fix: replace typehints '|' with Union for backwards compatability.

Author: zhenyuen

stonesoup/models/base.py

@@ -369,7 +369,7 @@ def _integrate(self, func: np.ndarray, a: np.ndarray, b: np.ndarray) -> np.ndarr
 
     def mean(
         self, latents: Latents, time_interval: timedelta, **kwargs
-    ) -> StateVector | StateVectors:
+    ) -> Union[StateVector, StateVectors]:
         """Model mean"""
         dt = time_interval.total_seconds()
         integrand_f = self._integrand
@@ -383,7 +383,7 @@ def mean(
             mu_W=self.mu_W,
         )
 
-    def covar(self, latents: Latents, time_interval: timedelta, **kwargs) -> CovarianceMatrix | CovarianceMatrices:
+    def covar(self, latents: Latents, time_interval: timedelta, **kwargs) -> Union[CovarianceMatrix, CovarianceMatrices]:
         """Model covariance"""
         dt = time_interval.total_seconds()
         integrand_f = self._integrand
@@ -461,11 +461,11 @@ def logcondpdf(self, state1: State, state2: State, latents: Optional[Latents] =
 
         return likelihood
 
-    def logpdf(self, state1: State, state2: State, **kwargs) -> Probability | np.ndarray:
+    def logpdf(self, state1: State, state2: State, **kwargs) -> Union[Probability, np.ndarray]:
         r"""Model log pdf/likelihood evaluation function"""
         return NotImplementedError
 
 
-    def pdf(self, state1: State, state2: State, **kwargs) -> Probability | np.ndarray:
+    def pdf(self, state1: State, state2: State, **kwargs) -> Union[Probability, np.ndarray]:
         r"""Model pdf/likelihood evaluation function"""
         return Probability.from_log_ufunc(self.logpdf(state1, state2, **kwargs))

stonesoup/models/base_driver.py

@@ -178,7 +178,7 @@ def mean(
         dt: float,
         mu_W: Optional[float]=None,
         **kwargs
-    ) -> StateVector | StateVectors:
+    ) -> Union[StateVector, StateVectors]:
         """Computes a num_samples of mean vectors"""
         mu_W = np.atleast_2d(self.mu_W) if mu_W is None else np.atleast_2d(mu_W)
 
@@ -209,7 +209,7 @@ def covar(
         mu_W: Optional[float] = None,
         sigma_W2: Optional[float] = None,
         **kwargs
-    ) -> CovarianceMatrix | CovarianceMatrices:
+    ) -> Union[CovarianceMatrix, CovarianceMatrices]:
         """Computes covariance matrix / matrices"""
         mu_W = np.atleast_2d(self.mu_W) if mu_W is None else np.atleast_2d(mu_W)
         sigma_W2 = np.atleast_2d(self.sigma_W2) if sigma_W2 is None else np.atleast_2d(sigma_W2)     

stonesoup/models/transition/base.py

@@ -178,7 +178,7 @@ def ndim_state(self):
         return sum(model.ndim_state for model in self.model_list)
 
 
-    def mean(self, **kwargs) -> StateVector | StateVectors:
+    def mean(self, **kwargs) -> Union[StateVector, StateVectors]:
         """Returns the transition model noise mean matrix.
 
         Returns
@@ -194,7 +194,7 @@ def mean(self, **kwargs) -> StateVector | StateVectors:
             return np.concatenate(mean_list, axis=1).view(StateVectors)
 
 
-    def covar(self, **kwargs) -> CovarianceMatrix | CovarianceMatrices:
+    def covar(self, **kwargs) -> Union[CovarianceMatrix, CovarianceMatrices]:
         """Returns the transition model noise covariance matrix.
 
         Returns

stonesoup/models/transition/levylinear.py

@@ -9,7 +9,7 @@
 from datetime import timedelta
 from scipy.integrate import quad_vec
 from scipy.linalg import expm, block_diag
-from typing import Optional
+from typing import Optional, Union
 
 
 class LinearLevyTransitionModel(TransitionModel, LinearModel, LevyModel):
@@ -93,11 +93,11 @@ def matrix(self, time_interval, **kwargs):
 
     def rvs(
         self,
-        latents: Latents | None = None,
+        latents: Optional[Latents] = None,
         num_samples: int = 1,
         random_state: RandomState = None,
         **kwargs
-    ) -> StateVector | StateVectors:
+    ) -> Union[StateVector, StateVectors]:
         coeff = self.noise_diff_coeff
         return super().rvs(latents, num_samples, random_state, **kwargs) * coeff
 
@@ -187,11 +187,11 @@ def matrix(self, time_interval, **kwargs):
 
     def rvs(
         self,
-        latents: Latents | None = None,
+        latents: Optional[Latents] = None,
         num_samples: int = 1,
         random_state: RandomState = None,
         **kwargs
-    ) -> StateVector | StateVectors:
+    ) -> Union[StateVector, StateVectors]:
         coeff = self.noise_diff_coeff
         return super().rvs(latents, num_samples, random_state, **kwargs) * coeff