fixing bug in copula rvs

sklarpy/copulas/_prefit_dists.py

@@ -2,6 +2,7 @@
 from typing import Union, Iterable, Callable, Dict, List
 import numpy as np
 import pandas as pd
+from collections import deque
 
 from sklarpy.copulas import MarginalFitter
 from sklarpy.utils._input_handlers import check_multivariate_data, get_mask
@@ -799,17 +800,32 @@ def copula_rvs(self, size: int, copula_params: Union[Params, tuple],
             distribution. These correspond to randomly sampled cdf /
             pseudo-observation values of the univariate marginals.
         """
-        # generating random variables from multivariate distribution
-        raw_mv_rvs: np.ndarray = self._mv_object.rvs(size, copula_params)
-
-        # bounding these above and below
-        # eps: float = 10 ** -5
-        # rvs_df: pd.DataFrame = pd.DataFrame(raw_mv_rvs)
-        # rvs_df[rvs_df < 0] = eps
-        # rvs_df[rvs_df > 1] = 1 - eps
-        # mv_rvs: np.ndarray = rvs_df.to_numpy()
-
-        return self._g_to_u(raw_mv_rvs, copula_params)
+        max_num_loops: int = 100
+        num_loops: int = 0
+        d: int = self._mv_object._get_dim(copula_params)
+        valid_copula_rvs: deque = deque()
+        while size > 0:
+            # generating random variables from multivariate distribution
+            mv_rvs: np.ndarray = self._mv_object.rvs(size, copula_params)
+
+            # converting to copula rvs
+            raw_copula_rvs: np.ndarray = self._g_to_u(mv_rvs, copula_params)
+
+            # filtering out invalid copula rvs (not in [0, 1]^d)
+            mask: np.ndarray = ((raw_copula_rvs > 0) & (raw_copula_rvs < 1)
+                                ).sum(axis=1) == d
+            copula_rvs = raw_copula_rvs[mask]
+            valid_copula_rvs.append(copula_rvs)
+
+            # repeating until sample size reached
+            size -= copula_rvs.shape[0]
+            num_loops += 1
+
+            if num_loops > max_num_loops:
+                raise ArithmeticError(f"Unable to generate valid copula rvs. "
+                                      f"number of retries reached "
+                                      f"{max_num_loops}")
+        return np.concatenate(valid_copula_rvs, axis=0)
 
     def _get_components_summary(self,
                                 fitted_mv_object: FittedContinuousMultivariate,