Skip to content

GH-55369: [C++][IPC] Map V4 union child fields when subsetting #55370

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
professor-moody wants to merge 2 commits into
base: main
Choose a base branch
from
Open

GH-55369: [C++][IPC] Map V4 union child fields when subsetting #55370

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
7abfa94
GH-55369: [C++][IPC] Map V4 union child fields when subsetting
professor-moody Nov 2, 2026
12cffc6
GH-55369: [C++][IPC] Map V4 union child fields when subsetting
professor-moody Nov 2, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 13 additions & 0 deletions cpp/src/arrow/ipc/message.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -306,6 +306,19 @@ Status ReadFieldsSubset(int64_t offset, io::RandomAccessFile* file,
}
internal::IoRecordedRandomAccessFile io_recorded_random_access_file(required_size);
RETURN_NOT_OK(fields_loader(batch, &io_recorded_random_access_file));
if (message->version() < flatbuf::MetadataVersion::MetadataVersion_V5) {
// The read ranges recorded by the fields loader follow the V5 buffer layout,
// which differs from earlier metadata versions for union arrays (they have
// a validity buffer slot before V5). Read the whole body instead.
ARROW_ASSIGN_OR_RAISE(
auto bytes_read,
file->ReadAt(offset + metadata->size(), required_size, body->mutable_data()));
if (bytes_read != required_size) {
return Status::IOError("Expected to read ", required_size,
" bytes for message body, got ", bytes_read);
}
return Status::OK();
}
const auto& read_ranges = io_recorded_random_access_file.GetReadRanges();
for (const auto& range : read_ranges) {
auto read_result = file->ReadAt(offset + metadata->size() + range.offset,
Expand Down
62 changes: 62 additions & 0 deletions cpp/src/arrow/ipc/read_write_test.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1623,6 +1623,43 @@ class ReaderWriterMixin : public ExtensionTypesMixin {
}
}

void TestReadSubsetOfFieldsWithUnions(MetadataVersion metadata_version) {
// Union arrays have a validity buffer slot in V4 metadata but not in V5;
// reading a subset of fields must use the buffer layout of the message.
auto sparse_type = sparse_union({field("i", int32()), field("s", utf8())}, {0, 1});
auto dense_type = dense_union({field("i", int32()), field("s", utf8())}, {0, 1});
auto a0 = ArrayFromJSON(sparse_type, R"([[0, 11], [1, "x"], [0, 33]])");
auto a1 = ArrayFromJSON(dense_type, R"([[1, "y"], [0, 22], [0, 44]])");
auto a2 = ArrayFromJSON(int32(), "[1, 2, 3]");
auto my_schema =
schema({field("a0", sparse_type), field("a1", dense_type), field("a2", int32())});
auto batch = RecordBatch::Make(my_schema, a0->length(), {a0, a1, a2});

auto write_options = IpcWriteOptions::Defaults();
write_options.metadata_version = metadata_version;
const std::vector<std::vector<int>> subsets = {{0}, {1}, {0, 2}, {1, 2}, {2}};
for (const auto& included_fields : subsets) {
ARROW_SCOPED_TRACE("included_fields = ", ::testing::PrintToString(included_fields));
IpcReadOptions read_options = IpcReadOptions::Defaults();
read_options.included_fields = included_fields;
WriterHelper writer_helper;
RecordBatchVector out_batches;
std::shared_ptr<Schema> out_schema;
ASSERT_OK(RoundTripHelper(writer_helper, {batch}, write_options, read_options,
&out_batches, &out_schema));
FieldVector ex_fields;
ArrayVector ex_columns;
for (int i : included_fields) {
ex_fields.push_back(my_schema->field(i));
ex_columns.push_back(batch->column(i));
}
auto ex_batch = RecordBatch::Make(schema(ex_fields), batch->num_rows(), ex_columns);
ASSERT_EQ(out_batches.size(), 1);
ASSERT_OK(out_batches[0]->ValidateFull());
AssertBatchesEqual(*ex_batch, *out_batches[0]);
}
}

void TestWriteAfterClose() {
// Part of GH-35095.
std::shared_ptr<RecordBatch> batch_ints;
Expand Down Expand Up @@ -2139,6 +2176,31 @@ TEST_F(TestFileFormat, ReadFieldSubset) { TestReadSubsetOfFields(); }
TEST_F(TestFileFormatGenerator, ReadFieldSubset) { TestReadSubsetOfFields(); }
TEST_F(TestFileFormatGeneratorCoalesced, ReadFieldSubset) { TestReadSubsetOfFields(); }

TEST_F(TestStreamFormat, ReadFieldSubsetWithUnionsV4RoundTrip) {
ASSERT_NO_FATAL_FAILURE(TestReadSubsetOfFieldsWithUnions(MetadataVersion::V4));
}
TEST_F(TestStreamFormat, ReadFieldSubsetWithUnionsV5RoundTrip) {
ASSERT_NO_FATAL_FAILURE(TestReadSubsetOfFieldsWithUnions(MetadataVersion::V5));
}
TEST_F(TestFileFormat, ReadFieldSubsetWithUnionsV4RoundTrip) {
ASSERT_NO_FATAL_FAILURE(TestReadSubsetOfFieldsWithUnions(MetadataVersion::V4));
}
TEST_F(TestFileFormat, ReadFieldSubsetWithUnionsV5RoundTrip) {
ASSERT_NO_FATAL_FAILURE(TestReadSubsetOfFieldsWithUnions(MetadataVersion::V5));
}
TEST_F(TestFileFormatGenerator, ReadFieldSubsetWithUnionsV4RoundTrip) {
ASSERT_NO_FATAL_FAILURE(TestReadSubsetOfFieldsWithUnions(MetadataVersion::V4));
}
TEST_F(TestFileFormatGenerator, ReadFieldSubsetWithUnionsV5RoundTrip) {
ASSERT_NO_FATAL_FAILURE(TestReadSubsetOfFieldsWithUnions(MetadataVersion::V5));
}
TEST_F(TestFileFormatGeneratorCoalesced, ReadFieldSubsetWithUnionsV4RoundTrip) {
ASSERT_NO_FATAL_FAILURE(TestReadSubsetOfFieldsWithUnions(MetadataVersion::V4));
}
TEST_F(TestFileFormatGeneratorCoalesced, ReadFieldSubsetWithUnionsV5RoundTrip) {
ASSERT_NO_FATAL_FAILURE(TestReadSubsetOfFieldsWithUnions(MetadataVersion::V5));
}

TEST_F(TestFileFormatGeneratorCoalesced, Errors) {
std::shared_ptr<RecordBatch> batch;
ASSERT_OK(MakeIntRecordBatch(&batch));
Expand Down
Loading