def overwrite_from_dates(asof, dense_dates, sparse_dates, asset_idx, value):
"""Construct an Overwrite with the correct
start and end date based on the asof date of the delta,
the dense_dates, and the dense_dates.
Parameters
----------
asof : datetime
The asof date of the delta.
dense_dates : pd.DatetimeIndex
The dates requested by the loader.
sparse_dates : pd.DatetimeIndex
The dates that appeared in the dataset.
asset_idx : tuple of int
The index of the asset in the block. If this is a tuple, then this
is treated as the first and last index to use.
value : any
The value to overwrite with.
Returns
-------
overwrite : Float64Overwrite
The overwrite that will apply the new value to the data.
Notes
-----
This is forward-filling all dense dates that are between the asof_date date
and the next sparse date after the asof_date.
For example:
let ``asof = pd.Timestamp('2014-01-02')``,
``dense_dates = pd.date_range('2014-01-01', '2014-01-05')``
``sparse_dates = pd.to_datetime(['2014-01', '2014-02', '2014-04'])``
Then the overwrite will apply to indexes: 1, 2, 3, 4
"""
if asof is pd.NaT:
# Not an actual delta.
# This happens due to the groupby we do on the deltas.
return
first_row = dense_dates.searchsorted(asof)
next_idx = sparse_dates.searchsorted(asof.asm8, 'right')
if next_idx == len(sparse_dates):
# There is no next date in the sparse, this overwrite should apply
# through the end of the dense dates.
last_row = len(dense_dates) - 1
else:
# There is a next date in sparse dates. This means that the overwrite
# should only apply until the index of this date in the dense dates.
last_row = dense_dates.searchsorted(sparse_dates[next_idx]) - 1
if first_row > last_row:
return
first, last = asset_idx
yield make_adjustment_from_indices(
first_row, last_row, first, last, OVERWRITE, value
)
def test_unsupported_type(self):
class SomeClass(object):
pass
with self.assertRaises(TypeError) as e:
adj.make_adjustment_from_indices(
1, 2, 3, 4,
adjustment_kind=adj.OVERWRITE,
value=SomeClass(),
)
exc = e.exception
expected_msg = (
"Don't know how to make overwrite adjustments for values of type "
"%r." % SomeClass
)
self.assertEqual(str(exc), expected_msg)
def overwrite_from_dates(asof, dense_dates, sparse_dates, asset_idx, value):
"""Construct an Overwrite with the correct
start and end date based on the asof date of the delta,
the dense_dates, and the dense_dates.
Parameters
----------
asof : datetime
The asof date of the delta.
dense_dates : pd.DatetimeIndex
The dates requested by the loader.
sparse_dates : pd.DatetimeIndex
The dates that appeared in the dataset.
asset_idx : tuple of int
The index of the asset in the block. If this is a tuple, then this
is treated as the first and last index to use.
value : any
The value to overwrite with.
Returns
-------
overwrite : Float64Overwrite
The overwrite that will apply the new value to the data.
Notes
-----
This is forward-filling all dense dates that are between the asof_date date
and the next sparse date after the asof_date.
For example:
let ``asof = pd.Timestamp('2014-01-02')``,
``dense_dates = pd.date_range('2014-01-01', '2014-01-05')``
``sparse_dates = pd.to_datetime(['2014-01', '2014-02', '2014-04'])``
Then the overwrite will apply to indexes: 1, 2, 3, 4
"""
if asof is pd.NaT:
# Not an actual delta.
# This happens due to the groupby we do on the deltas.
return
first_row = dense_dates.searchsorted(asof)
next_idx = sparse_dates.searchsorted(asof.asm8, 'right')
if next_idx == len(sparse_dates):
# There is no next date in the sparse, this overwrite should apply
# through the end of the dense dates.
last_row = len(dense_dates) - 1
else:
# There is a next date in sparse dates. This means that the overwrite
# should only apply until the index of this date in the dense dates.
last_row = dense_dates.searchsorted(sparse_dates[next_idx]) - 1
if first_row > last_row:
return
first, last = asset_idx
yield make_adjustment_from_indices(first_row, last_row, first, last,
OVERWRITE, value)
def test_unsupported_type(self):
class SomeClass(object):
pass
with self.assertRaises(TypeError) as e:
adj.make_adjustment_from_indices(
1,
2,
3,
4,
adjustment_kind=adj.OVERWRITE,
value=SomeClass(),
)
exc = e.exception
expected_msg = (
"Don't know how to make overwrite adjustments for values of type "
"%r." % SomeClass)
self.assertEqual(str(exc), expected_msg)
def test_make_datetime_adjustment(self):
overwrite_dt = make_datetime64ns(0)
result = adj.make_adjustment_from_indices(
1, 2, 3, 4,
adjustment_kind=adj.OVERWRITE,
value=overwrite_dt,
)
expected = adj.Datetime64Overwrite(
first_row=1,
last_row=2,
first_col=3,
last_col=4,
value=overwrite_dt,
)
self.assertEqual(result, expected)
def test_make_int_adjustment(self):
result = adj.make_adjustment_from_indices(
1,
2,
3,
4,
adjustment_kind=adj.OVERWRITE,
value=1,
)
expected = adj.Int64Overwrite(
first_row=1,
last_row=2,
first_col=3,
last_col=4,
value=1,
)
self.assertEqual(result, expected)
def test_make_datetime_adjustment(self):
overwrite_dt = make_datetime64ns(0)
result = adj.make_adjustment_from_indices(
1,
2,
3,
4,
adjustment_kind=adj.OVERWRITE,
value=overwrite_dt,
)
expected = adj.Datetime64Overwrite(
first_row=1,
last_row=2,
first_col=3,
last_col=4,
value=overwrite_dt,
)
self.assertEqual(result, expected)
def test_make_float_adjustment(self, name, adj_type):
expected_types = {
'add': adj.Float64Add,
'multiply': adj.Float64Multiply,
'overwrite': adj.Float64Overwrite,
}
result = adj.make_adjustment_from_indices(
1, 2, 3, 4,
adjustment_kind=adj_type,
value=0.5,
)
expected = expected_types[name](
first_row=1,
last_row=2,
first_col=3,
last_col=4,
value=0.5,
)
self.assertEqual(result, expected)
def test_make_float_adjustment(self, name, adj_type):
expected_types = {
"add": adj.Float64Add,
"multiply": adj.Float64Multiply,
"overwrite": adj.Float64Overwrite,
}
result = adj.make_adjustment_from_indices(
1,
2,
3,
4,
adjustment_kind=adj_type,
value=0.5,
)
expected = expected_types[name](
first_row=1,
last_row=2,
first_col=3,
last_col=4,
value=0.5,
)
self.assertEqual(result, expected)
