Fix encoding type inference for boolean columns when pyarrow is installed (#3210)

* Work around pandas-dev/pandas#55332

* Only test boolean column with pandas >= 1.0.0

Author: jonmmease

altair/utils/core.py

@@ -588,8 +588,10 @@ def parse_shorthand(
                     column = dfi.get_column_by_name(unescaped_field)
                     try:
                         attrs["type"] = infer_vegalite_type_for_dfi_column(column)
-                    except NotImplementedError:
-                        # Fall back to pandas-based inference
+                    except (NotImplementedError, AttributeError):
+                        # Fall back to pandas-based inference.
+                        # Note: The AttributeError catch is a workaround for
+                        # https://github.com/pandas-dev/pandas/issues/55332
                         if isinstance(data, pd.DataFrame):
                             attrs["type"] = infer_vegalite_type(data[unescaped_field])
                         else:

tests/utils/test_core.py

@@ -1,4 +1,6 @@
 import types
+from packaging.version import Version
+from importlib.metadata import version as importlib_version
 
 import numpy as np
 import pandas as pd
@@ -16,6 +18,8 @@
 except ImportError:
     pa = None
 
+PANDAS_VERSION = Version(importlib_version("pandas"))
+
 
 FAKE_CHANNELS_MODULE = f'''
 """Fake channels module for utility tests."""
@@ -160,6 +164,10 @@ def check(s, data, **kwargs):
     check("month(z)", data, timeUnit="month", field="z", type="temporal")
     check("month(t)", data, timeUnit="month", field="t", type="temporal")
 
+    if PANDAS_VERSION >= Version("1.0.0"):
+        data["b"] = pd.Series([True, False, True, False, None], dtype="boolean")
+        check("b", data, field="b", type="nominal")
+
 
 @pytest.mark.skipif(pa is None, reason="pyarrow not installed")
 def test_parse_shorthand_for_arrow_timestamp():