From 39db5dc0bbf1e4bfee2e99a532061e20e12a81ac Mon Sep 17 00:00:00 2001
From: Jeroen van Zundert <mail@jeroenvanzundert.nl>
Date: Wed, 31 May 2023 20:38:07 +0200
Subject: [PATCH] fix(python): Fix DataFrame.to_arrow() for 0x0 dataframes

No schema can be inferred from the batches, as `._df.to_arrow()` returns an empty list. `pa.Tables.from_batches()` requires a schema if no batches are passed in.

Closes #8884.
---
 py-polars/polars/dataframe/frame.py  |  7 +++++--
 py-polars/tests/unit/test_interop.py | 17 +++++++++++++++++
 2 files changed, 22 insertions(+), 2 deletions(-)

diff --git a/py-polars/polars/dataframe/frame.py b/py-polars/polars/dataframe/frame.py
index b8699c08d3c6..267d51460ac1 100644
--- a/py-polars/polars/dataframe/frame.py
+++ b/py-polars/polars/dataframe/frame.py
@@ -1845,8 +1845,11 @@ def to_arrow(self) -> pa.Table:
         bar: [["a","b","c","d","e","f"]]
 
         """
-        record_batches = self._df.to_arrow()
-        return pa.Table.from_batches(record_batches)
+        if self.shape[1]:  # all except 0x0 dataframe
+            record_batches = self._df.to_arrow()
+            return pa.Table.from_batches(record_batches)
+        else:  # 0x0 dataframe, cannot infer schema from batches
+            return pa.table({})
 
     @overload
     def to_dict(self, as_series: Literal[True] = ...) -> dict[str, Series]:
diff --git a/py-polars/tests/unit/test_interop.py b/py-polars/tests/unit/test_interop.py
index d1ff18876f85..059ad1d4b5ef 100644
--- a/py-polars/tests/unit/test_interop.py
+++ b/py-polars/tests/unit/test_interop.py
@@ -220,6 +220,23 @@ def test_arrow_null_roundtrip() -> None:
         assert c1.to_pylist() == c2.to_pylist()
 
 
+def test_arrow_empty_dataframe() -> None:
+    # 0x0 dataframe
+    df = pl.DataFrame({})
+    tbl = pa.table({})
+    assert df.to_arrow() == tbl
+    df2 = cast(pl.DataFrame, pl.from_arrow(df.to_arrow()))
+    assert_frame_equal(df2, df)
+
+    # 0 row dataframe
+    df = pl.DataFrame({}, schema={"a": pl.Int32})
+    tbl = pa.Table.from_batches([], pa.schema([pa.field("a", pa.int32())]))
+    assert df.to_arrow() == tbl
+    df2 = cast(pl.DataFrame, pl.from_arrow(df.to_arrow()))
+    assert df2.schema == {"a": pl.Int32}
+    assert df2.shape == (0, 1)
+
+
 def test_arrow_dict_to_polars() -> None:
     pa_dict = pa.DictionaryArray.from_arrays(
         indices=np.array([0, 1, 2, 3, 1, 0, 2, 3, 3, 2]),