class ExcludedList(object):
"""URL list for storing excluded URLs. As URLs are checked for exclusion
before seen check, there are (a lot of) duplicates.
read-out is not supported because current HQ makes no use of these URLs.
"""
# TODO: duplicated code with DivertQueue
def __init__(self, jobname, bufsize=20):
self.qdir = os.path.join(hqconfig.get('datadir'), jobname, 'ex')
if not os.path.isdir(self.qdir):
os.makedirs(self.qdir)
FileEnqueue.recover(self.qdir)
self.enq = FileEnqueue(self.qdir, buffer=bufsize, suffix='ex')
self.queuedcount = 0
def flush(self):
self.enq._flush()
return self.enq.close()
def shutdown(self):
self.flush()
def get_status(self):
r = dict(queued=self.queuedcount)
return r
def add(self, furi):
self.enq.queue(furi)
self.queuedcount += 1
class CrawlJob(object):
def __init__(self, jobconfig):
self.jobconfig = jobconfig
self.enq = FileEnqueue(qdir=hqconfig.inqdir(self.jobconfig.name),
suffix=os.getpid(),
buffer=1000,
executor=None,
gzip=9)
def discovered(self, curis):
self.enq.queue(curis)
return dict(processed=len(curis))
def shutdown(self):
self.flush()
def flush(self):
self.enq._flush()
self.enq.close()
class DivertQueue(object):
"""Workset compatible class for storing URIs for delivery to external
services (including other HQ)."""
def __init__(self, basedir, name, bufsize=500):
self.name = name
self.qdir = os.path.join(basedir, str(name))
FileEnqueue.recover(self.qdir)
self.enq = FileEnqueue(self.qdir, buffer=bufsize, suffix='d')
self.queuedcount = 0
def flush(self):
self.enq._flush()
return self.enq.close()
def shutdown(self):
self.flush()
def get_status(self):
r = dict(name=self.name, queued=self.queuedcount)
return r
def schedule(self, curi):
self.enq.queue(curi)
self.queuedcount += 1
def listqfiles(self):
try:
fns = os.listdir(self.qdir)
qfiles = []
for fn in fns:
if '0' <= fn[0] <= '9' and not fn.endswith('.open'):
qfiles.append(os.path.abspath(os.path.join(self.qdir, fn)))
return qfiles
except:
return []
class WorksetWriter(object):
"""writing side of Workset."""
def __init__(self, wsdir, wsid):
self.wsid = wsid
self.qdir = os.path.join(wsdir, str(self.wsid))
FileEnqueue.recover(self.qdir)
self.enq = FileEnqueue(self.qdir, buffer=500)
self.scheduledcount = 0
def flush(self):
self.enq._flush()
return self.enq.close()
def shutdown(self):
self.flush()
def get_status(self):
r = dict(id=self.wsid, running=True,
scheduled=self.scheduledcount)
return r
def schedule(self, curi):
self.enq.queue(curi)
self.scheduledcount += 1
