Previous Section  < Day Day Up >  Next Section

Hack 39. Meander Your Google Neighborhood

Google Neighborhood attempts to detangle the Web by building a "neighborhood" of sites around a URL.

It's called the World Wide Web, not the World Wide Straight Line. Sites link to other sites, building a web of sites. And what a tangled web we weave.

Google Neighborhood by the Python-wise Mark Pilgrim (http://diveintomark.org) attempts to detangle some small portion of the Web by using the Google API to find sites related to a URL that you provide, scraping the links on the sites returned and building a "neighborhood" of sites that link both the original URL and each other.

If you'd like to give this hack a whirl without having to run it yourself, there's a live version available at http://diveintomark.org/archives/2002/06/04/who_are_the_people_in_your_neighborhood. The source code (included in the following section) for Google Neighborhood is available for download from http://diveintomark.org/projects/misc/neighbor.py.txt.

2.21.1. The Code

Google Neighborhood is written in the Python (http://www.python.org) programming language. Your system will need to have Python installed for you to run this hack.

"""Blogroll finder and aggregator"""



__author_  _ = "Mark Pilgrim (f8dy@diveintomark.org)"

__copyright_  _ = "Copyright 2002, Mark Pilgrim"

__license_  _ = "Python"



try:

    import timeoutsocket # http://www.timo-tasi.org/python/timeoutsocket.py

    timeoutsocket.setDefaultSocketTimeout(10)

except:

    pass

import urllib, urlparse, os, time, operator, sys, pickle, re, cgi, time

from sgmllib import SGMLParser

from threading import *



BUFFERSIZE = 1024

IGNOREEXTS = ('.xml', '.opml', '.rss', '.rdf', '.pdf', '.doc')

INCLUDEEXTS = ('', '.html', '.htm', '.shtml', '.php', '.asp', '.jsp')

IGNOREDOMAINS = ('cgi.alexa.com', 'adserver1.backbeatmedia.com', 'ask.slashdot.org', 

'freshmeat.net', 'readroom.ipl.org', 'amazon.com', 'ringsurf.com')



def prettyURL(url):

    protocol, domain, path, params, query, fragment = urlparse.urlparse(url)

    if path == '/':

        path = ''

    return urlparse.urlunparse(('', domain, path, '', '', '')).replace('//', '')

    

def simplifyURL(url):

    url = url.replace('www.', '')

    url = url.replace('/coming.html', '/')

    protocol, domain, path, params, query, fragment = urlparse.urlparse(url)

    if path == '':

        url = url + '/'

    return url



class MinimalURLOpener(urllib.FancyURLopener):

    def __init_  _(self, *args):

        apply(urllib.FancyURLopener.__init_  _, (self,) + args)

        self.addheaders = [('User-agent', '')]

    def http_error_401(self, url, fp, errcode, errmsg, headers, data=None):

        pass



class BlogrollParser(SGMLParser):

    def __init_  _(self, url):

        SGMLParser.__init_  _(self)

        self.url = url

        self.reset( )

        

    def reset(self):

        SGMLParser.reset(self)

        self.possible = []

        self.blogroll = []

        self.ina = 0

    

    def _goodlink(self, href):

        protocol, domain, path, params, query, fragment = urlparse.urlparse(href)

        if protocol.lower( ) <> 'http': return 0

        if self.url.find(domain) <> -1: return 0

        if domain in IGNOREDOMAINS: return 0

        if domain.find(':5335') <> -1: return 0

        if domain.find('.google') <> -1: return 0

        if fragment: return 0

        shortpath, ext = os.path.splitext(path)

        ext = ext.lower( )

        if ext in INCLUDEEXTS: return 1

        if ext.lower( ) in IGNOREEXTS: return 0

        # more rules here?

        return 1

    

    def _confirmpossibles(self):

        if len(self.possible) >= 4:

            for url in self.possible:

                if url not in self.blogroll:

                    self.blogroll.append(url)

        self.possible = []

    

    def start_a(self, attrs):

        self.ina = 1

        hreflist = [e[1] for e in attrs if e[0]=='href']

        if not hreflist: return

        href = simplifyURL(hreflist[0])

        if self._goodlink(href):

            self.possible.append(href)



    def end_a(self):

        self.ina = 0

    

    def handle_data(self, data):

        if self.ina: return

        if data.strip( ):

            self._confirmpossibles( )

        

    def end_html(self, attrs):

        self.confirmpossibles( )



def getRadioBlogroll(url):

    try:

        usock = MinimalURLOpener( ).open('%s/gems/mySubscriptions.opml' % url)

        opmlSource = usock.read( )

        usock.close( )

    except:

        return []

    if opmlSource.find('<opml') == -1: return []

    radioBlogroll = []

    start = 0

    while 1:

        p = opmlSource.find('htmlUrl="', start)

        if p == -1: break

        refurl = opmlSource[p:p+100].split('"')[1]

        radioBlogroll.append(refurl)

        start = p + len(refurl) + 10

    return radioBlogroll



def getBlogroll(url):

    if url[:7] <> 'http://':

        url = 'http://' + url

    radioBlogroll = getRadioBlogroll(url)

    if radioBlogroll:

        return radioBlogroll

    parser = BlogrollParser(url)

    try:

        usock = MinimalURLOpener( ).open(url)

        htmlSource = usock.read( )

        usock.close( )

    except:

        return []

    parser.feed(htmlSource)

    return parser.blogroll



class BlogrollThread(Thread):

    def __init_  _(self, master, url):

        Thread.__init_  _(self)

        self.master = master

        self.url = url



    def run(self):

        self.master.callback(self.url, getBlogroll(self.url))



class BlogrollThreadMaster:

    def __init_  _(self, url, recurse):

        self.blogrollDict = {}

        self.done = 0

        if type(url)==type(''):

            blogroll = getBlogroll(url)

        else:

            blogroll = url

        self.run(blogroll, recurse)

    

    def callback(self, url, blogroll):

        if not self.done:

            self.blogrollDict[url] = blogroll

    

    def run(self, blogroll, recurse):

        start = 0

        end = 5

        while 1:

            threads = []

            for url in blogroll[start:end]:

                if not self.blogrollDict.has_key(url):

                    t = BlogrollThread(self, url)

                    threads.append(t)

            for t in threads:

                t.start( )

                time.sleep(0.000001)

            for t in threads:

                time.sleep(0.000001)

                t.join(10)

            start += 5

            end += 5

            if start > len(blogroll): break

        if recurse > 1:

            masterlist = reduce(operator.add, self.blogrollDict.values( ))

            newlist = [url for url in masterlist if not self.blogrollDict.            has_key(url)]

            self.run(newlist, recurse - 1)

        else:

            self.done = 1



def sortBlogrollData(blogrollDict):

    sortD = {}

    for blogroll in blogrollDict.values( ):

        for url in blogroll:

            sortD[url] = sortD.setdefault(url, 0) + 1

    sortI = [(v, k) for k, v in sortD.items( )]

    sortI.sort( )

    sortI.reverse( )

    return sortI



def trimdata(sortI, cutoff):

    return [(c, url) for c, url in sortI if c >= cutoff]



def getRelated(url):

    import google

    results = []

    start = 0

    for i in range(3):

        data = google.doGoogleSearch('related:%s' % url, start)

        results.extend([oneResult.URL for oneResult in data.results])

        start += 10

        if len(data.results) < 10: break

    return results



def getNeighborhood(baseURL):

    relatedList = getRelated(baseURL)

    blogrollDict = BlogrollThreadMaster(relatedList, 1).blogrollDict

    neighborhood = sortBlogrollData(blogrollDict)

    neighborhood = trimdata(neighborhood, 2)

    neighborhood = [(c,url, prettyURL(url)) for c,url in neighborhood]

    return neighborhood

    

def render_html(baseURL, data):

    output = []

    output.append("""

<table class="socialnetwork" summary="neighborhood for %s">

<caption>Neighborhood for %s</caption>

<thead>

<tr>

<th scope="col">Name</th>

<th scope="col">Links</th>

<th shope="col">Explore</th>

</tr>

</thead>

<tbody>""" % (cgi.escape(prettyURL(baseURL)), cgi.escape(prettyURL(baseURL))))

    for c, url, title in data:

        output.append("""<tr><td><a href="%s">%s</a></td>

<td>%s</td><td><a href="%s">explore</a></td></tr

>""" % (url, title, c, 'http://diveintomark.org/cgi-bin/neighborhood.cgi?url=%s' % 

cgi.escape(url)))

    output.append("""

</tbody>

</table>""")

    return "".join(output)



def render_rss(baseURL, data):

    title = prettyURL(baseURL)

    channeltitle = "%s neighborhood" % title

    localtime = time.strftime('%Y-%m-%dT%H:%M:%S-05:00', time.localtime( ))

    output = []

    output.append("""<?xml version="1.0"?>

<rdf:RDF xmlns="http://purl.org/rss/1.0/"

xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:dc="http://purl.org/dc/

elements/1.1/" xmlns:sy="http://purl.org/rss/1.0/modules/syndication/" xmlns:admin=

"http://webns.net/mvcb/">



<channel rdf:about="%(baseURL)s">

<title>%(channeltitle)s</title>

<link>%(baseURL)s</link>

<description>Sites in the virtual neighborhood of %(title)s</description>

<language>en-us</language>

<lastBuildDate>%(localtime)s</lastBuildDate>

<pubDate>%(localtime)s</pubDate>

<admin:generatorAgent rdf:resource="http://divintomark.org/cgi-bin/neighborhood.cgi/

?v=1.1" />

<admin:errorReportsTo rdf:resource="mailto:f8dy@diveintomark.org"/>

<sy:updatePeriod>weekly</sy:updatePeriod>

<sy:updateFrequency>1</sy:updateFrequency>

<sy:updateBase>2000-01-01T12:00+00:00</sy:updateBase>

<items>

<rdf:Seq>

""" % locals( ))

##"""

    for c, url, title in data:

        output.append("""<rdf:li rdf:resource="%s" />

""" % url)

    output.append("""</rdf:Seq>

</items>

</channel>

""")

    for c, url, title in data:

        output.append("""<item rdf:about="%(url)s">

<title>%(title)s</title>

<link>%(url)s</link>

<description>%(c)s links</description>

</item>

""" % locals( ))

    output.append("""</rdf:RDF>""")

    return "".join(output)

                      

if __name__ == '__main_  _':

    print render_html(getNeighborhood(sys.argv[1]))

You'll also need an HTML form to call the neighborhood.cgi script. Here's a simple one:

<form action="/cgi-bin/neighborhood.cgi" method="get">

URL: <input name="url" type="text" />

<br />

Output as: <input name="fl" type="radio" value="html" checked="true" /> HTML

<input name="fl" type="radio" value="rss" checked="true" /> RSS

<br />

<input type="submit" value="Meander" />

</form>

Save the form as neighborhood.html, being sure to alter the action= to point at the location in which you installed the CGI script ["How to Run the Scripts" in the Preface].

2.21.2. Running the Hack

Point your browser at the location of the form you saved just a moment ago. Provide it with the URL that you're interested in using as the center, select HTML or RSS output, and hit the Meander button.

Figure 2-13 shows a representation of Rael's (raelity.org's, to be precise) Google Neighborhood. Clicking on any of the links on the left transports you to the URL shown. More interestingly, the "explore" link shifts your point of view, centering the neighborhood on the associated URL. You can thus meander a neighborhood to your heart's content; don't be surprised, especially in the blogging world, if you keep coming across the same links. Speaking of links, the number listed beneath the "Links" heading represents the number of links the associated site has to the currently focused site.

Figure 2-13. raelity.org's Google Neighborhood


2.21.3. Hacking the Hack

If you want to hack this hack, concentrate your efforts on a small block of code, specifying what file extensions you want to include and exclude, as well as what domains you want to exclude when calculating your neighborhoods:

IGNOREEXTS = ('.xml', '.opml', '.rss', '.rdf', '.pdf', '.doc')

INCLUDEEXTS = ('', '.html', '.htm', '.shtml', '.php', '.asp', '.jsp')

IGNOREDOMAINS = ('cgi.alexa.com', 'adserver1.backbeatmedia.com', 'ask.

slashdot.org','freshmeat.net', 'readroom.ipl.org', 'amazon.com',

'ringsurf.com')

2.21.3.1 Noticing/ignoring file extensions

The way the hack is currently written, the neighborhood is built around pretty standard files. However, you could create a neighborhood of sites served by PHP (http://www.php.net), including only URLs with a PHP (.php) extension. Or perhaps your interest lies in Word documents and PDF files. You'd alter the code as follows:

IGNOREEXTS = ('.xml', '.opml', '.rss', '.rdf', '.html', '.htm', '.shtml', 

'.php', '.asp', '.jsp')

INCLUDEEXTS = ('', 

'.pdf', '.doc'

)

2.21.3.2 Ignoring domains

Sometimes, when building a neighborhood, you might notice that the same links are popping up again and again. They're not really part of the neighborhood but tend to be places that the web pages making up your neighborhood often link to. For example, most Blogger-based weblogs include a link to http://www.blogger.com as a matter of course.

Exclude domains that hold no interest to you by adding them to the IGNOREDOMAINS list:

IGNOREDOMAINS = ('cgi.alexa.com', 'adserver1.backbeatmedia.com', 

'ask.slashdot.org', 'freshmeat.net', 'readroom.ipl.org', 'amazon.com', 

'ringsurf.com', 'blogger.com')

    Previous Section  < Day Day Up >  Next Section