[MNT] Isolate scikit-learn dependency with checks (#674)

`scikit-learn` is a soft dependency - this PR fully isolates
`scikit-learn` with dependency checks.

It is not removed from the core dependency set, even though the
dependency is entirely localized in the `CovarianceShrinkage` object,
since it seems to be called frequently from the rest of the code base.

Author: fkiraly

pypfopt/risk_models.py

@@ -25,6 +25,7 @@
 
 import numpy as np
 import pandas as pd
+from skbase.utils.dependencies import _check_soft_dependencies
 
 from .expected_returns import returns_from_prices
 
@@ -298,11 +299,14 @@ def min_cov_determinant(
         warnings.warn("data is not in a dataframe", RuntimeWarning)
         prices = pd.DataFrame(prices)
 
-    # Extra dependency
-    try:
-        import sklearn.covariance
-    except (ModuleNotFoundError, ImportError):
-        raise ImportError("Please install scikit-learn via pip or poetry")
+    if not _check_soft_dependencies(["scikit-learn"], severity="none"):
+        raise ImportError(
+            "scikit-learn is required to use min_cov_determinant. "
+            "Please ensure that scikit-learn is installed in your environment,"
+            " e.g via pip install scikit-learn"
+        )
+
+    from sklearn.covariance import fast_mcd
 
     assets = prices.columns
 
@@ -312,7 +316,7 @@ def min_cov_determinant(
         X = returns_from_prices(prices, log_returns)
     # X = np.nan_to_num(X.values)
     X = X.dropna().values
-    raw_cov_array = sklearn.covariance.fast_mcd(X, random_state=random_state)[1]
+    raw_cov_array = fast_mcd(X, random_state=random_state)[1]
     cov = pd.DataFrame(raw_cov_array, index=assets, columns=assets) * frequency
     return fix_nonpositive_semidefinite(cov, kwargs.get("fix_method", "spectral"))
 
@@ -379,13 +383,16 @@ def __init__(self, prices, returns_data=False, frequency=252, log_returns=False)
         :param log_returns: whether to compute using log returns
         :type log_returns: bool, defaults to False
         """
-        # Optional import
-        try:
-            from sklearn import covariance
+        if not _check_soft_dependencies(["scikit-learn"], severity="none"):
+            raise ImportError(
+                "scikit-learn is required to use CovarianceShrinkage. "
+                "Please ensure that scikit-learn is installed in your environment,"
+                " e.g via pip install scikit-learn"
+            )
+
+        from sklearn import covariance
 
-            self.covariance = covariance
-        except (ModuleNotFoundError, ImportError):  # pragma: no cover
-            raise ImportError("Please install scikit-learn via pip or poetry")
+        self.covariance = covariance
 
         if not isinstance(prices, pd.DataFrame):
             warnings.warn("data is not in a dataframe", RuntimeWarning)

pyproject.toml

@@ -35,9 +35,9 @@ dependencies = [
     "cvxpy>=1.1.19",
     "numpy>=1.26.0",
     "pandas>=0.19",
+    "scikit-base<0.14.0",
     "scikit-learn>=0.24.1",
     "scipy>=1.3.0",
-    "scikit-base<0.14.0",
 ]
 
 [project.optional-dependencies]
@@ -54,7 +54,6 @@ dependencies = [
 all_extras = [
     "matplotlib>=3.2.0",
     "plotly>=5.0.0,<6",
-    "scikit-learn>=0.24.1",
     "ecos>=2.0.14,<2.1",
     "plotly>=5.0.0,<7",
     "cvxopt; python_version < '3.14'",