|
1
|
|
|
# |
|
2
|
|
|
# Copyright 2013 Quantopian, Inc. |
|
3
|
|
|
# |
|
4
|
|
|
# Licensed under the Apache License, Version 2.0 (the "License"); |
|
5
|
|
|
# you may not use this file except in compliance with the License. |
|
6
|
|
|
# You may obtain a copy of the License at |
|
7
|
|
|
# |
|
8
|
|
|
# http://www.apache.org/licenses/LICENSE-2.0 |
|
9
|
|
|
# |
|
10
|
|
|
# Unless required by applicable law or agreed to in writing, software |
|
11
|
|
|
# distributed under the License is distributed on an "AS IS" BASIS, |
|
12
|
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
13
|
|
|
# See the License for the specific language governing permissions and |
|
14
|
|
|
# limitations under the License. |
|
15
|
|
|
|
|
16
|
|
|
import pytz |
|
17
|
|
|
import numpy as np |
|
18
|
|
|
import pandas as pd |
|
19
|
|
|
import talib |
|
20
|
|
|
|
|
21
|
|
|
from datetime import timedelta, datetime |
|
22
|
|
|
from unittest import TestCase, skip |
|
23
|
|
|
|
|
24
|
|
|
from zipline.utils.test_utils import setup_logger, teardown_logger |
|
25
|
|
|
|
|
26
|
|
|
import zipline.utils.factory as factory |
|
27
|
|
|
from zipline.finance.trading import TradingEnvironment |
|
28
|
|
|
|
|
29
|
|
|
from zipline.test_algorithms import TALIBAlgorithm |
|
30
|
|
|
|
|
31
|
|
|
import zipline.transforms.ta as ta |
|
32
|
|
|
|
|
33
|
|
|
|
|
34
|
|
|
class TestTALIB(TestCase): |
|
35
|
|
|
|
|
36
|
|
|
@classmethod |
|
37
|
|
|
def setUpClass(cls): |
|
38
|
|
|
cls.env = TradingEnvironment() |
|
39
|
|
|
|
|
40
|
|
|
@classmethod |
|
41
|
|
|
def tearDownClass(cls): |
|
42
|
|
|
del cls.env |
|
43
|
|
|
|
|
44
|
|
|
def setUp(self): |
|
45
|
|
|
setup_logger(self) |
|
46
|
|
|
sim_params = factory.create_simulation_parameters( |
|
47
|
|
|
start=datetime(1990, 1, 1, tzinfo=pytz.utc), |
|
48
|
|
|
end=datetime(1990, 3, 30, tzinfo=pytz.utc)) |
|
49
|
|
|
self.source, self.panel = \ |
|
50
|
|
|
factory.create_test_panel_ohlc_source(sim_params, self.env) |
|
51
|
|
|
|
|
52
|
|
|
def tearDown(self): |
|
53
|
|
|
teardown_logger(self) |
|
54
|
|
|
|
|
55
|
|
|
@skip |
|
56
|
|
|
def test_talib_with_default_params(self): |
|
57
|
|
|
BLACKLIST = ['make_transform', 'BatchTransform', |
|
58
|
|
|
# TODO: Figure out why MAVP generates a KeyError |
|
59
|
|
|
'MAVP'] |
|
60
|
|
|
names = [name for name in dir(ta) |
|
61
|
|
|
if name[0].isupper() and name not in BLACKLIST] |
|
62
|
|
|
|
|
63
|
|
|
for name in names: |
|
64
|
|
|
print(name) |
|
65
|
|
|
zipline_transform = getattr(ta, name)(sid=0) |
|
66
|
|
|
talib_fn = getattr(talib.abstract, name) |
|
67
|
|
|
|
|
68
|
|
|
start = datetime(1990, 1, 1, tzinfo=pytz.utc) |
|
69
|
|
|
end = start + timedelta(days=zipline_transform.lookback + 10) |
|
70
|
|
|
sim_params = factory.create_simulation_parameters( |
|
71
|
|
|
start=start, end=end) |
|
72
|
|
|
source, panel = \ |
|
73
|
|
|
factory.create_test_panel_ohlc_source(sim_params, self.env) |
|
74
|
|
|
|
|
75
|
|
|
algo = TALIBAlgorithm(talib=zipline_transform) |
|
76
|
|
|
algo.run(source) |
|
77
|
|
|
|
|
78
|
|
|
zipline_result = np.array( |
|
79
|
|
|
algo.talib_results[zipline_transform][-1]) |
|
80
|
|
|
|
|
81
|
|
|
talib_data = dict() |
|
82
|
|
|
data = zipline_transform.window |
|
83
|
|
|
# TODO: Figure out if we are clobbering the tests by this |
|
84
|
|
|
# protection against empty windows |
|
85
|
|
|
if not data: |
|
86
|
|
|
continue |
|
87
|
|
|
for key in ['open', 'high', 'low', 'volume']: |
|
88
|
|
|
if key in data: |
|
89
|
|
|
talib_data[key] = data[key][0].values |
|
90
|
|
|
talib_data['close'] = data['price'][0].values |
|
91
|
|
|
expected_result = talib_fn(talib_data) |
|
92
|
|
|
|
|
93
|
|
|
if isinstance(expected_result, list): |
|
94
|
|
|
expected_result = np.array([e[-1] for e in expected_result]) |
|
95
|
|
|
else: |
|
96
|
|
|
expected_result = np.array(expected_result[-1]) |
|
97
|
|
|
if not (np.all(np.isnan(zipline_result)) and |
|
98
|
|
|
np.all(np.isnan(expected_result))): |
|
99
|
|
|
self.assertTrue(np.allclose(zipline_result, expected_result)) |
|
100
|
|
|
else: |
|
101
|
|
|
print('--- NAN') |
|
102
|
|
|
|
|
103
|
|
|
# reset generator so next iteration has data |
|
104
|
|
|
# self.source, self.panel = \ |
|
105
|
|
|
# factory.create_test_panel_ohlc_source(self.sim_params) |
|
106
|
|
|
|
|
107
|
|
|
def test_multiple_talib_with_args(self): |
|
108
|
|
|
zipline_transforms = [ta.MA(timeperiod=10), |
|
109
|
|
|
ta.MA(timeperiod=25)] |
|
110
|
|
|
talib_fn = talib.abstract.MA |
|
111
|
|
|
algo = TALIBAlgorithm(talib=zipline_transforms, identifiers=[0]) |
|
112
|
|
|
algo.run(self.source) |
|
113
|
|
|
# Test if computed values match those computed by pandas rolling mean. |
|
114
|
|
|
sid = 0 |
|
115
|
|
|
talib_values = np.array([x[sid] for x in |
|
116
|
|
|
algo.talib_results[zipline_transforms[0]]]) |
|
117
|
|
|
np.testing.assert_array_equal(talib_values, |
|
118
|
|
|
pd.rolling_mean(self.panel[0]['price'], |
|
119
|
|
|
10).values) |
|
120
|
|
|
talib_values = np.array([x[sid] for x in |
|
121
|
|
|
algo.talib_results[zipline_transforms[1]]]) |
|
122
|
|
|
np.testing.assert_array_equal(talib_values, |
|
123
|
|
|
pd.rolling_mean(self.panel[0]['price'], |
|
124
|
|
|
25).values) |
|
125
|
|
|
for t in zipline_transforms: |
|
126
|
|
|
talib_result = np.array(algo.talib_results[t][-1]) |
|
127
|
|
|
talib_data = dict() |
|
128
|
|
|
data = t.window |
|
129
|
|
|
# TODO: Figure out if we are clobbering the tests by this |
|
130
|
|
|
# protection against empty windows |
|
131
|
|
|
if not data: |
|
132
|
|
|
continue |
|
133
|
|
|
for key in ['open', 'high', 'low', 'volume']: |
|
134
|
|
|
if key in data: |
|
135
|
|
|
talib_data[key] = data[key][0].values |
|
136
|
|
|
talib_data['close'] = data['price'][0].values |
|
137
|
|
|
expected_result = talib_fn(talib_data, **t.call_kwargs)[-1] |
|
138
|
|
|
np.testing.assert_allclose(talib_result, expected_result) |
|
139
|
|
|
|
|
140
|
|
|
def test_talib_with_minute_data(self): |
|
141
|
|
|
|
|
142
|
|
|
ma_one_day_minutes = ta.MA(timeperiod=10, bars='minute') |
|
143
|
|
|
|
|
144
|
|
|
# Assert that the BatchTransform window length is enough to cover |
|
145
|
|
|
# the amount of minutes in the timeperiod. |
|
146
|
|
|
|
|
147
|
|
|
# Here, 10 minutes only needs a window length of 1. |
|
148
|
|
|
self.assertEquals(1, ma_one_day_minutes.window_length) |
|
149
|
|
|
|
|
150
|
|
|
# With minutes greater than the 390, i.e. one trading day, we should |
|
151
|
|
|
# have a window_length of two days. |
|
152
|
|
|
ma_two_day_minutes = ta.MA(timeperiod=490, bars='minute') |
|
153
|
|
|
self.assertEquals(2, ma_two_day_minutes.window_length) |
|
154
|
|
|
|
|
155
|
|
|
# TODO: Ensure that the lookback into the datapanel is returning |
|
156
|
|
|
# expected results. |
|
157
|
|
|
# Requires supplying minute instead of day data to the unit test. |
|
158
|
|
|
# When adding test data, should add more minute events than the |
|
159
|
|
|
# timeperiod to ensure that lookback is behaving properly. |
|
160
|
|
|
|
quantopian /
zipline
