[Python] Append struct after appending children

During conversion from Python to Arrow, when a struct's child
hits a capacity error and chunking is triggered, this can leave
the Finish'd chunk in an invalid state since the struct's length
does not match the length of its children.

This change simply tries to Append the children first, and only
if successful will Append the struct. This is safe because the
order of Append'ing between the struct and its child is not
specified. It is only specified that they must be consistent
with each other.

python/pyarrow/src/arrow/python/python_to_arrow.cc

@@ -926,14 +926,14 @@ class PyStructConverter : public StructConverter<PyConverter, PyConverterTrait>
     }
     switch (input_kind_) {
       case InputKind::DICT:
-        RETURN_NOT_OK(this->struct_builder_->Append());
-        return AppendDict(value);
+        RETURN_NOT_OK(AppendDict(value));
+        return this->struct_builder_->Append();
       case InputKind::TUPLE:
-        RETURN_NOT_OK(this->struct_builder_->Append());
-        return AppendTuple(value);
+        RETURN_NOT_OK(AppendTuple(value));
+        return this->struct_builder_->Append();
       case InputKind::ITEMS:
-        RETURN_NOT_OK(this->struct_builder_->Append());
-        return AppendItems(value);
+        RETURN_NOT_OK(AppendItems(value));
+        return this->struct_builder_->Append();
       default:
         RETURN_NOT_OK(InferInputKind(value));
         return Append(value);

python/pyarrow/tests/test_pandas.py

@@ -4936,3 +4936,62 @@ def test_array_conversion_for_datetime():
 
     result = arr.to_pandas()
     tm.assert_series_equal(result, series)
+
+
+def test_nested_chunking_valid():
+    # GH-32439
+    # https://github.com/apache/arrow/issues/32439
+    # Chunking can cause arrays to be in invalid state
+    # when nested types are involved.
+    # Here we simply ensure we validate correctly.
+
+    x = "0" * 1000000000
+    df = pd.DataFrame({"strings": [x, x, x]})
+    tab = pa.Table.from_pandas(df)
+    # we expect to trigger chunking internally
+    # an assertion failure here may just mean this threshold has changed
+    assert tab.column(0).num_chunks > 1
+    tm.assert_frame_equal(tab.to_pandas(self_destruct=True), df)
+
+    struct = {"struct_field": x}
+    df = pd.DataFrame({"structs": [struct, struct, struct]})
+    tab = pa.Table.from_pandas(df)
+    assert tab.column(0).num_chunks > 1
+    tm.assert_frame_equal(tab.to_pandas(self_destruct=True), df)
+
+    lists = [x]
+    del x
+    df = pd.DataFrame({"lists": [lists, lists, lists]})
+    tab = pa.Table.from_pandas(df)
+    assert tab.column(0).num_chunks > 1
+    tm.assert_frame_equal(tab.to_pandas(self_destruct=True), df)
+
+    los = [struct]
+    df = pd.DataFrame({"los": [los, los, los]})
+    tab = pa.Table.from_pandas(df)
+    assert tab.column(0).num_chunks > 1
+    tm.assert_frame_equal(tab.to_pandas(self_destruct=True), df)
+
+    sol = {"struct_field": lists}
+    df = pd.DataFrame({"sol": [sol, sol, sol]})
+    tab = pa.Table.from_pandas(df)
+    assert tab.column(0).num_chunks > 1
+    tm.assert_frame_equal(tab.to_pandas(self_destruct=True), df)
+
+    map_of_los = {"a": los}
+    df = pd.DataFrame({"maps": [map_of_los, map_of_los, map_of_los]})
+    tab = pa.Table.from_pandas(
+        df,
+        schema=pa.schema(
+            [(
+                "maps",
+                pa.map_(
+                    pa.string(),
+                    pa.list_(
+                        pa.struct([pa.field("struct_field", pa.string())])
+                    )
+                )
+            )]))
+    assert tab.column(0).num_chunks > 1
+    tm.assert_frame_equal(tab.to_pandas(
+        self_destruct=True, maps_as_pydicts="strict"), df)