How to fix error "'NoneType' object has no attribute 'split'"?

[Atrisa]

I have a file with lines like this:

Host: gateway.edge.messenger.live.com 
Referer: 13.gateway.edge.messenger.live.com/help 
Host: ads.lfstmedia.com
Referer: http://facebook.mafiawars.com:80/?
Host: gateway.edge.messenger.live.com 
Referer: 13.gateway.edge.messenger.live.com/help 
Host: ads.lfstmedia.com
Referer: 13.gateway.edge.messenger.live.com/help
Host: ads.google.com
Referer: 16.17.18.19

With the following code:

refhost = dict()

for line in open('referer.txt'):
    if line.startswith('Host'):
        host = line.split(': ')[1].strip('\n')
    elif line.startswith('Referer'):
        ref = line.split(': ')[1].strip('\n')
	refer = urlparse(ref).hostname.split('.')
        if (len(refer) > 3 and refer[-1].isalpha()):
            referrer = tuple(refer[-3:])       
            refhost.setdefault(referrer, dict()).setdefault(host, 0)
            refhost[referrer][host] +=1

I try to take the referer,the sites have been referred from the referrer and the count of each of the sites and put them in a dictionary. Then for sites that after resplitting by (.) is longer than 3 or ends in an alphabetical letter, take the last 3 words of it, to have a result like this:

messenger.live.com {'gateway.edge.messenger.live.com': 2, 'ads.lfstmedia.com': 1}
facebook.mafiawars.com {'ads.lfstmedia.com': 1} 
16.17.18.19 {'ads.google.com': 1}

But it gives me the following error:

AttributeError: 'NoneType' object has no attribute 'split'

Can anyone please tell what am I doing wrong here?
Thanks in advance, Atrisa

[bvdet]

urlparse(ref).h ostname returns None when hostname is not present. urlparse is interpreting the string "13.gateway.edg e.messenger.liv e.com/help" as path.

[Atrisa]

I changed

urlparse(ref).hostname

to

urlparse(ref).netloc

, now it neither gives error nor prints anything. Is there anything else that could do what I want

hostname

to do?

[bvdet]

Parse the string manually. What possible variations of the string do you expect?

[Atrisa]

I could solve the problem by using

netloc

instead of

hostname

and then some manual parsing. Thanks bvdet.