1
|
|
|
# |
2
|
|
|
# Copyright 2014 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
|
|
|
from copy import copy |
17
|
|
|
import six |
18
|
|
|
|
19
|
|
|
import numpy as np |
20
|
|
|
from datetime import timedelta |
21
|
|
|
import pandas as pd |
22
|
|
|
|
23
|
|
|
from zipline.sources.data_source import DataSource |
24
|
|
|
from zipline.utils import tradingcalendar as calendar_nyse |
25
|
|
|
from zipline.gens.utils import hash_args |
26
|
|
|
|
27
|
|
|
|
28
|
|
|
class RandomWalkSource(DataSource): |
29
|
|
|
"""RandomWalkSource that emits events with prices that follow a |
30
|
|
|
random walk. Will generate valid datetimes that match market hours |
31
|
|
|
of the supplied calendar and can generate emit events with |
32
|
|
|
user-defined frequencies (e.g. minutely). |
33
|
|
|
|
34
|
|
|
""" |
35
|
|
|
VALID_FREQS = frozenset(('daily', 'minute')) |
36
|
|
|
|
37
|
|
|
def __init__(self, start_prices=None, freq='minute', start=None, |
38
|
|
|
end=None, drift=0.1, sd=0.1, calendar=calendar_nyse): |
39
|
|
|
""" |
40
|
|
|
:Arguments: |
41
|
|
|
start_prices : dict |
42
|
|
|
sid -> starting price. |
43
|
|
|
Default: {0: 100, 1: 500} |
44
|
|
|
freq : str <default='minute'> |
45
|
|
|
Emits events according to freq. |
46
|
|
|
Can be 'daily' or 'minute' |
47
|
|
|
start : datetime <default=start of calendar> |
48
|
|
|
Start dt to emit events. |
49
|
|
|
end : datetime <default=end of calendar> |
50
|
|
|
End dt until to which emit events. |
51
|
|
|
drift: float <default=0.1> |
52
|
|
|
Constant drift of the price series. |
53
|
|
|
sd: float <default=0.1> |
54
|
|
|
Standard deviation of the price series. |
55
|
|
|
calendar : calendar object <default: NYSE> |
56
|
|
|
Calendar to use. |
57
|
|
|
See zipline.utils for different choices. |
58
|
|
|
|
59
|
|
|
:Example: |
60
|
|
|
# Assumes you have instantiated your Algorithm |
61
|
|
|
# as myalgo. |
62
|
|
|
myalgo = MyAlgo() |
63
|
|
|
source = RandomWalkSource() |
64
|
|
|
myalgo.run(source) |
65
|
|
|
|
66
|
|
|
""" |
67
|
|
|
# Hash_value for downstream sorting. |
68
|
|
|
self.arg_string = hash_args(start_prices, freq, start, end, |
69
|
|
|
calendar.__name__) |
70
|
|
|
|
71
|
|
|
if freq not in self.VALID_FREQS: |
72
|
|
|
raise ValueError('%s not in %s' % (freq, self.VALID_FREQS)) |
73
|
|
|
|
74
|
|
|
self.freq = freq |
75
|
|
|
if start_prices is None: |
76
|
|
|
self.start_prices = {0: 100, |
77
|
|
|
1: 500} |
78
|
|
|
else: |
79
|
|
|
self.start_prices = start_prices |
80
|
|
|
|
81
|
|
|
self.calendar = calendar |
82
|
|
|
if start is None: |
83
|
|
|
self.start = calendar.start |
84
|
|
|
else: |
85
|
|
|
self.start = start |
86
|
|
|
if end is None: |
87
|
|
|
self.end = calendar.end_base |
88
|
|
|
else: |
89
|
|
|
self.end = end |
90
|
|
|
|
91
|
|
|
self.drift = drift |
92
|
|
|
self.sd = sd |
93
|
|
|
|
94
|
|
|
self.sids = self.start_prices.keys() |
95
|
|
|
|
96
|
|
|
self.open_and_closes = \ |
97
|
|
|
calendar.open_and_closes[self.start:self.end] |
98
|
|
|
|
99
|
|
|
self._raw_data = None |
100
|
|
|
|
101
|
|
|
@property |
102
|
|
|
def instance_hash(self): |
103
|
|
|
return self.arg_string |
104
|
|
|
|
105
|
|
|
@property |
106
|
|
|
def mapping(self): |
107
|
|
|
return { |
108
|
|
|
'dt': (lambda x: x, 'dt'), |
109
|
|
|
'sid': (lambda x: x, 'sid'), |
110
|
|
|
'price': (float, 'price'), |
111
|
|
|
'volume': (int, 'volume'), |
112
|
|
|
'open_price': (float, 'open_price'), |
113
|
|
|
'high': (float, 'high'), |
114
|
|
|
'low': (float, 'low'), |
115
|
|
|
} |
116
|
|
|
|
117
|
|
|
def _gen_next_step(self, x): |
118
|
|
|
x += np.random.randn() * self.sd + self.drift |
119
|
|
|
return max(x, 0.1) |
120
|
|
|
|
121
|
|
|
def _gen_events(self, cur_prices, current_dt): |
122
|
|
|
for sid, price in six.iteritems(cur_prices): |
123
|
|
|
cur_prices[sid] = self._gen_next_step(cur_prices[sid]) |
124
|
|
|
|
125
|
|
|
event = { |
126
|
|
|
'dt': current_dt, |
127
|
|
|
'sid': sid, |
128
|
|
|
'price': cur_prices[sid], |
129
|
|
|
'volume': np.random.randint(1e5, 1e6), |
130
|
|
|
'open_price': cur_prices[sid], |
131
|
|
|
'high': cur_prices[sid] + .1, |
132
|
|
|
'low': cur_prices[sid] - .1, |
133
|
|
|
} |
134
|
|
|
|
135
|
|
|
yield event |
136
|
|
|
|
137
|
|
|
def raw_data_gen(self): |
138
|
|
|
cur_prices = copy(self.start_prices) |
139
|
|
|
for _, (open_dt, close_dt) in self.open_and_closes.iterrows(): |
140
|
|
|
current_dt = copy(open_dt) |
141
|
|
|
if self.freq == 'minute': |
142
|
|
|
# Emit minutely trade signals from open to close |
143
|
|
|
while current_dt <= close_dt: |
144
|
|
|
for event in self._gen_events(cur_prices, current_dt): |
145
|
|
|
yield event |
146
|
|
|
current_dt += timedelta(minutes=1) |
147
|
|
|
elif self.freq == 'daily': |
148
|
|
|
# Emit one signal per day at close |
149
|
|
|
for event in self._gen_events( |
150
|
|
|
cur_prices, pd.tslib.normalize_date(close_dt)): |
151
|
|
|
yield event |
152
|
|
|
|
153
|
|
|
@property |
154
|
|
|
def raw_data(self): |
155
|
|
|
if not self._raw_data: |
156
|
|
|
self._raw_data = self.raw_data_gen() |
157
|
|
|
return self._raw_data |
158
|
|
|
|