Merge pull request #225 from matthewwardrop/pandas_dict_recarray

Author: matthewwardrop

formulaic/materializers/pandas.py

@@ -2,6 +2,7 @@
 
 import functools
 import itertools
+from collections.abc import Mapping
 from typing import TYPE_CHECKING, Any, Dict, List, Sequence, Set, Tuple, cast
 
 import numpy
@@ -22,9 +23,24 @@
 
 class PandasMaterializer(FormulaMaterializer):
     REGISTER_NAME = "pandas"
-    REGISTER_INPUTS: Sequence[str] = ("pandas.core.frame.DataFrame", "pandas.DataFrame")
+    REGISTER_INPUTS: Sequence[str] = (
+        "pandas.core.frame.DataFrame",
+        "pandas.DataFrame",
+        "dict",
+        "numpy.rec.recarray",
+    )
     REGISTER_OUTPUTS: Sequence[str] = ("pandas", "numpy", "sparse")
 
+    @override
+    def _init(self) -> None:
+        if isinstance(self.data, (dict, Mapping)):
+            if all(numpy.isscalar(v) for v in self.data.values()):
+                self.data = pandas.DataFrame(self.data, index=[0])
+            else:
+                self.data = pandas.DataFrame(self.data)
+        elif isinstance(self.data, numpy.rec.recarray):
+            self.data = pandas.DataFrame.from_records(self.data)
+
     @override
     def _is_categorical(self, values: Any) -> bool:
         if isinstance(values, (pandas.Series, pandas.Categorical)):

tests/materializers/test_base.py

@@ -18,6 +18,8 @@ class TestFormulaMaterializer:
     def test_registrations(self):
         assert sorted(FormulaMaterializer.REGISTERED_NAMES) == ["arrow", "pandas"]
         assert sorted(FormulaMaterializer.REGISTERED_INPUTS) == [
+            "dict",
+            "numpy.rec.recarray",
             "pandas.DataFrame",
             "pandas.core.frame.DataFrame",
             "pyarrow.lib.Table",

tests/materializers/test_pandas.py

@@ -117,6 +117,23 @@ def data_with_nulls(self):
     def materializer(self, data):
         return PandasMaterializer(data)
 
+    def test_data_conversion(self):
+        df = PandasMaterializer({"a": [1, 2, 3]}).data
+        assert isinstance(df, pandas.DataFrame)
+        assert df.columns == ["a"]
+
+        df2 = PandasMaterializer({"a": 1}).data
+        assert isinstance(df2, pandas.DataFrame)
+        assert df2.columns == ["a"]
+        assert list(df2["a"]) == [1]
+
+        df3 = PandasMaterializer(
+            numpy.recarray((2,), dtype=[("x", int), ("y", float), ("z", int)])
+        ).data
+        assert isinstance(df3, pandas.DataFrame)
+        assert list(df3.columns) == ["x", "y", "z"]
+        assert len(df3["x"]) == 2
+
     @pytest.mark.parametrize("formula,tests", PANDAS_TESTS.items())
     def test_get_model_matrix(self, materializer, formula, tests):
         mm = materializer.get_model_matrix(formula, ensure_full_rank=True)