def keywords_from_url():
"""
URL : /keywords_from_url
Extract keywords from the text content of a web page.
Method : POST
Form data :
- url : the url to analyze [string, required]
- hits : limit number of keywords returned [int, optional, 100 by default]
Return a JSON dictionary : {"keywords":[list of keywords]}
"""
# get POST data and load language resources
data = dict((key, request.form.get(key)) for key in request.form.keys())
if "url" not in data :
raise InvalidUsage('No url specified in POST data')
# crawl url, detect main language and get main text from url
url_data = url.crawl(data["url"])
if not url_data :
raise InvalidUsage('No content to analyze')
text_content = url.extract_content(url_data.text)
# analyze text and extract keywords
keywords = language.keyword_mining(text_content)
# limit the number of keywords
total = len(keywords)
hits = int(data.get("hits", 100))
keywords = [kw for kw, score in keywords.most_common(hits)]
return jsonify(keywords=keywords, total=total)
def pipeline(response, spider) :
"""
Index a page.
"""
# skip rss or atom urls
if not response.css("html").extract_first() :
return
# get domain
domain = url.domain(response.url)
# extract title
title = response.css('title::text').extract_first()
title = title.strip() if title else ""
# extract description
description = response.css("meta[name=description]::attr(content)").extract_first()
description = description.strip() if description else ""
# get main language of page, and main content of page
lang = url.detect_language(response.body)
if lang not in languages :
raise InvalidUsage('Language not supported')
body, boilerplate = url.extract_content(response.body, languages.get(lang))
# weight of page
weight = 3
if not title and not description :
weight = 0
elif not title :
weight = 1
elif not description :
weight = 2
if body.count(" ") < boilerplate.count(" ") or not url.create_description(body) :
# probably bad content quality
weight -= 1
res = spider.es_client.index(index="web-%s"%lang, id=response.url, body={
"url":response.url,
"domain":domain,
"title":title,
"description":description,
"body":body,
"weight":weight
})
if response.status in spider.handle_httpstatus_list and 'Location' in response.headers:
newurl = response.headers['Location']
meta = {'dont_redirect': True, "handle_httpstatus_list" : spider.handle_httpstatus_list}
meta.update(response.request.meta)
return Request(url = newurl.decode("utf8"), meta = meta, callback=spider.parse)
def index():
"""
URL : /index
Index a new URL in search engine.
Method : POST
Form data :
- url : the url to index [string, required]
Return a success message.
"""
# get POST data
data = dict((key, request.form.get(key)) for key in request.form.keys())
if "url" not in data:
raise InvalidUsage('No url specified in POST data')
# crawl url
url_data = url.crawl(data["url"])
if not url_data:
raise InvalidUsage("URL is invalid or has no text inside")
# get main language of page
lang = url.detect_language(url_data.text)
if lang not in languages:
raise InvalidUsage('Language not supported')
# extract title of url
title = url.extract_title(url_data.text)
# extract description of url
description = url.extract_description(url_data.text)
# extract main content of url
body = url.extract_content(url_data.text, languages.get(lang))
# index url and data
res = client.index(index="web-%s" % lang,
doc_type='page',
id=data["url"],
body={
"title": title,
"description": description,
"body": body,
"url": data["url"]
})
return "Success"
def pipeline(response, spider) :
"""
Index a page.
"""
# skip rss or atom urls
if not response.css("html").extract_first() :
return
# get domain
domain = url.domain(response.url)
# extract title
title = response.css('title::text').extract_first()
title = title.strip() if title else ""
# extract description
description = response.css("meta[name=description]::attr(content)").extract_first()
description = description.strip() if description else ""
# get main language of page, and main content of page
lang = url.detect_language(response.body)
if lang not in languages :
raise InvalidUsage('Language not supported')
body, boilerplate = url.extract_content(response.body, languages.get(lang))
# weight of page
weight = 3
if not title and not description :
weight = 0
elif not title :
weight = 1
elif not description :
weight = 2
if body.count(" ") < boilerplate.count(" ") or not url.create_description(body) :
# probably bad content quality
weight -= 1
# -- TEST -- #
"""keywords = Counter()
text_for_keywords = "%s\n%s\n%s"%(title, description, bestbody)
r = requests.post('http://localhost:5001/keywords_from_text', data = {'text':text_for_keywords})
data = r.json()
#print(hit.url, data)
for k in data["keywords"] :
keywords[k] += 1
keywords = " ".join(["%s "%(kw)*score for kw, score in keywords.most_common(100)])"""
# index url and data
res = spider.es_client.index(index="web-%s"%lang, doc_type='page', id=response.url, body={
"url":response.url,
"domain":domain,
"title":title,
"description":description,
"body":body,
"weight":weight
})
# try to create thumbnail from page
img_link = response.css("meta[property='og:image']::attr(content)").extract_first()
if not img_link :
img_link = response.css("meta[name='twitter:image']::attr(content)").extract_first()
if img_link :
q = Queue(connection=spider.redis_conn)
q.enqueue(create_thumbnail, response.url, lang, img_link)
# check for redirect url
if response.status in spider.handle_httpstatus_list and 'Location' in response.headers:
newurl = response.headers['Location']
meta = {'dont_redirect': True, "handle_httpstatus_list" : spider.handle_httpstatus_list}
meta.update(response.request.meta)
return Request(url = newurl.decode("utf8"), meta = meta, callback=spider.parse)