def test_set_interval():
ts = TimeSeries()
nose.tools.assert_raises(KeyError, ts.get, 0)
nose.tools.assert_raises(KeyError, ts.set_interval, 2, 4, 5)
ts[1.2] = 1
ts[3] = 0
ts[6] = 2
assert ts[0] == 1
assert ts[5.5] == 0
assert ts[7] == 2
ts[2:4] = 5
assert ts.items() == [(1.2, 1), (2, 5), (4, 0), (6, 2)]
ts[3:5] = 4
assert ts.items() == [(1.2, 1), (2, 5), (3, 4), (5, 0), (6, 2)]
tsc = TimeSeries(ts)
ts.set_interval(3, 4, 4)
assert ts.items() == [(1.2, 1), (2, 5), (3, 4), (4, 4), (5, 0), (6, 2)]
tsc.set_interval(3, 4, 4, compact=True)
assert tsc.items() == [(1.2, 1), (2, 5), (3, 4), (5, 0), (6, 2)]
def test_difference():
a = TimeSeries(data=[(0, 0), (2, 2)], default=0)
b = TimeSeries(data=[(1, 1), (3, 2)], default=0)
c = a - b
nose.tools.eq_(list(c.items()), [(0, 0), (1, -1), (2, 1), (3, 0)])
def test_mean():
ts = TimeSeries([[1, 2], [2, 3], [6, 1], [8, 4]], domain=Domain(0, 9))
assert ts.mean() == 22. / 9
ts = TimeSeries([[1, 2], [2, 3], [6, 1], [8, 4]],
domain=Domain([1, 2], [3, 5], [6, 8]))
nose.tools.assert_raises(NotImplementedError, ts.mean)
def test_init_data():
ts = TimeSeries([(1, 2), (2, 3), (6, 1), (8, 4)])
assert ts[0] == 2
assert ts[1] == 2
assert ts[1.5] == 2
assert ts[6] == 1
assert ts[7] == 1
assert ts[8] == 4
assert ts[10] == 4
ts = TimeSeries([[1, 2], [2, 3], [6, 1], [8, 4]])
assert ts[0] == 2
assert ts[1] == 2
assert ts[1.5] == 2
assert ts[6] == 1
assert ts[7] == 1
assert ts[8] == 4
assert ts[10] == 4
ts = TimeSeries({1: 2, 2: 3, 6: 1, 8: 4})
assert ts[0] == 2
assert ts[1] == 2
assert ts[1.5] == 2
assert ts[6] == 1
assert ts[7] == 1
assert ts[8] == 4
assert ts[10] == 4
def test_set_domain():
ts = TimeSeries()
assert ts.domain == Domain([-inf, inf])
ts.domain = None
assert ts.domain == Domain([-inf, inf])
ts.domain = [-inf, 5]
assert ts.domain == Domain([-inf, 5])
ts.domain = [5, inf]
assert ts.domain == Domain([5, inf])
ts.domain = [-inf, inf]
assert ts.domain == Domain([-inf, inf])
ts.domain = [2, 5]
assert ts.domain == Domain([2, 5])
ts.domain = [[2, 5], [9, 10]]
assert ts.domain == Domain([2, 5], [9, 10])
ts = TimeSeries(data=[(1, 2), (2, 3), (6, 1), (8, 4)])
nose.tools.assert_raises(ValueError, setattr, ts, 'domain', [1.5, 7])
nose.tools.assert_raises(ValueError, setattr, ts, 'domain', [0, 7])
nose.tools.assert_raises(ValueError, setattr, ts, 'domain', [1.5, 9])
def test_mask():
start = datetime.datetime(2015, 3, 1)
# v. simple
a = TimeSeries()
a.set(start, 1)
a.set(datetime.datetime(2015, 4, 2), 0)
a.set(datetime.datetime(2015, 4, 3), 1)
a.set(datetime.datetime(2015, 4, 4), 0)
end = datetime.datetime(2015, 4, 5)
mask = TimeSeries(default=False)
mask[datetime.datetime(2015, 4, 1)] = True
mask[datetime.datetime(2015, 4, 3)] = False
# not normalized
distribution = a.distribution(
start=start,
end=end,
normalized=False,
mask=mask,
)
assert distribution[0] == 24 * 60 * 60 # one day
assert distribution[1] == 24 * 60 * 60
# normalized
distribution = a.distribution(
start=start,
end=end,
mask=mask,
)
assert distribution[0] == 0.5
assert distribution[1] == 0.5
def test_distribution_empty():
ts = TimeSeries()
mask = TimeSeries(default=0)
mask[0] = 1
mask[2] = 0
# distribution with default args and no default value on empty
# TimeSeries doesn't know what to do
nose.tools.assert_raises(ValueError, ts.distribution)
# distribution with start and end, but no default value on empty
# TimeSeries doesn't know what to do
nose.tools.assert_raises(ValueError, ts.distribution, 0, 10)
# no matter what is passed in to distribution, if the default
# value is not set on an empty TimeSeries this should be an error
nose.tools.assert_raises(ValueError, ts.distribution, mask=mask)
ts = TimeSeries(default=0)
# no mask or start/end on empty TimeSeries, don't know what to do
nose.tools.assert_raises(ValueError, ts.distribution)
# start and end or mask given, is fine
distribution = ts.distribution(0, 10)
assert distribution[0] == 1.0
distribution = ts.distribution(mask=mask)
assert distribution[0] == 1.0
def test_iterperiods():
# timeseries with a default value and no points returns one period
# with the default value from -infinity to infinity, or
ts = TimeSeries(default=0)
assert list(ts.iterperiods()) == [(-inf, inf, 0)]
# timeseries with no default value and no points raises error
ts = TimeSeries()
with nose.tools.assert_raises(KeyError):
list(ts.iterperiods())
ts = TimeSeries()
ts.set(datetime.datetime(2015, 3, 1), 1)
ts.set(datetime.datetime(2015, 3, 2), 0)
ts.set(datetime.datetime(2015, 3, 3), 1)
ts.set(datetime.datetime(2015, 3, 4), 2)
answer = [
(datetime.datetime(2015, 3, 1), datetime.datetime(2015, 3, 2), 1),
(datetime.datetime(2015, 3, 2), datetime.datetime(2015, 3, 3), 0),
(datetime.datetime(2015, 3, 3), datetime.datetime(2015, 3, 4), 1)]
result = []
for (t0, t1, v0) in ts.iterperiods(
start=datetime.datetime(2015, 3, 1),
end=datetime.datetime(2015, 3, 4)
):
result.append((t0, t1, v0))
assert answer == result
answer = [
(datetime.datetime(2015, 3, 1), datetime.datetime(2015, 3, 2), 1),
(datetime.datetime(2015, 3, 3), datetime.datetime(2015, 3, 4), 1),
]
result = []
for (t0, t1, v0) in ts.iterperiods(
start=datetime.datetime(2015, 3, 1),
end=datetime.datetime(2015, 3, 4),
value=1,
):
result.append((t0, t1, v0))
assert answer == result
def filter(t0, t1, value):
return True if not value else False
answer = [
(datetime.datetime(2015, 3, 2), datetime.datetime(2015, 3, 3), 0),
]
result = []
for (t0, t1, v0) in ts.iterperiods(
start=datetime.datetime(2015, 3, 1),
end=datetime.datetime(2015, 3, 4),
value=filter,
):
result.append((t0, t1, v0))
assert answer == result
def test_merge():
ts_a = TimeSeries(default=None)
ts_b = TimeSeries(default=None)
ts_a[0] = None
ts_b[0] = True
ts_merge = TimeSeries.merge([ts_a, ts_b])
assert True in ts_merge[0]
assert None in ts_merge[0]
def test_merge():
ts_a = TimeSeries()
ts_b = TimeSeries()
# ts_a[0] = None
ts_b[0] = True
ts_merge = TimeSeries.merge([ts_a, ts_b])
assert True in ts_merge[0]
assert None in ts_merge[0]
ts_c = TimeSeries.merge([])
assert list(ts_c.items()) == []
def test_sample():
# Check using int
ts = TimeSeries([[1, 2], [2, 3], [6, 1], [8, 4]], domain=Domain(1, 9))
assert dict(ts.sample(1)) == {i: ts[i] for i in range(1, 10)}
assert dict(
ts.sample(0.5)) == {1 + i / 2.: ts[1 + i / 2.]
for i in range(0, 17)}
ts = TimeSeries(
[[1, 2], [2, 3], [6, 1], [8, 4]],
domain=Domain([1, 2], [3, 5], [6, 8]),
)
nose.tools.assert_raises(NotImplementedError, ts.sample, 0.5)
def test_single_merges():
# a single empty time series
ts = TimeSeries()
assert TimeSeries.merge([ts]) == TimeSeries(default=[None])
# multiple empty time series
ts_a = TimeSeries()
ts_b = TimeSeries()
assert TimeSeries.merge([ts_a, ts_b]) == \
TimeSeries(default=[None, None])
# test a single time series with only one measurement
ts = TimeSeries()
ts[21] = 42
merged = TimeSeries.merge([ts])
assert list(merged.items()) == [(21, [42])]
# test an empty time series and a time series with one measurement
ts_a = TimeSeries()
ts_a[21] = 42
ts_b = TimeSeries()
ts = TimeSeries.merge([ts_a, ts_b])
assert ts[20] == [None, None]
assert ts[23] == [42, None]
# test an empty time series and a time series with one entry
ts_a = TimeSeries()
ts_a[21] = 42
ts_a[22] = 41
ts_a[23] = 40
ts_b = TimeSeries()
ts_b[20] = 1
ts_b[22] = 2
ts_b[24] = 3
merged = TimeSeries.merge([ts_a, ts_b])
assert list(merged.items()) == [
(20, [None, 1]),
(21, [42, 1]),
(22, [41, 2]),
(23, [40, 2]),
(24, [40, 3]),
]
def test_remove_points_from_interval():
ts = TimeSeries(default=0)
ts[0] = 0
ts[1] = 2
ts[3] = 1
ts[4] = 0
assert ts[5] == 0
del ts[3.5:4.5]
assert ts[5] == 1
ts[4] = 0
del ts[3:4.5]
assert ts[5] == 2
ts[3] = 1
ts[4] = 0
del ts[3.5:4]
assert ts[5] == 0
def example_dictlike():
# test overwriting keys
l = TimeSeries()
l[datetime.datetime(2010, 1, 1)] = 5
l[datetime.datetime(2010, 1, 2)] = 4
l[datetime.datetime(2010, 1, 3)] = 3
l[datetime.datetime(2010, 1, 7)] = 2
l[datetime.datetime(2010, 1, 4)] = 1
l[datetime.datetime(2010, 1, 4)] = 10
l[datetime.datetime(2010, 1, 4)] = 5
l[datetime.datetime(2010, 1, 1)] = 1
l[datetime.datetime(2010, 1, 7)] = 1.2
l[datetime.datetime(2010, 1, 8)] = 1.3
l[datetime.datetime(2010, 1, 12)] = 1.3
# do some wackiness with a bunch of points
dt = datetime.datetime(2010, 1, 12)
for i in range(1000):
dt += datetime.timedelta(hours=random.random())
l[dt] = math.sin(i / float(math.pi))
dt -= datetime.timedelta(hours=500)
dt -= datetime.timedelta(minutes=30)
for i in range(1000):
dt += datetime.timedelta(hours=random.random())
l[dt] = math.cos(i / float(math.pi))
# what does this get?
print >> sys.stderr, l[datetime.datetime(2010, 1, 3, 23, 59, 59)]
# output the time series
for i, j in l:
print(i.isoformat(), j)
def test_scalar_ops():
a = TimeSeries()
a.set(datetime.datetime(2015, 3, 1), 1)
a.set(datetime.datetime(2015, 3, 2), 0)
a.set(datetime.datetime(2015, 3, 3), 3)
a.set(datetime.datetime(2015, 3, 4), 2)
ts_half = a.multiply(0.5)
ts_bool = a.to_bool(invert=False)
ts_threshold = a.threshold(value=1.1)
# test before domain, should give default value
assert ts_half[datetime.datetime(2015, 2, 24)] is None
assert ts_bool[datetime.datetime(2015, 2, 24)] == None
assert ts_threshold[datetime.datetime(2015, 2, 24)] == None
# test values throughout series
assert ts_half[datetime.datetime(2015, 3, 1, 6)] == 0.5
assert ts_bool[datetime.datetime(2015, 3, 1, 6)] == True
assert ts_threshold[datetime.datetime(2015, 3, 1, 6)] == False
assert ts_half[datetime.datetime(2015, 3, 2, 6)] == 0
assert ts_bool[datetime.datetime(2015, 3, 2, 6)] == False
assert ts_threshold[datetime.datetime(2015, 3, 2, 6)] == False
assert ts_half[datetime.datetime(2015, 3, 3, 6)] == 1.5
assert ts_bool[datetime.datetime(2015, 3, 3, 6)] == True
assert ts_threshold[datetime.datetime(2015, 3, 3, 6)] == True
# test after domain, should give last value
assert ts_half[datetime.datetime(2015, 3, 4, 18)] == 1
assert ts_bool[datetime.datetime(2015, 3, 4, 18)] == True
assert ts_threshold[datetime.datetime(2015, 3, 4, 18)] == True
def test_domain():
ts = TimeSeries()
assert ts.domain == Domain([-inf, inf])
ts.domain = None
assert ts.domain == Domain([-inf, inf])
ts.domain = [-inf, 5]
assert ts.domain == Domain([-inf, 5])
ts.domain = [5, inf]
assert ts.domain == Domain([5, inf])
ts.domain = [-inf, inf]
assert ts.domain == Domain(-inf, inf)
ts.domain = [2, 5]
assert ts.domain == Domain([2, 5])
ts.domain = [[2, 5], [9, 10]]
assert ts.domain == Domain([[2, 5], [9, 10]])
ts.domain = [[-inf, 1], [2, 5], [9, 10]]
assert ts.domain == Domain([[-inf, 1], [2, 5], [9, 10]])
ts.domain = [[2, 5], [9, 10], [11, inf]]
assert ts.domain == Domain([[2, 5], [9, 10], [11, inf]])
ts.domain = [[-inf, 1], [2, 5], [9, 10], [11, inf]]
assert ts.domain == Domain([[-inf, 1], [2, 5], [9, 10], [11, inf]])
def example_arrow():
l = TimeSeries()
l[arrow.Arrow(2010, 1, 1)] = 0
l[arrow.Arrow(2010, 1, 3, 10)] = 1
l[arrow.Arrow(2010, 1, 5)] = 0
l[arrow.Arrow(2010, 1, 8)] = 1
l[arrow.Arrow(2010, 1, 17)] = 0
l[arrow.Arrow(2010, 1, 19)] = 1
l[arrow.Arrow(2010, 1, 23)] = 0
l[arrow.Arrow(2010, 1, 26)] = 1
l[arrow.Arrow(2010, 1, 28)] = 0
l[arrow.Arrow(2010, 1, 31)] = 1
l[arrow.Arrow(2010, 2, 5)] = 0
for time, value in l:
print(time.naive.isoformat(), 0.1 * value + 1.1)
print('')
start = arrow.Arrow(2010, 1, 1)
end = arrow.Arrow(2010, 2, 5)
unit = {'hours': 25}
for start, end in span_range(start, end, unit):
print(start.naive.isoformat(), l.mean(start, end))
print('')
for start, end in span_range(start, end, unit):
print(start.naive.isoformat(), -0.2)
print(start.naive.isoformat(), 1.2)
def example_mean():
l = TimeSeries()
l[datetime.datetime(2010, 1, 1)] = 0
l[datetime.datetime(2010, 1, 3, 10)] = 1
l[datetime.datetime(2010, 1, 5)] = 0
l[datetime.datetime(2010, 1, 8)] = 1
l[datetime.datetime(2010, 1, 17)] = 0
l[datetime.datetime(2010, 1, 19)] = 1
l[datetime.datetime(2010, 1, 23)] = 0
l[datetime.datetime(2010, 1, 26)] = 1
l[datetime.datetime(2010, 1, 28)] = 0
l[datetime.datetime(2010, 1, 31)] = 1
l[datetime.datetime(2010, 2, 5)] = 0
for time, value in l:
print(time.isoformat(), 0.1 * value + 1.1)
print('')
timestep = {'hours': 25}
start = datetime.datetime(2010, 1, 1)
while start <= datetime.datetime(2010, 2, 5):
end = start + datetime.timedelta(**timestep)
print(start.isoformat(), l.mean(start, end))
start = end
print('')
start = datetime.datetime(2010, 1, 1)
while start <= datetime.datetime(2010, 2, 5):
end = start + datetime.timedelta(**timestep)
print(start.isoformat(), -0.2)
print(start.isoformat(), 1.2)
start = end
def add_annotation(self, file: os.PathLike, **kwargs):
self.file_list.append(file)
# setup default kwargs #
if self.ordinal:
kwargs.update({'value_transform': lambda cls: self.order.get(cls)})
if not kwargs.get('time_transform', None):
kwargs.update({'time_transform': lambda t: float(t)})
if not kwargs.get('skip_header', None):
kwargs.update({'skip_header': False})
if not os.path.isfile(file):
raise FileNotFoundError(f'File {file} not found.')
if os.path.getsize(file):
ts = TimeSeries.from_csv(file,
**kwargs).slice(self.start_time,
self.end_time)
meta = pd.read_csv(file, names=COLUMNS.keys())
meta.drop(meta[meta['time'] < self.start_time].index)
meta.drop(meta[meta['time'] > self.end_time].index)
else:
ts = TimeSeries()
meta = pd.DataFrame()
self.metadata.append(meta)
self.series.append(ts)
self._data = self._report()
def test_mask():
start = datetime.datetime(2015, 3, 1)
# v. simple
a = TimeSeries()
a.set(start, 1)
a.set(datetime.datetime(2015, 3, 2), 0)
a.set(datetime.datetime(2015, 3, 3), 1)
a.set(datetime.datetime(2015, 3, 4), 0)
end = datetime.datetime(2015, 3, 5)
mask = Domain(datetime.datetime(2015, 3, 1), datetime.datetime(2015, 3, 3))
# not normalized
distribution = a.distribution(
start=start,
end=end,
normalized=False,
mask=mask,
)
assert distribution[0] == 24 * 60 * 60 # one day
assert distribution[1] == 24 * 60 * 60
# normalized
distribution = a.distribution(
start=start,
end=end,
mask=mask,
)
assert distribution[0] == 0.5
assert distribution[1] == 0.5
def example_dictlike():
# test overwriting keys
ts = TimeSeries()
ts[datetime.datetime(2010, 1, 1)] = 5
ts[datetime.datetime(2010, 1, 2)] = 4
ts[datetime.datetime(2010, 1, 3)] = 3
ts[datetime.datetime(2010, 1, 7)] = 2
ts[datetime.datetime(2010, 1, 4)] = 1
ts[datetime.datetime(2010, 1, 4)] = 10
ts[datetime.datetime(2010, 1, 4)] = 5
ts[datetime.datetime(2010, 1, 1)] = 1
ts[datetime.datetime(2010, 1, 7)] = 1.2
ts[datetime.datetime(2010, 1, 8)] = 1.3
ts[datetime.datetime(2010, 1, 12)] = 1.3
# do some wackiness with a bunch of points
dt = datetime.datetime(2010, 1, 12)
for i in range(1000):
dt += datetime.timedelta(hours=random.random())
ts[dt] = math.sin(i / float(math.pi))
dt -= datetime.timedelta(hours=500)
dt -= datetime.timedelta(minutes=30)
for i in range(1000):
dt += datetime.timedelta(hours=random.random())
ts[dt] = math.cos(i / float(math.pi))
# output the time series
for i, j in ts:
print(i.isoformat(), j)
def test_distribution_set():
time_series = TimeSeries()
time_series[1.2] = {'broccoli'}
time_series[1.4] = {'broccoli', 'orange'}
time_series[1.7] = {'broccoli', 'orange', 'banana'}
time_series[2.2] = {'orange', 'banana'}
time_series[3.5] = {'orange', 'banana', 'beets'}
def test_distribution():
start = datetime.datetime(2015, 3, 1)
# v. simple
a = TimeSeries()
a.set(start, 1)
a.set(datetime.datetime(2015, 3, 2), 0)
a.set(datetime.datetime(2015, 3, 3), 1)
a.set(datetime.datetime(2015, 3, 4), 0)
end = datetime.datetime(2015, 3, 5)
# not normalized
distribution = a.distribution(
start=start,
end=end,
normalized=False,
)
assert distribution[0] == 24 * 60 * 60 * 2 # two days
assert distribution[1] == 24 * 60 * 60 * 2
# normalized
distribution = a.distribution(start=start, end=end)
assert distribution[0] == 0.5
assert distribution[1] == 0.5
def from_pandas(df, compact=True):
'''TODO'''
domain = Domain(df.index[0], df.index[-1])
data = {col: TimeSeries(df[col].T, domain=domain) for col in df.columns}
if compact:
for val in data.values():
val.compact()
return data
def test_exists():
ts = TimeSeries([(-5, 0), (0, 23), (5, None)])
ts_exists = ts.exists()
assert ts_exists[-10] is False
assert ts_exists[-2] is True
assert ts_exists[3] is True
assert ts_exists[10] is False
def example_sum():
a = TimeSeries()
a.set(datetime.datetime(2015, 3, 1), 1)
a.set(datetime.datetime(2015, 3, 2), 0)
a.set(datetime.datetime(2015, 3, 3), 1)
a.set(datetime.datetime(2015, 3, 5), 0)
a.set(datetime.datetime(2015, 3, 6), 0)
b = TimeSeries()
b.set(datetime.datetime(2015, 3, 1), 0)
b.set(datetime.datetime(2015, 3, 2, 12), 1)
b.set(datetime.datetime(2015, 3, 3, 13, 13), 0)
b.set(datetime.datetime(2015, 3, 4), 1)
b.set(datetime.datetime(2015, 3, 5), 0)
b.set(datetime.datetime(2015, 3, 5, 12), 1)
b.set(datetime.datetime(2015, 3, 5, 19), 0)
c = TimeSeries()
c.set(datetime.datetime(2015, 3, 1, 17), 0)
c.set(datetime.datetime(2015, 3, 1, 21), 1)
c.set(datetime.datetime(2015, 3, 2, 13, 13), 0)
c.set(datetime.datetime(2015, 3, 4, 18), 1)
c.set(datetime.datetime(2015, 3, 5, 4), 0)
# output the three time series
for i, ts in enumerate([a, b, c]):
for (t0, v0), (t1, v1) in ts.iterintervals(1):
print(t0.isoformat(), i)
print(t1.isoformat(), i)
print('')
for (t0, v0), (t1, v1) in ts.iterintervals(0):
print(t0.isoformat(), i)
print(t1.isoformat(), i)
print('')
# output the sum
# for dt, i in sum([a, b, c]):
# print dt.isoformat(), i
# print ''
for dt, i in TimeSeries.merge([a, b, c], operation=sum):
print(dt.isoformat(), i)
def make_random_timeseries():
length = random.randint(1, 10)
result = TimeSeries()
t = 0
for i in range(length):
t += random.randint(0, 5)
x = random.randint(0, 5)
result[t] = x
return result
def test_none_handling():
ts = TimeSeries()
ts[1] = (0, 1)
ts[2] = (None, 0)
ts[3] = (2, 0)
# print(ts.distribution(normalized=False))
assert (ts.distribution()[(0, 1)] == 0.5)
assert (ts.distribution()[(None, 0)] == 0.5)
def test_set_interval_datetime():
ts = TimeSeries(default=400)
ts[datetime(2012, 1, 4, 12)] = 5
ts[datetime(2012, 1, 9, 18)] = 10
ts[datetime(2012, 1, 8):datetime(2012, 1, 10)] = 100
assert ts.items() == [(datetime(2012, 1, 4, 12, 0), 5),
(datetime(2012, 1, 8, 0, 0), 100),
(datetime(2012, 1, 10, 0, 0), 10)]
def test_moving_average():
ts = TimeSeries([[1, 2], [2, 3], [6, 1], [8, 4]], domain=Domain(1, 9))
ts2 = TimeSeries(
[[1, 2], [2, 3], [6, 1], [8, 4]],
domain=Domain(1 - 1, 9 + 1),
)
assert dict(ts.moving_average(
1, 2)) == {i: ts2.mean(i - 1, i + 1)
for i in range(1, 10)}
assert dict(ts.moving_average(0.5, 2)) == {
1 + i / 2.: ts2.mean(1 + i / 2. - 1, 1 + i / 2. + 1)
for i in range(0, 17)
}
ts = TimeSeries([[1, 2], [2, 3], [6, 1], [8, 4]],
domain=Domain([1, 2], [3, 5], [6, 8]))
nose.tools.assert_raises(NotImplementedError, ts.moving_average, 1, 0.5)
