1
|
|
|
import datetime |
2
|
|
|
|
3
|
|
|
import numpy as np |
4
|
|
|
import pandas as pd |
5
|
|
|
from six import iteritems |
6
|
|
|
from six.moves import zip |
7
|
|
|
|
8
|
|
|
from zipline.utils.numpy_utils import np_NaT |
9
|
|
|
|
10
|
|
|
|
11
|
|
|
def next_date_frame(dates, events_by_sid): |
12
|
|
|
""" |
13
|
|
|
Make a DataFrame representing the simulated next known date for an event. |
14
|
|
|
|
15
|
|
|
Parameters |
16
|
|
|
---------- |
17
|
|
|
dates : pd.DatetimeIndex. |
18
|
|
|
The index of the returned DataFrame. |
19
|
|
|
events_by_sid : dict[int -> pd.Series] |
20
|
|
|
Dict mapping sids to a series of dates. Each k:v pair of the series |
21
|
|
|
represents the date we learned of the event mapping to the date the |
22
|
|
|
event will occur. |
23
|
|
|
Returns |
24
|
|
|
------- |
25
|
|
|
next_events: pd.DataFrame |
26
|
|
|
A DataFrame where each column is a security from `events_by_sid` where |
27
|
|
|
the values are the dates of the next known event with the knowledge we |
28
|
|
|
had on the date of the index. Entries falling after the last date will |
29
|
|
|
have `NaT` as the result in the output. |
30
|
|
|
|
31
|
|
|
|
32
|
|
|
See Also |
33
|
|
|
-------- |
34
|
|
|
previous_date_frame |
35
|
|
|
""" |
36
|
|
|
cols = { |
37
|
|
|
equity: np.full_like(dates, np_NaT) for equity in events_by_sid |
38
|
|
|
} |
39
|
|
|
raw_dates = dates.values |
40
|
|
|
for equity, event_dates in iteritems(events_by_sid): |
41
|
|
|
data = cols[equity] |
42
|
|
|
if not event_dates.index.is_monotonic_increasing: |
43
|
|
|
event_dates = event_dates.sort_index() |
44
|
|
|
|
45
|
|
|
# Iterate over the raw Series values, since we're comparing against |
46
|
|
|
# numpy arrays anyway. |
47
|
|
|
iterkv = zip(event_dates.index.values, event_dates.values) |
48
|
|
|
for knowledge_date, event_date in iterkv: |
49
|
|
|
date_mask = ( |
50
|
|
|
(knowledge_date <= raw_dates) & |
51
|
|
|
(raw_dates <= event_date) |
52
|
|
|
) |
53
|
|
|
value_mask = (event_date <= data) | (data == np_NaT) |
54
|
|
|
data[date_mask & value_mask] = event_date |
55
|
|
|
|
56
|
|
|
return pd.DataFrame(index=dates, data=cols) |
57
|
|
|
|
58
|
|
|
|
59
|
|
|
def previous_date_frame(date_index, events_by_sid): |
60
|
|
|
""" |
61
|
|
|
Make a DataFrame representing simulated next earnings date_index. |
62
|
|
|
|
63
|
|
|
Parameters |
64
|
|
|
---------- |
65
|
|
|
date_index : DatetimeIndex. |
66
|
|
|
The index of the returned DataFrame. |
67
|
|
|
events_by_sid : dict[int -> DatetimeIndex] |
68
|
|
|
Dict mapping sids to a series of dates. Each k:v pair of the series |
69
|
|
|
represents the date we learned of the event mapping to the date the |
70
|
|
|
event will occur. |
71
|
|
|
|
72
|
|
|
Returns |
73
|
|
|
------- |
74
|
|
|
previous_events: pd.DataFrame |
75
|
|
|
A DataFrame where each column is a security from `events_by_sid` where |
76
|
|
|
the values are the dates of the previous event that occured on the date |
77
|
|
|
of the index. Entries falling before the first date will have `NaT` as |
78
|
|
|
the result in the output. |
79
|
|
|
|
80
|
|
|
See Also |
81
|
|
|
-------- |
82
|
|
|
next_date_frame |
83
|
|
|
""" |
84
|
|
|
sids = list(events_by_sid) |
85
|
|
|
out = np.full((len(date_index), len(sids)), np_NaT, dtype='datetime64[ns]') |
86
|
|
|
dn = date_index[-1].asm8 |
87
|
|
|
for col_idx, sid in enumerate(sids): |
88
|
|
|
# events_by_sid[sid] is Series mapping knowledge_date to actual |
89
|
|
|
# event_date. We don't care about the knowledge date for |
90
|
|
|
# computing previous earnings. |
91
|
|
|
values = events_by_sid[sid].values |
92
|
|
|
values = values[values <= dn] |
93
|
|
|
out[date_index.searchsorted(values), col_idx] = values |
94
|
|
|
|
95
|
|
|
frame = pd.DataFrame(out, index=date_index, columns=sids) |
96
|
|
|
frame.ffill(inplace=True) |
97
|
|
|
return frame |
98
|
|
|
|
99
|
|
|
|
100
|
|
|
def normalize_data_query_time(dt, time, tz): |
101
|
|
|
"""Apply the correct time and timezone to a date. |
102
|
|
|
|
103
|
|
|
Parameters |
104
|
|
|
---------- |
105
|
|
|
dt : pd.Timestamp |
106
|
|
|
The original datetime that represents the date. |
107
|
|
|
time : datetime.time |
108
|
|
|
The time of day to use as the cutoff point for new data. Data points |
109
|
|
|
that you learn about after this time will become available to your |
110
|
|
|
algorithm on the next trading day. |
111
|
|
|
tz : tzinfo |
112
|
|
|
The timezone to normalize your dates to before comparing against |
113
|
|
|
`time`. |
114
|
|
|
|
115
|
|
|
Returns |
116
|
|
|
------- |
117
|
|
|
query_dt : pd.Timestamp |
118
|
|
|
The timestamp with the correct time and date in utc. |
119
|
|
|
""" |
120
|
|
|
# merge the correct date with the time in the given timezone then convert |
121
|
|
|
# back to utc |
122
|
|
|
return pd.Timestamp( |
123
|
|
|
datetime.datetime.combine(dt.date(), time), |
124
|
|
|
tz=tz, |
125
|
|
|
).tz_convert('utc') |
126
|
|
|
|
127
|
|
|
|
128
|
|
|
def normalize_timestamp_to_query_time(df, |
129
|
|
|
time, |
130
|
|
|
tz, |
131
|
|
|
inplace=False, |
132
|
|
|
ts_field='timestamp'): |
133
|
|
|
"""Update the timestamp field of a dataframe to normalize dates around |
134
|
|
|
some data query time/timezone. |
135
|
|
|
|
136
|
|
|
Parameters |
137
|
|
|
---------- |
138
|
|
|
df : pd.DataFrame |
139
|
|
|
The dataframe to update. This needs a column named ``ts_field``. |
140
|
|
|
time : datetime.time |
141
|
|
|
The time of day to use as the cutoff point for new data. Data points |
142
|
|
|
that you learn about after this time will become available to your |
143
|
|
|
algorithm on the next trading day. |
144
|
|
|
tz : tzinfo |
145
|
|
|
The timezone to normalize your dates to before comparing against |
146
|
|
|
`time`. |
147
|
|
|
inplace : bool, optional |
148
|
|
|
Update the dataframe in place. |
149
|
|
|
ts_field : str, optional |
150
|
|
|
The name of the timestamp field in ``df``. |
151
|
|
|
|
152
|
|
|
Returns |
153
|
|
|
------- |
154
|
|
|
df : pd.DataFrame |
155
|
|
|
The dataframe with the timestamp field normalized. If ``inplace`` is |
156
|
|
|
true, then this will be the same object as ``df`` otherwise this will |
157
|
|
|
be a copy. |
158
|
|
|
""" |
159
|
|
|
if not inplace: |
160
|
|
|
# don't mutate the dataframe in place |
161
|
|
|
df = df.copy() |
162
|
|
|
|
163
|
|
|
dtidx = pd.DatetimeIndex(df.loc[:, ts_field], tz='utc') |
164
|
|
|
dtidx_local_time = dtidx.tz_convert(tz) |
165
|
|
|
to_roll_forward = dtidx_local_time.time > time |
166
|
|
|
# for all of the times that are greater than our query time add 1 |
167
|
|
|
# day and truncate to the date |
168
|
|
|
df.loc[to_roll_forward, ts_field] = ( |
169
|
|
|
dtidx_local_time[to_roll_forward] + datetime.timedelta(days=1) |
170
|
|
|
).normalize().tz_localize(None).tz_localize('utc') # cast back to utc |
171
|
|
|
df.loc[~to_roll_forward, ts_field] = dtidx[~to_roll_forward].normalize() |
172
|
|
|
return df |
173
|
|
|
|