wassname/catalyst
1
0
Fork 0
mirror of https://github.com/wassname/catalyst.git synced 2026-08-15 12:15:22 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #1358 from quantopian/smoothing

Browse Source 
ENH: added smoothing to zipline
This commit is contained in:
Scott Sanderson
2016-08-02 10:32:11 -04:00
committed by GitHub
parent 7761c736da 499703d227
commit 129d16fd3d
4 changed files with 146 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/pipeline/test_filter.py
+103
View File
@@ -30,6 +30,7 @@ from zipline.errors import BadPercentileBounds
from zipline.pipeline import Filter, Factor, TermGraph
from zipline.pipeline.classifiers import Classifier
from zipline.pipeline.factors import CustomFactor
from zipline.pipeline.filters import All, Any
from zipline.testing import check_arrays, parameter_space, permute_rows
from zipline.utils.numpy_utils import float64_dtype, int64_dtype
from .base import BasePipelineTestCase, with_default_shape
@@ -395,6 +396,108 @@ class FilterTestCase(BasePipelineTestCase):
)
check_arrays(results['isfinite'], isfinite(data))
def test_all(self):
data = array([[1, 1, 1, 1, 1, 1],
[0, 1, 1, 1, 1, 1],
[1, 0, 1, 1, 1, 1],
[1, 1, 0, 1, 1, 1],
[1, 1, 1, 0, 1, 1],
[1, 1, 1, 1, 0, 1],
[1, 1, 1, 1, 1, 0]], dtype=bool)
# With a window_length of N, 0's should be "sticky" for the (N - 1)
# days after the 0 in the base data.
# Note that, the way ``self.run_graph`` works, we compute the same
# number of output rows for all inputs, so we only get the last 4
# outputs for expected_3 even though we have enought input data to
# compute 5 rows.
expected_3 = array([[0, 0, 0, 1, 1, 1],
[1, 0, 0, 0, 1, 1],
[1, 1, 0, 0, 0, 1],
[1, 1, 1, 0, 0, 0]], dtype=bool)
expected_4 = array([[0, 0, 0, 1, 1, 1],
[0, 0, 0, 0, 1, 1],
[1, 0, 0, 0, 0, 1],
[1, 1, 0, 0, 0, 0]], dtype=bool)
class Input(Filter):
inputs = ()
window_length = 0
results = self.run_graph(
TermGraph({
'3': All(inputs=[Input()], window_length=3),
'4': All(inputs=[Input()], window_length=4),
}),
initial_workspace={Input(): data},
mask=self.build_mask(ones(shape=data.shape)),
)
check_arrays(results['3'], expected_3)
check_arrays(results['4'], expected_4)
def test_any(self):
# FUN FACT: The inputs and outputs here are exactly the negation of
# the inputs and outputs for test_all above. This isn't a coincidence.
#
# By de Morgan's Laws, we have::
#
# ~(a & b) == (~a | ~b)
#
# negating both sides, we have::
#
# (a & b) == ~(a | ~b)
#
# Since all(a, b) is isomorphic to (a & b), and any(a, b) is isomorphic
# to (a | b), we have::
#
# all(a, b) == ~(any(~a, ~b))
#
data = array([[0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0],
[0, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 1, 0],
[0, 0, 0, 0, 0, 1]], dtype=bool)
# With a window_length of N, 1's should be "sticky" for the (N - 1)
# days after the 1 in the base data.
# Note that, the way ``self.run_graph`` works, we compute the same
# number of output rows for all inputs, so we only get the last 4
# outputs for expected_3 even though we have enought input data to
# compute 5 rows.
expected_3 = array([[1, 1, 1, 0, 0, 0],
[0, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 0],
[0, 0, 0, 1, 1, 1]], dtype=bool)
expected_4 = array([[1, 1, 1, 0, 0, 0],
[1, 1, 1, 1, 0, 0],
[0, 1, 1, 1, 1, 0],
[0, 0, 1, 1, 1, 1]], dtype=bool)
class Input(Filter):
inputs = ()
window_length = 0
results = self.run_graph(
TermGraph({
'3': Any(inputs=[Input()], window_length=3),
'4': Any(inputs=[Input()], window_length=4),
}),
initial_workspace={Input(): data},
mask=self.build_mask(ones(shape=data.shape)),
)
check_arrays(results['3'], expected_3)
check_arrays(results['4'], expected_4)
@parameter_space(factor_len=[2, 3, 4])
def test_window_safe(self, factor_len):
# all true data set of (days, securities)