| 1 |
nino.borges |
633 |
"""
|
| 2 |
|
|
|
| 3 |
|
|
sender domain extractor
|
| 4 |
|
|
|
| 5 |
|
|
"""
|
| 6 |
|
|
import re
|
| 7 |
|
|
|
| 8 |
|
|
|
| 9 |
|
|
if __name__ == '__main__':
|
| 10 |
|
|
contents = open(r"C:\Test-PY\GDWN-MBTA-SD_export-clean.dat").readlines()
|
| 11 |
|
|
domainsOutputFile = open(r"C:\Test-PY\GDWN-MBTA-SenderDomains.txt",'w')
|
| 12 |
|
|
emailsOutputFile = open(r"C:\Test-PY\GDWN-MBTA-SenderEmailAddresses.txt",'w')
|
| 13 |
|
|
|
| 14 |
|
|
rawMatrix = {}
|
| 15 |
|
|
delim = ''
|
| 16 |
|
|
emailAddressesMatrix = {}
|
| 17 |
|
|
domainMatrix = {}
|
| 18 |
|
|
count = 0
|
| 19 |
|
|
|
| 20 |
|
|
print "creating the raw list..."
|
| 21 |
|
|
for line in contents:
|
| 22 |
|
|
count = count +1
|
| 23 |
|
|
if count > 5000:
|
| 24 |
|
|
print "Heartbeat"
|
| 25 |
|
|
count = 0
|
| 26 |
|
|
line = line.replace("\n","")
|
| 27 |
|
|
line = line.split(delim)
|
| 28 |
|
|
if len(line) >1:
|
| 29 |
|
|
bates = line[0]
|
| 30 |
|
|
sender = line[1]
|
| 31 |
|
|
if sender:
|
| 32 |
|
|
rawMatrix[sender] = 1
|
| 33 |
|
|
print "raw matrix made."
|
| 34 |
|
|
rawList = rawMatrix.keys()
|
| 35 |
|
|
rawList.sort()
|
| 36 |
|
|
print "processing list..."
|
| 37 |
|
|
for i in rawList:
|
| 38 |
|
|
print i
|
| 39 |
|
|
if '@' in i:
|
| 40 |
|
|
x = re.search('(\w+[.|\w])*@(\w+[.])*\w+',i)
|
| 41 |
|
|
if x:
|
| 42 |
|
|
emailAddy = x.group()
|
| 43 |
|
|
emailAddressesMatrix[emailAddy.lower()] = 1
|
| 44 |
|
|
domain = "@%s"%emailAddy.split("@")[1]
|
| 45 |
|
|
domainMatrix[domain.lower()] = 1
|
| 46 |
|
|
print "done."
|
| 47 |
|
|
print "Unpacking..."
|
| 48 |
|
|
domainsList = domainMatrix.keys()
|
| 49 |
|
|
domainsList.sort()
|
| 50 |
|
|
emailsList = emailAddressesMatrix.keys()
|
| 51 |
|
|
emailsList.sort()
|
| 52 |
|
|
for i in domainsList:
|
| 53 |
|
|
domainsOutputFile.write("%s\n"%i)
|
| 54 |
|
|
domainsOutputFile.close()
|
| 55 |
|
|
for i in emailsList:
|
| 56 |
|
|
emailsOutputFile.write("%s\n"%i)
|
| 57 |
|
|
emailsOutputFile.close()
|
| 58 |
|
|
|
