feat: Add asdict()

fgpyo/util/metric.py

@@ -116,8 +116,11 @@
 
 """
 
+import dataclasses
+import sys
 from abc import ABC
 from enum import Enum
+from inspect import isclass
 from pathlib import Path
 from typing import Any
 from typing import Callable
@@ -127,8 +130,17 @@
 from typing import List
 from typing import TypeVar
 
+if sys.version_info >= (3, 10):
+    from typing import TypeGuard
+else:
+    from typing_extensions import TypeGuard
+
+import attr
+
 from fgpyo import io
 from fgpyo.util import inspect
+from fgpyo.util.inspect import AttrsInstance
+from fgpyo.util.inspect import DataclassInstance
 
 MetricType = TypeVar("MetricType", bound="Metric")
 
@@ -334,3 +346,65 @@ def fast_concat(*inputs: Path, output: Path) -> None:
             io.write_lines(
                 path=output, lines_to_write=list(io.read_lines(input_path))[1:], append=True
             )
+
+
+def _is_dataclass_instance(metric: Metric) -> TypeGuard[DataclassInstance]:
+    """
+    Test if the given metric is a dataclass instance.
+
+    NB: `dataclasses.is_dataclass` returns True for both dataclass instances and class objects, and
+    we need to override the built-in function's `TypeGuard`.
+
+    Args:
+        metric: An instance of a Metric.
+
+    Returns:
+        True if the given metric is an instance of a dataclass-decorated Metric.
+        False otherwise.
+    """
+    return not isclass(metric) and dataclasses.is_dataclass(metric)
+
+
+def _is_attrs_instance(metric: Metric) -> TypeGuard[AttrsInstance]:
+    """
+    Test if the given metric is an attr.s instance.
+
+    NB: `attr.has` provides a type guard, but only on the class object - we want to narrow the type
+    of the metric instance, so we implement a guard here.
+
+    Args:
+        metric: An instance of a Metric.
+
+    Returns:
+        True if the given metric is an instance of an attr.s-decorated Metric.
+        False otherwise.
+    """
+    return not isclass(metric) and attr.has(metric.__class__)
+
+
+def asdict(metric: Metric) -> Dict[str, Any]:
+    """
+    Convert a Metric instance to a dictionary.
+
+    No formatting is performed on the values, and they are returned as contained (and typed) in the
+    underlying dataclass. Use `Metric.format_value` to format the values to string.
+
+    Args:
+        metric: An instance of a Metric.
+
+    Returns:
+        A dictionary representation of the given metric.
+
+    Raises:
+        TypeError: If the given metric is not an instance of a `dataclass` or `attr.s`-decorated
+        Metric.
+    """
+    if _is_dataclass_instance(metric):
+        return dataclasses.asdict(metric)
+    elif _is_attrs_instance(metric):
+        return attr.asdict(metric)
+    else:
+        raise TypeError(
+            "The provided metric is not an instance of a `dataclass` or `attr.s`-decorated Metric "
+            f"class: {metric.__class__}"
+        )

fgpyo/util/tests/test_metric.py

@@ -29,6 +29,9 @@
 from fgpyo.util.inspect import is_attr_class
 from fgpyo.util.inspect import is_dataclasses_class
 from fgpyo.util.metric import Metric
+from fgpyo.util.metric import _is_attrs_instance
+from fgpyo.util.metric import _is_dataclass_instance
+from fgpyo.util.metric import asdict
 
 
 class EnumTest(enum.Enum):
@@ -519,3 +522,49 @@ def test_metric_columns_out_of_order(tmp_path: Path, data_and_classes: DataBuild
     names = list(NameMetric.read(path=path))
     assert len(names) == 1
     assert names[0] == name
+
+
+def test_is_dataclass_instance() -> None:
+    """Test that _is_dataclass_instance works as expected."""
+
+    # True for `dataclass`-decorated instances but not `attr.s`-decorated instances
+    assert _is_dataclass_instance(dataclasses_data_and_classes.Person(name="name", age=42))
+    assert not _is_dataclass_instance(attr_data_and_classes.Person(name="name", age=42))
+
+    # And False for both classes
+    assert not _is_dataclass_instance(dataclasses_data_and_classes.Person)
+    assert not _is_dataclass_instance(attr_data_and_classes.Person)
+
+
+def test_is_attrs_instance() -> None:
+    """Test that _is_attrs_instance works as expected."""
+
+    # True for `attr.s`-decorated instances but not `dataclass`-decorated instances
+    assert not _is_attrs_instance(dataclasses_data_and_classes.Person(name="name", age=42))
+    assert _is_attrs_instance(attr_data_and_classes.Person(name="name", age=42))
+
+    # And False for both classes
+    assert not _is_attrs_instance(dataclasses_data_and_classes.Person)
+    assert not _is_attrs_instance(attr_data_and_classes.Person)
+
+
+@pytest.mark.parametrize("data_and_classes", (attr_data_and_classes, dataclasses_data_and_classes))
+def test_asdict(data_and_classes: DataBuilder) -> None:
+    """Test that asdict works as expected on both dataclass and attr.s decoreated metrics."""
+
+    assert asdict(data_and_classes.Person(name="name", age=42)) == {"name": "name", "age": 42}
+
+
+def test_asdict_raises() -> None:
+    """Test that we raise a TypeError when asdict is called on a non-metric class."""
+
+    class UndecoratedMetric(Metric["UndecoratedMetric"]):
+        foo: int
+        bar: str
+
+        def __init__(self, foo: int, bar: str):
+            self.foo = foo
+            self.bar = bar
+
+    with pytest.raises(TypeError, match="The provided metric is not an instance"):
+        asdict(UndecoratedMetric(foo=1, bar="a"))