|
1
|
|
|
from abc import ( |
|
2
|
|
|
ABCMeta, |
|
3
|
|
|
abstractmethod, |
|
4
|
|
|
) |
|
5
|
|
|
import os |
|
6
|
|
|
from bcolz import ctable |
|
7
|
|
|
from datetime import datetime |
|
8
|
|
|
import numpy as np |
|
9
|
|
|
from numpy import float64 |
|
10
|
|
|
from os.path import join |
|
11
|
|
|
import pandas as pd |
|
12
|
|
|
from pandas import read_csv |
|
13
|
|
|
from six import with_metaclass |
|
14
|
|
|
|
|
15
|
|
|
from zipline.finance.trading import TradingEnvironment |
|
16
|
|
|
|
|
17
|
|
|
MINUTES_PER_DAY = 390 |
|
18
|
|
|
|
|
19
|
|
|
_writer_env = TradingEnvironment() |
|
20
|
|
|
|
|
21
|
|
|
|
|
22
|
|
|
class BcolzMinuteBarWriter(with_metaclass(ABCMeta)): |
|
23
|
|
|
""" |
|
24
|
|
|
Class capable of writing minute OHLCV data to disk into bcolz format. |
|
25
|
|
|
""" |
|
26
|
|
|
@abstractmethod |
|
27
|
|
|
def gen_frames(self, assets): |
|
28
|
|
|
""" |
|
29
|
|
|
Return an iterator of pairs of (asset_id, pd.dataframe). |
|
30
|
|
|
""" |
|
31
|
|
|
raise NotImplementedError() |
|
32
|
|
|
|
|
33
|
|
|
def write(self, directory, assets, sid_path_func=None): |
|
34
|
|
|
_iterator = self.gen_frames(assets) |
|
35
|
|
|
|
|
36
|
|
|
return self._write_internal(directory, _iterator, |
|
37
|
|
|
sid_path_func=sid_path_func) |
|
38
|
|
|
|
|
39
|
|
|
@staticmethod |
|
40
|
|
|
def full_minutes_for_days(env, dt1, dt2): |
|
41
|
|
|
start_date = env.normalize_date(dt1) |
|
42
|
|
|
end_date = env.normalize_date(dt2) |
|
43
|
|
|
|
|
44
|
|
|
all_minutes = [] |
|
45
|
|
|
|
|
46
|
|
|
for day in env.days_in_range(start_date, end_date): |
|
47
|
|
|
minutes_in_day = pd.date_range( |
|
48
|
|
|
start=pd.Timestamp( |
|
49
|
|
|
datetime( |
|
50
|
|
|
year=day.year, |
|
51
|
|
|
month=day.month, |
|
52
|
|
|
day=day.day, |
|
53
|
|
|
hour=9, |
|
54
|
|
|
minute=31), |
|
55
|
|
|
tz='US/Eastern').tz_convert('UTC'), |
|
56
|
|
|
periods=390, |
|
57
|
|
|
freq="min" |
|
58
|
|
|
) |
|
59
|
|
|
|
|
60
|
|
|
all_minutes.append(minutes_in_day) |
|
61
|
|
|
|
|
62
|
|
|
# flatten |
|
63
|
|
|
return pd.DatetimeIndex( |
|
64
|
|
|
np.concatenate(all_minutes), copy=False, tz='UTC' |
|
65
|
|
|
) |
|
66
|
|
|
|
|
67
|
|
|
def _write_internal(self, directory, iterator, sid_path_func=None): |
|
68
|
|
|
first_open = pd.Timestamp( |
|
69
|
|
|
datetime( |
|
70
|
|
|
year=2002, |
|
71
|
|
|
month=1, |
|
72
|
|
|
day=2, |
|
73
|
|
|
hour=9, |
|
74
|
|
|
minute=31 |
|
75
|
|
|
), tz='US/Eastern').tz_convert('UTC') |
|
76
|
|
|
|
|
77
|
|
|
for asset_id, df in iterator: |
|
78
|
|
|
if sid_path_func is None: |
|
79
|
|
|
path = join(directory, "{0}.bcolz".format(asset_id)) |
|
80
|
|
|
else: |
|
81
|
|
|
path = sid_path_func(directory, asset_id) |
|
82
|
|
|
|
|
83
|
|
|
os.makedirs(path) |
|
84
|
|
|
|
|
85
|
|
|
minutes = self.full_minutes_for_days(_writer_env, |
|
86
|
|
|
first_open, df.index[-1]) |
|
87
|
|
|
minutes_count = len(minutes) |
|
88
|
|
|
|
|
89
|
|
|
dt_col = np.zeros(minutes_count, dtype=np.uint32) |
|
90
|
|
|
open_col = np.zeros(minutes_count, dtype=np.uint32) |
|
91
|
|
|
high_col = np.zeros(minutes_count, dtype=np.uint32) |
|
92
|
|
|
low_col = np.zeros(minutes_count, dtype=np.uint32) |
|
93
|
|
|
close_col = np.zeros(minutes_count, dtype=np.uint32) |
|
94
|
|
|
vol_col = np.zeros(minutes_count, dtype=np.uint32) |
|
95
|
|
|
|
|
96
|
|
|
for row in df.iterrows(): |
|
97
|
|
|
dt = row[0] |
|
98
|
|
|
idx = minutes.searchsorted(dt) |
|
99
|
|
|
|
|
100
|
|
|
dt_col[idx] = dt.value / 1e9 |
|
101
|
|
|
open_col[idx] = row[1].loc["open"] |
|
102
|
|
|
high_col[idx] = row[1].loc["high"] |
|
103
|
|
|
low_col[idx] = row[1].loc["low"] |
|
104
|
|
|
close_col[idx] = row[1].loc["close"] |
|
105
|
|
|
vol_col[idx] = row[1].loc["volume"] |
|
106
|
|
|
|
|
107
|
|
|
ctable( |
|
108
|
|
|
columns=[ |
|
109
|
|
|
open_col, |
|
110
|
|
|
high_col, |
|
111
|
|
|
low_col, |
|
112
|
|
|
close_col, |
|
113
|
|
|
vol_col, |
|
114
|
|
|
dt_col |
|
115
|
|
|
], |
|
116
|
|
|
names=[ |
|
117
|
|
|
"open", |
|
118
|
|
|
"high", |
|
119
|
|
|
"low", |
|
120
|
|
|
"close", |
|
121
|
|
|
"volume", |
|
122
|
|
|
"dt" |
|
123
|
|
|
], |
|
124
|
|
|
rootdir=path, |
|
125
|
|
|
mode='w' |
|
126
|
|
|
) |
|
127
|
|
|
|
|
128
|
|
|
|
|
129
|
|
|
class MinuteBarWriterFromDataFrames(BcolzMinuteBarWriter): |
|
130
|
|
|
_csv_dtypes = { |
|
131
|
|
|
'open': float64, |
|
132
|
|
|
'high': float64, |
|
133
|
|
|
'low': float64, |
|
134
|
|
|
'close': float64, |
|
135
|
|
|
'volume': float64, |
|
136
|
|
|
} |
|
137
|
|
|
|
|
138
|
|
|
def gen_frames(self, assets): |
|
139
|
|
|
for asset in assets: |
|
140
|
|
|
df = assets[asset] |
|
141
|
|
|
yield asset, df.set_index("minute") |
|
142
|
|
|
|
|
143
|
|
|
|
|
144
|
|
|
class MinuteBarWriterFromCSVs(BcolzMinuteBarWriter): |
|
145
|
|
|
""" |
|
146
|
|
|
BcolzMinuteBarWriter constructed from a map of CSVs to assets. |
|
147
|
|
|
|
|
148
|
|
|
Parameters |
|
149
|
|
|
---------- |
|
150
|
|
|
asset_map: dict |
|
151
|
|
|
A map from asset_id -> path to csv with data for that asset. |
|
152
|
|
|
|
|
153
|
|
|
CSVs should have the following columns: |
|
154
|
|
|
minute : datetime64 |
|
155
|
|
|
open : float64 |
|
156
|
|
|
high : float64 |
|
157
|
|
|
low : float64 |
|
158
|
|
|
close : float64 |
|
159
|
|
|
volume : int64 |
|
160
|
|
|
""" |
|
161
|
|
|
_csv_dtypes = { |
|
162
|
|
|
'open': float64, |
|
163
|
|
|
'high': float64, |
|
164
|
|
|
'low': float64, |
|
165
|
|
|
'close': float64, |
|
166
|
|
|
'volume': float64, |
|
167
|
|
|
} |
|
168
|
|
|
|
|
169
|
|
|
def __init__(self, asset_map): |
|
170
|
|
|
self._asset_map = asset_map |
|
171
|
|
|
|
|
172
|
|
|
def gen_frames(self, assets): |
|
173
|
|
|
""" |
|
174
|
|
|
Read CSVs as DataFrames from our asset map. |
|
175
|
|
|
""" |
|
176
|
|
|
dtypes = self._csv_dtypes |
|
177
|
|
|
|
|
178
|
|
|
for asset in assets: |
|
179
|
|
|
path = self._asset_map.get(asset) |
|
180
|
|
|
if path is None: |
|
181
|
|
|
raise KeyError("No path supplied for asset %s" % asset) |
|
182
|
|
|
df = read_csv(path, parse_dates=['minute'], dtype=dtypes) |
|
183
|
|
|
df = df.set_index("minute").tz_localize("UTC") |
|
184
|
|
|
|
|
185
|
|
|
yield asset, df |
|
186
|
|
|
|
quantopian /
zipline
