def test_npoints():
ts = traces.TimeSeries()
ts[0] = 4
ts[1] = 2
ts[2] = 1
ts[5] = 2
ts[8] = 4
nose.tools.eq_(ts.n_points(), 5)
nose.tools.eq_(
ts.n_points(start=0, end=8, include_start=False, include_end=False), 3)
nose.tools.eq_(
ts.n_points(start=0, end=8, include_start=False, include_end=True), 4)
nose.tools.eq_(
ts.n_points(start=0, end=8, include_start=True, include_end=False), 4)
nose.tools.eq_(
ts.n_points(start=0, end=8, include_start=True, include_end=True), 5)
nose.tools.eq_(
ts.n_points(start=1, end=8, include_start=False, include_end=False), 2)
nose.tools.eq_(
ts.n_points(start=1, end=8, include_start=False, include_end=True), 3)
nose.tools.eq_(
ts.n_points(start=1, end=8, include_start=True, include_end=False), 3)
nose.tools.eq_(
ts.n_points(start=1, end=8, include_start=True, include_end=True), 4)
ts = traces.TimeSeries()
nose.tools.eq_(ts.n_points(), 0)
nose.tools.eq_(ts.n_points(include_start=False), 0)
nose.tools.eq_(ts.n_points(include_end=False), 0)
def test_traces(self):
ts = traces.TimeSeries(default=0)
ts[datetime(2042, 2, 1, 6, 0, 0)] = 10
ts[datetime(2042, 2, 1, 7, 45, 56)] = 10
print(ts)
f = ts[datetime(2042, 2, 1, 6, 0, 0)]
f2 = ts[datetime(2042, 2, 1, 6, 10, 0)]
print(f, f2)
# del ts[ts.first_key()]
print(ts)
ts2 = traces.TimeSeries(default=0)
ts2[datetime(2042, 2, 1, 5, 0, 0)] = 5
ts2[datetime(2042, 2, 1, 6, 45, 0)] = 5
ts2[datetime(2042, 2, 1, 7, 40, 0)] = 2
ts2[datetime(2042, 2, 1, 8, 0, 0)] = 10
# h = traces.Histogram().median()
count = ts * ts2
# d = count.plot(figure_width=6,interpolate=)
count.compact()
count.plot()
l = list(count.iterintervals(n=1))
# count = traces.TimeSeries.merge([ts, ts2], operation=sum)
# hist = count.distribution()
# a, b= hist.items()
# plt.hist(hist.items())
# plt.show()
self.assertEqual(count[datetime(2042, 2, 1, 7, 50)], 20)
self.assertEqual(count[datetime(2042, 2, 1, 8)], 100)
def test_mean_interpolate():
ts = traces.TimeSeries()
ts[0] = 0
ts[1] = 0
ts[3] = 20
nose.tools.assert_almost_equal(
ts.mean(0, 2, interpolate='linear'),
2.5,
)
assert ts.mean(0, 2, interpolate='linear') == 2.5
mask = traces.TimeSeries(default=False)
mask[0] = True
mask[0.5] = False
mask[1] = True
mask[3] = False
nose.tools.assert_almost_equal(
ts.mean(0, 2, mask=mask, interpolate='linear'),
10/3.0,
)
nose.tools.assert_almost_equal(
ts.mean(0, 3, mask=mask, interpolate='linear'),
8.0,
)
def test_encode_refuted_rec_sync():
from magnum.examples.feasible_example2 import feasible_example as g
from magnum.solvers import smt
from magnum.solvers import milp
from stl.boolean_eval import pointwise_sat
dt = g.model.dt
refuted = {'u': traces.TimeSeries([(0, 0.5)])}
phi = encode_refuted_rec(refuted, 0.1, g.times, dt=dt)
g = bind(g).specs.learned.set(phi)
res = smt.encode_and_run(g)
assert pointwise_sat(phi, dt=dt)(res.solution)
res = milp.encode_and_run(g)
assert pointwise_sat(phi, dt=dt)(res.solution)
refuted = {'u': traces.TimeSeries([(0, 1), (0.4, 1), (1, 0)])}
phi = encode_refuted_rec(refuted, 0.1, g.times, dt=dt)
g = bind(g).specs.learned.set(phi)
res = smt.encode_and_run(g)
assert pointwise_sat(phi, dt=dt)(res.solution)
res = milp.encode_and_run(g)
assert pointwise_sat(phi, dt=dt)(res.solution)
def test_bin():
# make some random dates
start = datetime.datetime(2018, 12, 13, 7, 43, 15)
end = datetime.datetime(2019, 2, 3, 8, 45, 10)
# make a timeseries
span = end-start
ts = traces.TimeSeries()
ts[start-span/2] = 2
ts[start] = 12
ts[start+span/3] = 5
ts[end - span/4] = 14
ts[end+span] = None
# nose.tools.assert_raises(KeyError, ts.bin, 'days')
# make a mask
mask = traces.TimeSeries(default=False)
mask[start] = True
mask[end] = False
mask[start + 3*span/10] = False
mask[start + 5*span/10] = True
binned = ts.bin('weeks', mask=mask)
first = binned.peekitem(0)
last = binned.peekitem()
assert len(binned) == 7
assert first[0] == datetime.datetime(2018, 12, 10, 0, 0)
assert last[0] == datetime.datetime(2019, 1, 21, 0, 0)
assert int(last[1][5]) == 30581
def test_traces_am(self):
ts = traces.TimeSeries(default=0)
ts2 = traces.TimeSeries(default=0)
ts.set_interval(start=datetime(year=2000, month=1, day=1, hour=4), end=datetime(year=2000, month=1, day=1, hour=5), value=5)
ts.set_interval(start=datetime(year=2000, month=1, day=1, hour=5), end=datetime(year=2000, month=1, day=1, hour=6), value=10)
ts.set_interval(start=datetime(year=2000, month=1, day=1, hour=6), end=datetime(year=2000, month=1, day=1, hour=7), value=3)
ts2.set_interval(start=datetime(year=2000, month=1, day=1, hour=3, minute=30), end=datetime(year=2000, month=1, day=1, hour=4, minute=30), value=8)
ts2.set_interval(start=datetime(year=2000, month=1, day=1, hour=4, minute=30), end=datetime(year=2000, month=1, day=1, hour=5, minute=30), value=5)
ts2.set_interval(start=datetime(year=2000, month=1, day=1, hour=5, minute=30), end=datetime(year=2000, month=1, day=1, hour=6, minute=30), value=3)
ts2.set_interval(start=datetime(year=2000, month=1, day=1, hour=6, minute=30), end=datetime(year=2000, month=1, day=1, hour=7, minute=30), value=5)
# ts.plot()
# ts2.plot()
ts3 = ts * ts2
# ts3.plot()
# plt.show()
self.assertEqual(ts3[datetime(year=2000, month=1, day=1, hour=3, minute=50)], 0)
self.assertEqual(ts3[datetime(year=2000, month=1, day=1, hour=4, minute=10)], 40)
self.assertEqual(ts3[datetime(year=2000, month=1, day=1, hour=4, minute=40)], 25)
self.assertEqual(ts3[datetime(year=2000, month=1, day=1, hour=5, minute=10)], 50)
self.assertEqual(ts3[datetime(year=2000, month=1, day=1, hour=5, minute=40)], 30)
self.assertEqual(ts3[datetime(year=2000, month=1, day=1, hour=6, minute=10)], 9)
self.assertEqual(ts3[datetime(year=2000, month=1, day=1, hour=6, minute=40)], 15)
self.assertEqual(ts3[datetime(year=2000, month=1, day=1, hour=7, minute=10)], 0)
def test_missing():
"""example code for dealing with missing datapoints"""
router_a = traces.TimeSeries([(-10, 0), (-7, 1), (-5, None), (0, 3),
(1, 3), (5, None)])
assert router_a[-6] == 1
assert router_a[-15] is None
assert not router_a[15]
router_b = traces.TimeSeries([(-8, 0), (-5, 0), (-2, 1), (5, 3)],
default=0)
assert router_b[7] == 3
# a separate signal tells us that the router went down at -10, -1 and 10
# we have no extra data about when it came online (other than the readings)
for timestamp in [-10, -1, 10]:
if timestamp < router_b.first_key():
router_b.default = None
for start, end, value in router_b.iterperiods():
if timestamp >= start and timestamp < end:
router_b[timestamp] = None
if timestamp >= router_b.last_key():
router_b[timestamp] = None
assert router_b[-15] is None
assert router_b[-6] == 0
assert router_b[0] is None
assert router_b[7] == 3
router_list = [router_a, router_b]
# the default here should be the element returned by `count_merge([])`
clients = traces.TimeSeries.merge(router_list,
operation=traces.operations.strict_sum)
assert clients[-15] is None
assert clients[-6] == 1
assert clients[-0.5] is None
assert clients[3] is None
assert clients[15] is None
system_uptime = clients.exists()
assert system_uptime[-15] is False
assert system_uptime[-6] is True
assert system_uptime[-1] is False
clients = traces.TimeSeries.merge(router_list,
operation=traces.operations.ignorant_sum)
assert clients[-15] == 0
assert clients[-6] == 1
assert clients[-0.5] == 0
assert clients[3] == 3
assert clients[15] == 0
system_uptime = clients.exists()
assert system_uptime[-15] is True
assert system_uptime[-5] is True
assert system_uptime[-1] is True
def test_reference():
cart = traces.TimeSeries()
cart[1.2] = {'broccoli'}
cart[1.7] = {'broccoli', 'apple'}
cart[2.2] = {'apple'}
cart[3.5] = {'apple', 'beets'}
assert cart[2] == {'broccoli', 'apple'}
assert cart[-1] is None
cart = traces.TimeSeries(default=set())
assert cart[-1] == set([])
def test_deltas(self):
filter = Filters.DepthFilter(3)
reader = TimeLimitedReader('resources/orderbook/_orderbooks.csv.gz', skip_time='530 sec', limit_time='10 sec')
bids = []
asks = []
for item, _ in reader:
if item.symbol == 'XBTUSD':
res = filter.process(item)
if res is not None and res[-2] != -1:
v = np.sum(res[-1][1,:])
if res[-2] % 2 == 0 and v < 0:
bids.append((res[0], -v))
elif v > 0:
asks.append((res[0], v))
# tf, symbol, side? __delta-values
bid_ts, bid_deltas = zip(*bids)
ask_ts, ask_deltas = zip(*asks)
bid_deltas = np.array(bid_deltas)
ask_deltas = np.array(ask_deltas)
bid_deltas[1:] -= bid_deltas[:-1]
ask_deltas[1:] -= ask_deltas[:-1]
# bids = zip(bid_ts, np.log(bid_deltas))
# asks = zip(ask_ts, np.log(ask_deltas))
# bids = zip(bid_ts[1:], np.clip(bid_deltas[1:], -1000., 1000.))
# asks = zip(ask_ts[1:], np.clip(ask_deltas[1:], -1000., 1000.))
bids = zip(bid_ts[1:], bid_deltas[1:])
asks = zip(ask_ts[1:], ask_deltas[1:])
print(f'bid deltas (neg) {len(bid_ts)}, ask deltas (pos) = {len(ask_ts)}')
t1 = time.time()
ts_bid = traces.TimeSeries(bids, default=0)
ts_ask = traces.TimeSeries(asks, default=0)
ts = ts_ask * ts_bid
t2 = time.time() - t1
print(f'Time to compute timeseries {t2}')
ts.plot()
plt.show()
t1 = time.time()
sq1 = np.sqrt(np.sum(np.square(ask_deltas[1:])))
sq2 = np.sqrt(np.sum(np.square(bid_deltas[1:])))
values = sum(list(ts._d.values()))
hy = values / sq1 / sq2
t2 = time.time() - t1
print(f'Time to compute Hoyashi-Yoshido cor {t2}')
print(hy)
def test_find_refuted_radius():
from magnum.examples.rock_paper_scissors import rps as g
play = {'u': traces.TimeSeries([(0, 1)])}
counter = {'w': traces.TimeSeries([(0, 20 / 60)])}
r = find_refuted_radius(g, play, counter, tol=1e-6)
assert approx(10 / 60, abs=1e-5) == r
play = {'u': traces.TimeSeries([(0, 0)])}
counter = {'w': traces.TimeSeries([(0, 20 / 60)])}
r = find_refuted_radius(g, play, counter, tol=1e-6)
assert approx(10 / 60, abs=1e-5) == r
play = {'u': traces.TimeSeries([(0, 20 / 60)])}
counter = {'w': traces.TimeSeries([(0, 40 / 60)])}
r = find_refuted_radius(g, play, counter, tol=1e-6)
assert approx(10 / 60, abs=1e-5) == r
play = {'u': traces.TimeSeries([(0, 40 / 60)])}
counter = {'w': traces.TimeSeries([(0, 0)])}
r = find_refuted_radius(g, play, counter, tol=1e-6)
assert approx(10 / 60, abs=1e-5) == r
def test_sample():
time_list = [
datetime.datetime(2016, 1, 1, 1, 1, 2),
datetime.datetime(2016, 1, 1, 1, 1, 3),
datetime.datetime(2016, 1, 1, 1, 1, 8),
datetime.datetime(2016, 1, 1, 1, 1, 10)
]
ts = _make_ts(int, time_list, [1, 2, 3, 0])
def curr_time(i):
return datetime.datetime(2016, 1, 1, 1, 1, i)
# Check first arguments
assert dict(ts.sample(1, time_list[0], time_list[-1])) == {
curr_time(i): ts[curr_time(i)]
for i in range(2, 11)
}
assert dict(ts.sample(2, time_list[0], time_list[-1])) == {
curr_time(i): ts[curr_time(i)]
for i in range(2, 11, 2)
}
nose.tools.assert_raises(ValueError, ts.sample, -1, time_list[0],
time_list[-1])
nose.tools.assert_raises(ValueError, ts.sample, 20, time_list[0],
time_list[-1])
# Check second and third arguments
nose.tools.assert_raises(ValueError, ts.sample, 1, time_list[3],
time_list[0])
assert dict(ts.sample(1, curr_time(5), curr_time(10))) == {
curr_time(i): ts[curr_time(i)]
for i in range(5, 11)
}
assert dict(ts.sample(1, curr_time(2), curr_time(5))) == {
curr_time(i): ts[curr_time(i)]
for i in range(2, 6)
}
assert dict(ts.sample(1, curr_time(0), curr_time(13))) == {
curr_time(i): ts[curr_time(i)]
for i in range(0, 14)
}
# Check using int
ts = traces.TimeSeries([[1, 2], [2, 3], [6, 1], [8, 4]])
assert dict(ts.sample(1, 1, 8)) == {i: ts[i] for i in range(1, 9)}
assert dict(ts.sample(
0.5, 1, 8)) == {1 + i / 2.: ts[1 + i / 2.]
for i in range(0, 15)}
nose.tools.assert_raises(ValueError, ts.sample, 0.5, -traces.inf, 8)
nose.tools.assert_raises(ValueError, ts.sample, 0.5, 1, traces.inf)
# Test pandas compatibility
pd_ts = pd.Series(dict(ts.sample(1, 1, 8)))
assert all(pd_ts.index[i - 1] == i for i in range(1, 9))
assert all(pd_ts.values[i - 1] == ts[i] for i in range(1, 9))
def _eval(x):
y = f(x)
out = traces.TimeSeries(((t - dt, v) for t, v in y))
out = out.slice(y.domain.start(), y.domain.end())
out.compact()
return out
def test_histogram_stats_with_nones():
histogram = traces.Histogram()
nose.tools.eq_(histogram.mean(), None)
nose.tools.eq_(histogram.variance(), None)
nose.tools.eq_(histogram.standard_deviation(), None)
nose.tools.eq_(histogram.min(), None)
nose.tools.eq_(histogram.max(), None)
nose.tools.eq_(histogram.median(), None)
histogram = traces.Histogram.from_dict({None: 1}, key=hash)
nose.tools.eq_(histogram.mean(), None)
nose.tools.eq_(histogram.variance(), None)
nose.tools.eq_(histogram.standard_deviation(), None)
nose.tools.eq_(histogram.min(), None)
nose.tools.eq_(histogram.max(), None)
nose.tools.eq_(histogram.median(), None)
ts = traces.TimeSeries()
ts[0] = None
ts[1] = 5
ts[2] = 6
ts[3] = None
ts[9] = 7
ts[10] = None
histogram = ts.distribution(start=0, end=10)
nose.tools.eq_(histogram.mean(), 6)
def _eval(x):
y1, y2 = f1(x), f2(x)
y = y1.operation(y2, lambda a, b: (a, b))
out = traces.TimeSeries(apply_until(y), domain=y1.domain)
out.compact()
return out
def test_encode_refuted_rec():
refuted = {
'u1': traces.TimeSeries([(0, 0), (1, 1)]),
'u2': traces.TimeSeries([(0, 0.5)])
}
phi = encode_refuted_rec(refuted, 0.2, [0])
psi1 = stl.parse('(u1 < -0.2) | (u1 > 0.2)')
psi2 = stl.parse('(u2 < 0.3) | (u2 > 0.7)')
assert phi == psi1 | psi2
psi3 = stl.parse('X(u1 < 0.8)')
psi4 = stl.parse('(X(u2 < 0.3)) | (X(u2 > 0.7))')
phi = encode_refuted_rec(refuted, 0.2, [1])
assert phi == psi3 | psi4
phi = encode_refuted_rec(refuted, 0.2, [0, 1])
assert set(phi.args) == set((psi1 | psi2 | psi3 | psi4).args)
def test_compact():
# since this is random, do it a bunch of times
for n_trial in range(100):
# make two time series, one compact and one not
test_ts = traces.TimeSeries()
compact_ts = traces.TimeSeries()
for t in range(100):
value = random.randint(0, 2)
test_ts.set(t, value)
compact_ts.set(t, value, compact=True)
# make test_ts compact
test_ts.compact()
# items should be exactly the same
assert test_ts.items() == compact_ts.items()
def _eval(x):
y = f(x)
if len(y) <= 1:
return y
out = traces.TimeSeries(process_intervals(y)).slice(
y.domain.start(), y.domain.end())
out.compact()
return out
def extract_ts(name, model, g, store):
dt = g.model.dt
# TODO: hack to have to eval this
# TODO: support extracting H=0 timeseries
ts = traces.TimeSeries(
((dt * t, eval(str(model[store[name, t]]))) for t in g.times
if (name, t) in store and store[name, t] in model),
domain=(0, g.scope + dt))
ts.compact()
return ts
def get_stats_from_ss(ss):
df = pd.concat([
pd.DataFrame(dict(t=ss.start, dx=1)),
pd.DataFrame(dict(t=ss.stop, dx=-1))
])
w = df.sort_values('t')['dx'].cumsum().pipe(
lambda s: set_index(s, df.t - df.t.iloc[0])).sort_index()
x = traces.TimeSeries(data=w)
s = pd.Series(*reversed(list(zip(*x.sample(0.1)))))
return x.mean(), (s == 0).mean()
def test_plot():
ts = traces.TimeSeries()
ts[0] = 0
ts[1] = 2
ts[3] = 1
ts[5] = 0
figure, axes = ts.plot()
return figure
def extract_ts(name, model, g, store):
dt = g.model.dt
model = {k: v.primal for k, v in model.variables.items()}
ts = traces.TimeSeries(
((dt * t, model[store[name, t][0].name])
for t in g.times if not isinstance(store[name, t][0], (float, int))
and store[name, t][0].name in model),
domain=(0, g.scaled_scope + dt))
ts.compact()
return ts
def test_smt_radius_oracle():
from magnum.examples.rock_paper_scissors import rps as g
play = {'u': traces.TimeSeries([(0, 1)])}
counter = {'w': traces.TimeSeries([(0, 20 / 60)])}
oracle = smt_radius_oracle(g=g, play=play, counter=counter)
assert not oracle(r=9 / 60)
assert oracle(r=20 / 60)
assert oracle(r=1)
play = {'u': traces.TimeSeries([(0, 0)])}
counter = {'w': traces.TimeSeries([(0, 20 / 60)])}
oracle = smt_radius_oracle(g=g, play=play, counter=counter)
assert not oracle(r=5 / 60)
assert not oracle(r=9 / 60)
assert oracle(r=11 / 60)
assert oracle(r=1)
def test_invalid_call():
ts = traces.TimeSeries()
ts[0] = 0
ts[1] = 1
ts.plot(interpolate='previous')
ts.plot(interpolate='linear')
with pytest.raises(ValueError):
ts.plot(interpolate='yomama')
def test_radd():
ts1 = traces.TimeSeries(default=0)
ts1[0] = 1
ts1[2] = 0
ts1[3] = 1
ts1[4] = 0
ts2 = traces.TimeSeries(default=0)
ts2[-1] = 1
ts2[2] = 0
ts2[3] = 1
ts2[4] = 0
ts3 = ts1 + ts2
nose.tools.eq_(list(ts3.items()), [(-1, 1), (0, 2), (2, 0), (3, 2),
(4, 0)])
nose.tools.assert_raises(TypeError, ts3.__radd__, 1)
def slice(self, series, start, end, default=0):
"""Technically TimeSeries has a method for this, but it wasn't working and it
was easier to write something of my own than to try to fix that.
"""
result = traces.TimeSeries(default=default)
for t0, t1, value in series.iterperiods(start, end):
result[t0] = value
result[t1] = series[t1]
return result
def get_trace(df, date):
end_date = date + pd.Timedelta(days=1)
single_date_df = df[(df['Date'] >= date) & (df['Date'] < end_date)]
per_car_timeseries = []
for x in single_date_df[["Date", "duration"]].values:
ts = traces.TimeSeries(default=0)
ticket_start = x[0]
ticket_end = get_ticket_end(ticket_start, x[1])
ts[ticket_start], ts[ticket_end] = 1, 0
per_car_timeseries.append(ts)
ret = traces.TimeSeries.merge(per_car_timeseries, operation=sum, compact=True)
return ret
def test_counter_examples():
from magnum.examples.feasible_example2 import feasible_example as g
res = milp.encode_and_run(g)
assert res.feasible
ces = [{'w': traces.TimeSeries([(0, 0)])}]
res = milp.encode_and_run(g, counter_examples=ces)
assert res.feasible
ces = [{'w': traces.TimeSeries([(0, 1)])}]
res = milp.encode_and_run(g, counter_examples=ces)
assert not res.feasible
ces = [{
'w': traces.TimeSeries([(0, 1)])
}, {
'w': traces.TimeSeries([(0, 0)])
}]
res = milp.encode_and_run(g, counter_examples=ces)
assert not res.feasible
def stack(self, duration, start_times, operation):
ts_list = []
for start in start_times:
end = start + duration
ts = traces.TimeSeries()
for t0, dur, value in self.iterperiods(start, end):
offset = t0 - start
if isinstance(offset, datetime.timedelta):
offset = offset.total_seconds()
ts.set(offset, value, compact=True)
ts_list.append(ts)
return traces.TimeSeries.merge(ts_list, operation=operation)
def test_rps_counter_examples():
from magnum.examples.rock_paper_scissors import rps as g
from stl.boolean_eval import pointwise_sat
# Respond to Paper
ces = [{'w': traces.TimeSeries([(0, 20 / 60)])}]
res = milp.encode_and_run(g, counter_examples=ces)
assert res.feasible
assert pytest.approx(res.cost) == 10
phi = stl.parse('X((x >= 10) & (x <= 50))')
assert pointwise_sat(phi, dt=g.model.dt)(res.solution)
# Respond to Scissors and Paper
ces.append({'w': traces.TimeSeries([(0, 40 / 60)])})
res = milp.encode_and_run(g, counter_examples=ces)
assert res.feasible
assert pytest.approx(res.cost) == 10
phi = stl.parse('X((x >= 30) & (x <= 50))')
assert pointwise_sat(phi, dt=g.model.dt)(res.solution)
ces.append({'w': traces.TimeSeries([(0, 0)])})
res = milp.encode_and_run(g, counter_examples=ces)
assert not res.feasible
assert pytest.approx(res.cost) == 0
phi = stl.parse('X((x = 10) | (x = 30) | (x = 50))')
assert pointwise_sat(phi, dt=g.model.dt)(res.solution)
g = g.invert()
res = milp.encode_and_run(g)
assert res.feasible
assert pytest.approx(res.cost) == 10
ces = [{'u': traces.TimeSeries([(0, 20 / 60)])}]
res = milp.encode_and_run(g, counter_examples=ces)
assert res.feasible
assert pytest.approx(res.cost) == 10
ces = [{'u': traces.TimeSeries([(0, 40 / 60)])}]
res = milp.encode_and_run(g, counter_examples=ces)
assert res.feasible
assert pytest.approx(res.cost) == 10
ces = [({'u': traces.TimeSeries([(0, 0)])})]
res = milp.encode_and_run(g, counter_examples=ces)
assert res.feasible
ces = [({'u': traces.TimeSeries([(0, 1)])})]
res = milp.encode_and_run(g, counter_examples=ces)
assert res.feasible
def get_time(self, triad):
credentials = self.get_credentials()
http = credentials.authorize(httplib2.Http())
service = discovery.build('calendar', 'v3', http=http)
now = datetime.datetime.utcnow().isoformat() + 'Z' # 'Z' indicates UTC time
# Got freebusy code from https://gist.github.com/cwurld/9b4e10dbeecab28345a3
body = {
"timeMin": now,
"timeMax": (datetime.datetime.utcnow() + self.time_window).isoformat() + 'Z',
"timeZone": 'US/Central',
"items": [{"id": email} for email in triad]
}
eventsResult = service.freebusy().query(body=body).execute()
cal_dict = eventsResult[u'calendars']
busy_times_list = []
for cal_name in cal_dict:
busy_times = traces.TimeSeries(default=0) # default is free (0),
for busy_window in cal_dict[cal_name]['busy']: # manually add busy times
busy_times[busy_window['start']] = 1
busy_times[busy_window['end']] = 0
busy_times_list.append(busy_times)
# combine all of the calendars to find when everyone is free
try:
combined_free_times = traces.TimeSeries.merge(busy_times_list, operation=sum)
except ValueError:
print('One of the time series is empty. One of the emails is most likely wrong.')
return
all_start_times = self.interim_periods(combined_free_times)
all_intervals = self.entire_interval_free(all_start_times).to_bool(invert=True)
eligible_times = [i[0] for i in all_intervals.items() if i[1] is True \
and self.within_timebox(i[0])\
and not self.is_weekend(i[0])]
if not eligible_times:
print('There are no available times for that group in that time range.')
return
else:
event_time = eligible_times[0] # for now, take first time that works. we can refine
print(event_time)
return event_time
