def get_state_probs(self):
"""
Returns the join probability of all next-calls and the number of
probabilities counted.
"""
js_events = sal.get_calls(seq=self.js)
app = self.parent()
for row in app.distribution_state_iter(self.spec, js_events, cache=app.cache):
yield app.dist_adapter(row)
def get_term_frequency(doc, nprocs, min_seq_len=1):
result = State()
with common.finish(
concurrent.futures.ThreadPoolExecutor(max_workers=nprocs),
accumulator=result) as executor:
for pkg in sal.get_packages(doc=doc):
for seq in sal.get_sequences(pkg=pkg):
if len(sal.get_calls(seq=seq)) >= min_seq_len:
executor.submit(seq_term_frequency, seq)
return result.get()
def filter_anomalies(self, seq):
calls = sal.get_calls(seq=seq)
for evt, call in zip(calls, get_state_probs(self.app, self.spec, calls)):
for idx, st in get_anomalous_states(call, self.threshold):
if not self.accept_state(st):
continue
loc = sal.get_call_location(evt)
cid = (loc, call.name, idx)
if cid in self.visited:
continue
self.visited.add(cid)
yield loc, call.name, idx, st.name
def seq_term_frequency(seq):
c = Counter()
for call in sal.get_calls(seq=seq):
c[call['call']] += 1
return c
def call_dist(self):
js_events = sal.get_calls(seq=self.js)
app = self.parent()
return app.distribution_call_iter(self.spec,
js_events,
cache=app.cache)