Implement safe_ea_cast to avoid catching Exception (pandas-dev#29293)

Author: jbrockmendel

pandas/core/arrays/__init__.py

@@ -2,6 +2,7 @@
     ExtensionArray,
     ExtensionOpsMixin,
     ExtensionScalarOpsMixin,
+    try_cast_to_ea,
 )
 from .categorical import Categorical  # noqa: F401
 from .datetimes import DatetimeArray  # noqa: F401

pandas/core/arrays/base.py

@@ -32,6 +32,29 @@
 _extension_array_shared_docs = dict()  # type: Dict[str, str]
 
 
+def try_cast_to_ea(cls_or_instance, obj, dtype=None):
+    """
+    Call to `_from_sequence` that returns the object unchanged on Exception.
+
+    Parameters
+    ----------
+    cls_or_instance : ExtensionArray subclass or instance
+    obj : arraylike
+        Values to pass to cls._from_sequence
+    dtype : ExtensionDtype, optional
+
+    Returns
+    -------
+    ExtensionArray or obj
+    """
+    try:
+        result = cls_or_instance._from_sequence(obj, dtype=dtype)
+    except Exception:
+        # We can't predict what downstream EA constructors may raise
+        result = obj
+    return result
+
+
 class ExtensionArray:
     """
     Abstract base class for custom 1-D array types.
@@ -1156,9 +1179,9 @@ def _maybe_convert(arr):
                     # https://github.com/pandas-dev/pandas/issues/22850
                     # We catch all regular exceptions here, and fall back
                     # to an ndarray.
-                    try:
-                        res = self._from_sequence(arr)
-                    except Exception:
+                    res = try_cast_to_ea(self, arr)
+                    if not isinstance(res, type(self)):
+                        # exception raised in _from_sequence; ensure we have ndarray
                         res = np.asarray(arr)
                 else:
                     res = np.asarray(arr)

pandas/core/arrays/categorical.py

@@ -56,7 +56,7 @@
 
 from pandas.io.formats import console
 
-from .base import ExtensionArray, _extension_array_shared_docs
+from .base import ExtensionArray, _extension_array_shared_docs, try_cast_to_ea
 
 _take_msg = textwrap.dedent(
     """\
@@ -2613,10 +2613,10 @@ def _get_codes_for_values(values, categories):
         # Support inferring the correct extension dtype from an array of
         # scalar objects. e.g.
         # Categorical(array[Period, Period], categories=PeriodIndex(...))
-        try:
-            values = categories.dtype.construct_array_type()._from_sequence(values)
-        except Exception:
-            # but that may fail for any reason, so fall back to object
+        cls = categories.dtype.construct_array_type()
+        values = try_cast_to_ea(cls, values)
+        if not isinstance(values, cls):
+            # exception raised in _from_sequence
             values = ensure_object(values)
             categories = ensure_object(categories)
     else:

pandas/core/groupby/groupby.py

@@ -43,7 +43,7 @@ class providing the base-class of operations.
 
 from pandas.core import nanops
 import pandas.core.algorithms as algorithms
-from pandas.core.arrays import Categorical
+from pandas.core.arrays import Categorical, try_cast_to_ea
 from pandas.core.base import DataError, PandasObject, SelectionMixin
 import pandas.core.common as com
 from pandas.core.construction import extract_array
@@ -819,14 +819,8 @@ def _try_cast(self, result, obj, numeric_only=False):
                 # if the type is compatible with the calling EA.
 
                 # return the same type (Series) as our caller
-                try:
-                    result = obj._values._from_sequence(result, dtype=dtype)
-                except Exception:
-                    # https://github.com/pandas-dev/pandas/issues/22850
-                    # pandas has no control over what 3rd-party ExtensionArrays
-                    # do in _values_from_sequence. We still want ops to work
-                    # though, so we catch any regular Exception.
-                    pass
+                cls = dtype.construct_array_type()
+                result = try_cast_to_ea(cls, result, dtype=dtype)
             elif numeric_only and is_numeric_dtype(dtype) or not numeric_only:
                 result = maybe_downcast_to_dtype(result, dtype)
 

pandas/core/series.py

@@ -55,7 +55,7 @@
 import pandas as pd
 from pandas.core import algorithms, base, generic, nanops, ops
 from pandas.core.accessor import CachedAccessor
-from pandas.core.arrays import ExtensionArray
+from pandas.core.arrays import ExtensionArray, try_cast_to_ea
 from pandas.core.arrays.categorical import Categorical, CategoricalAccessor
 from pandas.core.arrays.sparse import SparseAccessor
 import pandas.core.common as com
@@ -2849,14 +2849,7 @@ def combine(self, other, func, fill_value=None):
         elif is_extension_array_dtype(self.values):
             # The function can return something of any type, so check
             # if the type is compatible with the calling EA.
-            try:
-                new_values = self._values._from_sequence(new_values)
-            except Exception:
-                # https://github.com/pandas-dev/pandas/issues/22850
-                # pandas has no control over what 3rd-party ExtensionArrays
-                # do in _values_from_sequence. We still want ops to work
-                # though, so we catch any regular Exception.
-                pass
+            new_values = try_cast_to_ea(self._values, new_values)
         return self._constructor(new_values, index=new_index, name=new_name)
 
     def combine_first(self, other):