diff --git a/hn.py b/hn.py deleted file mode 100644 index 71dd4c2..0000000 --- a/hn.py +++ /dev/null @@ -1,231 +0,0 @@ -""" -This program is free software: you can redistribute it and/or modify -it under the terms of the GNU General Public License as published by -the Free Software Foundation, either version 3 of the License, or -(at your option) any later version. - -This program is distributed in the hope that it will be useful, -but WITHOUT ANY WARRANTY; without even the implied warranty of -MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the -GNU General Public License for more details. - -You should have received a copy of the GNU General Public License -along with this program. If not, see . -""" - - -from xml.sax.saxutils import escape - -import urllib, re, os, urlparse -import HTMLParser, feedparser -from BeautifulSoup import BeautifulSoup -from pprint import pprint - -import codecs -import sys -streamWriter = codecs.lookup('utf-8')[-1] -sys.stdout = streamWriter(sys.stdout) - - -HN_RSS_FEED = "http://news.ycombinator.com/rss" - -NEGATIVE = re.compile("comment|meta|footer|footnote|foot") -POSITIVE = re.compile("post|hentry|entry|content|text|body|article") -PUNCTUATION = re.compile("""[!"#$%&\'()*+,-./:;<=>?@[\\]^_`{|}~]""") - - -def grabContent(link, html): - - replaceBrs = re.compile("
[ \r\n]*
") - html = re.sub(replaceBrs, "

", html) - - try: - soup = BeautifulSoup(html) - except HTMLParser.HTMLParseError: - return "" - - # REMOVE SCRIPTS - for s in soup.findAll("script"): - s.extract() - - allParagraphs = soup.findAll("p") - topParent = None - - parents = [] - for paragraph in allParagraphs: - - parent = paragraph.parent - - if (parent not in parents): - parents.append(parent) - parent.score = 0 - - if (parent.has_key("class")): - if (NEGATIVE.match(parent["class"])): - parent.score -= 50 - if (POSITIVE.match(parent["class"])): - parent.score += 25 - - if (parent.has_key("id")): - if (NEGATIVE.match(parent["id"])): - parent.score -= 50 - if (POSITIVE.match(parent["id"])): - parent.score += 25 - - if (parent.score == None): - parent.score = 0 - - innerText = paragraph.renderContents() #"".join(paragraph.findAll(text=True)) - if (len(innerText) > 10): - parent.score += 1 - - parent.score += innerText.count(",") - - for parent in parents: - if ((not topParent) or (parent.score > topParent.score)): - topParent = parent - - if (not topParent): - return "" - - # REMOVE LINK'D STYLES - styleLinks = soup.findAll("link", attrs={"type" : "text/css"}) - for s in styleLinks: - s.extract() - - # REMOVE ON PAGE STYLES - for s in soup.findAll("style"): - s.extract() - - # CLEAN STYLES FROM ELEMENTS IN TOP PARENT - for ele in topParent.findAll(True): - del(ele['style']) - del(ele['class']) - - killDivs(topParent) - clean(topParent, "form") - clean(topParent, "object") - clean(topParent, "iframe") - - fixLinks(topParent, link) - - return topParent.renderContents() - - -def fixLinks(parent, link): - tags = parent.findAll(True) - - for t in tags: - if (t.has_key("href")): - t["href"] = urlparse.urljoin(link, t["href"]) - if (t.has_key("src")): - t["src"] = urlparse.urljoin(link, t["src"]) - - -def clean(top, tag, minWords=10000): - tags = top.findAll(tag) - - for t in tags: - if (t.renderContents().count(" ") < minWords): - t.extract() - - -def killDivs(parent): - - divs = parent.findAll("div") - for d in divs: - p = len(d.findAll("p")) - img = len(d.findAll("img")) - li = len(d.findAll("li")) - a = len(d.findAll("a")) - embed = len(d.findAll("embed")) - pre = len(d.findAll("pre")) - code = len(d.findAll("code")) - - if (d.renderContents().count(",") < 10): - if ((pre == 0) and (code == 0)): - if ((img > p ) or (li > p) or (a > p) or (p == 0) or (embed > 0)): - d.extract() - - -def upgradeLink(link): - - link = link.encode('utf-8') - - if (not (link.startswith("http://news.ycombinator.com") or link.endswith(".pdf"))): - linkFile = "upgraded/" + re.sub(PUNCTUATION, "_", link) - if (os.path.exists(linkFile)): - return open(linkFile).read() - else: - content = "" - try: - html = urllib.urlopen(link).read() - content = grabContent(link, html) - filp = open(linkFile, "w") - filp.write(content) - filp.close() - except IOError: - pass - return content - else: - return "" - - - -def upgradeFeed(feedUrl): - - feedData = urllib.urlopen(feedUrl).read() - - upgradedLinks = [] - parsedFeed = feedparser.parse(feedData) - - for entry in parsedFeed.entries: - upgradedLinks.append((entry, upgradeLink(entry.link))) - - rss = """ - - Hacker News - http://news.ycombinator.com/ - Links for the intellectually curious, ranked by readers. - - """ - - for entry, content in upgradedLinks: - rss += u""" - - %s - %s - %s - - Comments
%s
Comments]]> - - -""" % (entry.title, escape(entry.link), escape(entry.comments), entry.comments, content.decode('utf-8'), entry.comments) - - rss += """ - -""" - - - return rss - -if __name__ == "__main__": - print upgradeFeed(HN_RSS_FEED) - - - - - - - - - - - - - - - - - - diff --git a/models/entry.py b/models/entry.py index db460a4..f8ef0dd 100644 --- a/models/entry.py +++ b/models/entry.py @@ -2,15 +2,22 @@ from sqlalchemy import Column, Integer, ForeignKey, String, Text, DateTime from datetime import datetime from time import mktime import urllib2 -#import hn +from readability.readability import Document import html2text import HTMLParser from models import Base def fetch_readability(link): - text = hn.upgradeLink(link) - text = text.decode('utf8') + h2t = html2text.HTML2Text() + h2t.body_width = 0 + h2t.inline_links = False + h2t.ignore_links = True + h2t.ignore_images = True + response = urllib2.urlopen(link) + text = response.read() + text = Document(text).summary() + text = h2t.handle(text) return text def fetch_full_page(link): diff --git a/readability/__init__.py b/readability/__init__.py new file mode 100644 index 0000000..8822a51 --- /dev/null +++ b/readability/__init__.py @@ -0,0 +1 @@ +from .readability import Document diff --git a/readability/cleaners.py b/readability/cleaners.py new file mode 100644 index 0000000..9b158c5 --- /dev/null +++ b/readability/cleaners.py @@ -0,0 +1,32 @@ +# strip out a set of nuisance html attributes that can mess up rendering in RSS feeds +import re +from lxml.html.clean import Cleaner + +bad_attrs = ['width', 'height', 'style', '[-a-z]*color', 'background[-a-z]*', 'on*'] +single_quoted = "'[^']+'" +double_quoted = '"[^"]+"' +non_space = '[^ "\'>]+' +htmlstrip = re.compile("<" # open + "([^>]+) " # prefix + "(?:%s) *" % ('|'.join(bad_attrs),) + # undesirable attributes + '= *(?:%s|%s|%s)' % (non_space, single_quoted, double_quoted) + # value + "([^>]*)" # postfix + ">" # end +, re.I) + +def clean_attributes(html): + while htmlstrip.search(html): + html = htmlstrip.sub('<\\1\\2>', html) + return html + +def normalize_spaces(s): + if not s: return '' + """replace any sequence of whitespace + characters with a single space""" + return ' '.join(s.split()) + +html_cleaner = Cleaner(scripts=True, javascript=True, comments=True, + style=True, links=True, meta=False, add_nofollow=False, + page_structure=False, processing_instructions=True, embedded=False, + frames=False, forms=False, annoying_tags=False, remove_tags=None, + remove_unknown_tags=False, safe_attrs_only=False) diff --git a/readability/debug.py b/readability/debug.py new file mode 100644 index 0000000..a5e644d --- /dev/null +++ b/readability/debug.py @@ -0,0 +1,25 @@ +def save_to_file(text, filename): + f = open(filename, 'wt') + f.write('') + f.write(text.encode('utf-8')) + f.close() + +uids = {} +def describe(node, depth=2): + if not hasattr(node, 'tag'): + return "[%s]" % type(node) + name = node.tag + if node.get('id', ''): name += '#'+node.get('id') + if node.get('class', ''): + name += '.' + node.get('class').replace(' ','.') + if name[:4] in ['div#', 'div.']: + name = name[3:] + if name in ['tr', 'td', 'div', 'p']: + if not node in uids: + uid = uids[node] = len(uids)+1 + else: + uid = uids.get(node) + name += "%02d" % (uid) + if depth and node.getparent() is not None: + return name+' - '+describe(node.getparent(), depth-1) + return name diff --git a/readability/encoding.py b/readability/encoding.py new file mode 100644 index 0000000..d05b7f4 --- /dev/null +++ b/readability/encoding.py @@ -0,0 +1,21 @@ +import re +import chardet + +def get_encoding(page): + text = re.sub(']*>\s*', ' ', page) + enc = 'utf-8' + if not text.strip() or len(text) < 10: + return enc # can't guess + try: + diff = text.decode(enc, 'ignore').encode(enc) + sizes = len(diff), len(text) + if abs(len(text) - len(diff)) < max(sizes) * 0.01: # 99% of utf-8 + return enc + except UnicodeDecodeError: + pass + res = chardet.detect(text) + enc = res['encoding'] + #print '->', enc, "%.2f" % res['confidence'] + if enc == 'MacCyrillic': + enc = 'cp1251' + return enc diff --git a/readability/htmls.py b/readability/htmls.py new file mode 100644 index 0000000..9b59993 --- /dev/null +++ b/readability/htmls.py @@ -0,0 +1,115 @@ +from cleaners import normalize_spaces, clean_attributes +from encoding import get_encoding +from lxml.html import tostring +import logging +import lxml.html +import re + +logging.getLogger().setLevel(logging.DEBUG) + +utf8_parser = lxml.html.HTMLParser(encoding='utf-8') + +def build_doc(page): + if isinstance(page, unicode): + page_unicode = page + else: + enc = get_encoding(page) + page_unicode = page.decode(enc, 'replace') + doc = lxml.html.document_fromstring(page_unicode.encode('utf-8', 'replace'), parser=utf8_parser) + return doc + +def js_re(src, pattern, flags, repl): + return re.compile(pattern, flags).sub(src, repl.replace('$', '\\')) + + +def normalize_entities(cur_title): + entities = { + u'\u2014':'-', + u'\u2013':'-', + u'—': '-', + u'–': '-', + u'\u00A0': ' ', + u'\u00AB': '"', + u'\u00BB': '"', + u'"': '"', + } + for c, r in entities.iteritems(): + if c in cur_title: + cur_title = cur_title.replace(c, r) + + return cur_title + +def norm_title(title): + return normalize_entities(normalize_spaces(title)) + +def get_title(doc): + title = doc.find('.//title') + if title is None or len(title.text) == 0: + return '[no-title]' + + return norm_title(title.text) + +def add_match(collection, text, orig): + text = norm_title(text) + if len(text.split()) >= 2 and len(text) >= 15: + if text.replace('"', '') in orig.replace('"', ''): + collection.add(text) + +def shorten_title(doc): + title = doc.find('.//title') + if title is None or title.text is None or len(title.text) == 0: + return '' + + title = orig = norm_title(title.text) + + candidates = set() + + for item in ['.//h1', './/h2', './/h3']: + for e in list(doc.iterfind(item)): + if e.text: + add_match(candidates, e.text, orig) + if e.text_content(): + add_match(candidates, e.text_content(), orig) + + for item in ['#title', '#head', '#heading', '.pageTitle', '.news_title', '.title', '.head', '.heading', '.contentheading', '.small_header_red']: + for e in doc.cssselect(item): + if e.text: + add_match(candidates, e.text, orig) + if e.text_content(): + add_match(candidates, e.text_content(), orig) + + if candidates: + title = sorted(candidates, key=len)[-1] + else: + for delimiter in [' | ', ' - ', ' :: ', ' / ']: + if delimiter in title: + parts = orig.split(delimiter) + if len(parts[0].split()) >= 4: + title = parts[0] + break + elif len(parts[-1].split()) >= 4: + title = parts[-1] + break + else: + if ': ' in title: + parts = orig.split(': ') + if len(parts[-1].split()) >= 4: + title = parts[-1] + else: + title = orig.split(': ', 1)[1] + + if not 15 < len(title) < 150: + return orig + + return title + +def get_body(doc): + [ elem.drop_tree() for elem in doc.xpath('.//script | .//link | .//style') ] + raw_html = unicode(tostring(doc.body or doc)) + cleaned = clean_attributes(raw_html) + try: + #BeautifulSoup(cleaned) #FIXME do we really need to try loading it? + return cleaned + except Exception: #FIXME find the equivalent lxml error + logging.error("cleansing broke html content: %s\n---------\n%s" % (raw_html, cleaned)) + return raw_html diff --git a/readability/readability.py b/readability/readability.py new file mode 100755 index 0000000..fc37636 --- /dev/null +++ b/readability/readability.py @@ -0,0 +1,593 @@ +#!/usr/bin/env python +import logging +import re +import sys + +from collections import defaultdict +from lxml.etree import tostring +from lxml.etree import tounicode +from lxml.html import document_fromstring +from lxml.html import fragment_fromstring + +from cleaners import clean_attributes +from cleaners import html_cleaner +from htmls import build_doc +from htmls import get_body +from htmls import get_title +from htmls import shorten_title + + +logging.basicConfig(level=logging.INFO) +log = logging.getLogger() + + +REGEXES = { + 'unlikelyCandidatesRe': re.compile('combx|comment|community|disqus|extra|foot|header|menu|remark|rss|shoutbox|sidebar|sponsor|ad-break|agegate|pagination|pager|popup|tweet|twitter', re.I), + 'okMaybeItsACandidateRe': re.compile('and|article|body|column|main|shadow', re.I), + 'positiveRe': re.compile('article|body|content|entry|hentry|main|page|pagination|post|text|blog|story', re.I), + 'negativeRe': re.compile('combx|comment|com-|contact|foot|footer|footnote|masthead|media|meta|outbrain|promo|related|scroll|shoutbox|sidebar|sponsor|shopping|tags|tool|widget', re.I), + 'divToPElementsRe': re.compile('<(a|blockquote|dl|div|img|ol|p|pre|table|ul)', re.I), + #'replaceBrsRe': re.compile('(]*>[ \n\r\t]*){2,}',re.I), + #'replaceFontsRe': re.compile('<(\/?)font[^>]*>',re.I), + #'trimRe': re.compile('^\s+|\s+$/'), + #'normalizeRe': re.compile('\s{2,}/'), + #'killBreaksRe': re.compile('((\s| ?)*){1,}/'), + #'videoRe': re.compile('http:\/\/(www\.)?(youtube|vimeo)\.com', re.I), + #skipFootnoteLink: /^\s*(\[?[a-z0-9]{1,2}\]?|^|edit|citation needed)\s*$/i, +} + + +class Unparseable(ValueError): + pass + + +def describe(node, depth=1): + if not hasattr(node, 'tag'): + return "[%s]" % type(node) + name = node.tag + if node.get('id', ''): + name += '#' + node.get('id') + if node.get('class', ''): + name += '.' + node.get('class').replace(' ', '.') + if name[:4] in ['div#', 'div.']: + name = name[3:] + if depth and node.getparent() is not None: + return name + ' - ' + describe(node.getparent(), depth - 1) + return name + + +def to_int(x): + if not x: + return None + x = x.strip() + if x.endswith('px'): + return int(x[:-2]) + if x.endswith('em'): + return int(x[:-2]) * 12 + return int(x) + + +def clean(text): + text = re.sub('\s*\n\s*', '\n', text) + text = re.sub('[ \t]{2,}', ' ', text) + return text.strip() + + +def text_length(i): + return len(clean(i.text_content() or "")) + + +class Document: + """Class to build a etree document out of html.""" + TEXT_LENGTH_THRESHOLD = 25 + RETRY_LENGTH = 250 + + def __init__(self, input, **options): + """Generate the document + + :param input: string of the html content. + + kwargs: + - attributes: + - debug: output debug messages + - min_text_length: + - retry_length: + - url: will allow adjusting links to be absolute + + """ + self.input = input + self.options = options + self.html = None + + def _html(self, force=False): + if force or self.html is None: + self.html = self._parse(self.input) + return self.html + + def _parse(self, input): + doc = build_doc(input) + doc = html_cleaner.clean_html(doc) + base_href = self.options.get('url', None) + if base_href: + doc.make_links_absolute(base_href, resolve_base_href=True) + else: + doc.resolve_base_href() + return doc + + def content(self): + return get_body(self._html(True)) + + def title(self): + return get_title(self._html(True)) + + def short_title(self): + return shorten_title(self._html(True)) + + def get_clean_html(self): + return clean_attributes(tounicode(self.html)) + + def summary(self, html_partial=False): + """Generate the summary of the html docuemnt + + :param html_partial: return only the div of the document, don't wrap + in html and body tags. + + """ + try: + ruthless = True + while True: + self._html(True) + for i in self.tags(self.html, 'script', 'style'): + i.drop_tree() + for i in self.tags(self.html, 'body'): + i.set('id', 'readabilityBody') + if ruthless: + self.remove_unlikely_candidates() + self.transform_misused_divs_into_paragraphs() + candidates = self.score_paragraphs() + + best_candidate = self.select_best_candidate(candidates) + + if best_candidate: + article = self.get_article(candidates, best_candidate, + html_partial=html_partial) + else: + if ruthless: + log.debug("ruthless removal did not work. ") + ruthless = False + self.debug( + ("ended up stripping too much - " + "going for a safer _parse")) + # try again + continue + else: + log.debug( + ("Ruthless and lenient parsing did not work. " + "Returning raw html")) + article = self.html.find('body') + if article is None: + article = self.html + cleaned_article = self.sanitize(article, candidates) + article_length = len(cleaned_article or '') + retry_length = self.options.get( + 'retry_length', + self.RETRY_LENGTH) + of_acceptable_length = article_length >= retry_length + if ruthless and not of_acceptable_length: + ruthless = False + # Loop through and try again. + continue + else: + return cleaned_article + except StandardError, e: + log.exception('error getting summary: ') + raise Unparseable(str(e)), None, sys.exc_info()[2] + + def get_article(self, candidates, best_candidate, html_partial=False): + # Now that we have the top candidate, look through its siblings for + # content that might also be related. + # Things like preambles, content split by ads that we removed, etc. + sibling_score_threshold = max([ + 10, + best_candidate['content_score'] * 0.2]) + # create a new html document with a html->body->div + if html_partial: + output = fragment_fromstring('

') + else: + output = document_fromstring('
') + best_elem = best_candidate['elem'] + for sibling in best_elem.getparent().getchildren(): + # in lxml there no concept of simple text + # if isinstance(sibling, NavigableString): continue + append = False + if sibling is best_elem: + append = True + sibling_key = sibling # HashableElement(sibling) + if sibling_key in candidates and \ + candidates[sibling_key]['content_score'] >= sibling_score_threshold: + append = True + + if sibling.tag == "p": + link_density = self.get_link_density(sibling) + node_content = sibling.text or "" + node_length = len(node_content) + + if node_length > 80 and link_density < 0.25: + append = True + elif node_length <= 80 \ + and link_density == 0 \ + and re.search('\.( |$)', node_content): + append = True + + if append: + # We don't want to append directly to output, but the div + # in html->body->div + if html_partial: + output.append(sibling) + else: + output.getchildren()[0].getchildren()[0].append(sibling) + #if output is not None: + # output.append(best_elem) + return output + + def select_best_candidate(self, candidates): + sorted_candidates = sorted(candidates.values(), key=lambda x: x['content_score'], reverse=True) + for candidate in sorted_candidates[:5]: + elem = candidate['elem'] + self.debug("Top 5 : %6.3f %s" % ( + candidate['content_score'], + describe(elem))) + + if len(sorted_candidates) == 0: + return None + + best_candidate = sorted_candidates[0] + return best_candidate + + def get_link_density(self, elem): + link_length = 0 + for i in elem.findall(".//a"): + link_length += text_length(i) + #if len(elem.findall(".//div") or elem.findall(".//p")): + # link_length = link_length + total_length = text_length(elem) + return float(link_length) / max(total_length, 1) + + def score_paragraphs(self, ): + MIN_LEN = self.options.get( + 'min_text_length', + self.TEXT_LENGTH_THRESHOLD) + candidates = {} + ordered = [] + for elem in self.tags(self._html(), "p", "pre", "td"): + parent_node = elem.getparent() + if parent_node is None: + continue + grand_parent_node = parent_node.getparent() + + inner_text = clean(elem.text_content() or "") + inner_text_len = len(inner_text) + + # If this paragraph is less than 25 characters + # don't even count it. + if inner_text_len < MIN_LEN: + continue + + if parent_node not in candidates: + candidates[parent_node] = self.score_node(parent_node) + ordered.append(parent_node) + + if grand_parent_node is not None and grand_parent_node not in candidates: + candidates[grand_parent_node] = self.score_node( + grand_parent_node) + ordered.append(grand_parent_node) + + content_score = 1 + content_score += len(inner_text.split(',')) + content_score += min((inner_text_len / 100), 3) + #if elem not in candidates: + # candidates[elem] = self.score_node(elem) + + #WTF? candidates[elem]['content_score'] += content_score + candidates[parent_node]['content_score'] += content_score + if grand_parent_node is not None: + candidates[grand_parent_node]['content_score'] += content_score / 2.0 + + # Scale the final candidates score based on link density. Good content + # should have a relatively small link density (5% or less) and be + # mostly unaffected by this operation. + for elem in ordered: + candidate = candidates[elem] + ld = self.get_link_density(elem) + score = candidate['content_score'] + self.debug("Candid: %6.3f %s link density %.3f -> %6.3f" % ( + score, + describe(elem), + ld, + score * (1 - ld))) + candidate['content_score'] *= (1 - ld) + + return candidates + + def class_weight(self, e): + weight = 0 + if e.get('class', None): + if REGEXES['negativeRe'].search(e.get('class')): + weight -= 25 + + if REGEXES['positiveRe'].search(e.get('class')): + weight += 25 + + if e.get('id', None): + if REGEXES['negativeRe'].search(e.get('id')): + weight -= 25 + + if REGEXES['positiveRe'].search(e.get('id')): + weight += 25 + + return weight + + def score_node(self, elem): + content_score = self.class_weight(elem) + name = elem.tag.lower() + if name == "div": + content_score += 5 + elif name in ["pre", "td", "blockquote"]: + content_score += 3 + elif name in ["address", "ol", "ul", "dl", "dd", "dt", "li", "form"]: + content_score -= 3 + elif name in ["h1", "h2", "h3", "h4", "h5", "h6", "th"]: + content_score -= 5 + return { + 'content_score': content_score, + 'elem': elem + } + + def debug(self, *a): + if self.options.get('debug', False): + log.debug(*a) + + def remove_unlikely_candidates(self): + for elem in self.html.iter(): + s = "%s %s" % (elem.get('class', ''), elem.get('id', '')) + if len(s) < 2: + continue + #self.debug(s) + if REGEXES['unlikelyCandidatesRe'].search(s) and (not REGEXES['okMaybeItsACandidateRe'].search(s)) and elem.tag not in ['html', 'body']: + self.debug("Removing unlikely candidate - %s" % describe(elem)) + elem.drop_tree() + + def transform_misused_divs_into_paragraphs(self): + for elem in self.tags(self.html, 'div'): + # transform
s that do not contain other block elements into + #

s + #FIXME: The current implementation ignores all descendants that + # are not direct children of elem + # This results in incorrect results in case there is an + # buried within an for example + if not REGEXES['divToPElementsRe'].search( + unicode(''.join(map(tostring, list(elem))))): + #self.debug("Altering %s to p" % (describe(elem))) + elem.tag = "p" + #print "Fixed element "+describe(elem) + + for elem in self.tags(self.html, 'div'): + if elem.text and elem.text.strip(): + p = fragment_fromstring('

') + p.text = elem.text + elem.text = None + elem.insert(0, p) + #print "Appended "+tounicode(p)+" to "+describe(elem) + + for pos, child in reversed(list(enumerate(elem))): + if child.tail and child.tail.strip(): + p = fragment_fromstring('

') + p.text = child.tail + child.tail = None + elem.insert(pos + 1, p) + #print "Inserted "+tounicode(p)+" to "+describe(elem) + if child.tag == 'br': + #print 'Dropped
at '+describe(elem) + child.drop_tree() + + def tags(self, node, *tag_names): + for tag_name in tag_names: + for e in node.findall('.//%s' % tag_name): + yield e + + def reverse_tags(self, node, *tag_names): + for tag_name in tag_names: + for e in reversed(node.findall('.//%s' % tag_name)): + yield e + + 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 s,
    s, and
    s + for el in self.reverse_tags(node, "table", "ul", "div"): + if el in allowed: + 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 + + # Count the text length excluding any surrounding whitespace + content_length = text_length(el) + 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
  • s than

    s" + to_remove = True + elif counts["input"] > (counts["p"] / 3): + reason = "less than 3x

    s than 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 = "s with too short content length, or too many 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() + + for el in ([node] + [n for n in node.iter()]): + if not self.options.get('attributes', None): + #el.attrib = {} #FIXME:Checkout the effects of disabling this + pass + + self.html = node + return self.get_clean_html() + + +class HashableElement(): + def __init__(self, node): + self.node = node + self._path = None + + def _get_path(self): + if self._path is None: + reverse_path = [] + node = self.node + while node is not None: + node_id = (node.tag, tuple(node.attrib.items()), node.text) + reverse_path.append(node_id) + node = node.getparent() + self._path = tuple(reverse_path) + return self._path + path = property(_get_path) + + def __hash__(self): + return hash(self.path) + + def __eq__(self, other): + return self.path == other.path + + def __getattr__(self, tag): + return getattr(self.node, tag) + + +def main(): + from optparse import OptionParser + parser = OptionParser(usage="%prog: [options] [file]") + parser.add_option('-v', '--verbose', action='store_true') + parser.add_option('-u', '--url', default=None, help="use URL instead of a local file") + (options, args) = parser.parse_args() + + if not (len(args) == 1 or options.url): + parser.print_help() + sys.exit(1) + + file = None + if options.url: + import urllib + file = urllib.urlopen(options.url) + else: + file = open(args[0], 'rt') + enc = sys.__stdout__.encoding or 'utf-8' + try: + print Document(file.read(), + debug=options.verbose, + url=options.url).summary().encode(enc, 'replace') + finally: + file.close() + +if __name__ == '__main__': + main()