feat: add pyarrow list and struct to pandas engine (#1699)

* feat: add pyarrow list and struct to pandas engine

Signed-off-by: Ajith Aravind <[email protected]>

* test: add tests for pyarrow list and struct

Signed-off-by: Ajith Aravind <[email protected]>

* fix: linting errors

Signed-off-by: Ajith Aravind <[email protected]>

---------

Signed-off-by: Ajith Aravind <[email protected]>

Author: aaravind100

pandera/engines/pandas_engine.py

@@ -22,6 +22,7 @@
     List,
     NamedTuple,
     Optional,
+    Tuple,
     Type,
     Union,
     cast,
@@ -1765,7 +1766,7 @@ class ArrowDecimal128(DataType, dtypes.Decimal):
         precision: int = 28
         scale: int = 0
 
-        def __post_init__(self) -> None:
+        def __post_init__(self):
             type_ = pd.ArrowDtype(
                 pyarrow.decimal128(self.precision, self.scale)
             )
@@ -1832,3 +1833,67 @@ def from_parametrized_dtype(
                 value_type=pyarrow_dtype.value_type,  # type: ignore
                 ordered=pyarrow_dtype.ordered,  # type: ignore
             )
+
+    @Engine.register_dtype(
+        equivalents=[
+            pyarrow.list_,
+            pyarrow.ListType,
+            pyarrow.FixedSizeListType,
+        ]
+    )
+    @immutable(init=True)
+    class ArrowList(DataType):
+        """Semantic representation of a :class:`pyarrow.list_`."""
+
+        type: Optional[pd.ArrowDtype] = dataclasses.field(
+            default=None, init=False
+        )
+        value_type: Optional[Union[pyarrow.DataType, pyarrow.Field]] = (
+            pyarrow.string()
+        )
+        list_size: Optional[int] = -1
+
+        def __post_init__(self):
+            type_ = pd.ArrowDtype(
+                pyarrow.list_(self.value_type, self.list_size)
+            )
+            object.__setattr__(self, "type", type_)
+
+        @classmethod
+        def from_parametrized_dtype(
+            cls,
+            pyarrow_dtype: Union[pyarrow.ListType, pyarrow.FixedSizeListType],
+        ):
+            try:
+                _dtype = cls(
+                    value_type=pyarrow_dtype.value_type,  # type: ignore
+                    list_size=pyarrow_dtype.list_size,  # type: ignore
+                )
+            except AttributeError:
+                _dtype = cls(value_type=pyarrow_dtype.value_type)  # type: ignore
+            return _dtype
+
+    @Engine.register_dtype(equivalents=[pyarrow.struct, pyarrow.StructType])
+    @immutable(init=True)
+    class ArrowStruct(DataType):
+        """Semantic representation of a :class:`pyarrow.struct`."""
+
+        type: Optional[pd.ArrowDtype] = dataclasses.field(
+            default=None, init=False
+        )
+        fields: Optional[
+            Union[
+                Iterable[Union[pyarrow.Field, Tuple[str, pyarrow.DataType]]],
+                Dict[str, pyarrow.DataType],
+            ]
+        ] = tuple()
+
+        def __post_init__(self):
+            type_ = pd.ArrowDtype(pyarrow.struct(self.fields))
+            object.__setattr__(self, "type", type_)
+
+        @classmethod
+        def from_parametrized_dtype(cls, pyarrow_dtype: pyarrow.StructType):
+            return cls(
+                fields=[pyarrow_dtype.field(i) for i in range(pyarrow_dtype.num_fields)]  # type: ignore
+            )

tests/core/test_pandas_engine.py

@@ -8,6 +8,7 @@
 import hypothesis.strategies as st
 import numpy as np
 import pandas as pd
+import pyarrow
 import pytest
 import pytz
 from hypothesis import given
@@ -237,3 +238,62 @@ def test_pandas_date_coerce_dtype(to_df, data):
     assert (
         coerced_data.map(lambda x: isinstance(x, date)) | coerced_data.isna()
     ).all()
+
+
+pandas_arrow_dtype_cases = (
+    (
+        pd.Series([["a", "b", "c"]]),
+        pyarrow.list_(pyarrow.string()),
+    ),
+    (
+        pd.Series([["a", "b"]]),
+        pyarrow.list_(pyarrow.string(), 2),
+    ),
+    (
+        pd.Series([{"foo": 1, "bar": "a"}]),
+        pyarrow.struct(
+            [
+                ("foo", pyarrow.int64()),
+                ("bar", pyarrow.string()),
+            ]
+        ),
+    ),
+)
+
+
+@pytest.mark.parametrize(("data", "dtype"), pandas_arrow_dtype_cases)
+def test_pandas_arrow_dtype(data, dtype):
+    """Test pyarrow dtype."""
+    dtype = pandas_engine.Engine.dtype(dtype)
+
+    dtype.coerce(data)
+
+
+pandas_arrow_dtype_errors_cases = (
+    (
+        pd.Series([["a", "b", "c"]]),
+        pyarrow.list_(pyarrow.int64()),
+    ),
+    (
+        pd.Series([["a", "b"]]),
+        pyarrow.list_(pyarrow.string(), 3),
+    ),
+    (
+        pd.Series([{"foo": 1, "bar": "a"}]),
+        pyarrow.struct(
+            [
+                ("foo", pyarrow.string()),
+                ("bar", pyarrow.int64()),
+            ]
+        ),
+    ),
+)
+
+
+@pytest.mark.parametrize(("data", "dtype"), pandas_arrow_dtype_errors_cases)
+def test_pandas_arrow_dtype_errors(data, dtype):
+    """Test pyarrow dtype raises ArrowInvalid or ArrowTypeError on bad data."""
+    dtype = pandas_engine.Engine.dtype(dtype)
+
+    with pytest.raises((pyarrow.ArrowInvalid, pyarrow.ArrowTypeError)):
+        dtype.coerce(data)

tests/strategies/test_strategies.py

@@ -64,6 +64,8 @@
             pandas_engine.ArrowUInt16,
             pandas_engine.ArrowUInt32,
             pandas_engine.ArrowUInt64,
+            pandas_engine.ArrowList,
+            pandas_engine.ArrowStruct,
         ]
     )