[ENH] deduplicate dependency checkers and isolate soft dependency imports by using skbase

pycaret/containers/models/classification.py

@@ -263,7 +263,9 @@ def __init__(self, experiment):
         if self.engine == "sklearn":
             from sklearn.linear_model import LogisticRegression
         elif self.engine == "sklearnex":
-            if _check_soft_dependencies("sklearnex", extra=None, severity="warning"):
+            if _check_soft_dependencies(
+                "scikit-learn-intelex", extra=None, severity="warning"
+            ):
                 from sklearnex.linear_model import LogisticRegression
             else:
                 from sklearn.linear_model import LogisticRegression
@@ -324,7 +326,9 @@ def __init__(self, experiment):
         if self.engine == "sklearn":
             from sklearn.neighbors import KNeighborsClassifier
         elif self.engine == "sklearnex":
-            if _check_soft_dependencies("sklearnex", extra=None, severity="warning"):
+            if _check_soft_dependencies(
+                "scikit-learn-intelex", extra=None, severity="warning"
+            ):
                 from sklearnex.neighbors import KNeighborsClassifier
             else:
                 from sklearn.neighbors import KNeighborsClassifier
@@ -580,7 +584,9 @@ def __init__(self, experiment):
         if self.engine == "sklearn":
             from sklearn.svm import SVC
         elif self.engine == "sklearnex":
-            if _check_soft_dependencies("sklearnex", extra=None, severity="warning"):
+            if _check_soft_dependencies(
+                "scikit-learn-intelex", extra=None, severity="warning"
+            ):
                 from sklearnex.svm import SVC
             else:
                 from sklearn.svm import SVC

pycaret/containers/models/clustering.py

@@ -191,7 +191,9 @@ def __init__(self, experiment):
         if self.engine == "sklearn":
             from sklearn.cluster import KMeans
         elif self.engine == "sklearnex":
-            if _check_soft_dependencies("sklearnex", extra=None, severity="warning"):
+            if _check_soft_dependencies(
+                "scikit-learn-intelex", extra=None, severity="warning"
+            ):
                 from sklearnex.cluster import KMeans
             else:
                 from sklearn.cluster import KMeans
@@ -340,7 +342,9 @@ def __init__(self, experiment):
         if self.engine == "sklearn":
             from sklearn.cluster import DBSCAN
         elif self.engine == "sklearnex":
-            if _check_soft_dependencies("sklearnex", extra=None, severity="warning"):
+            if _check_soft_dependencies(
+                "scikit-learn-intelex", extra=None, severity="warning"
+            ):
                 from sklearnex.cluster import DBSCAN
             else:
                 from sklearn.cluster import DBSCAN

pycaret/containers/models/regression.py

@@ -233,7 +233,9 @@ def __init__(self, experiment):
         if self.engine == "sklearn":
             from sklearn.linear_model import LinearRegression
         elif self.engine == "sklearnex":
-            if _check_soft_dependencies("sklearnex", extra=None, severity="warning"):
+            if _check_soft_dependencies(
+                "scikit-learn-intelex", extra=None, severity="warning"
+            ):
                 from sklearnex.linear_model import LinearRegression
             else:
                 from sklearn.linear_model import LinearRegression
@@ -287,7 +289,9 @@ def __init__(self, experiment):
         if self.engine == "sklearn":
             from sklearn.linear_model import Lasso
         elif self.engine == "sklearnex":
-            if _check_soft_dependencies("sklearnex", extra=None, severity="warning"):
+            if _check_soft_dependencies(
+                "scikit-learn-intelex", extra=None, severity="warning"
+            ):
                 from sklearnex.linear_model import Lasso
             else:
                 from sklearn.linear_model import Lasso
@@ -344,7 +348,9 @@ def __init__(self, experiment):
         if self.engine == "sklearn":
             from sklearn.linear_model import Ridge
         elif self.engine == "sklearnex":
-            if _check_soft_dependencies("sklearnex", extra=None, severity="warning"):
+            if _check_soft_dependencies(
+                "scikit-learn-intelex", extra=None, severity="warning"
+            ):
                 from sklearnex.linear_model import Ridge
             else:
                 from sklearn.linear_model import Ridge
@@ -401,7 +407,9 @@ def __init__(self, experiment):
         if self.engine == "sklearn":
             from sklearn.linear_model import ElasticNet
         elif self.engine == "sklearnex":
-            if _check_soft_dependencies("sklearnex", extra=None, severity="warning"):
+            if _check_soft_dependencies(
+                "scikit-learn-intelex", extra=None, severity="warning"
+            ):
                 from sklearnex.linear_model import ElasticNet
             else:
                 from sklearn.linear_model import ElasticNet
@@ -1004,7 +1012,9 @@ def __init__(self, experiment):
         if self.engine == "sklearn":
             from sklearn.svm import SVR
         elif self.engine == "sklearnex":
-            if _check_soft_dependencies("sklearnex", extra=None, severity="warning"):
+            if _check_soft_dependencies(
+                "scikit-learn-intelex", extra=None, severity="warning"
+            ):
                 from sklearnex.svm import SVR
             else:
                 from sklearn.svm import SVR
@@ -1065,7 +1075,9 @@ def __init__(self, experiment):
         if self.engine == "sklearn":
             from sklearn.neighbors import KNeighborsRegressor
         elif self.engine == "sklearnex":
-            if _check_soft_dependencies("sklearnex", extra=None, severity="warning"):
+            if _check_soft_dependencies(
+                "scikit-learn-intelex", extra=None, severity="warning"
+            ):
                 from sklearnex.neighbors import KNeighborsRegressor
             else:
                 from sklearn.neighbors import KNeighborsRegressor

pycaret/containers/models/time_series.py

@@ -1495,7 +1495,7 @@ def return_regressor_class(self):
         if self.engine == "sklearn":
             from sklearn.linear_model import LinearRegression
         elif self.engine == "sklearnex":
-            _check_soft_dependencies("sklearnex", extra=None, severity="error")
+            _check_soft_dependencies("scikit-learn-intelex", extra=None, severity="error")  # noqa: E501
             from sklearnex.linear_model import LinearRegression
 
         if self.gpu_param == "force":
@@ -1552,7 +1552,7 @@ def return_regressor_class(self):
         if self.engine == "sklearn":
             from sklearn.linear_model import ElasticNet
         elif self.engine == "sklearnex":
-            _check_soft_dependencies("sklearnex", extra=None, severity="error")
+            _check_soft_dependencies("scikit-learn-intelex", extra=None, severity="error")  # noqa: E501
             from sklearnex.linear_model import ElasticNet
 
         if self.gpu_param == "force":
@@ -1613,7 +1613,7 @@ def return_regressor_class(self):
         if self.engine == "sklearn":
             from sklearn.linear_model import Ridge
         elif self.engine == "sklearnex":
-            _check_soft_dependencies("sklearnex", extra=None, severity="error")
+            _check_soft_dependencies("scikit-learn-intelex", extra=None, severity="error")  # noqa: E501
             from sklearnex.linear_model import Ridge
 
         if self.gpu_param == "force":
@@ -1673,7 +1673,7 @@ def return_regressor_class(self):
         if self.engine == "sklearn":
             from sklearn.linear_model import Lasso
         elif self.engine == "sklearnex":
-            _check_soft_dependencies("sklearnex", extra=None, severity="error")
+            _check_soft_dependencies("scikit-learn-intelex", extra=None, severity="error")  # noqa: E501
             from sklearnex.linear_model import Lasso
 
         if self.gpu_param == "force":
@@ -2040,7 +2040,7 @@ def return_regressor_class(self):
         if self.engine == "sklearn":
             from sklearn.neighbors import KNeighborsRegressor
         elif self.engine == "sklearnex":
-            _check_soft_dependencies("sklearnex", extra=None, severity="error")
+            _check_soft_dependencies("scikit-learn-intelex", extra=None, severity="error")  # noqa: E501
             from sklearnex.neighbors import KNeighborsRegressor
 
         if self.gpu_param == "force":

pycaret/utils/_dependencies.py

@@ -1,95 +1,20 @@
 # Adapted from
 # https://github.com/sktime/sktime/blob/v0.11.0/sktime/utils/validation/_dependencies.py
 
-import sys
-from distutils.version import LooseVersion
-from importlib import import_module
-from typing import Dict, Optional, Union
+from typing import Optional
 
-from importlib_metadata import distributions
+from skbase.utils.dependencies import _check_soft_dependencies as _skbase_csd
+from skbase.utils.dependencies._dependencies import _get_installed_packages
 
-from pycaret.internal.logging import get_logger, redirect_output
+from pycaret.internal.logging import get_logger
 
 logger = get_logger()
 
-INSTALLED_MODULES = None
-
-
-def _try_import_and_get_module_version(
-    modname: str,
-) -> Optional[Union[LooseVersion, bool]]:
-    """Returns False if module is not installed, None if version is not available"""
-    try:
-        if modname in sys.modules:
-            mod = sys.modules[modname]
-        else:
-            if logger:
-                with redirect_output(logger):
-                    mod = import_module(modname)
-            else:
-                mod = import_module(modname)
-        try:
-            ver = mod.__version__
-        except AttributeError:
-            # Version could not be obtained
-            ver = None
-    except ImportError:
-        ver = False
-    if ver:
-        ver = LooseVersion(ver)
-    return ver
-
-
-# Based on packages_distributions() from importlib_metadata
-def get_installed_modules() -> Dict[str, Optional[LooseVersion]]:
-    """
-    Get installed modules and their versions from pip metadata.
-    """
-    global INSTALLED_MODULES
-    if not INSTALLED_MODULES:
-        # Get all installed modules and their versions without
-        # needing to import them.
-        module_versions = {}
-        # top_level.txt contains information about modules
-        # in the package. It is not always present, in which case
-        # the assumption is that the package name is the module name.
-        # https://setuptools.pypa.io/en/latest/deprecated/python_eggs.html
-        for dist in distributions():
-            for pkg in (dist.read_text("top_level.txt") or "").split():
-                try:
-                    ver = LooseVersion(dist.metadata["Version"])
-                except Exception:
-                    ver = None
-                module_versions[pkg] = ver
-        INSTALLED_MODULES = module_versions
-    return INSTALLED_MODULES
-
-
-def _get_module_version(modname: str) -> Optional[Union[LooseVersion, bool]]:
-    """Will cache the version in INSTALLED_MODULES
-
-    Returns False if module is not installed."""
-    installed_modules = get_installed_modules()
-    if modname not in installed_modules:
-        # Fallback. This should never happen unless module is not present
-        installed_modules[modname] = _try_import_and_get_module_version(modname)
-    return installed_modules[modname]
-
-
-def get_module_version(modname: str) -> Optional[LooseVersion]:
-    """Raises a ValueError if module is not installed"""
-    version = _get_module_version(modname)
-    if version is False:
-        raise ValueError(f"Module '{modname}' is not installed.")
-    return version
 
-
-def is_module_installed(modname: str) -> bool:
-    try:
-        get_module_version(modname)
-        return True
-    except ValueError:
-        return False
+def get_module_version_str(modname: str) -> str:
+    """Raises a ValueError if module is not installed"""
+    versions = _get_installed_packages()
+    return versions.get(modname, "Not installed")
 
 
 def _check_soft_dependencies(
@@ -103,18 +28,28 @@ def _check_soft_dependencies(
 
     Parameters
     ----------
-    package : str
-        Package to check
-    severity : str, optional
-        Whether to raise an error ("error") or just a warning message ("warning"),
-        by default "error"
+    packages : str
+        package name to check
+        str should be package name and/or package version specifications to check.
+        Must be a PEP 440 compatible specifier string, for a single package.
+        For instance, the PEP 440 compatible package name such as ``"pandas"``;
+        or a package requirement specifier string such as ``"pandas>1.2.3"``.
+
+    severity : str, "error" (default), "warning", "none"
+        whether the check should raise an error, a warning, or nothing
+
+        * "error" - raises a ``ModuleNotFoundError`` if one of packages is not installed
+        * "warning" - raises a warning if one of packages is not installed
+          function returns False if one of packages is not installed, otherwise True
+        * "none" - does not raise exception or warning
+          function returns False if one of packages is not installed, otherwise True
+
     extra : Optional[str], optional
         The 'extras' that will install this package, by default "all_extras".
         If None, it means that the dependency is not available in optional
         requirements file and must be installed by the user on their own.
-    install_name : Optional[str], optional
-        The package name to install, by default None
-        If none, the name in `package` argument is used
+
+    install_name : ignored, present only for backwards compatibility
 
     Returns
     -------
@@ -130,33 +65,22 @@ def _check_soft_dependencies(
     RuntimeError
         Is the severity argument is not one of the allowed values
     """
-    install_name = install_name or package
+    msg = (
+        f"\n'{package}' is a soft dependency and not included in the "
+        f"pycaret installation. Please run: `pip install {package}` to install. "
+    )
+    if extra is not None:
+        msg += (
+            f"\nAlternately, you can install {package} by running "
+            f"`pip install pycaret[{extra}]`"
+        )
 
-    package_available = is_module_installed(package)
+    package_available = _skbase_csd(package, severity=severity, msg=msg)
 
     if package_available:
-        ver = get_module_version(package)
+        ver = get_module_version_str(package)
         logger.info(
             "Soft dependency imported: {k}: {stat}".format(k=package, stat=str(ver))
         )
-    else:
-        msg = (
-            f"\n'{package}' is a soft dependency and not included in the "
-            f"pycaret installation. Please run: `pip install {install_name}` to install."
-        )
-        if extra is not None:
-            msg += f"\nAlternately, you can install this by running `pip install pycaret[{extra}]`"
-
-        if severity == "error":
-            logger.exception(f"{msg}")
-            raise ModuleNotFoundError(msg)
-        elif severity == "warning":
-            logger.warning(f"{msg}")
-            package_available = False
-        else:
-            raise RuntimeError(
-                "Error in calling _check_soft_dependencies, severity "
-                f'argument must be "error" or "warning", found "{severity}".'
-            )
 
     return package_available

pycaret/utils/_show_versions.py

@@ -11,7 +11,7 @@
 import sys
 from typing import Optional
 
-from pycaret.utils._dependencies import get_module_version
+from pycaret.utils._dependencies import get_module_version_str
 
 required_deps = [
     "pip",
@@ -127,7 +127,7 @@ def _get_deps_info(optional: bool = False, logger: Optional[logging.Logger] = No
 
     for modname in deps:
         try:
-            ver = get_module_version(modname)
+            ver = get_module_version_str(modname)
             if not ver:
                 ver = "Installed but version unavailable"
         except ValueError:

pyproject.toml

@@ -46,6 +46,8 @@ classifiers = [
 # this set should be kept minimal!
 dependencies = [
   # Base
+  "packaging",
+  "scikit-base>=0.6.1",
   "ipython>=5.5.0",
   "ipywidgets>=7.6.5",
   "tqdm>=4.62.0",