def __init__(self, search_url):
self.base_url = MutableURL(MutableURL(search_url).baseurl())
self.search_url = MutableURL(search_url)
class CraigsCrawler(object):
def __init__(self, search_url):
self.base_url = MutableURL(MutableURL(search_url).baseurl())
self.search_url = MutableURL(search_url)
def find_all_posts(self,
max_records=MAX_RESULTS,
records_per_page=RESULTS_PER_PAGE):
"""
Finds all craigslist links for the specified search parameters
It iterates through all pages of the search result.
The number of iterations of search pages is determined by the total
count in the page. For example, If there are 2500 results, at 100 results
per page, that 25 pages. Thus we need to pull 25 pages.
"""
current_url = self.search_url
start_at_record = 0
links = []
while True:
# Set the starting point for the records returned
current_url.query['s'] = [str(start_at_record)]
# Get page with 100 records after starting point
# import pdb;pdb.set_trace()
templinks = [
link for link in self.extract_links_from_search_page(
str(current_url))
]
# Chain results with the latest set of records
links = itertools.chain(links, templinks)
# Increment counters and break if there are no more records left
start_at_record += records_per_page
if start_at_record >= max_records or templinks == []:
break
return links
def extract_links_from_search_page(self, url):
"""
Extract all links from a url that is built to return search results for
a specific product
"""
# Build response object from url content
body = requests.get(url)
response = HtmlResponse(url=url, body=body.content)
body.connection.close()
# Build selector from response
selector = Selector(response=response)
# Select all the rows returned on that page.
rows = selector.css('p[class=row]')
# Extract all the links in the rows
links = [
row.css('a[class="hdrlnk"]::attr(href)').extract_first()
for row in rows
]
# Expand links
links = [self.base_url.joinurl(link) for link in links]
return links
def extract_info_from_post(self, post):
"""
Determines if the specified craigslist link has contact info. This is
done based of the fact that there is a hyperlink created in the html
page with the class showcontact when a page is hiding user contact info
We are not interested in the users number, we only want to make sure
that the listing is associated with a phone number
"""
baseurl = MutableURL(post)
# Get link content & build response object from url content
body = requests.get(post)
response = HtmlResponse(url=post, body=body.content)
body.connection.close()
# Build selector from response
selector = Selector(response=response)
# Extract the price from the link
price = selector.css('span[class="price"]').xpath('text()').re(
'[$]\d+')
# Create the response
post = {'link': post}
# Attach the link that contains the full content of the page
post['contact_info_link'] = selector.css(
'a[class="showcontact"]::attr(href)').extract_first()
# Expand the link
post['contact_info_link'] = self.base_url.joinurl(
post['contact_info_link']) if post['contact_info_link'] else None
post['price'] = int(price[0][1:]) if price else None
return post
def scrape_average_value(self):
"""
Extract all the urls for posts that search url returns.
"""
pool = ThreadPool(100)
post_links = self.find_all_posts()
posts = pool.map(self.extract_info_from_post, post_links)
pool.close()
pool.join()
# Extract the price from each of the posts
prices = [
post['price'] for post in posts
if post.get('price', None) and post.get('contact_info_link', None)
]
# Get some statistics on the prices
stddev = numpy.std(prices)
average = numpy.average(prices)
# Construct and return the response
response = {}
# response['urls'] = posts
response['prices'] = sorted(prices)
response['url'] = str(self.search_url)
response['used'] = average - stddev
response['good'] = average
response['new'] = average + stddev
response['stddev'] = stddev
return response
