) + content_score
#pname = describe(parent_node)
#else:
#pweight = 0
#pname = "no parent"
to_remove = False
reason = ""
#if el.tag == 'div' and counts["img"] >= 1:
# continue
if counts["p"] and counts["img"] > counts["p"]:
reason = "too many images (%s)" % counts["img"]
to_remove = True
elif counts["li"] > counts["p"] and tag != "ul" and tag != "ol":
reason = "more <li>s than <p>s"
to_remove = True
elif counts["input"] > (counts["p"] / 3):
reason = "less than 3x <p>s than <input>s"
to_remove = True
elif content_length < (MIN_LEN) and (counts["img"] == 0 or counts["img"] > 2):
reason = "too short content length %s without a single image" % content_length
to_remove = True
elif weight < 25 and link_density > 0.2:
reason = "too many links %.3f for its weight %s" % (link_density, weight)
to_remove = True
elif weight >= 25 and link_density > 0.5:
reason = "too many links %.3f for its weight %s" % (link_density, weight)
to_remove = True
elif (counts["embed"] == 1 and content_length < 75) or counts["embed"] > 1:
reason = "<embed>s with too short content length, or too many <embed>s"
to_remove = True
# if el.tag == 'div' and counts['img'] >= 1 and to_remove:
# imgs = el.findall('.//img')
# valid_img = False
# self.debug(tounicode(el))
# for img in imgs:
#
# height = img.get('height')
# text_length = img.get('text_length')
# self.debug ("height %s text_length %s" %(repr(height), repr(text_length)))
# if to_int(height) >= 100 or to_int(text_length) >= 100:
# valid_img = True
# self.debug("valid image" + tounicode(img))
# break
# if valid_img:
# to_remove = False
# self.debug("Allowing %s" %el.text_content())
# for desnode in self.tags(el, "table", "ul", "div"):
# allowed[desnode] = True
#find x non empty preceding and succeeding siblings
i, j = 0, 0
x = 1
siblings = []
for sib in el.itersiblings():
#self.debug(sib.text_content())
sib_content_length = text_length(sib)
if sib_content_length:
i =+ 1
siblings.append(sib_content_length)
if i == x:
break
for sib in el.itersiblings(preceding=True):
#self.debug(sib.text_content())
sib_content_length = text_length(sib)
if sib_content_length:
j =+ 1
siblings.append(sib_content_length)
if j == x:
break
#self.debug(str(siblings))
if siblings and sum(siblings) > 1000 :
to_remove = False
self.debug("Allowing %s" % describe(el))
for desnode in self.tags(el, "table", "ul", "div"):
allowed[desnode] = True
if to_remove:
self.debug("Cleaned %6.3f %s with weight %s cause it has %s." %
(content_score, describe(el), weight, reason))
#print tounicode(el)
#self.debug("pname %s pweight %.3f" %(pname, pweight))
el.drop_tree()
return clean_attributes(tounicode(node))
