Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 19 additions & 0 deletions chainladder/core/common.py
Original file line number Diff line number Diff line change
Expand Up @@ -435,3 +435,22 @@ def _validate_assumption(
if axis == 2:
arr = arr[None, None, :, None]
return arr

def _assumption_values(self, value, X: TriangleLike | None, axis: int = 3) -> list:
"""
Return the individual values of an assumption that can be checked.

A scalar gives itself and a list, tuple, set or array gives its
elements. A dict or callable gives the values it resolves to against
``X`` along ``axis``, or nothing when ``X`` is ``None``.
"""
if isinstance(value, dict) or callable(value):
if X is None:
return []
values = self._validate_assumption(X, value, axis=axis).flatten()
elif isinstance(value, (list, tuple, set, np.ndarray)):
values = np.asarray(list(value) if isinstance(value, set) else value)
values = values.flatten()
else:
return [value]
return [v.item() if isinstance(v, np.generic) else v for v in values]
57 changes: 51 additions & 6 deletions chainladder/development/development.py
Original file line number Diff line number Diff line change
Expand Up @@ -16,6 +16,7 @@
Literal,
# Self, # Make use of this once Python 3.10 is deprecated.
TYPE_CHECKING,
get_args,
)

if TYPE_CHECKING:
Expand All @@ -24,6 +25,18 @@
from pandas import Series
from types import ModuleType

# Allowed values of Development's string parameters. The annotations in
# Development.__init__ use them, and _check_param_values reads them with
# get_args.
Average = Literal["volume", "simple", "regression", "geometric"]
SigmaInterpolation = Literal["log-linear", "mack"]


def _one_of(values: tuple) -> str:
"""Return ``values`` as "'a', 'b' or 'c'" for an error message."""
quoted = [repr(v) for v in values]
return ", ".join(quoted[:-1]) + " or " + quoted[-1]


class Development(DevelopmentBase):
"""
Expand All @@ -37,11 +50,9 @@ class Development(DevelopmentBase):

Setting n_periods to 1 does not allow enough degrees of freedom to support
calculation of all regression statistics; only LDFs would be calculated.
average: literal (or list of literals), or float, optional (default = 'volume')
average: literal (or list of literals), optional (default = 'volume')
type of averaging to use for ldf average calculation.
Options include 'volume', 'simple', 'regression', and 'geometric'. If numeric values are supplied,
then (2-average) in the style of Zehnwirth & Barnett is used
for the exponent of the regression weights.
Options include 'volume', 'simple', 'regression', and 'geometric'.
sigma_interpolation: string optional (default = 'log-linear')
Options include 'log-linear' and 'mack'
drop: tuple or list of tuples
Expand Down Expand Up @@ -328,8 +339,8 @@ class Development(DevelopmentBase):
def __init__(
self,
n_periods: int = -1,
average: Literal["volume", "simple", "regression", "geometric"] = "volume",
sigma_interpolation: Literal["log-linear", "mack"] = "log-linear",
average: Average = "volume",
sigma_interpolation: SigmaInterpolation = "log-linear",
drop: tuple | list[tuple] | None = None,
drop_high: bool | int | list[bool] | list[int] | None = None,
drop_low: bool | int | list[bool] | list[int] | None = None,
Expand All @@ -356,6 +367,39 @@ def __init__(
# Undeclared until fitted attributes - scikit-learn convention.
self.average_: np.ndarray

def _check_param_values(self, X: TriangleLike | None = None) -> None:
"""
Check ``average``, ``n_periods`` and ``sigma_interpolation`` and raise
a ``ValueError`` naming the first one with a value ``fit`` cannot use.

``average`` takes the values of ``Average``, ``n_periods`` takes -1 or
a positive integer (0 is kept as an alias for -1), and
``sigma_interpolation`` takes the values of ``SigmaInterpolation``. A
list is checked value by value. ``fit`` calls the method with the
triangle, so a dict or callable is checked once resolved against it;
``read_json`` calls it with no triangle and skips those.
"""
averages = get_args(Average)
for value in self._assumption_values(self.average, X):

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

somewhere in this module or development/base.py, there's also checking for length of the parameters. do you think that should be in scope for this method? (doesn't have to be this PR)

if not (isinstance(value, str) and value in averages):
raise ValueError(f"average must be {_one_of(averages)}, got {value!r}.")
for value in self._assumption_values(self.n_periods, X):
if (
isinstance(value, (bool, np.bool_))
or not isinstance(value, (int, np.integer))
or not (value == -1 or value >= 0)
):
raise ValueError(
"n_periods must be -1 or a positive integer (0 is read as "
f"-1), got {value!r}."
)
interpolations = get_args(SigmaInterpolation)
if self.sigma_interpolation not in interpolations:
raise ValueError(
f"sigma_interpolation must be {_one_of(interpolations)}, "
f"got {self.sigma_interpolation!r}."
)

def fit(self, X: TriangleLike, y: None = None, sample_weight: None = None):
"""
Fit the model with X.
Expand All @@ -375,6 +419,7 @@ def fit(self, X: TriangleLike, y: None = None, sample_weight: None = None):
Returns the instance itself.
"""

self._check_param_values(X)
# Triangle must be cumulative and in "development" mode.
obj: TriangleLike = self._set_fit_groups(X).incr_to_cum().val_to_dev().copy()
xp: ModuleType = obj.get_array_module()
Expand Down
42 changes: 42 additions & 0 deletions chainladder/development/tests/test_development.py
Original file line number Diff line number Diff line change
Expand Up @@ -1090,3 +1090,45 @@ def test_pct_reported_requires_ldf(raa):
_ = raa.pct_reported_
with pytest.raises(AttributeError):
_ = raa.pct_unreported_


@pytest.mark.parametrize(
"params, name",
[
({"average": "bogus"}, "average"),
({"average": 1.5}, "average"),
({"average": ["volume"] * 8 + ["bogus"]}, "average"),
({"average": lambda age: "bogus" if age == 12 else "volume"}, "average"),
({"n_periods": -5}, "n_periods"),
({"n_periods": 2.5}, "n_periods"),
({"n_periods": True}, "n_periods"),
({"sigma_interpolation": "bogus"}, "sigma_interpolation"),
],
)
def test_fit_rejects_invalid_params(raa, params, name):
"""
Check that ``fit`` raises a ``ValueError`` naming a parameter whose value
it cannot use, instead of failing later or fitting silently (#1569).
"""
with pytest.raises(ValueError, match=name):
cl.Development(**params).fit(raa)


def test_n_periods_zero_fits_as_all_periods(raa):
"""Check that ``n_periods=0`` still fits as ``n_periods=-1`` (#1606)."""
zero = cl.Development(n_periods=0).fit(raa)
all_periods = cl.Development(n_periods=-1).fit(raa)
assert np.allclose(zero.ldf_.values, all_periods.ldf_.values)
assert np.allclose(zero.sigma_.values, all_periods.sigma_.values)


def test_valid_params_pass_validation(raa):
"""Check that lists, dicts and callables of valid values still fit."""
cl.Development(
average=["volume", "simple"] * 4 + ["regression"],
n_periods={age: -1 if age < 60 else 3 for age in raa.development},
sigma_interpolation="mack",
).fit(raa)
cl.Development(average=lambda age: "geometric", n_periods=[-1] * 9).fit(raa)
dev = cl.Development(n_periods=4, average="simple")
assert cl.read_json(dev.to_json()).get_params() == dev.get_params()
41 changes: 41 additions & 0 deletions chainladder/development/tests/test_io.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,41 @@
"""Tests for saving and loading development estimators with JSON."""

import json

import pytest

import chainladder as cl


@pytest.mark.parametrize(
"params, name",
[({"average": "bogus"}, "average"), ({"n_periods": -5}, "n_periods")],
)
def test_read_json_rejects_invalid_params(params, name):
"""Check that ``read_json`` checks the parameter values it loads (#1569)."""
json_str = json.dumps({"params": params, "__class__": "Development"})
with pytest.raises(ValueError, match=name):
cl.read_json(json_str)
with pytest.raises(ValueError, match=name):
cl.read_json(json.dumps([{"name": "dev", **json.loads(json_str)}]))


@pytest.mark.parametrize("name", ["Trend", "BerquistSherman", "ParallelogramOLF"])
def test_read_json_loads_estimators_without_the_hook(name):
"""
Check that ``read_json`` still loads estimators that do not define
``_check_param_values``.
"""
estimator = getattr(cl, name)()
assert cl.read_json(estimator.to_json()).get_params() == estimator.get_params()


def test_read_json_leaves_dict_params_to_fit():
"""
Check that ``read_json`` loads a dict assumption without a triangle to
resolve it against, leaving its values to be checked in ``fit``.
"""
dev = cl.Development(n_periods={12: 2, 24: -1})
loaded = cl.read_json(dev.to_json())
assert isinstance(loaded, cl.Development)
assert list(loaded.n_periods.values()) == [2, -1]
20 changes: 15 additions & 5 deletions chainladder/utils/utility_functions.py
Original file line number Diff line number Diff line change
Expand Up @@ -393,13 +393,11 @@ def read_json(json_str, array_backend=None):
array_backend = options.ARRAY_BACKEND
json_dict = json.loads(json_str)
if type(json_dict) is list:
import chainladder as cl

return Pipeline(
steps=[
(
item["name"],
cl.__dict__[item["__class__"]]().set_params(**item["params"]),
_estimator_from_json(item["__class__"], item["params"]),
)
for item in json_dict
]
Expand Down Expand Up @@ -444,9 +442,21 @@ def read_json(json_str, array_backend=None):
else:
return tri
else:
import chainladder as cl
return _estimator_from_json(json_dict["__class__"], json_dict["params"])


def _estimator_from_json(class_name, params):

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

is this a from_dict is disguise?

"""
Build the chainladder estimator named ``class_name`` with ``params`` and
check the parameter values it can check without a triangle.
"""
import chainladder as cl

return cl.__dict__[json_dict["__class__"]]().set_params(**json_dict["params"])
estimator = cl.__dict__[class_name]().set_params(**params)
# Only estimators with parameter checks define the hook.
if hasattr(estimator, "_check_param_values"):
estimator._check_param_values()
return estimator


def _origin_periods(index, grain):
Expand Down
Loading