feat: support Index.to_frame()

bigframes/core/indexes/base.py

@@ -115,6 +115,19 @@ def from_frame(
         index._linked_frame = frame
         return index
 
+    def to_frame(
+        self, index: bool = True, name: blocks.Label | None = None
+    ) -> bigframes.dataframe.DataFrame:
+        provided_name = name if name else self.name
+        series = self.to_series()
+        series.name = provided_name
+        frame = series.to_frame()
+        if index:  # matching pandas behavior
+            frame.index.name = self.name
+        else:
+            frame = frame.reset_index(drop=True)
+        return frame
+
     @property
     def _session(self):
         return self._block.session

bigframes/core/indexes/multi.py

@@ -19,7 +19,9 @@
 import bigframes_vendored.pandas.core.indexes.multi as vendored_pandas_multindex
 import pandas
 
+import bigframes.core.blocks as blocks
 from bigframes.core.indexes.base import Index
+import bigframes.dataframe
 
 
 class MultiIndex(Index, vendored_pandas_multindex.MultiIndex):
@@ -46,3 +48,28 @@ def from_arrays(
         pd_index = pandas.MultiIndex.from_arrays(arrays, sortorder, names)
         # Index.__new__ should detect multiple levels and properly create a multiindex
         return cast(MultiIndex, Index(pd_index))
+
+    def to_frame(
+        self,
+        index: bool = True,
+        name: Sequence[blocks.Label] | blocks.Label | None = None,
+    ) -> bigframes.dataframe.DataFrame:
+        columns = [
+            [self.values[j][i] for j in range(len(self.values))]
+            for i in range(len(self.values[0]))
+        ]
+        if isinstance(name, Sequence):
+            if len(name) != len(columns):
+                raise ValueError(
+                    "Length of provided names must match length of MultiIndex columns"
+                )
+            data = {name[i]: column for i, column in enumerate(columns)}
+        elif name is None:
+            data = {i: column for i, column in enumerate(columns)}
+        else:
+            raise ValueError("'name' parameter must be of type Sequence")
+        original_index = columns
+        result = bigframes.dataframe.DataFrame(
+            data, index=original_index if index else None
+        )
+        return result

tests/system/small/test_index.py

@@ -16,6 +16,7 @@
 import pandas as pd
 import pytest
 
+import bigframes.core.indexes as indexes
 import bigframes.pandas as bpd
 from tests.system.utils import assert_pandas_index_equal_ignore_index_type
 
@@ -320,6 +321,30 @@ def test_index_to_series(
     pd.testing.assert_series_equal(bf_result, pd_result)
 
 
+@pytest.mark.parametrize("index_arg", [True, False])
+@pytest.mark.parametrize("name_arg", [None, "food"])
+def test_index_to_frame(index_arg, name_arg):
+    pd_idx: pd.Index = pd.Index(
+        ["Ant", "Bear", "Cow"], name="animal", dtype="string[pyarrow]"
+    )
+    bf_idx = indexes.Index(["Ant", "Bear", "Cow"], name="animal")
+
+    if name_arg is None:
+        pd_df = pd_idx.to_frame(index=index_arg)
+        bf_df = bf_idx.to_frame(index=index_arg)
+    else:
+        pd_df = pd_idx.to_frame(index=index_arg, name=name_arg)
+        bf_df = bf_idx.to_frame(index=index_arg, name=name_arg)
+    pd.testing.assert_frame_equal(
+        pd_df, bf_df.to_pandas(), check_column_type=False, check_index_type=False
+    )
+    # BigFrames type casting is weird
+    # automatically casts dtype to string whereas pandas dtype is object
+    # additionally, pandas uses string[python] and BigFrames uses string[pyarrow]
+    # so we set dtype in pandas index creation
+    # similarly, pandas uses int64 dtype for numerical index and BigFrames uses Int64
+
+
 @pytest.mark.parametrize(
     ("how",),
     [

tests/system/small/test_multiindex.py

@@ -16,6 +16,7 @@
 import pandas
 import pytest
 
+import bigframes.core.indexes as indexes
 import bigframes.pandas as bpd
 from tests.system.utils import assert_pandas_df_equal, skip_legacy_pandas
 
@@ -45,6 +46,27 @@ def test_multi_index_from_arrays():
     pandas.testing.assert_index_equal(bf_idx.to_pandas(), pd_idx)
 
 
+@pytest.mark.parametrize("index_arg", [True, False])
+@pytest.mark.parametrize("name_arg", [None, ["x", "y"]])
+def test_multi_index_to_frame(index_arg, name_arg):
+
+    pd_idx = pandas.MultiIndex.from_arrays([["a", "b", "c"], ["d", "e", "f"]])
+    bf_idx = indexes.MultiIndex.from_arrays([["a", "b", "c"], ["d", "e", "f"]])
+    if name_arg is None:
+        pd_df = pd_idx.to_frame(index=index_arg)
+        bf_df = bf_idx.to_frame(index=index_arg)
+    else:
+        pd_df = pd_idx.to_frame(index=index_arg, name=name_arg)
+        bf_df = bf_idx.to_frame(index=index_arg, name=name_arg)
+    pandas.testing.assert_frame_equal(
+        pd_df,
+        bf_df.to_pandas(),
+        check_dtype=False,
+        check_column_type=False,
+        check_index_type=False,
+    )
+
+
 @skip_legacy_pandas
 def test_read_pandas_multi_index_axes():
     index = pandas.MultiIndex.from_arrays(