def test_is_valley_free_false(self):
peerup = [0, 5, 6, 4] # DPU
peerpeer = [0, 5, 6, 5, 1] # UPPD
downpeer = [4, 6, 5, 0] # DPD
downup = [4, 5, 4] # DU
self.assertFalse(vft.is_valley_free(self.sample_graph, peerup))
self.assertFalse(vft.is_valley_free(self.sample_graph, peerpeer))
self.assertFalse(vft.is_valley_free(self.sample_graph, downpeer))
self.assertFalse(vft.is_valley_free(self.sample_graph, downup))
def filter(g,
traceroutes,
filters=['sh', 'loop', 'ex', 'vf', 'lp'],
first_edge=True):
logger.info('Traceroutes: %d', len(traceroutes))
# remove empty traces
traceroutes = [x for x in traceroutes if len(x) > 0]
logger.info('Non empty traceroutes: %d', (len(traceroutes)))
traceroutes = [x for x in traceroutes if len(x) > 1]
logger.info('Larger than one hop traceroutes: %d', (len(traceroutes)))
# remove traces with unknown nodes
traceroutes, _ = vft.trace_clean(g, traceroutes)
logger.info('Ignored: %d', _)
traceroutes = vft.trace_in_vertex_id(g, traceroutes)
logger.info('Trace count: %d', len(traceroutes))
progress = progressbar1.AnimatedProgressBar(end=len(traceroutes), width=15)
good_traceroutes = traceroutes[:]
if 'sh' in filters:
logger.debug('Remove short traces')
good_traceroutes = [x for x in good_traceroutes if len(x) >= 3]
logger.debug('Remained: %d', len(good_traceroutes))
if 'loop' in filters:
logger.debug('Remove traces with loops')
good_traceroutes = [
x for x in good_traceroutes if len(set(x)) == len(x)
]
logger.debug('Remained: %d' % len(good_traceroutes))
if 'ex' in filters:
logger.debug('Remove non existent traces')
good_traceroutes = [
x for x in good_traceroutes if vft.trace_exists(g, x)
]
logger.debug('Remained: %d', len(good_traceroutes))
if 'vf' in filters:
logger.debug('Remove non vf traces')
good_traceroutes = [
x for x in good_traceroutes if vft.is_valley_free(g, x)
]
logger.debug('Remained: %d' % len(good_traceroutes))
if 'lp' in filters:
logger.debug('Remove non lp traces')
good_traceroutes = [
x for x in good_traceroutes
if vft.is_local_preferenced(g, x, first_edge=first_edge)
]
logger.debug('Remained: %d' % len(good_traceroutes))
# convert back node ids to node names
good_traceroutes = [[g.vs[id]["name"] for id in trace]
for trace in good_traceroutes]
logger.debug(len(good_traceroutes))
return good_traceroutes
def ba_generator(ba_graph, sh_paths, stretch, vf_g, progressbar=False):
vf_count = 0
trace_count = 0
lp_count = 0
progress = progressbar1.DummyProgressBar(end=10, width=15)
if progressbar:
progress = progressbar1.AnimatedProgressBar(end=len(sh_paths),
width=15)
for (s, t), shl in sh_paths:
progress += 1
progress.show_progress()
logger.debug('SH from %s to %s is %d' % (s, t, shl))
random_route = helpers.random_route_walk(ba_graph, s, t, shl + stretch)
logger.debug('Random route: %s' % random_route)
real_stretch = len(random_route) - shl
if real_stretch != stretch:
continue
trace_count += 1
is_vf = vft.is_valley_free(ba_graph,
random_route,
vfmode=vft.CLOSENESS)
logger.debug(
'Trace edge dir: %s' %
vft.trace_to_string(ba_graph, random_route, vfmode=vft.CLOSENESS))
logger.debug('Is VF: %s' % is_vf)
if is_vf:
is_lp = vft.is_local_preferenced(ba_graph,
random_route,
first_edge=True,
vfmode=vft.CLOSENESS,
vf_g=vf_g)
else:
is_lp = 0
logger.debug('Is LP: %s' % is_lp)
vf_count += int(is_vf)
lp_count += int(is_lp)
logger.info('Stretch %d' % stretch)
logger.info('Trace count: %d' % trace_count)
logger.info('VF count: %d' % vf_count)
logger.info('LP count: %d' % lp_count)
return (stretch, trace_count, vf_count, lp_count)
def test_is_valley_free_true(self):
simple_vf = [0, 5, 4, 6, 2] # UUDD - vf
middle_peer_vf = [0, 5, 6, 2] # UPD - vf
just_down = [4, 6, 3] # DD - vf
just_up = [1, 5, 4] # UU - vf
start_peer = [5, 6, 3] # PD - vf
just_peer = [5, 6] # P - vf
self.assertTrue(vft.is_valley_free(self.sample_graph, simple_vf))
self.assertTrue(vft.is_valley_free(self.sample_graph, middle_peer_vf))
self.assertTrue(vft.is_valley_free(self.sample_graph, just_down))
self.assertTrue(vft.is_valley_free(self.sample_graph, just_up))
self.assertTrue(vft.is_valley_free(self.sample_graph, start_peer))
self.assertTrue(vft.is_valley_free(self.sample_graph, just_peer))
def vf_attributes(g, trace, vfmode, get_lp_soft, get_lp_hard, vf_g=None):
is_vf = int(vft.is_valley_free(g, trace, vfmode))
is_lp_soft = -1
is_lp_hard = -1
if is_vf:
if get_lp_soft:
lp_soft = vft.is_local_preferenced(g, trace,
vf_g=vf_g,
first_edge=True,
vfmode=vfmode)
is_lp_soft = int(lp_soft)
else:
is_lp_prelabeled_soft = -1
if get_lp_hard:
lp_hard = vft.is_local_preferenced(g, trace,
vf_g=vf_g,
first_edge=False,
vfmode=vfmode)
is_lp_hard = int(lp_hard)
else:
is_lp_hard = -1
return (is_vf, is_lp_soft, is_lp_hard)
def purify(g, meta, out, count=1000):
results = list()
results2 = list()
results3 = list()
all_vf = 0
all_nonvf = 0
all_vf_closeness = 0
all_nonvf_closeness = 0
short_results = list()
short_results2 = list()
short_results3 = list()
all_short_vf = 0
all_short_nonvf = 0
all_short_vf_closeness = 0
all_short_nonvf_closeness = 0
long_results = list()
long_results2 = list()
long_results3 = list()
all_long_vf = 0
all_long_nonvf = 0
all_long_vf_closeness = 0
all_long_nonvf_closeness = 0
# remove traces with already calculated all_path
logger.warn('[r]ONLY NOT FILLED PATHS[/]')
meta = [x for x in meta if not helpers.ALL_PATH_COUNT in x]
# traces with a maximum stretch
logger.warn('[r]!!!ONLY WITH LOW STRETCH[/]')
meta = [x for x in meta if x[helpers.STRETCH] < 4]
# shorter meta records
logger.warn('[r]!!!ONLY SHORT TRACES[/]')
meta = [x for x in meta if len(x[helpers.TRACE]) < 5]
meta_map = {tuple(x[helpers.TRACE]): x for x in meta}
# traceroutes = [x for x in meta if x[TRACE_LEN] == x[SH_LEN]]
logger.info('All trace count: %d' % len(meta))
tr_count = min(len(meta), count)
meta = random.sample(meta, tr_count)
logger.info('Chosen trace count: %d' % len(meta))
real_vf = [x for x in meta if x[helpers.IS_VF] == 1]
real_nonvf = [x for x in meta if x[helpers.IS_VF] == 0]
real_vf_closeness = [x for x in meta if x[helpers.IS_VF_CLOSENESS] == 1]
real_nonvf_closeness = [x for x in meta if x[helpers.IS_VF_CLOSENESS] == 0]
logger.info('Real vf: %f[%d]' % ((len(real_vf)/float(len(meta)), len(real_vf))))
logger.info('Real nonvf: %f[%d]' % ((len(real_nonvf)/float(len(meta)), len(real_nonvf))))
logger.info('Real vf closeness: %f[%d]' % ((len(real_vf_closeness)/float(len(meta)), len(real_vf_closeness))))
logger.info('Real nonvf closeness: %f[%d]' % ((len(real_nonvf_closeness)/float(len(meta)), len(real_nonvf_closeness))))
logger.info('Remove unknown traces. Trace count before: %d' % len(meta))
traceroutes = [x[helpers.TRACE] for x in meta]
traceroutes, ignored = vft.trace_clean(g, traceroutes)
logger.info('Traceroutes after: %d. Ignored: %d' % (len(traceroutes), ignored))
traceroutes = vft.trace_in_vertex_id(g, traceroutes)
progress = progressbar1.AnimatedProgressBar(end=len(traceroutes), width=15)
for trace in traceroutes:
progress += 1
progress.show_progress()
for x in range(0, g.vcount()):
g.vs[x]['traces'] = dict()
s, t = trace[0], trace[-1]
sh_path = g.get_all_shortest_paths(s, t, mode=i.OUT)
all_path = helpers.dfs_mark(copy.deepcopy(g), s, t, len(trace))
# if len(sh_path) != len(all_path):
# print len(sh_path)
# print len(all_path)
# print s, t
# sanity check
for x in all_path:
if x[0] != s or x[-1] != t:
logger.error('ALERT')
if len(set([tuple(x) for x in all_path])) != len(all_path):
logger.error('LENGTH ALERT')
logger.error('%s' % len(all_path))
logger.error('%s' % len(set([tuple(x) for x in all_path])))
logger.error('%s' % sorted(all_path))
long_path = [x for x in all_path if len(x) == len(trace)]
short_path = [x for x in all_path if len(x) < len(trace)]
named_trace = [g.vs[x]['name'] for x in trace]
extra_meta = {
helpers.ALL_PATH_COUNT: len(all_path),
helpers.SAME_LONG_PATH_COUNT: len(long_path),
helpers.SHORTER_PATH_COUNT: len(short_path)
}
meta_map[tuple(named_trace)].update(extra_meta)
vf_count = sum([1 if vft.is_valley_free(g, x, vfmode=vft.PRELABELED) else 0 for x in all_path])
nonvf = len(all_path) - vf_count
vf_closeness_count = sum([1 if vft.is_valley_free(g, x, vfmode=vft.CLOSENESS) else 0 for x in all_path])
nonvf_closeness = len(all_path) - vf_closeness_count
tmp = [1 if vft.is_valley_free(g, x, vfmode=vft.PRELABELED) else 0 for x in short_path]
short_vf_count = sum(tmp)
short_nonvf = len(tmp) - short_vf_count
tmp = [1 if vft.is_valley_free(g, x, vfmode=vft.CLOSENESS) else 0 for x in short_path]
short_vf_closeness_count = sum(tmp)
short_nonvf_closeness = len(tmp) - short_vf_closeness_count
tmp = [1 if vft.is_valley_free(g, x, vfmode=vft.PRELABELED) else 0 for x in long_path]
long_vf_count = sum(tmp)
long_nonvf = len(tmp) - long_vf_count
tmp = [1 if vft.is_valley_free(g, x, vfmode=vft.CLOSENESS) else 0 for x in long_path]
long_vf_closeness_count = sum(tmp)
long_nonvf_closeness = len(tmp) - long_vf_closeness_count
extra_meta = {
helpers.ALL_PATH_VF_COUNT: vf_closeness_count,
helpers.SAME_LONG_PATH_VF_COUNT: long_vf_closeness_count,
helpers.SHORTER_PATH_VF_COUNT: short_vf_closeness_count
}
meta_map[tuple(named_trace)].update(extra_meta)
all_vf += vf_count
all_nonvf += nonvf
all_vf_closeness += vf_closeness_count
all_nonvf_closeness += nonvf_closeness
all_long_vf += long_vf_count
all_long_nonvf += long_nonvf
all_long_vf_closeness += long_vf_closeness_count
all_long_nonvf_closeness += long_nonvf_closeness
all_short_vf += short_vf_count
all_short_nonvf += short_nonvf
all_short_vf_closeness += short_vf_closeness_count
all_short_nonvf_closeness += short_nonvf_closeness
results.append(vf_count / float(len(all_path)))
results3.append(vf_closeness_count / float(len(all_path)))
if len(all_path) > 1: results2.append(vf_count / float(len(all_path)))
long_results.append(long_vf_count / float(len(long_path)))
long_results3.append(long_vf_closeness_count / float(len(long_path)))
if len(long_path) > 1: long_results2.append(long_vf_count / float(len(long_path)))
if len(short_path) > 0:
short_results.append(short_vf_count / float(len(short_path)))
short_results3.append(short_vf_closeness_count / float(len(short_path)))
else:
pass
# short_results.append(0)
# short_results3.append(0)
if len(short_path) > 1: short_results2.append(short_vf_count / float(len(short_path)))
# save mofified meta
meta_mod = [x for x in meta_map.itervalues()]
helpers.save_to_json(out, meta_mod)
# print results
print 'ALL'
print 'VF count: %d' % all_vf
print 'VF CLOSENESS count: %d' % all_vf_closeness
print 'Non vf count: %d' % all_nonvf
print 'Non vf CLOSENESS count: %d' % all_nonvf_closeness
print 'VF perc: %f' % (all_vf/float(all_vf + all_nonvf))
print 'VF CLOSENESS perc: %f' % (all_vf_closeness/float(all_vf_closeness + all_nonvf_closeness))
print 'Mean VF prob: %f' % np.mean(results)
print 'Mean VF CLOSENESS prob: %f' % np.mean(results3)
print 'Mean VF2 prob: %f' % np.mean(results2)
print '=========='
print 'SHORT'
print 'VF count: %d' % all_short_vf
print 'VF CLOSENESS count: %d' % all_short_vf_closeness
print 'Non vf count: %d' % all_short_nonvf
print 'Non vf CLOSENESS count: %d' % all_short_nonvf_closeness
if all_short_vf + all_short_nonvf > 0:
print 'VF perc: %f' % (all_short_vf/float(all_short_vf + all_short_nonvf))
if all_short_vf_closeness + all_short_nonvf_closeness > 0:
print 'VF CLOSENESS perc: %f' % (all_short_vf_closeness/float(all_short_vf_closeness + all_short_nonvf_closeness))
print 'Mean VF prob: %f' % np.mean(short_results)
print 'Mean VF CLOSENESS prob: %f' % np.mean(short_results3)
print 'Mean VF2 prob: %f' % np.mean(short_results2)
print '=-----------------'
print 'LONG'
print 'VF count: %d' % all_long_vf
print 'VF CLOSENESS count: %d' % all_long_vf_closeness
print 'Non vf count: %d' % all_long_nonvf
print 'Non vf CLOSENESS count: %d' % all_long_nonvf_closeness
print 'VF perc: %f' % (all_long_vf/float(all_long_vf + all_long_nonvf))
print 'VF CLOSENESS perc: %f' % (all_long_vf_closeness/float(all_long_vf_closeness + all_long_nonvf_closeness))
print 'Mean VF prob: %f' % np.mean(long_results)
print 'Mean VF CLOSENESS prob: %f' % np.mean(long_results3)
print 'Mean VF2 prob: %f' % np.mean(long_results2)
def purify(labeled_g, out, network_path, extra_hop=0):
vs = [x.index for x in labeled_g.vs]
## Jus like in R
# print '================'
# for x in orig_vs:
# shp = labeled_g.get_all_shortest_paths(x, orig_vs, mode=i.ALL)
# res = []
# mes = 0
# for p in shp:
# mes += 1
# # print [labeled_g.vs[u]['name'] for u in p]
# vf_indicator = 1 if vft.is_valley_free(labeled_g, p) else 0
# # if vf_indicator == 0:
# # print [labeled_g.vs[u]['name'] for u in [p[0], p[-1]]]
# # print [labeled_g.degree(u) for u in p]
# # print vft.trace_to_string(labeled_g, p)
# # print vf_indicator == 1
# res.append(vf_indicator)
# # raw_input()
# # print mes
# print np.mean(res)
# raw_input()
# print '///////////////////////////'
pairs = random_pairs(vs, NODE_PAIRS)
print 'Random pairs: %d' % len(pairs)
probed_pairs = 0
all_vf = 0
all_nonvf = 0
all_vf_closeness = 0
all_nonvf_closeness = 0
results = []
results2 = []
results3 = []
short_results = list()
short_results2 = list()
short_results3 = list()
all_short_vf = 0
all_short_nonvf = 0
all_short_vf_closeness = 0
all_short_nonvf_closeness = 0
long_results = list()
long_results2 = list()
long_results3 = list()
all_long_vf = 0
all_long_nonvf = 0
all_long_vf_closeness = 0
all_long_nonvf_closeness = 0
results_closeness = []
results3_closeness = []
progress = progressbar1.AnimatedProgressBar(end=len(pairs), width=15)
for s, t in pairs:
progress += 1
progress.show_progress()
for x in range(0, labeled_g.vcount()):
labeled_g.vs[x]['traces'] = dict()
# all_path = labeled_g.get_all_shortest_paths(s, t, mode=i.ALL)
sh_len = labeled_g.shortest_paths(s, t, mode=i.ALL)[0][0]
sh_len += 1 # convert to hop count
all_path = helpers.dfs_mark(copy.deepcopy(labeled_g), s, t,
sh_len + extra_hop)
if all_path is None or len(all_path) < 1:
print 'No path between %s %s' % (s, t)
continue
probed_pairs += 1
vf_indicator = [
1 if vft.is_valley_free(labeled_g, x) else 0 for x in all_path
]
vf_closeness_indicator = [
1 if vft.is_valley_free(labeled_g, x, vfmode=vft.ORDER_CLOSENESS)
else 0 for x in all_path
]
vf_count = sum(vf_indicator)
vf_closeness_count = sum(vf_closeness_indicator)
nonvf = len(all_path) - vf_count
nonvf_closeness = len(all_path) - vf_closeness_count
all_vf += vf_count
all_nonvf += nonvf
all_vf_closeness += vf_closeness_count
all_nonvf_closeness += nonvf_closeness
long_path = [x for x in all_path if len(x) == sh_len + extra_hop]
short_path = [x for x in all_path if len(x) < sh_len + extra_hop]
tmp = [
1 if vft.is_valley_free(labeled_g, x, vfmode=vft.ORDER_PRELABELED)
else 0 for x in all_path if len(x) < sh_len + extra_hop
]
short_vf_count = sum(tmp)
short_nonvf = len(tmp) - short_vf_count
tmp = [
1 if vft.is_valley_free(labeled_g, x, vfmode=vft.ORDER_CLOSENESS)
else 0 for x in all_path if len(x) < sh_len + extra_hop
]
short_vf_closeness_count = sum(tmp)
short_nonvf_closeness = len(tmp) - short_vf_closeness_count
tmp = [
1 if vft.is_valley_free(labeled_g, x, vfmode=vft.ORDER_PRELABELED)
else 0 for x in all_path if len(x) >= sh_len + extra_hop
]
long_vf_count = sum(tmp)
long_nonvf = len(tmp) - long_vf_count
tmp = [
1 if vft.is_valley_free(labeled_g, x, vfmode=vft.ORDER_CLOSENESS)
else 0 for x in all_path if len(x) >= sh_len + extra_hop
]
long_vf_closeness_count = sum(tmp)
long_nonvf_closeness = len(tmp) - long_vf_closeness_count
if len(all_path) > 0:
results.append(vf_count / float(len(all_path)))
results_closeness.append(vf_closeness_count / float(len(all_path)))
else:
results.append(0)
results_closeness.append(0)
results3.append([vf_count, nonvf])
results3_closeness.append([vf_closeness_count, nonvf_closeness])
if len(all_path) > 1: results2.append(vf_count / float(len(all_path)))
all_long_vf += long_vf_count
all_long_nonvf += long_nonvf
all_long_vf_closeness += long_vf_closeness_count
all_long_nonvf_closeness += long_nonvf_closeness
all_short_vf += short_vf_count
all_short_nonvf += short_nonvf
all_short_vf_closeness += short_vf_closeness_count
all_short_nonvf_closeness += short_nonvf_closeness
if len(long_path) > 0:
long_results.append(long_vf_count / float(len(long_path)))
long_results3.append(long_vf_closeness_count /
float(len(long_path)))
else:
long_results.append(0)
long_results3.append(0)
if len(long_path) > 1:
long_results2.append(long_vf_count / float(len(long_path)))
if len(short_path) > 0:
short_results.append(short_vf_count / float(len(short_path)))
short_results3.append(short_vf_closeness_count /
float(len(short_path)))
else:
short_results.append(0)
short_results3.append(0)
print
with open(out, 'w') as f:
f.write('%s\n' % network_path)
f.write('Probed pairs: %d\n' % probed_pairs)
f.write('VF count: %d\n' % all_vf)
f.write('Non vf count: %d\n' % all_nonvf)
f.write('VF perc: %f\n' % (all_vf / float(all_vf + all_nonvf)))
f.write('Mean VF prob: %f\n' % np.mean(results))
f.write('Mean VF2 prob: %f\n' % np.mean(results2))
f.write('\n')
f.write('VF CLOSENESS count: %d\n' % all_vf_closeness)
f.write('Non vf CLOSENESS count: %d\n' % all_nonvf_closeness)
f.write(
'VF CLOSENESS perc: %f\n' %
(all_vf_closeness / float(all_vf_closeness + all_nonvf_closeness)))
f.write('Mean VF CLOSENESS prob: %f\n' % np.mean(results_closeness))
f.write('\n')
f.write('==========\n')
f.write('VF count: %d\n' % all_short_vf)
f.write('VF CLOSENESS count: %d\n' % all_short_vf_closeness)
f.write('Non vf count: %d\n' % all_short_nonvf)
f.write('Non vf CLOSENESS count: %d\n' % all_short_nonvf_closeness)
if all_short_vf + all_short_nonvf > 0:
f.write('VF perc: %f\n' %
(all_short_vf / float(all_short_vf + all_short_nonvf)))
if all_short_vf_closeness + all_short_nonvf_closeness > 0:
f.write(
'VF CLOSENESS perc: %f\n' %
(all_short_vf_closeness /
float(all_short_vf_closeness + all_short_nonvf_closeness)))
f.write('Mean VF prob: %f\n' % np.mean(short_results))
f.write('Mean VF CLOSENESS prob: %f\n' % np.mean(short_results3))
f.write('Mean VF2 prob: %f\n' % np.mean(short_results2))
f.write('=-----------------\n')
f.write('VF count: %d\n' % all_long_vf)
f.write('VF CLOSENESS count: %d\n' % all_long_vf_closeness)
f.write('Non vf count: %d\n' % all_long_nonvf)
f.write('Non vf CLOSENESS count: %d\n' % all_long_nonvf_closeness)
f.write('VF perc: %f\n' %
(all_long_vf / float(all_long_vf + all_long_nonvf)))
f.write('VF CLOSENESS perc: %f\n' %
(all_long_vf_closeness /
float(all_long_vf_closeness + all_long_nonvf_closeness)))
f.write('Mean VF prob: %f\n' % np.mean(long_results))
f.write('Mean VF CLOSENESS prob: %f\n' % np.mean(long_results3))
f.write('Mean VF2 prob: %f\n' % np.mean(long_results2))
def filter(g, traceroutes):
results = list()
# remove traces with unknown nodes
traceroutes = vft.trace_in_vertex_id(g, traceroutes)
progress = progressbar1.AnimatedProgressBar(end=len(traceroutes), width=15)
for trace in traceroutes:
progress += 1
progress.show_progress()
if not vft.trace_exists(g, trace):
print 'BUG?'
continue
for x in range(0, g.vcount()):
g.vs[x]['traces'] = dict()
trace = tuple(trace)
s, t = trace[0], trace[-1]
sh_len = g.shortest_paths(s, t, mode=i.ALL)[0][0]
sh_len += 1 # igraph's hop count to node count
all_routes = helpers.dfs_mark(g, s, t, sh_len + 1)
# all_routes2 = helpers.dfs_simple(g, s, t, sh_len + 1, ())
# if set(all_routes) - set(all_routes2) != set(all_routes2) - set(all_routes):
# print 'AJAJAJ'
# print all_routes
# print '----------'
# print all_routes2
sh_routes = [x for x in all_routes if len(x) == sh_len]
all_vf_routes = [x for x in all_routes if vft.is_valley_free(g, x)]
prediction_set = set(sh_routes) | set(all_vf_routes)
result = [
trace,
len(trace),
sh_len,
len(sh_routes),
trace in sh_routes,
len(all_vf_routes),
trace in all_vf_routes,
len(all_routes),
trace in all_routes,
len(prediction_set),
trace in prediction_set,
vft.is_valley_free(g, trace),
# sh_routes, all_vf_routes, all_routes,
vft.trace_to_string(g, trace)
]
results.append(result)
print >> sys.stderr, (
'TRACE\tTRACE_LEN\tSH_LEN',
'\t#SH_ROUTE\tOK',
'\t#ALL_VF\tOK',
'\t#ALL_ROUTE\tOK',
'\t#PREDICTION_SET\tOK',
'\tIS_VF',
# '\tSH_ROUTES\tALL_VF_ROUTES\tALL_ROUTE',
'\tTRACE_STR')
for result in results:
result = [str(r) for r in result]
print >> sys.stderr, '\t'.join(result)
return results
def main():
parser = argparse.ArgumentParser(
description=('SANDBOX mode. ', 'Write something ', 'useful here'),
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('--progressbar', action='store_true')
parser.add_argument('--verbose', '-v', action='count', default=0)
parser.add_argument('--edge-drop',
dest='edge_drop',
type=float,
default=0.0)
parser.add_argument('--closeness-limit',
dest='closeness_limit',
type=float,
default=0.0)
parser.add_argument('network')
parser.add_argument('traceroutes')
arguments = parser.parse_args()
show_progress = arguments.progressbar
arguments.verbose = min(len(helpers.LEVELS), arguments.verbose)
logging.getLogger('compnet').setLevel(helpers.LEVELS[arguments.verbose])
g = helpers.load_network(arguments.network)
traceroutes = helpers.load_from_json(arguments.traceroutes)
logger.info('ecount: %d' % g.ecount())
logger.info('vcount: %d' % g.vcount())
logger.info('trace count: %d' % len(traceroutes))
g_dummy = g.copy()
progress = progressbar1.DummyProgressBar(end=10, width=15)
if show_progress:
progress = progressbar1.AnimatedProgressBar(end=len(traceroutes),
width=15)
closeness_list = []
for x in g_dummy.vs:
progress += 1
progress.show_progress()
closeness_list.append((x.index, g_dummy.closeness(x)))
end = int(arguments.closeness_limit * g_dummy.vcount())
logger.debug('Top node count: %d' % end)
top_nodes = sorted(closeness_list, key=lambda x: x[1], reverse=True)[:end]
top_nodes_index = [x[0] for x in top_nodes]
top_nodes_name = [g_dummy.vs[x[0]]['name'] for x in top_nodes]
top_edges = [
e for e in g_dummy.es
if e.source in top_nodes_index and e.target in top_nodes_index
]
logger.debug('Top edge count: %d' % len(top_edges))
random.shuffle(top_edges)
edge_drop = top_edges[:int(len(top_edges) * arguments.edge_drop)]
logger.debug('Dropped edge count: %d' % len(edge_drop))
# edges = [x.index for x in g_dummy.es]
# random.shuffle(edges)
# edge_drop = edges[:int(g.ecount() * arguments.edge_drop)]
g_dummy.delete_edges(edge_drop)
traceroutes = traceroutes[:10000]
all_edges = []
for trace in traceroutes:
edges = zip(trace, trace[1:])
edges = [tuple(sorted(e)) for e in edges]
all_edges.extend(edges)
all_edges = list(set(all_edges))
top_edges = [
e for e in all_edges
if e[0] in top_nodes_name and e[1] in top_nodes_name
]
logger.info('TOP edge count in real traceroutes: %d' % len(top_edges))
found_top_edges = []
increments = []
for trace in traceroutes:
edges = zip(trace, trace[1:])
edges = [tuple(sorted(e)) for e in edges]
top_edges = [
x for x in edges
if x[0] in top_nodes_name and x[1] in top_nodes_name
]
found_top_edges.extend(top_edges)
found_top_edges = list(set(found_top_edges))
increments.append(len(found_top_edges))
logger.info('Found top edge count: %d' % len(found_top_edges))
dummy_sh_traceroutes_meta = []
original_sh_traceroutes_meta = []
stretches = []
progress = progressbar1.DummyProgressBar(end=10, width=15)
if show_progress:
progress = progressbar1.AnimatedProgressBar(end=len(traceroutes),
width=15)
for trace in traceroutes:
progress += 1
progress.show_progress()
s, t = trace[0], trace[-1]
# logger.debug('Get shortest paths from {s} to {t}'.format(s=s, t=t))
sh_dummy = random.choice(g_dummy.get_shortest_paths(s, t))
sh_original = random.choice(g.get_shortest_paths(s, t))
stretch = len(sh_dummy) - len(sh_original)
dummy_sh_traceroutes_meta.append((sh_dummy, stretch))
original_sh_traceroutes_meta.append((sh_original, 0))
stretches.append(stretch)
# logger.debug('Stretch: %d' % stretch)
# logger.debug('SH DUMMY: %s' % [g_dummy.vs[x]['name'] for x in sh_dummy])
# logger.debug('SH ORIG: %s' % [g.vs[x]['name'] for x in sh_original])
dummy_sh_meta = [(x[0], x[1],
vft.is_valley_free(g_dummy, x[0], vft.CLOSENESS))
for x in dummy_sh_traceroutes_meta]
dummy_sh_len_hist = collections.Counter(
[len(x[0]) for x in dummy_sh_traceroutes_meta])
original_sh_len_hist = collections.Counter(
[len(x[0]) for x in original_sh_traceroutes_meta])
original_len_hist = collections.Counter([len(x) for x in traceroutes])
stretches = [x for x in stretches if x >= 0]
stretch_hist = collections.Counter(stretches)
import matplotlib.pyplot as plt
print
print[(x, 100 * y / float(len(traceroutes)), y)
for x, y in stretch_hist.iteritems()]
plt.plot([x for x in stretch_hist.iterkeys()],
[x for x in stretch_hist.itervalues()], 'g^')
plt.ylabel('some numbers')
# plt.show()
logger.info('Dummy VF stat')
max_stretch = max(dummy_sh_meta, key=lambda x: x[1])[1]
for stretch in range(0, max_stretch + 1):
stretched_traces = [x for x in dummy_sh_meta if x[1] == stretch]
count = len(stretched_traces)
vf_count = len([x for x in stretched_traces if x[2]])
vf_perc = vf_count / float(count)
nonvf_count = count - vf_count
nonvf_perc = nonvf_count / float(count)
logger.info(
'{stretch} -- {vf_perc}[{vf_count}]\t{nonvf_perc}[{nonvf_count}]'.
format(stretch=stretch,
vf_perc=vf_perc,
vf_count=vf_count,
nonvf_perc=nonvf_perc,
nonvf_count=nonvf_count))
import matplotlib.pyplot as plt
plt.plot(increments, 'g^')
plt.ylabel('some numbers')
plt.show()
def random_nonvf_route(g, s, t, hop_count,
path=None, vfmode=None):
if path is None:
try:
if isinstance(s, str):
s = g.vs.find(s).index
if isinstance(t, str):
t = g.vs.find(t).index
# check if s and t are valid inidecies
_, _ = g.vs[s], g.vs[t]
except (ValueError, IndexError):
raise IndexError('Vertex index out of range or not exists')
# some special case
if hop_count < 1:
# print 'HOP COUNT: %d' % hop_count
return (False, [])
if s == t: return (True, [s, ])
# if s != t then length must be larger than 1
if hop_count == 1:
# print 'S: %s, T: %s, HC: %d' % (s, t, hop_count)
return (False, [])
shortest_route = g.shortest_paths(s, t, mode=i.OUT)[0][0] + 1
if hop_count < shortest_route:
# print 'TOO SHORT %d' % (hop_count)
return (False, [])
path = [s, ]
hop_count -= 1
if vfmode is None: vfmode = vft.CLOSENESS
if s == t:
return (vft.is_valley_free(g, path, vfmode=vfmode), path)
logger.debug('Hop count remained: %d' % hop_count)
logger.debug('Current node: %s' % s)
neighbors = [x for x in g.neighbors(s, mode=i.OUT) if x not in path]
distances = [x[0] for x in g.shortest_paths(neighbors, t, mode=i.OUT)]
candidates = filter(lambda x: x[1] + 1 <= hop_count, # +, mert az igraph
# azt mondja meg,
# s-bol hany hopp t
zip(neighbors, distances))
weights = [-1 if not vft.is_valley_free(g, path + [x[0], ], vfmode=vfmode)
else vft.edge_dir(g, [s, x[0]], vfmode=vfmode).value
for x in candidates]
# create a list where every columnt is neighbors, distances, weights
# respectevly
candidates = zip(*(zip(*candidates) + [weights, ]))
# sort by weights
candidates = sorted(candidates, key=itemgetter(2))
if len(candidates) == 0:
return (False, [])
logger.debug('Valid candidates: %s' % candidates)
first_route = (False, []) # by default there was no route to T
for next_hop in candidates:
logger.debug('Chosen one: %s' % next_hop[0])
isvf, r = random_nonvf_route(g, next_hop[0], t,
hop_count - 1,
path + [next_hop[0], ], vfmode)
if len(r) == 0: continue
if not isvf:
# we are done, we found a nonVF route
return (isvf, r)
# our first guess a vf route. save it for later use (e.g. there is no
# nonVF rotue) but lets try again with another candidate
if len(first_route[1]) == 0: # first save
first_route = (isvf, r)
return first_route
def purify(g,
meta_original,
out,
count=1000,
try_per_race=1,
show_progress=False,
with_lp=True):
empty = 0
# remove traces with already calculated random paths
logger.warn('[r]ONLY NOT FILLED PATHS[/]')
meta_filled = [
x for x in meta_original if helpers.RANDOM_WALK_RUN_COUNT not in x
]
# Filter if interested only in routes of stretch 1
# meta_filled = [x for x in meta_original
# if x[helpers.TRACE_LEN]-x[helpers.SH_LEN] == 1]
## traces with a maximum stretch
# logger.warn('[r]!!!ONLY WITH STRETCH[/]')
# meta = [x for x in meta if x[helpers.STRETCH] > -1]
# # shorter meta records
# logger.warn('[r]!!!ONLY SHORT TRACES[/]')
# meta = [x for x in meta if len(x[helpers.TRACE]) < 5]
# meta_map = {tuple(x[helpers.TRACE]): x for x in meta_filled}
logger.info('All trace count: %d' % len(meta_filled))
tr_count = min(len(meta_filled), count)
meta_random = random.sample(meta_filled, tr_count)
logger.info('Chosen subset count: %d' % len(meta_random))
# real_vf_degree = [x for x in meta_random if x[helpers.IS_VF_DEGREE] == 1]
# real_nonvf_degree = [x for x in meta_random if x[helpers.IS_VF_DEGREE] == 0]
# assert len(real_nonvf_degree) == tr_count - len(real_vf_degree)
# real_vf_prelabeled = [x for x in meta_random if x[helpers.IS_VF_PRELABELED] == 1]
# real_nonvf_prelabeled = [x for x in meta_random if x[helpers.IS_VF_PRELABELED] == 0]
# assert len(real_nonvf_prelabeled) == tr_count - len(real_vf_prelabeled)
# real_vf_closeness = [x for x in meta_random if x[helpers.IS_VF_CLOSENESS] == 1]
# real_nonvf_closeness = [x for x in meta_random if x[helpers.IS_VF_CLOSENESS] == 0]
# assert len(real_nonvf_closeness) == tr_count - len(real_vf_closeness)
# logger.info('Real vf degree: %f[%d]' % ((len(real_vf_degree) / float(tr_count),
# len(real_vf_degree))))
# logger.info('Real nonvf degree: %f[%d]' % ((len(real_nonvf_degree) / float(tr_count),
# len(real_nonvf_degree))))
# logger.info('Real vf prelabeled: %f[%d]' % ((len(real_vf_prelabeled) / float(tr_count),
# len(real_vf_prelabeled))))
# logger.info('Real nonvf prelabeled: %f[%d]' % ((len(real_nonvf_prelabeled) / float(tr_count),
# len(real_nonvf_prelabeled))))
# logger.info('Real vf closeness: %f[%d]' % ((len(real_vf_closeness)/float(tr_count), len(real_vf_closeness))))
# logger.info('Real nonvf closeness: %f[%d]' % ((len(real_nonvf_closeness)/float(tr_count), len(real_nonvf_closeness))))
# traceroutes = [x[helpers.TRACE] for x in meta_random]
# traceroutes = vft.trace_in_vertex_id(g, traceroutes)
try:
meta_random[0][helpers.TRACE]
except Exception:
meta_random = [{helpers.TRACE: x} for x in meta_random]
progress = progressbar1.DummyProgressBar(end=10, width=15)
if show_progress:
progress = progressbar1.AnimatedProgressBar(end=len(meta_random),
width=15)
stretch_list = []
max_stretch = max(
[x[helpers.TRACE_LEN] - x[helpers.SH_LEN] for x in meta_random])
for stretch in range(0, max_stretch + 1):
metas = [
x for x in meta_random
if x[helpers.TRACE_LEN] - x[helpers.SH_LEN] == stretch
]
stretch_list.extend(list(repeat(stretch, len(metas))))
# print(stretch_list)
lenghts = random.shuffle(stretch_list)
strx_array = []
for idx, trace_meta in enumerate(meta_random):
progress += 1
progress.show_progress()
# print(trace_meta[helpers.TRACE])
shl = trace_meta[helpers.SH_LEN]
trace = vft.trace_in_vertex_id(g, [
trace_meta[helpers.TRACE],
])
if len(trace) != 1:
print 'PROBLEM'
print trace_meta
continue
trace = trace[0]
# print(trace)
random_walk_closeness_route_vf = 0
random_walk_closeness_route_lp_soft = 0
random_walk_closeness_route_lp_hard = 0
random_walk_degree_route_vf = 0
random_walk_degree_route_lp_soft = 0
random_walk_degree_route_lp_hard = 0
random_walk_prelabeled_route_vf = 0
random_walk_prelabeled_route_lp_soft = 0
random_walk_prelabeled_route_lp_hard = 0
s, t = trace[0], trace[-1]
for counter in xrange(0, try_per_race):
# random_path = helpers.random_route_walk(g, s, t, len(trace)) # Modified
random_path = helpers.random_route_walk(
g, s, t, shl + stretch_list[idx]) # Modified
if len(random_path) == 0:
empty += 1
if vft.is_valley_free(g, random_path, vfmode=vft.CLOSENESS):
random_walk_closeness_route_vf += 1
if (len(random_path) == shl + 1):
strx_array.append(1)
if with_lp:
lp_soft = vft.is_local_preferenced(g,
random_path,
first_edge=True,
vfmode=vft.CLOSENESS)
lp_hard = vft.is_local_preferenced(g,
random_path,
first_edge=False,
vfmode=vft.CLOSENESS)
if lp_soft:
random_walk_closeness_route_lp_soft += 1
if lp_hard:
random_walk_closeness_route_lp_hard += 1
else:
if (len(random_path) == shl + 1):
strx_array.append(0)
# if vft.is_valley_free(g, random_path, vfmode=vft.DEGREE):
# random_walk_degree_route_vf += 1
# if with_lp:
# lp_soft = vft.is_local_preferenced(g, random_path,
# first_edge=True,
# vfmode=vft.DEGREE)
# lp_hard = vft.is_local_preferenced(g, random_path,
# first_edge=False,
# vfmode=vft.DEGREE)
# if lp_soft:
# random_walk_degree_route_lp_soft += 1
# if lp_hard:
# random_walk_degree_route_lp_hard += 1
# if vft.is_valley_free(g, random_path, vfmode=vft.PRELABELED):
# random_walk_prelabeled_route_vf += 1
# if with_lp:
# lp_soft = vft.is_local_preferenced(g, random_path,
# first_edge=True,
# vfmode=vft.PRELABELED)
# lp_hard = vft.is_local_preferenced(g, random_path,
# first_edge=False,
# vfmode=vft.PRELABELED)
# if lp_soft:
# random_walk_prelabeled_route_lp_soft += 1
# if lp_hard:
# random_walk_prelabeled_route_lp_hard += 1
# sanity check
# if random_path[0] != s or random_path[-1] != t:
# logger.error('ALERT')
if len(random_path) != len(set(random_path)):
logger.error('LENGTH ERROR')
extra_meta = {
helpers.RANDOM_WALK_RUN_COUNT:
try_per_race,
helpers.RANDOM_WALK_VF_CLOSENESS_ROUTE:
random_walk_closeness_route_vf,
helpers.RANDOM_WALK_VF_DEGREE_ROUTE:
random_walk_degree_route_vf,
helpers.RANDOM_WALK_VF_PRELABELED_ROUTE:
random_walk_prelabeled_route_vf,
}
if with_lp:
extra_meta.update({
helpers.RANDOM_WALK_LP_SOFT_CLOSENESS_ROUTE:
random_walk_closeness_route_lp_soft,
helpers.RANDOM_WALK_LP_HARD_CLOSENESS_ROUTE:
random_walk_closeness_route_lp_hard,
helpers.RANDOM_WALK_LP_SOFT_DEGREE_ROUTE:
random_walk_degree_route_lp_soft,
helpers.RANDOM_WALK_LP_HARD_DEGREE_ROUTE:
random_walk_degree_route_lp_hard,
helpers.RANDOM_WALK_LP_SOFT_PRELABELED_ROUTE:
random_walk_prelabeled_route_lp_soft,
helpers.RANDOM_WALK_LP_HARD_PRELABELED_ROUTE:
random_walk_prelabeled_route_lp_hard
})
trace_meta.update(extra_meta)
## save modified meta
# all meta_* get only references from meta_original
helpers.save_to_json(out, meta_random)
# meta_mod = [x for x in meta_map.itervalues()]
# helpers.save_to_json(out, meta_mod)
# calculate results
# real_vf = [x[helpers.IS_VF_CLOSENESS] for x in meta_random]
# real_vf_ratio = np.mean(real_vf)
random_walk_vf_ratio_per_element = [
x[helpers.RANDOM_WALK_VF_CLOSENESS_ROUTE] /
x[helpers.RANDOM_WALK_RUN_COUNT] for x in meta_random
]
random_walk_vf_ratio = np.mean(random_walk_vf_ratio_per_element)
# print results
logger.info('')
logger.info('Empty: %d' % empty)
logger.info('Tested trace count: %d' % len(meta_random))
# logger.info('VF ratio in tested traces: %f' % real_vf_ratio)
logger.info('VF ratio in random walks: %f' % random_walk_vf_ratio)
logger.info('VF ratio in random walks for path stretch 1: %f' %
np.mean(strx_array))
def purify(g, traceroutes, flags, show_progress=False):
results = list()
# remove traces with unknown nodes
traceroutes = vft.trace_in_vertex_id(g, traceroutes)
# generate valley-free graph
if flags[FLAG_PRELABELED]:
logger.info('Generate VF_G_PRE')
vf_g_pre = vft.convert_to_vf(g, vfmode=vft.PRELABELED)
else:
logger.info('Skip prelabeled graph')
if flags[FLAG_DEGREE]:
logger.info('Generate VF_G_DEGREE')
vf_g_degree = vft.convert_to_vf(g, vfmode=vft.DEGREE)
else:
logger.info('Skip degree graph')
if flags[FLAG_CLOSENESS]:
logger.info('Generate VF_G_CLOSENESS')
vf_g_closeness = vft.convert_to_vf(g, vfmode=vft.CLOSENESS)
else:
logger.info('Skip closeness graph')
progress = progressbar1.DummyProgressBar(end=10, width=15)
if show_progress:
progress = progressbar1.AnimatedProgressBar(end=len(traceroutes),
width=15)
for trace in traceroutes:
progress += 1
progress.show_progress()
logger.debug('Current trace: %s' % ([g.vs[x]['name'] for x in trace]))
if len(trace) == 1: continue
s, t = trace[0], trace[-1]
is_vf_prelabeled = -1
is_lp_prelabeled_hard = -1
is_lp_prelabeled_soft = -1
is_vf_degree = -1
is_lp_degree_hard = -1
is_lp_degree_soft = -1
is_vf_closeness = -1
is_lp_closeness_hard = -1
is_lp_closeness_soft = -1
trace_len = len(trace)
sh_len = g.shortest_paths(s, t, mode=i.OUT)[0][0]
sh_len += 1 # convert hop count to node Counter
if flags[FLAG_PRELABELED]:
is_vf_prelabeled = vft.is_valley_free(g, trace, vft.PRELABELED)
is_vf_prelabeled = int(is_vf_prelabeled)
if is_vf_prelabeled:
if flags[FLAG_LP_SOFT]:
lp_soft = vft.is_local_preferenced(g,
trace,
vf_g=vf_g_pre,
first_edge=True,
vfmode=vft.PRELABELED)
is_lp_prelabeled_soft = 1 if lp_soft else 0
else:
is_lp_prelabeled_soft = -1
if flags[FLAG_LP_HARD]:
lp_hard = vft.is_local_preferenced(g,
trace,
vf_g=vf_g_pre,
first_edge=False,
vfmode=vft.PRELABELED)
is_lp_prelabeled_hard = 1 if lp_hard else 0
else:
is_lp_prelabeled_hard = -1
if flags[FLAG_DEGREE]:
is_vf_degree = vft.is_valley_free(g, trace, vft.DEGREE)
is_vf_degree = int(is_vf_degree)
if is_vf_degree:
if flags[FLAG_LP_SOFT]:
lp_soft = vft.is_local_preferenced(g,
trace,
vf_g=vf_g_degree,
first_edge=True,
vfmode=vft.DEGREE)
is_lp_degree_soft = 1 if lp_soft else 0
else:
is_lp_degree_soft = -1
if flags[FLAG_LP_HARD]:
lp_hard = vft.is_local_preferenced(g,
trace,
vf_g=vf_g_degree,
first_edge=False,
vfmode=vft.DEGREE)
is_lp_degree_hard = 1 if lp_hard else 0
else:
is_lp_degree_hard = -1
if flags[FLAG_CLOSENESS]:
is_vf_closeness = vft.is_valley_free(g, trace, vft.CLOSENESS)
is_vf_closeness = int(is_vf_closeness)
if is_vf_closeness:
if flags[FLAG_LP_SOFT]:
lp_soft = vft.is_local_preferenced(g,
trace,
vf_g=vf_g_closeness,
first_edge=True,
vfmode=vft.CLOSENESS)
is_lp_closeness_soft = 1 if lp_soft else 0
else:
is_lp_closeness_soft = -1
if flags[FLAG_LP_HARD]:
lp_hard = vft.is_local_preferenced(g,
trace,
vf_g=vf_g_closeness,
first_edge=False,
vfmode=vft.CLOSENESS)
is_lp_closeness_hard = 1 if lp_hard else 0
else:
is_lp_closeness_hard = -1
if False:
sh_vf_len = vft.get_shortest_vf_route(g,
s,
t,
mode='vf',
vf_g=vf_g_pre,
_all=True,
vfmode=vft.PRELABELED)
# ugy tunik, mintha nem mindig lenne pontos? fentartassal kezelendo
# ez az ertek azert is kerult bele, hogy ellenorizzuk
in_vf_prediction = 1 if sh_vf_len and trace in sh_vf_len else 0
else:
sh_vf_len = -1
in_vf_prediction = -1
sh_vf_len = len(sh_vf_len[0]) if isinstance(sh_vf_len, list) else -1
percentage_stretch = trace_len / float(sh_len)
named_trace = [g.vs[_id]["name"] for _id in trace]
result = {
helpers.TRACE: named_trace,
helpers.TRACE_LEN: trace_len,
helpers.SH_LEN: sh_len,
helpers.SH_VF_LEN: sh_vf_len,
helpers.IS_VF_PRELABELED: is_vf_prelabeled,
helpers.IS_VF_DEGREE: is_vf_degree,
helpers.IS_VF_CLOSENESS: is_vf_closeness,
helpers.HOP_STRETCH: trace_len - sh_len,
helpers.PERC_STRETCH: percentage_stretch,
helpers.IN_VF_PRED: in_vf_prediction,
helpers.IS_LP_SOFT_PRELABELED: is_lp_prelabeled_soft,
helpers.IS_LP_HARD_PRELABELED: is_lp_prelabeled_hard,
helpers.IS_LP_SOFT_DEGREE: is_lp_degree_soft,
helpers.IS_LP_HARD_DEGREE: is_lp_degree_hard,
helpers.IS_LP_SOFT_CLOSENESS: is_lp_closeness_soft,
helpers.IS_LP_HARD_CLOSENESS: is_lp_closeness_hard,
}
results.append(result)
# print >> sys.stderr, ('TRACE\tTRACE_LEN\tSH_LEN\tSH_VF_LEN\tIS_VF',
# '\tSTRETCH\tIN_VF_PREDICTION\tIS_LP_F\tIS_LP_ALL')
# for result in results:
# result = [str(r) for r in result]
# print >> sys.stderr, '\t'.join(result)
# statistic = statistics.purify(g, results,
# 'nc+ec+tc+rt+vf+vf_closeness+pred+lp_soft_prelabeled+lp_hard_prelabeled+lp_soft_degree+lp_hard_degree+lp_soft_closeness+lp_hard_closeness'.split('+'))
# statistics.stat_printer(statistic)
return results