1
0
mirror of https://github.com/gsi-upm/soil synced 2024-11-13 06:52:28 +00:00
soil/tests/test_analysis.py
J. Fernando Sánchez 6c4f44b4cb Partial MESA compatibility and several fixes
Documentation for the new APIs is still a work in progress :)
2021-10-15 20:16:49 +02:00

91 lines
2.5 KiB
Python

from unittest import TestCase
import os
import pandas as pd
import yaml
from functools import partial
from os.path import join
from soil import simulation, analysis, agents
ROOT = os.path.abspath(os.path.dirname(__file__))
class Ping(agents.FSM):
defaults = {
'count': 0,
}
@agents.default_state
@agents.state
def even(self):
self.debug(f'Even {self["count"]}')
self['count'] += 1
return self.odd
@agents.state
def odd(self):
self.debug(f'Odd {self["count"]}')
self['count'] += 1
return self.even
class TestAnalysis(TestCase):
# Code to generate a simple sqlite history
def setUp(self):
"""
The initial states should be applied to the agent and the
agent should be able to update its state."""
config = {
'name': 'analysis',
'seed': 'seed',
'network_params': {
'generator': 'complete_graph',
'n': 2
},
'agent_type': Ping,
'states': [{'interval': 1}, {'interval': 2}],
'max_time': 30,
'num_trials': 1,
'environment_params': {
}
}
s = simulation.from_config(config)
self.env = s.run_simulation(dry_run=True)[0]
def test_saved(self):
env = self.env
assert env.get_agent(0)['count', 0] == 1
assert env.get_agent(0)['count', 29] == 30
assert env.get_agent(1)['count', 0] == 1
assert env.get_agent(1)['count', 29] == 15
assert env['env', 29, None]['SEED'] == env['env', 29, 'SEED']
def test_count(self):
env = self.env
df = analysis.read_sql(env._history.db_path)
res = analysis.get_count(df, 'SEED', 'state_id')
assert res['SEED'][self.env['SEED']].iloc[0] == 1
assert res['SEED'][self.env['SEED']].iloc[-1] == 1
assert res['state_id']['odd'].iloc[0] == 2
assert res['state_id']['even'].iloc[0] == 0
assert res['state_id']['odd'].iloc[-1] == 1
assert res['state_id']['even'].iloc[-1] == 1
def test_value(self):
env = self.env
df = analysis.read_sql(env._history.db_path)
res_sum = analysis.get_value(df, 'count')
assert res_sum['count'].iloc[0] == 2
import numpy as np
res_mean = analysis.get_value(df, 'count', aggfunc=np.mean)
assert res_mean['count'].iloc[15] == (16+8)/2
res_total = analysis.get_majority(df)
res_total['SEED'].iloc[0] == self.env['SEED']