Simplify _extract_codes_and_categories

Author: stanmart

src/tabmat/categorical_matrix.py

@@ -246,7 +246,7 @@ def _extract_codes_and_categories_polars(cat_vec) -> tuple[np.ndarray, np.ndarra
 
 
 def _extract_codes_and_categories_numpy(cat_vec) -> tuple[np.ndarray, np.ndarray]:
-    if pd and pd.__version__ >= "1.5.0":
+    if pd:
         indices, categories = pd.factorize(cat_vec, sort=True)
     else:
         indices, categories = _factorize(cat_vec)
@@ -261,25 +261,18 @@ def _extract_codes_and_categories(cat_vec) -> tuple[np.ndarray, np.ndarray]:
     can be converted to a numpy array. Pandas and polars inputs are special
     cased for performance considerations.
     """
-    if pd and isinstance(cat_vec, pd.Categorical):
-        # Narwhals can only handle series, not bare Categoricals
-        cat_vec = pd.Series(cat_vec, copy=False)
-    # We convert to narwhals first so we handle narwhalized and non-narwhalized
-    # pandas and polars inputs the same way.
-    cat_vec = nw.from_native(cat_vec, series_only=True, strict=False)
-
-    if isinstance(cat_vec, nw.Series):
-        package = nw.get_native_namespace(cat_vec).__name__
-    else:
-        package = None
+    cat_vec_native = nw.to_native(cat_vec, strict=False)
 
-    if package == "pandas":
-        return _extract_codes_and_categories_pandas(nw.to_native(cat_vec))
-    elif package == "polars":
-        return _extract_codes_and_categories_polars(nw.to_native(cat_vec))
+    if pd and isinstance(cat_vec_native, (pd.Series, pd.Categorical)):
+        return _extract_codes_and_categories_pandas(cat_vec_native)
+    elif pl and isinstance(cat_vec_native, pl.Series):
+        return _extract_codes_and_categories_polars(cat_vec_native)
     else:
-        if isinstance(cat_vec, nw.Series):
-            cat_vec = cat_vec.cast(nw.String).to_numpy()
+        if isinstance(
+            cat_vec_narwhals := nw.from_native(cat_vec, series_only=True, strict=False),
+            nw.Series,
+        ):
+            cat_vec = cat_vec_narwhals.cast(nw.String).to_numpy()
         else:
             cat_vec = np.asarray(cat_vec)
         return _extract_codes_and_categories_numpy(cat_vec)