# Pull URL
# Strip query string
# Query exists
# IFN save to sqlite
# IFN push to queue
# IFY do nothing
seen = {}
if __name__ == '__main__':
from helpers import client
ingest = client.queue('ingest')
scrape = client.queue('scrape')
while True:
claimed = ingest.claim(ttl=180, grace=60)
send = []
for msg in claimed:
msg.delete()
if seen.get(msg.body):
print "skipping %s, seen %d pages" % (msg.body, len(seen.keys()))
continue
print "Sending along %s" % msg.body
seen[msg.body] = True
send.append({'body': msg.body, 'ttl': 180})
if len(send): scrape.post(send)
else:
yield parent.scheme + '://' + parent.netloc + parent.path + '/' + link
seen = {}
def scrapable(uri):
if not uri.startswith(sys.argv[1]):
return False
if seen.get(uri):
return False
seen[uri] = True
return True
if __name__ == '__main__':
scrape = client.queue('scrape')
ingest = client.queue('ingest')
complete = client.queue('completed')
while True:
claimed = scrape.claim(ttl=180, grace=60, limit=1)
for msg in claimed:
messages = [
{'body': u, 'ttl': 180}
for u in scrape_generator(msg.body)
if scrapable(u)
]
if len(messages):
ingest.post(messages)
complete.post({'body': msg.body, 'ttl': 300})
msg.delete()
# 1. take a URL from the CLI
# 2. Push to zaqar
# 3. Listen on `collected` queue and print
import sys
import datetime
from helpers import client
if __name__ == '__main__':
q = client.queue('ingest')
q.post({'body': sys.argv[1], 'ttl': 300})
complete = client.queue('completed')
while True:
claimed = complete.claim(ttl=180, grace=60)
for msg in claimed:
print "%s %s" % (datetime.datetime.now(), msg.body)
msg.delete()
