Skip to content

Add row_groups_per_file to to_parquet/write_parquet #620

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
joaquinhuigomez wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Add row_groups_per_file to to_parquet/write_parquet #620

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 2 additions & 0 deletions _duckdb-stubs/__init__.pyi
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -734,6 +734,7 @@ class DuckDBPyRelation:
field_ids: ParquetFieldsOptions | None = None,
row_group_size_bytes: int | str | None = None,
row_group_size: int | None = None,
row_groups_per_file: int | None = None,
overwrite: bool | None = None,
per_thread_output: bool | None = None,
use_tmp_file: bool | None = None,
Expand Down Expand Up @@ -790,6 +791,7 @@ class DuckDBPyRelation:
field_ids: ParquetFieldsOptions | None = None,
row_group_size_bytes: str | int | None = None,
row_group_size: int | None = None,
row_groups_per_file: int | None = None,
overwrite: bool | None = None,
per_thread_output: bool | None = None,
use_tmp_file: bool | None = None,
Expand Down
10 changes: 5 additions & 5 deletions src/include/duckdb_python/pyrelation.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -213,11 +213,11 @@ struct DuckDBPyRelation {

void ToParquet(const string &filename, const nb::object &compression = nb::none(),
const nb::object &field_ids = nb::none(), const nb::object &row_group_size_bytes = nb::none(),
const nb::object &row_group_size = nb::none(), const nb::object &overwrite = nb::none(),
const nb::object &per_thread_output = nb::none(), const nb::object &use_tmp_file = nb::none(),
const nb::object &partition_by = nb::none(), const nb::object &write_partition_columns = nb::none(),
const nb::object &append = nb::none(), const nb::object &filename_pattern = nb::none(),
const nb::object &file_size_bytes = nb::none());
const nb::object &row_group_size = nb::none(), const nb::object &row_groups_per_file = nb::none(),
const nb::object &overwrite = nb::none(), const nb::object &per_thread_output = nb::none(),
const nb::object &use_tmp_file = nb::none(), const nb::object &partition_by = nb::none(),
const nb::object &write_partition_columns = nb::none(), const nb::object &append = nb::none(),
const nb::object &filename_pattern = nb::none(), const nb::object &file_size_bytes = nb::none());

void ToCSV(const string &filename, const nb::object &sep = nb::none(), const nb::object &na_rep = nb::none(),
const nb::object &header = nb::none(), const nb::object &quotechar = nb::none(),
Expand Down
17 changes: 13 additions & 4 deletions src/pyrelation.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1262,10 +1262,11 @@ static Value NestedDictToStruct(const nb::object &dictionary) {

void DuckDBPyRelation::ToParquet(const string &filename, const nb::object &compression, const nb::object &field_ids,
const nb::object &row_group_size_bytes, const nb::object &row_group_size,
const nb::object &overwrite, const nb::object &per_thread_output,
const nb::object &use_tmp_file, const nb::object &partition_by,
const nb::object &write_partition_columns, const nb::object &append,
const nb::object &filename_pattern, const nb::object &file_size_bytes) {
const nb::object &row_groups_per_file, const nb::object &overwrite,
const nb::object &per_thread_output, const nb::object &use_tmp_file,
const nb::object &partition_by, const nb::object &write_partition_columns,
const nb::object &append, const nb::object &filename_pattern,
const nb::object &file_size_bytes) {
identifier_map_t<vector<Value>> options;

if (!nb::none().is(compression)) {
Expand Down Expand Up @@ -1306,6 +1307,14 @@ void DuckDBPyRelation::ToParquet(const string &filename, const nb::object &compr
options["row_group_size"] = {Value(row_group_size_int)};
}

if (!nb::none().is(row_groups_per_file)) {
if (!nb::isinstance<nb::int_>(row_groups_per_file)) {
throw InvalidInputException("to_parquet only accepts 'row_groups_per_file' as an integer");
}
int64_t row_groups_per_file_int = (int64_t)nb::int_(row_groups_per_file);
options["row_groups_per_file"] = {Value(row_groups_per_file_int)};
}

if (!nb::none().is(partition_by)) {
if (!nb::isinstance<nb::list>(partition_by)) {
throw InvalidInputException("to_parquet only accepts 'partition_by' as a list of strings");
Expand Down
9 changes: 5 additions & 4 deletions src/pyrelation/initialize.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,10 +39,11 @@ static void InitializeConsumers(nb::class_<DuckDBPyRelation> &m) {
"Write the relation object to a Parquet file in 'file_name'", nb::arg("file_name"), nb::kw_only(),
nb::arg("compression") = nb::none(), nb::arg("field_ids") = nb::none(),
nb::arg("row_group_size_bytes") = nb::none(), nb::arg("row_group_size") = nb::none(),
nb::arg("overwrite") = nb::none(), nb::arg("per_thread_output") = nb::none(),
nb::arg("use_tmp_file") = nb::none(), nb::arg("partition_by") = nb::none(),
nb::arg("write_partition_columns") = nb::none(), nb::arg("append") = nb::none(),
nb::arg("filename_pattern") = nb::none(), nb::arg("file_size_bytes") = nb::none());
nb::arg("row_groups_per_file") = nb::none(), nb::arg("overwrite") = nb::none(),
nb::arg("per_thread_output") = nb::none(), nb::arg("use_tmp_file") = nb::none(),
nb::arg("partition_by") = nb::none(), nb::arg("write_partition_columns") = nb::none(),
nb::arg("append") = nb::none(), nb::arg("filename_pattern") = nb::none(),
nb::arg("file_size_bytes") = nb::none());

DefineMethod(
{"to_csv", "write_csv"}, m, &DuckDBPyRelation::ToCSV, "Write the relation object to a CSV file in 'file_name'",
Expand Down
25 changes: 25 additions & 0 deletions tests/fast/api/test_to_parquet.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,6 +68,31 @@ def test_row_group_size(self):
parquet_rel = duckdb.read_parquet(temp_file_name)
assert rel.execute().fetchall() == parquet_rel.execute().fetchall()

def test_row_groups_per_file(self):
temp_file_name = os.path.join(tempfile.mkdtemp(), next(tempfile._get_candidate_names())) # noqa: PTH118

# use same test data as external/duckdb/test/sql/copy/row_groups_per_file.test, which also
# pins threads to 1: file rotation is best-effort and only exact on a single thread
con = duckdb.connect(config={"threads": 1})
rel = con.from_query("SELECT i AS col_a, i AS col_b FROM range(0,10000) tbl(i);")
rel.to_parquet(temp_file_name, row_group_size=2000, row_groups_per_file=1)

# 5 row groups of 2000 rows, one row group per file
files = list(pathlib.Path(temp_file_name).iterdir())
assert len(files) == 5, f"Expected 5 files, got {len(files)}"

# Verify data integrity
result = con.read_parquet(f"{temp_file_name}/*.parquet")
assert len(result.execute().fetchall()) == 10000

def test_row_groups_per_file_invalid_type(self):
temp_file_name = os.path.join(tempfile.mkdtemp(), next(tempfile._get_candidate_names())) # noqa: PTH118
rel = duckdb.sql("SELECT 1 AS i")
with pytest.raises(
duckdb.InvalidInputException, match="to_parquet only accepts 'row_groups_per_file' as an integer"
):
rel.to_parquet(temp_file_name, row_groups_per_file="x")

@pytest.mark.parametrize("write_columns", [None, True, False])
def test_partition(self, write_columns):
temp_file_name = os.path.join(tempfile.mkdtemp(), next(tempfile._get_candidate_names())) # noqa: PTH118
Expand Down