Merge pull request #660 from bashtage/upgrade-minimums

MAINT: Bump Python to 3.9

Author: bashtage

README.md

@@ -200,13 +200,13 @@ cov_est.cov.long_run
 These requirements reflect the testing environment. It is possible
 that arch will work with older versions.
 
-- Python (3.8+)
+- Python (3.9+)
 - NumPy (1.19+)
 - SciPy (1.5+)
 - Pandas (1.1+)
 - statsmodels (0.12+)
 - matplotlib (3+), optional
-- property-cached (1.6.4+), optional
+
 
 ### Optional Requirements
 

arch/bootstrap/_samplers.pyx

@@ -3,11 +3,11 @@
 
 import numpy as np
 
-cimport cython
 cimport numpy as np
 
 np.import_array()
 
+
 def stationary_bootstrap_sample(np.int64_t[:] indices,
                                 double[:] u,
                                 double p):

arch/bootstrap/base.py

@@ -1,17 +1,7 @@
 from __future__ import annotations
 
-from typing import (
-    Any,
-    Callable,
-    Dict,
-    Generator as PyGenerator,
-    List,
-    Mapping,
-    Sequence,
-    Tuple,
-    Union,
-    cast,
-)
+from collections.abc import Generator as PyGenerator, Mapping, Sequence
+from typing import Any, Callable, Union, cast
 import warnings
 
 import numpy as np
@@ -1290,7 +1280,7 @@ def _resample(self) -> tuple[tuple[ArrayLike, ...], dict[str, ArrayLike]]:
         """
         Resample all data using the values in _index
         """
-        indices = cast(Union[Int64Array, Tuple[Int64Array, ...]], self._index)
+        indices = cast(Union[Int64Array, tuple[Int64Array, ...]], self._index)
         pos_data: list[NDArray | pd.Series | pd.DataFrame] = []
         for values in self._args:
             if isinstance(values, (pd.Series, pd.DataFrame)):
@@ -1458,7 +1448,7 @@ def _resample(self) -> tuple[tuple[ArrayLike, ...], dict[str, ArrayLike]]:
         Resample all data using the values in _index
         """
         pos_indices, kw_indices = cast(
-            Tuple[List[Int64Array], Dict[str, Int64Array]], self._index
+            tuple[list[Int64Array], dict[str, Int64Array]], self._index
         )
         pos_data: list[np.ndarray | pd.DataFrame | pd.Series] = []
         for i, values in enumerate(self._args):

arch/bootstrap/multiple_comparison.py

@@ -1,7 +1,8 @@
 from __future__ import annotations
 
+from collections.abc import Hashable, Sequence
 import copy
-from typing import Hashable, Sequence, cast
+from typing import cast
 import warnings
 
 import numpy as np

arch/compat/statsmodels.py

@@ -1,10 +1,12 @@
-from typing import Any, Union
+from __future__ import annotations
+
+from typing import Any
 
 from numpy import recarray
 from pandas import DataFrame
 
 
-def dataset_loader(dataset: Any) -> Union[recarray, DataFrame]:
+def dataset_loader(dataset: Any) -> recarray | DataFrame:
     """Load a dataset using the new syntax is possible"""
     try:
         return dataset.load(as_pandas=True).data

arch/tests/bootstrap/test_bootstrap.py

@@ -1,5 +1,7 @@
+from __future__ import annotations
+
 import copy
-from typing import Callable, NamedTuple, Union
+from typing import Callable, NamedTuple
 import warnings
 
 import numpy as np
@@ -42,7 +44,7 @@ class BSData(NamedTuple):
     y_series: pd.Series
     z_df: pd.DataFrame
 
-    func: Callable[[np.ndarray, int], Union[float, np.ndarray]]
+    func: Callable[[np.ndarray, int], float | np.ndarray]
 
 
 @pytest.fixture(scope="function", params=[1234, "gen", "rs"])

arch/tests/covariance/test_covariance.py

@@ -1,5 +1,4 @@
 from itertools import product
-from typing import Tuple, Type
 
 import numpy as np
 from numpy.testing import assert_allclose
@@ -64,7 +63,7 @@ def data(request) -> ArrayLike:
     rs = np.random.RandomState([3894830, 432841, 323297, 8927821])
     ndim, use_pandas = request.param
     burn = 100
-    size: Tuple[int, ...] = (500 + burn,)
+    size: tuple[int, ...] = (500 + burn,)
     if ndim == 2:
         size += (3,)
     e = rs.standard_normal(size)
@@ -90,7 +89,7 @@ def estimator(request) -> CovarianceEstimator:
     return request.param
 
 
-def test_covariance_smoke(data: ArrayLike, estimator: Type[CovarianceEstimator]):
+def test_covariance_smoke(data: ArrayLike, estimator: type[CovarianceEstimator]):
     cov = estimator(data)
     est_cov = cov.cov
     ndim = data.shape[1] if data.ndim > 1 else 1
@@ -103,7 +102,7 @@ def test_covariance_smoke(data: ArrayLike, estimator: Type[CovarianceEstimator])
     assert isinstance(repr(cov), str)
 
 
-def test_covariance_errors(data: ArrayLike, estimator: Type[CovarianceEstimator]):
+def test_covariance_errors(data: ArrayLike, estimator: type[CovarianceEstimator]):
     with pytest.raises(ValueError, match="Degrees of freedom is <= 0"):
         estimator(data, df_adjust=data.shape[0] + 1)
     with pytest.raises(ValueError, match="df_adjust must be a non-negative"):
@@ -139,10 +138,10 @@ def test_bartlett_auto(data: ArrayLike):
     assert_allclose(expected_oss, np.squeeze(nw.cov.one_sided_strict))
     assert_allclose(expected_cov, np.squeeze(nw.cov.long_run))
     ce = CovarianceEstimate(
-        short_run=nw.cov.short_run,
-        one_sided_strict=nw.cov.one_sided_strict,
-        long_run=nw.cov.long_run,
-        one_sided=nw.cov.one_sided,
+        short_run=np.asarray(nw.cov.short_run),
+        one_sided_strict=np.asarray(nw.cov.one_sided_strict),
+        long_run=np.asarray(nw.cov.long_run),
+        one_sided=np.asarray(nw.cov.one_sided),
     )
     assert_allclose(ce.short_run, nw.cov.short_run)
     assert_allclose(ce.one_sided_strict, nw.cov.one_sided_strict)
@@ -155,7 +154,7 @@ def test_parzen_auto(data: ArrayLike):
 
     if data.ndim == 1:
         # This test is noisy
-        expected_bw: Tuple[int, ...] = (18, 19)
+        expected_bw: tuple[int, ...] = (18, 19)
         expected_weights = [
             1.0000e00,
             9.8575e-01,
@@ -239,27 +238,27 @@ def test_qs_auto(data: ArrayLike):
     assert_allclose(qs.kernel_weights[:10], np.array(expected_weights))
 
 
-def test_force_int(data: ArrayLike, estimator: Type[CovarianceEstimator]):
+def test_force_int(data: ArrayLike, estimator: type[CovarianceEstimator]):
     bw = estimator(data, force_int=False).bandwidth
     bw_int = estimator(data, force_int=True).bandwidth
     assert bw_int >= bw
     assert bw_int == int(bw_int)
 
 
-def test_first_weights(data: ArrayLike, estimator: Type[CovarianceEstimator]):
+def test_first_weights(data: ArrayLike, estimator: type[CovarianceEstimator]):
     w = estimator(data).kernel_weights
     assert w[0] == 1.0
 
 
-def test_constants(data: ArrayLike, estimator: Type[CovarianceEstimator]):
+def test_constants(data: ArrayLike, estimator: type[CovarianceEstimator]):
     cov_est = estimator(data)
     kc, bs, rate = KERNEL_PARAMS[estimator.__name__]
     assert_allclose(cov_est.kernel_const, kc)
     assert cov_est.bandwidth_scale == bs
     assert_allclose(cov_est.rate, rate)
 
 
-def test_weight_len(data: ArrayLike, estimator: Type[CovarianceEstimator]):
+def test_weight_len(data: ArrayLike, estimator: type[CovarianceEstimator]):
     cov_est = estimator(data, force_int=True)
     name = estimator.__name__
     is_qs = name in ("QuadraticSpectral", "Andrews")
@@ -270,7 +269,7 @@ def test_weight_len(data: ArrayLike, estimator: Type[CovarianceEstimator]):
     assert cov_est.kernel_weights.shape[0] == exp_len
 
 
-def test_kernel_weights(data: ArrayLike, estimator: Type[CovarianceEstimator]):
+def test_kernel_weights(data: ArrayLike, estimator: type[CovarianceEstimator]):
     if data.ndim == 1:
         return
     weights = np.arange(1, data.shape[1] + 1)
@@ -280,7 +279,7 @@ def test_kernel_weights(data: ArrayLike, estimator: Type[CovarianceEstimator]):
     assert wcov.bandwidth != cov.bandwidth
 
 
-def test_center(data: ArrayLike, estimator: Type[CovarianceEstimator]):
+def test_center(data: ArrayLike, estimator: type[CovarianceEstimator]):
     centered_cov = estimator(data, center=False, force_int=False)
     cov = estimator(data, force_int=False)
     assert centered_cov.bandwidth != cov.bandwidth

arch/tests/unitroot/cointegration_data.py

@@ -1,5 +1,3 @@
-from typing import Tuple
-
 import numpy as np
 import pandas as pd
 import pytest
@@ -8,7 +6,7 @@
 
 
 @pytest.fixture(scope="module", params=[True, False])
-def data(request) -> Tuple[Float64Array, Float64Array]:
+def data(request) -> tuple[Float64Array, Float64Array]:
     g = np.random.RandomState([12839028, 3092183, 902813])
     e = g.standard_normal((2000, 2))
     phi = g.random_sample((3, 2, 2))
@@ -31,7 +29,7 @@ def data(request) -> Tuple[Float64Array, Float64Array]:
 
 
 @pytest.fixture(scope="module", params=[True, False], ids=["pandas", "numpy"])
-def trivariate_data(request) -> Tuple[ArrayLike2D, ArrayLike2D]:
+def trivariate_data(request) -> tuple[ArrayLike2D, ArrayLike2D]:
     rs = np.random.RandomState([922019, 12882912, 192010, 10189, 109981])
     nobs = 1000
     burn = 100

arch/tests/unitroot/test_phillips_ouliaris.py

@@ -1,5 +1,3 @@
-from typing import Tuple
-
 import numpy as np
 from numpy.testing import assert_allclose
 import pytest
@@ -15,7 +13,7 @@
 from arch.utility.timeseries import add_trend
 
 
-def z_tests(z: Float64Array, lag: int, trend: UnitRootTrend) -> Tuple[float, float]:
+def z_tests(z: Float64Array, lag: int, trend: UnitRootTrend) -> tuple[float, float]:
     z = add_trend(z, trend=trend)
     u = z
     if z.shape[1] > 1:
@@ -37,7 +35,7 @@ def z_tests(z: Float64Array, lag: int, trend: UnitRootTrend) -> Tuple[float, flo
     return float(z_a), float(z_t)
 
 
-def p_tests(z: Float64Array, lag: int, trend: UnitRootTrend) -> Tuple[float, float]:
+def p_tests(z: Float64Array, lag: int, trend: UnitRootTrend) -> tuple[float, float]:
     x, y = z[:, 1:], z[:, 0]
     nobs = x.shape[0]
     x = add_trend(x, trend=trend)

arch/tests/unitroot/test_unitroot.py

@@ -1,5 +1,7 @@
 # TODO: Tests for features that are just called
 # TODO: Test for trend='ctt'
+from __future__ import annotations
+
 from arch.compat.statsmodels import dataset_loader
 
 import os

arch/tests/univariate/test_mean.py

@@ -72,7 +72,6 @@
 RTOL = 1e-4 if struct.calcsize("P") < 8 else 1e-6
 DISPLAY: Literal["off", "final"] = "off"
 UPDATE_FREQ = 0 if DISPLAY == "off" else 3
-SP_LT_14 = parse(scipy.__version__) < parse("1.4")
 SP500 = 100 * sp500.load()["Adj Close"].pct_change().dropna()
 
 
@@ -929,14 +928,9 @@ def test_convergence_warning(self):
         )
         am = arch_model(y, mean="ARX", lags=10, p=5, q=0)
 
-        if SP_LT_14:
-            with pytest.warns(ConvergenceWarning):
-                am.fit(disp=DISPLAY)
-                am.fit(show_warning=True, disp=DISPLAY)
-        else:
-            with pytest.warns(DataScaleWarning):
-                am.fit(disp=DISPLAY)
-                am.fit(show_warning=True, disp=DISPLAY)
+        with pytest.warns(DataScaleWarning):
+            am.fit(disp=DISPLAY)
+            am.fit(show_warning=True, disp=DISPLAY)
 
         with pytest.warns(DataScaleWarning):
             am.fit(show_warning=False, disp=DISPLAY)

arch/tests/univariate/test_recursions.py

@@ -2,7 +2,6 @@
 import pickle
 import timeit
 import types
-from typing import List
 
 import numpy as np
 from numpy.random import RandomState
@@ -58,7 +57,7 @@ def __init__(
         self.repeat = repeat
         self.number = number
         self._run = False
-        self.times: List[float] = []
+        self.times: list[float] = []
         self._codes = [first, second]
         self.ratio = np.inf
 
@@ -1242,10 +1241,12 @@ def test_figarch_performance(self):
 """
 
         midas_first = """
-recpy.figarch_recursion(parameters, fresids, sigma2, p, q, nobs, trunc_lag, backcast, var_bounds)
-                """
+recpy.figarch_recursion(parameters, fresids, sigma2, p, q,
+                        nobs, trunc_lag, backcast, var_bounds)
+"""
         midas_second = """
-rec.figarch_recursion(parameters, fresids, sigma2, p, q, nobs, trunc_lag, backcast, var_bounds)
+rec.figarch_recursion(parameters, fresids, sigma2, p, q,
+                      nobs, trunc_lag, backcast, var_bounds)
 """
         timer = Timer(
             midas_first,

arch/tests/univariate/test_variance_forecasting.py

@@ -45,8 +45,8 @@ def _compare_truncated_forecasts(full, trunc, start):
 
 class PreservedState:
     """
-    Context manager that will save NumPy's random generator's state when entering and restore
-    the original state when exiting.
+    Context manager that will save NumPy's random generator's state when entering and
+    restore the original state when exiting.
     """
 
     def __init__(self, random_state):

arch/tests/utility/test_cov.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 from arch.compat.statsmodels import dataset_loader
 
 from numpy import diff, log

arch/typing.py

@@ -1,19 +1,8 @@
 from __future__ import annotations
 
+from collections.abc import Hashable
 import datetime as dt
-from typing import (
-    TYPE_CHECKING,
-    Any,
-    Callable,
-    Dict,
-    Hashable,
-    List,
-    Literal,
-    Optional,
-    Tuple,
-    TypeVar,
-    Union,
-)
+from typing import TYPE_CHECKING, Any, Callable, Literal, Optional, TypeVar, Union
 
 import numpy as np
 from pandas import DataFrame, Series, Timestamp
@@ -61,11 +50,11 @@
     ) = Float64Array = Int64Array = Int32Array = BoolArray = AnyArray = NDArray
 
 BootstrapIndexT = Union[
-    Int64Array, Tuple[Int64Array, ...], Tuple[List[Int64Array], Dict[str, Int64Array]]
+    Int64Array, tuple[Int64Array, ...], tuple[list[Int64Array], dict[str, Int64Array]]
 ]
-RandomStateState = Tuple[str, Uint32Array, int, int, float]
+RandomStateState = tuple[str, Uint32Array, int, int, float]
 
-RNGType = Callable[[Union[int, Tuple[int, ...]]], Float64Array]
+RNGType = Callable[[Union[int, tuple[int, ...]]], Float64Array]
 ArrayLike1D = Union[NDArray, Series]
 ArrayLike2D = Union[NDArray, DataFrame]
 ArrayLike = Union[NDArray, DataFrame, Series]

arch/unitroot/_phillips_ouliaris.py

@@ -241,7 +241,8 @@ def phillips_ouliaris(
     .. math::
 
        \hat{\omega}_{11\cdot 2} = \hat{\omega}_{11}
-                                 - \hat{\omega}'_{21} \hat{\Omega}_{22}^{-1} \hat{\omega}_{21}
+                                 - \hat{\omega}'_{21} \hat{\Omega}_{22}^{-1}
+                                   \hat{\omega}_{21}
 
     and