mirror of
https://github.com/gsi-upm/soil
synced 2024-11-14 15:32:29 +00:00
164 lines
5.3 KiB
Python
164 lines
5.3 KiB
Python
from unittest import TestCase
|
|
import pytest
|
|
|
|
from soil import agents, environment
|
|
from soil import time as stime
|
|
|
|
|
|
class Dead(agents.FSM):
|
|
@agents.default_state
|
|
@agents.state
|
|
def only(self):
|
|
return self.die()
|
|
|
|
|
|
class TestAgents(TestCase):
|
|
def test_die_returns_infinity(self):
|
|
"""The last step of a dead agent should return time.INFINITY"""
|
|
d = Dead(unique_id=0, model=environment.Environment())
|
|
ret = d.step()
|
|
assert ret == stime.NEVER
|
|
|
|
def test_die_raises_exception(self):
|
|
"""A dead agent should raise an exception if it is stepped after death"""
|
|
d = Dead(unique_id=0, model=environment.Environment())
|
|
d.step()
|
|
with pytest.raises(stime.DeadAgent):
|
|
d.step()
|
|
|
|
def test_agent_generator(self):
|
|
"""
|
|
The step function of an agent could be a generator. In that case, the state of the
|
|
agent will be resumed after every call to step.
|
|
"""
|
|
a = 0
|
|
|
|
class Gen(agents.BaseAgent):
|
|
def step(self):
|
|
nonlocal a
|
|
for i in range(5):
|
|
yield
|
|
a += 1
|
|
|
|
e = environment.Environment()
|
|
g = Gen(model=e, unique_id=e.next_id())
|
|
e.schedule.add(g)
|
|
|
|
for i in range(5):
|
|
e.step()
|
|
assert a == i
|
|
|
|
def test_state_decorator(self):
|
|
class MyAgent(agents.FSM):
|
|
run = 0
|
|
|
|
@agents.default_state
|
|
@agents.state("original")
|
|
def root(self):
|
|
self.run += 1
|
|
return self.other
|
|
|
|
@agents.state
|
|
def other(self):
|
|
self.run += 1
|
|
|
|
e = environment.Environment()
|
|
a = MyAgent(model=e, unique_id=e.next_id())
|
|
a.step()
|
|
assert a.run == 1
|
|
a.step()
|
|
|
|
def test_broadcast(self):
|
|
"""
|
|
An agent should be able to broadcast messages to every other agent, AND each receiver should be able
|
|
to process it
|
|
"""
|
|
|
|
class BCast(agents.Evented):
|
|
pings_received = 0
|
|
|
|
def step(self):
|
|
print(self.model.broadcast)
|
|
try:
|
|
self.model.broadcast("PING")
|
|
except Exception as ex:
|
|
print(ex)
|
|
while True:
|
|
self.check_messages()
|
|
yield
|
|
|
|
def on_receive(self, msg, sender=None):
|
|
self.pings_received += 1
|
|
|
|
e = environment.EventedEnvironment()
|
|
|
|
for i in range(10):
|
|
e.add_agent(agent_class=BCast)
|
|
e.step()
|
|
pings_received = lambda: [a.pings_received for a in e.agents]
|
|
assert sorted(pings_received()) == list(range(1, 11))
|
|
e.step()
|
|
assert all(x == 10 for x in pings_received())
|
|
|
|
def test_ask_messages(self):
|
|
"""
|
|
An agent should be able to ask another agent, and wait for a response.
|
|
"""
|
|
|
|
# There are two agents, they try to send pings
|
|
# This is arguably a very contrived example. In practice, the or
|
|
# There should be a delay of one step between agent 0 and 1
|
|
# On the first step:
|
|
# Agent 0 sends a PING, but blocks before a PONG
|
|
# Agent 1 detects the PING, responds with a PONG, and blocks after its own PING
|
|
# After that step, every agent can both receive (there are pending messages) and send.
|
|
# In each step, for each agent, one message is sent, and another one is received
|
|
# (although not necessarily in that order).
|
|
|
|
# Results depend on ordering (agents are normally shuffled)
|
|
# so we force the timedactivation not to be shuffled
|
|
|
|
pings = []
|
|
pongs = []
|
|
responses = []
|
|
|
|
class Ping(agents.EventedAgent):
|
|
def step(self):
|
|
target_id = (self.unique_id + 1) % self.count_agents()
|
|
target = self.model.agents[target_id]
|
|
print("starting")
|
|
while True:
|
|
if pongs or not pings: # First agent, or anyone after that
|
|
pings.append(self.now)
|
|
response = yield target.ask("PING")
|
|
responses.append(response)
|
|
else:
|
|
print("NOT sending ping")
|
|
print("Checking msgs")
|
|
# Do not block if we have already received a PING
|
|
if not self.check_messages():
|
|
yield self.received()
|
|
print("done")
|
|
|
|
def on_receive(self, msg, sender=None):
|
|
if msg == "PING":
|
|
pongs.append(self.now)
|
|
return "PONG"
|
|
raise Exception("This should never happen")
|
|
|
|
e = environment.EventedEnvironment(schedule_class=stime.OrderedTimedActivation)
|
|
for i in range(2):
|
|
e.add_agent(agent_class=Ping)
|
|
assert e.now == 0
|
|
|
|
for i in range(5):
|
|
e.step()
|
|
time = i + 1
|
|
assert e.now == time
|
|
assert len(pings) == 2 * time
|
|
assert len(pongs) == (2 * time) - 1
|
|
# Every step between 0 and t appears twice
|
|
assert sum(pings) == sum(range(time)) * 2
|
|
# It is the same as pings, without the leading 0
|
|
assert sum(pongs) == sum(range(time)) * 2
|