def test_rolling_max():
for array, config in product(arrays, configurations):
s = Stream.source(array, dtype="float")
w = s.rolling(**config).max().rename("w")
expected = list(pd.Series(array).rolling(**config).max())
assert_op([w], expected)
def test_cumsum():
for array in arrays:
s = Stream.source(array, dtype="float")
w = s.cumsum().rename("w")
expected = list(pd.Series(array).cumsum())
assert_op([w], expected)
def test_div():
expected = [1, 2, 3, 4, 5, 6]
# (left, right) : (Stream, Stream)
s1 = Stream.source([2, 4, 6, 8, 10, 12], dtype="float")
s2 = Stream.source([2, 2, 2, 2, 2, 2], dtype="float")
w1 = s1.div(s2).rename("w1")
w2 = (s1 / s2).rename("w2")
assert_op([w1, w2], expected)
# (left, right) : (Stream, float)
w1 = s1.div(2).rename("w1")
w2 = (s1 / 2).rename("w2")
assert_op([w1, w2], expected)
def test_rsub():
# (left, right) : (float, Stream)
s1 = 6
s2 = Stream.source([1, 2, 3, 4, 5, 6], dtype="float")
w = (s1 - s2).rename("w")
assert_op([w], [5, 4, 3, 2, 1, 0])
def test_sub():
expected = [0, 1, 2, 3, 4, 5]
# (left, right) : (Stream, Stream)
s1 = Stream.source([1, 2, 3, 4, 5, 6], dtype="float")
s2 = Stream.source([1, 1, 1, 1, 1, 1], dtype="float")
w1 = s1.sub(s2).rename("w1")
w2 = (s1 - s2).rename("w2")
assert_op([w1, w2], expected)
# (left, right) : (Stream, float)
w1 = s1.sub(1).rename("w1")
w2 = (s1 - 1).rename("w2")
assert_op([w1, w2], expected)
def test_radd():
# (left, right) : (float, Stream)
s1 = 1
s2 = Stream.source([1, 2, 3, 4, 5, 6], dtype="float")
w = (s1 + s2).rename("w")
assert_op([w], [2, 3, 4, 5, 6, 7])
def test_clamp_min():
array = [-1, 2, -3, 4, -5]
s = Stream.source(array, dtype="float")
w = s.clamp_min(0).rename("w")
expected = list(pd.Series(array).apply(lambda x: x if x > 0 else 0))
assert_op([w], expected)
def test_ffill():
array = [-1, np.nan, -3, 4, np.nan]
s = Stream.source(array, dtype="float")
w = s.ffill().rename("w")
expected = list(pd.Series(array).ffill())
assert_op([w], expected)
def test_add():
# (left, right) : (Stream, Stream)
s1 = Stream.source([3, -4, 6, -7, 2, -6], dtype="float")
s2 = Stream.source([-3, 4, -6, 7, -2, 6], dtype="float")
w1 = s1.add(s2).rename("w1")
w2 = (s1 + s2).rename("w2")
assert_op([w1, w2], 6 * [0])
# (left, right) : (Stream, float)
s1 = Stream.source([1, 2, 3, 4, 5, 6], dtype="float")
s2 = 1
w1 = s1.add(s2).rename("w1")
w2 = (s1 + s2).rename("w2")
assert_op([w1, w2], [2, 3, 4, 5, 6, 7])
def test_rdiv():
expected = [6, 3, 2, 3 / 2, 6 / 5, 1]
# (left, right) : (Stream, Stream)
s = Stream.source([2, 4, 6, 8, 10, 12], dtype="float")
# (left, right) : (Stream, float)
w = (12 / s).rename("w")
assert_op([w], expected)
def test_rmul():
expected = [2, 4, 6, 8, 10, 12]
# (left, right) : (Stream, Stream)
s = Stream.source([1, 2, 3, 4, 5, 6], dtype="float")
# (left, right) : (Stream, float)
w = (2 * s).rename("w")
assert_op([w], expected)
def test_pow():
array = [1, -2, 3, -4, 5, -6]
s = Stream.source(array, dtype="float")
s1 = s.pow(3).rename("s1")
s2 = (s**3).rename("s2")
expected = list(pd.Series(array)**3)
assert_op([s1, s2], expected)
def test_ewm_mean():
array = [1, np.nan, 3, 4, 5, 6, np.nan, 7]
s = Stream.source(array, dtype="float")
for config in configurations:
w = s.ewm(**config).mean().rename("w")
expected = list(pd.Series(array).ewm(**config).mean())
assert_op([w], expected)
def test_push_one_datum():
s1 = Stream.placeholder(dtype="float").rename("s1")
s2 = Stream.placeholder(dtype="float").rename("s2")
feed = PushFeed([s1, s2])
assert isinstance(feed, PushFeed)
output = feed.push({"s1": 1, "s2": 6})
assert output == {"s1": 1, "s2": 6}
m1 = s1.clamp_min(0)
m2 = s2.clamp_max(0)
feed = PushFeed(
[s1.rename("v1"),
s2.rename("v2"),
m1.rename("v3"),
m2.rename("v4")])
arr1 = [1, -1, 2, -2, 3, -3, 4, -4, 5, -5]
arr2 = [-5, 5, -4, 4, -3, 3, -2, 2, -1, 1]
expected = {
"v1": [1, -1, 2, -2, 3, -3, 4, -4, 5, -5],
"v2": [-5, 5, -4, 4, -3, 3, -2, 2, -1, 1],
"v3": [1, 0, 2, 0, 3, 0, 4, 0, 5, 0],
"v4": [-5, 0, -4, 0, -3, 0, -2, 0, -1, 0]
}
for i in range(len(expected)):
output = feed.push({"v1": arr1[i], "v2": arr2[i]})
assert output == {
"v1": expected["v1"][i],
"v2": expected["v2"][i],
"v3": expected["v3"][i],
"v4": expected["v4"][i]
}
def test_bool_methods():
s = Stream.source(
["hello", "my", "name", "i", "am", "the", "data", "feed"],
dtype="string")
w1 = s.upper().rename("w1")
w2 = s.lower().rename("w2")
w3 = w1.endswith("E").invert().rename("w3")
feed = DataFeed([w1, w2, w3])
feed.compile()
assert feed.next() == {"w1": "HELLO", "w2": "hello", "w3": True}
def test_str_accessor():
s = Stream.source(
["hello", "my", "name", "i", "am", "the", "data", "feed"])
w1 = s.str.upper().rename("w1")
w2 = s.str.lower().rename("w2")
w3 = w1.str.endswith("E").rename("w3")
feed = DataFeed([w1, w2, w3])
feed.compile()
assert feed.next() == {"w1": "HELLO", "w2": "hello", "w3": False}
def create_env_with_price_series(config: dict, price_stream: Stream[np.array]):
np.seterr(divide="ignore")
action_scheme = PortfolioAllocation(num_assets=config["num_assets"],
commission=config["commission"])
total_steps = config["total_steps"] + config["min_periods"]
m = config["num_assets"]
p_streams = [
price_stream.apply(index(i), dtype="float").rename(f"p{i}")
for i in range(m)
]
lag_streams = [p_streams[i].lag().rename(f"l{i}") for i in range(m)]
prices = Stream.group(lag_streams).apply(
lambda x: np.array([x[f"l{i}"] for i in range(m)])).rename("prices")
features = []
for i in range(m):
with NameSpace(f"p{i}"):
features += make_features(p_streams[i])
r_streams = [(p_streams[i] / lag_streams[i]).rename(f"r{i}")
for i in range(m)]
y1 = Stream.group(r_streams).apply(
lambda x: np.array([x[f"r{i}"] for i in range(m)])).rename("y(t+1)")
weights = Stream.sensor(action_scheme,
lambda x: x.weights).rename("weights")
pv = Stream.sensor(action_scheme, lambda x: x.p,
dtype="float").rename("pv")
obs_group = Stream.group(features).rename("obs")
state_group = Stream.group([y1]).rename("state")
perform_group = Stream.group([prices, weights, pv]).rename("renderer")
feed = DataFeed([obs_group, state_group, perform_group])
env = TradingEnv(
action_scheme=action_scheme,
reward_scheme=Profit(),
observer=Observer(feed=feed,
window_size=config["window_size"],
num_assets=config["num_assets"],
min_periods=config["min_periods"]),
stopper=MaxStepsOrNoMoney(),
informer=RebalanceInformer(),
renderer=ReallocationChart(num_assets=config["num_assets"]))
return env
def test_pct_change():
configs = [
{"periods": 1, "fill_method": None},
{"periods": 1, "fill_method": "pad"},
{"periods": 1, "fill_method": "ffill"},
{"periods": 2, "fill_method": None},
{"periods": 2, "fill_method": "pad"},
{"periods": 2, "fill_method": "ffill"},
]
for array, config in product(arrays, configs):
s = Stream.source(array, dtype="float")
w = s.pct_change(**config).rename("w")
expected = list(pd.Series(array).pct_change(**config))
print(config)
assert_op([w], expected)
def test_boolean():
s = Stream.source(["hello", "my", "name", "is", "slim", "shady"],
dtype="string")
s.startswith("m").invert()
