mirror of
https://github.com/wassname/catalyst.git
synced 2026-06-29 05:15:44 +08:00
Dale Jung
38e8d5214d
Limited use of `pandas` data structures in both `HistoryContainer` and `RollingPanel`. Where possible, methods were amended to return raw `ndarrays` with the indexing logic done separately. This allows us to cut down the number of times pandas objects are created both as returns and intermediate values. The separation of indexing from data access allowed us to minimize the times we’d make use of pandas indexes. This required that that certain methods like `NDFrame.ffill` be replaced with versions that work with `ndarrays`. Some of this was done via straight numpy methods and others by access pandas internal machinery. Outside of allowing us to use faster ndarrays, many of these function provided speedups over their pandas counterparts as we didn’t require the extra features like handling multiple dtypes. i.e. np.isnan is faster than pd.isnull, but only works with certain dtypes.
394 lines
12 KiB
Python
394 lines
12 KiB
Python
#
|
|
# Copyright 2013 Quantopian, Inc.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
|
|
import datetime
|
|
|
|
import numpy as np
|
|
import pandas as pd
|
|
from copy import deepcopy
|
|
|
|
|
|
def _ensure_index(x):
|
|
if not isinstance(x, pd.Index):
|
|
x = pd.Index(sorted(x))
|
|
|
|
return x
|
|
|
|
|
|
class RollingPanel(object):
|
|
"""
|
|
Preallocation strategies for rolling window over expanding data set
|
|
|
|
Restrictions: major_axis can only be a DatetimeIndex for now
|
|
"""
|
|
|
|
def __init__(self,
|
|
window,
|
|
items,
|
|
sids,
|
|
cap_multiple=2,
|
|
dtype=np.float64,
|
|
initial_dates=None):
|
|
|
|
self._pos = window
|
|
self._window = window
|
|
|
|
self.items = _ensure_index(items)
|
|
self.minor_axis = _ensure_index(sids)
|
|
|
|
self.cap_multiple = cap_multiple
|
|
|
|
self.dtype = dtype
|
|
if initial_dates is None:
|
|
self.date_buf = np.empty(self.cap, dtype='M8[ns]') * pd.NaT
|
|
elif len(initial_dates) != window:
|
|
raise ValueError('initial_dates must be of length window')
|
|
else:
|
|
self.date_buf = np.hstack(
|
|
(
|
|
initial_dates,
|
|
np.empty(
|
|
window * (cap_multiple - 1),
|
|
dtype='datetime64[ns]',
|
|
),
|
|
),
|
|
)
|
|
|
|
self.buffer = self._create_buffer()
|
|
|
|
@property
|
|
def cap(self):
|
|
return self.cap_multiple * self._window
|
|
|
|
@property
|
|
def _start_index(self):
|
|
return self._pos - self._window
|
|
|
|
@property
|
|
def start_date(self):
|
|
return self.date_buf[self._start_index]
|
|
|
|
def oldest_frame(self, raw=False):
|
|
"""
|
|
Get the oldest frame in the panel.
|
|
"""
|
|
if raw:
|
|
return self.buffer.values[:, self._start_index, :]
|
|
return self.buffer.iloc[:, self._start_index, :]
|
|
|
|
def set_minor_axis(self, minor_axis):
|
|
self.minor_axis = _ensure_index(minor_axis)
|
|
self.buffer = self.buffer.reindex(minor_axis=self.minor_axis)
|
|
|
|
def set_items(self, items):
|
|
self.items = _ensure_index(items)
|
|
self.buffer = self.buffer.reindex(items=self.items)
|
|
|
|
def _create_buffer(self):
|
|
panel = pd.Panel(
|
|
items=self.items,
|
|
minor_axis=self.minor_axis,
|
|
major_axis=range(self.cap),
|
|
dtype=self.dtype,
|
|
)
|
|
return panel
|
|
|
|
def extend_back(self, missing_dts):
|
|
"""
|
|
Resizes the buffer to hold a new window with a new cap_multiple.
|
|
If cap_multiple is None, then the old cap_multiple is used.
|
|
"""
|
|
delta = len(missing_dts)
|
|
|
|
if not delta:
|
|
raise ValueError(
|
|
'missing_dts must be a non-empty index',
|
|
)
|
|
|
|
self._window += delta
|
|
|
|
self._pos += delta
|
|
|
|
self.date_buf = self.date_buf.copy()
|
|
self.date_buf.resize(self.cap)
|
|
self.date_buf = np.roll(self.date_buf, delta)
|
|
|
|
old_vals = self.buffer.values
|
|
shape = old_vals.shape
|
|
nan_arr = np.empty((shape[0], delta, shape[2]))
|
|
nan_arr.fill(np.nan)
|
|
|
|
new_vals = np.column_stack(
|
|
(nan_arr,
|
|
old_vals,
|
|
np.empty((shape[0], delta * (self.cap_multiple - 1), shape[2]))),
|
|
)
|
|
|
|
self.buffer = pd.Panel(
|
|
data=new_vals,
|
|
items=self.items,
|
|
minor_axis=self.minor_axis,
|
|
major_axis=np.arange(self.cap),
|
|
dtype=self.dtype,
|
|
)
|
|
|
|
# Fill the delta with the dates we calculated.
|
|
where = slice(self._start_index, self._start_index + delta)
|
|
self.date_buf[where] = missing_dts
|
|
|
|
def add_frame(self, tick, frame, minor_axis=None, items=None):
|
|
"""
|
|
"""
|
|
if self._pos == self.cap:
|
|
self._roll_data()
|
|
|
|
values = frame
|
|
if isinstance(frame, pd.DataFrame):
|
|
values = frame.values
|
|
|
|
self.buffer.values[:, self._pos, :] = values.astype(self.dtype)
|
|
self.date_buf[self._pos] = tick
|
|
|
|
self._pos += 1
|
|
|
|
def get_current(self, item=None, raw=False, start=None, end=None):
|
|
"""
|
|
Get a Panel that is the current data in view. It is not safe to persist
|
|
these objects because internal data might change
|
|
"""
|
|
item_indexer = slice(None)
|
|
if item:
|
|
item_indexer = self.items.get_loc(item)
|
|
|
|
start_index = self._start_index
|
|
end_index = self._pos
|
|
|
|
# get inital date window
|
|
where = slice(start_index, end_index)
|
|
current_dates = self.date_buf[where]
|
|
|
|
def convert_datelike_to_long(dt):
|
|
if isinstance(dt, pd.Timestamp):
|
|
return dt.asm8
|
|
if isinstance(dt, datetime.datetime):
|
|
return np.datetime64(dt)
|
|
return dt
|
|
|
|
# constrict further by date
|
|
if start:
|
|
start = convert_datelike_to_long(start)
|
|
start_index += current_dates.searchsorted(start)
|
|
|
|
if end:
|
|
end = convert_datelike_to_long(end)
|
|
_end = current_dates.searchsorted(end, 'right')
|
|
end_index -= len(current_dates) - _end
|
|
|
|
where = slice(start_index, end_index)
|
|
|
|
values = self.buffer.values[item_indexer, where, :]
|
|
current_dates = self.date_buf[where]
|
|
|
|
if raw:
|
|
# return copy so we can change it without side effects here
|
|
return values.copy()
|
|
|
|
major_axis = pd.DatetimeIndex(deepcopy(current_dates), tz='utc')
|
|
if values.ndim == 3:
|
|
return pd.Panel(values, self.items, major_axis, self.minor_axis,
|
|
dtype=self.dtype)
|
|
|
|
elif values.ndim == 2:
|
|
return pd.DataFrame(values, major_axis, self.minor_axis,
|
|
dtype=self.dtype)
|
|
|
|
def set_current(self, panel):
|
|
"""
|
|
Set the values stored in our current in-view data to be values of the
|
|
passed panel. The passed panel must have the same indices as the panel
|
|
that would be returned by self.get_current.
|
|
"""
|
|
where = slice(self._start_index, self._pos)
|
|
self.buffer.values[:, where, :] = panel.values
|
|
|
|
def current_dates(self):
|
|
where = slice(self._start_index, self._pos)
|
|
return pd.DatetimeIndex(deepcopy(self.date_buf[where]), tz='utc')
|
|
|
|
def _roll_data(self):
|
|
"""
|
|
Roll window worth of data up to position zero.
|
|
Save the effort of having to expensively roll at each iteration
|
|
"""
|
|
|
|
self.buffer.values[:, :self._window, :] = \
|
|
self.buffer.values[:, -self._window:, :]
|
|
self.date_buf[:self._window] = self.date_buf[-self._window:]
|
|
self._pos = self._window
|
|
|
|
@property
|
|
def window_length(self):
|
|
return self._window
|
|
|
|
|
|
class MutableIndexRollingPanel(object):
|
|
"""
|
|
A version of RollingPanel that exists for backwards compatibility with
|
|
batch_transform. This is a copy to allow behavior of RollingPanel to drift
|
|
away from this without breaking this class.
|
|
|
|
This code should be considered frozen, and should not be used in the
|
|
future. Instead, see RollingPanel.
|
|
"""
|
|
def __init__(self, window, items, sids, cap_multiple=2, dtype=np.float64):
|
|
|
|
self._pos = 0
|
|
self._window = window
|
|
|
|
self.items = _ensure_index(items)
|
|
self.minor_axis = _ensure_index(sids)
|
|
|
|
self.cap_multiple = cap_multiple
|
|
self.cap = cap_multiple * window
|
|
|
|
self.dtype = dtype
|
|
self.date_buf = np.empty(self.cap, dtype='M8[ns]')
|
|
|
|
self.buffer = self._create_buffer()
|
|
|
|
def _oldest_frame_idx(self):
|
|
return max(self._pos - self._window, 0)
|
|
|
|
def oldest_frame(self, raw=False):
|
|
"""
|
|
Get the oldest frame in the panel.
|
|
"""
|
|
if raw:
|
|
return self.buffer.values[:, self._oldest_frame_idx(), :]
|
|
return self.buffer.iloc[:, self._oldest_frame_idx(), :]
|
|
|
|
def set_sids(self, sids):
|
|
self.minor_axis = _ensure_index(sids)
|
|
self.buffer = self.buffer.reindex(minor_axis=self.minor_axis)
|
|
|
|
def _create_buffer(self):
|
|
panel = pd.Panel(
|
|
items=self.items,
|
|
minor_axis=self.minor_axis,
|
|
major_axis=range(self.cap),
|
|
dtype=self.dtype,
|
|
)
|
|
return panel
|
|
|
|
def get_current(self):
|
|
"""
|
|
Get a Panel that is the current data in view. It is not safe to persist
|
|
these objects because internal data might change
|
|
"""
|
|
|
|
where = slice(self._oldest_frame_idx(), self._pos)
|
|
major_axis = pd.DatetimeIndex(deepcopy(self.date_buf[where]), tz='utc')
|
|
return pd.Panel(self.buffer.values[:, where, :], self.items,
|
|
major_axis, self.minor_axis, dtype=self.dtype)
|
|
|
|
def set_current(self, panel):
|
|
"""
|
|
Set the values stored in our current in-view data to be values of the
|
|
passed panel. The passed panel must have the same indices as the panel
|
|
that would be returned by self.get_current.
|
|
"""
|
|
where = slice(self._oldest_frame_idx(), self._pos)
|
|
self.buffer.values[:, where, :] = panel.values
|
|
|
|
def current_dates(self):
|
|
where = slice(self._oldest_frame_idx(), self._pos)
|
|
return pd.DatetimeIndex(deepcopy(self.date_buf[where]), tz='utc')
|
|
|
|
def _roll_data(self):
|
|
"""
|
|
Roll window worth of data up to position zero.
|
|
Save the effort of having to expensively roll at each iteration
|
|
"""
|
|
|
|
self.buffer.values[:, :self._window, :] = \
|
|
self.buffer.values[:, -self._window:, :]
|
|
self.date_buf[:self._window] = self.date_buf[-self._window:]
|
|
self._pos = self._window
|
|
|
|
def add_frame(self, tick, frame, minor_axis=None, items=None):
|
|
"""
|
|
"""
|
|
if self._pos == self.cap:
|
|
self._roll_data()
|
|
|
|
if isinstance(frame, pd.DataFrame):
|
|
minor_axis = frame.columns
|
|
items = frame.index
|
|
|
|
if set(minor_axis).difference(set(self.minor_axis)) or \
|
|
set(items).difference(set(self.items)):
|
|
self._update_buffer(frame)
|
|
|
|
vals = frame.T.astype(self.dtype)
|
|
self.buffer.loc[:, self._pos, :] = vals
|
|
self.date_buf[self._pos] = tick
|
|
|
|
self._pos += 1
|
|
|
|
def _update_buffer(self, frame):
|
|
|
|
# Get current frame as we only need to care about the data that is in
|
|
# the active window
|
|
old_buffer = self.get_current()
|
|
if self._pos >= self._window:
|
|
# Don't count the last major_axis entry if we're past our window,
|
|
# since it's about to roll off the end of the panel.
|
|
old_buffer = old_buffer.iloc[:, 1:, :]
|
|
|
|
nans = pd.isnull(old_buffer)
|
|
|
|
# Find minor_axes that have only nans
|
|
# Note that minor is axis 2
|
|
non_nan_cols = set(old_buffer.minor_axis[~np.all(nans, axis=(0, 1))])
|
|
# Determine new columns to be added
|
|
new_cols = set(frame.columns).difference(non_nan_cols)
|
|
# Update internal minor axis
|
|
self.minor_axis = _ensure_index(new_cols.union(non_nan_cols))
|
|
|
|
# Same for items (fields)
|
|
# Find items axes that have only nans
|
|
# Note that items is axis 0
|
|
non_nan_items = set(old_buffer.items[~np.all(nans, axis=(1, 2))])
|
|
new_items = set(frame.index).difference(non_nan_items)
|
|
self.items = _ensure_index(new_items.union(non_nan_items))
|
|
|
|
# :NOTE:
|
|
# There is a simpler and 10x faster way to do this:
|
|
#
|
|
# Reindex buffer to update axes (automatically adds nans)
|
|
# self.buffer = self.buffer.reindex(items=self.items,
|
|
# major_axis=np.arange(self.cap),
|
|
# minor_axis=self.minor_axis)
|
|
#
|
|
# However, pandas==0.12.0, for which we remain backwards compatible,
|
|
# has a bug in .reindex() that this triggers. Using .update() as before
|
|
# seems to work fine.
|
|
|
|
new_buffer = self._create_buffer()
|
|
new_buffer.update(
|
|
self.buffer.loc[non_nan_items, :, non_nan_cols])
|
|
|
|
self.buffer = new_buffer
|