"""Add to descriptors contents of descriptor archive in descriptor_dir."""

num_descriptors = 0

num_relays = 0

print('Reading descriptors from: {0}'.format(descriptor_dir))

reader = stem.descriptor.reader.DescriptorReader(descriptor_dir,

validate=True)

reader.register_skip_listener(skip_listener)

# use read listener to store metrics type annotation, which is otherwise discarded

cur_type_annotation = [None]

def read_listener(path):

f = open(path)

# store initial metrics type annotation

initial_position = f.tell()

first_line = f.readline()

f.seek(initial_position)

if (first_line[0:5] == '@type'):

cur_type_annotation[0] = first_line

else:

cur_type_annotation[0] = None

reader.register_read_listener(read_listener)

with reader:

for desc in reader:

if (num_descriptors % 10000 == 0):

print('{0} descriptors processed.'.format(num_descriptors))

num_descriptors += 1

if (desc.fingerprint not in descriptors):

descriptors[desc.fingerprint] = {}

num_relays += 1

# stuff type annotation into stem object

desc.type_annotation = cur_type_annotation[0]

descriptors[desc.fingerprint]\

[pathsim.timestamp(desc.published)] = desc

print('#descriptors: {0}; #relays:{1}'.\

"""For every input consensus, finds the descriptors published most recently before the descriptor times listed for the relays in that consensus, records state changes indicated by descriptors published during the consensus fresh period, and writes out pickled consensus and descriptor objects with the relevant information.

Inputs:

in_dirs: list of (consensus in dir, descriptor in dir,

processed descriptor out dir) triples *in order*

slim: Whether to use custom slim classes or Stem classes.

initial_descriptor_dir: Contains descriptors to initialize processing.

"""

descriptors = {}

def skip_listener(path, exception):

print('ERROR [{0}]: {1}'.format(path.encode('ascii', 'ignore'), exception.__unicode__().encode('ascii','ignore')))

if slim:

print('Outputting slim classes.')

# initialize descriptors

if (initial_descriptor_dir is not None):

read_descriptors(descriptors, initial_descriptor_dir, skip_listener)

for in_consensuses_dir, in_descriptors, desc_out_dir in in_dirs:

# read all descriptors into memory

read_descriptors(descriptors, in_descriptors, skip_listener)

# output pickled consensuses, dict of most recent descriptors, and

# list of hibernation status changes

num_consensuses = 0

pathnames = []

for dirpath, dirnames, fnames in os.walk(in_consensuses_dir):

for fname in fnames:

pathnames.append(os.path.join(dirpath,fname))

pathnames.sort()

for pathname in pathnames:

filename = os.path.basename(pathname)

if (filename[0] == '.'):

continue

print('Processing consensus file {0}'.format(filename))

cons_f = open(pathname, 'rb')

# store metrics type annotation line

initial_position = cons_f.tell()

first_line = cons_f.readline()

if (first_line[0:5] == '@type'):

type_annotation = first_line

else:

type_annotation = None

cons_f.seek(initial_position)

descriptors_out = dict()

hibernating_statuses = [] # (time, fprint, hibernating)

cons_valid_after = None

cons_fresh_until = None

if slim:

cons_bw_weights = None

cons_bwweightscale = None

relays = {}

num_not_found = 0

num_found = 0

# read in consensus document

i = 0

for document in stem.descriptor.parse_file(cons_f, validate=True,

document_handler='DOCUMENT'):

if (i > 0):

raise ValueError('Unexpectedly found more than one consensus in file: {}'.\

format(pathname))

if (cons_valid_after == None):

cons_valid_after = document.valid_after

# compute timestamp version once here

valid_after_ts = pathsim.timestamp(cons_valid_after)

if (cons_fresh_until == None):

cons_fresh_until = document.fresh_until

# compute timestamp version once here

fresh_until_ts = pathsim.timestamp(cons_fresh_until)

if slim:

if (cons_bw_weights == None):

cons_bw_weights = document.bandwidth_weights

if (cons_bwweightscale == None) and \

('bwweightscale' in document.params):

cons_bwweightscale = document.params[\

'bwweightscale']

for r_stat in document.routers:

relays[r_stat.fingerprint] = pathsim.RouterStatusEntry(\

r_stat.fingerprint, r_stat.nickname, r_stat.flags, r_stat.bandwidth)

consensus = document

i += 1

# find relays' most recent unexpired descriptor published

# before the publication time in the consensus

# and status changes in fresh period (i.e. hibernation)

for fprint, r_stat in consensus.routers.iteritems():

pub_time = pathsim.timestamp(r_stat.published)

desc_time = 0

descs_while_fresh = []

desc_time_fresh = None

# get all descriptors with this fingerprint

if (r_stat.fingerprint in descriptors):

for t,d in descriptors[r_stat.fingerprint].items():

# update most recent desc seen before cons pubtime

# allow pubtime after valid_after but not fresh_until

if (valid_after_ts-t <\

pathsim.TorOptions.router_max_age) and\

(t <= pub_time) and (t > desc_time) and\

(t <= fresh_until_ts):

desc_time = t

# store fresh-period descs for hibernation tracking

if (t >= valid_after_ts) and \

(t <= fresh_until_ts):

descs_while_fresh.append((t,d))

# find most recent hibernating stat before fresh period

# prefer most-recent descriptor before fresh period

# but use oldest after valid_after if necessary

if (desc_time_fresh == None):

desc_time_fresh = t

elif (desc_time_fresh < valid_after_ts):

if (t > desc_time_fresh) and\

(t <= valid_after_ts):

desc_time_fresh = t

else:

if (t < desc_time_fresh):

desc_time_fresh = t

# output best descriptor if found

if (desc_time != 0):

num_found += 1

# store discovered recent descriptor

desc = descriptors[r_stat.fingerprint][desc_time]

if slim:

descriptors_out[r_stat.fingerprint] = \

pathsim.ServerDescriptor(desc.fingerprint, \

desc.hibernating, desc.nickname, \

desc.family, desc.address, \

desc.exit_policy, desc.ntor_onion_key)

else:

if (desc.type_annotation is not None):

descriptors_out[r_stat.fingerprint] = desc.type_annotation + str(desc)

else:

descriptors_out[r_stat.fingerprint] = str(desc)

# store hibernating statuses

if (desc_time_fresh == None):

raise ValueError('Descriptor error for {0}:{1}.\n Found descriptor before published date {2}: {3}\nDid not find descriptor for initial hibernation status for fresh period starting {4}.'.format(r_stat.nickname, r_stat.fingerprint, pub_time, desc_time, valid_after_ts))

desc = descriptors[r_stat.fingerprint][desc_time_fresh]

cur_hibernating = desc.hibernating

# setting initial status

hibernating_statuses.append((0, desc.fingerprint,\

cur_hibernating))

if (cur_hibernating):

print('{0}:{1} was hibernating at consenses period start'.format(desc.nickname, desc.fingerprint))

descs_while_fresh.sort(key = lambda x: x[0])

for (t,d) in descs_while_fresh:

if (d.hibernating != cur_hibernating):

cur_hibernating = d.hibernating

hibernating_statuses.append(\

(t, d.fingerprint, cur_hibernating))

if (cur_hibernating):

print('{0}:{1} started hibernating at {2}'\

.format(d.nickname, d.fingerprint, t))

else:

print('{0}:{1} stopped hibernating at {2}'\

.format(d.nickname, d.fingerprint, t))

else:

# print(\

# 'Descriptor not found for {0}:{1}:{2}'.format(\

# r_stat.nickname,r_stat.fingerprint, pub_time))

num_not_found += 1

# output pickled consensus, recent descriptors, and

# hibernating status changes

if (cons_valid_after != None) and\

(cons_fresh_until != None):

if slim:

consensus_out = pathsim.NetworkStatusDocument(\

cons_valid_after, cons_fresh_until, cons_bw_weights,\

cons_bwweightscale, relays)

else:

if (type_annotation is not None):

consensus_out = type_annotation + str(consensus)

else:

consensus_out = str(consensus)

hibernating_statuses.sort(key = lambda x: x[0],\

reverse=True)

outpath = os.path.join(desc_out_dir,\

cons_valid_after.strftime(\

'%Y-%m-%d-%H-%M-%S-network_state'))

f = open(outpath, 'wb')

pickle.dump(consensus_out, f, pickle.HIGHEST_PROTOCOL)

pickle.dump(descriptors_out,f,pickle.HIGHEST_PROTOCOL)

pickle.dump(hibernating_statuses,f,pickle.HIGHEST_PROTOCOL)

f.close()

print('Wrote descriptors for {0} relays.'.\

format(num_found))

print('Did not find descriptors for {0} relays\n'.\

format(num_not_found))

else:

print('Problem parsing {0}.'.format(filename))

num_consensuses += 1

cons_f.close()