github-actions[bot] commented on code in PR #63192:
URL: https://github.com/apache/doris/pull/63192#discussion_r3232488095
##########
be/src/format/parquet/vparquet_column_reader.cpp:
##########
@@ -103,6 +118,677 @@ static void fill_array_offset(FieldSchema* field,
ColumnArray::Offsets64& offset
}
}
+static constexpr int64_t UNIX_EPOCH_DAYNR = 719528;
+static constexpr int64_t MICROS_PER_SECOND = 1000000;
+
+static int64_t variant_date_value(const VecDateTimeValue& value) {
+ return value.daynr() - UNIX_EPOCH_DAYNR;
+}
+
+static int64_t variant_date_value(const DateV2Value<DateV2ValueType>& value) {
+ return value.daynr() - UNIX_EPOCH_DAYNR;
+}
+
+static int64_t variant_datetime_value(const VecDateTimeValue& value) {
+ int64_t timestamp = 0;
+ value.unix_timestamp(×tamp, cctz::utc_time_zone());
+ return timestamp * MICROS_PER_SECOND;
+}
+
+static int64_t variant_datetime_value(const DateV2Value<DateTimeV2ValueType>&
value) {
+ int64_t timestamp = 0;
+ value.unix_timestamp(×tamp, cctz::utc_time_zone());
+ return timestamp * MICROS_PER_SECOND + value.microsecond();
+}
+
+static int64_t variant_datetime_value(const TimestampTzValue& value) {
+ int64_t timestamp = 0;
+ value.unix_timestamp(×tamp, cctz::utc_time_zone());
+ return timestamp * MICROS_PER_SECOND + value.microsecond();
+}
+
+static int find_child_idx(const FieldSchema& field, std::string_view name) {
+ for (int i = 0; i < field.children.size(); ++i) {
+ if (field.children[i].lower_case_name == name) {
+ return i;
+ }
+ }
+ return -1;
+}
+
+static bool is_variant_wrapper_field(const FieldSchema& field) {
+ auto type = remove_nullable(field.data_type);
+ if (type->get_primitive_type() != TYPE_STRUCT &&
type->get_primitive_type() != TYPE_VARIANT) {
+ return false;
+ }
+
+ bool has_metadata = false;
+ bool has_value = false;
+ bool has_typed_value = false;
+ for (const auto& child : field.children) {
+ if (child.lower_case_name == "metadata") {
+ if (child.physical_type != tparquet::Type::BYTE_ARRAY) {
+ return false;
+ }
+ has_metadata = true;
+ continue;
+ }
+ if (child.lower_case_name == "value") {
+ if (child.physical_type != tparquet::Type::BYTE_ARRAY) {
+ return false;
+ }
+ has_value = true;
+ continue;
+ }
+ if (child.lower_case_name == "typed_value") {
+ has_typed_value = true;
+ continue;
+ }
+ return false;
+ }
+ return has_typed_value || (has_metadata && has_value);
+}
+
+static const IColumn& remove_nullable_column(const IColumn& column) {
+ if (const auto* nullable = check_and_get_column<ColumnNullable>(&column)) {
+ return nullable->get_nested_column();
+ }
+ return column;
+}
+
+static Status get_binary_field(const Field& field, std::string* value, bool*
present) {
+ if (field.is_null()) {
+ *present = false;
+ return Status::OK();
+ }
+ *present = true;
+ switch (field.get_type()) {
+ case TYPE_STRING:
+ *value = field.get<TYPE_STRING>();
+ return Status::OK();
+ case TYPE_CHAR:
+ *value = field.get<TYPE_CHAR>();
+ return Status::OK();
+ case TYPE_VARCHAR:
+ *value = field.get<TYPE_VARCHAR>();
+ return Status::OK();
+ case TYPE_VARBINARY: {
+ auto ref = field.get<TYPE_VARBINARY>().to_string_ref();
+ value->assign(ref.data, ref.size);
+ return Status::OK();
+ }
+ default:
+ return Status::Corruption("Parquet VARIANT binary field has unexpected
Doris type {}",
+ field.get_type_name());
+ }
+}
+
+static PathInData append_path(const PathInData& prefix, const PathInData&
suffix) {
+ if (prefix.empty()) {
+ return suffix;
+ }
+ if (suffix.empty()) {
+ return prefix;
+ }
+ PathInDataBuilder builder;
+ builder.append(prefix.get_parts(), false);
+ builder.append(suffix.get_parts(), false);
+ return builder.build();
+}
+
+static Status parse_json_to_variant_map(const std::string& json, const
PathInData& prefix,
+ VariantMap* values) {
+ auto parsed_column = ColumnVariant::create(0, false);
+ ParseConfig parse_config;
+ StringRef json_ref(json.data(), json.size());
+ RETURN_IF_CATCH_EXCEPTION(
+ variant_util::parse_json_to_variant(*parsed_column, json_ref,
nullptr, parse_config));
+ Field parsed = (*parsed_column)[0];
+ auto& parsed_values = parsed.get<TYPE_VARIANT>();
+ for (auto& [path, value] : parsed_values) {
+ (*values)[append_path(prefix, path)] = std::move(value);
+ }
+ return Status::OK();
+}
+
+static void append_json_string(std::string_view value, std::string* json) {
+ auto column = ColumnString::create();
+ VectorBufferWriter writer(*column);
+ writer.write_json_string(value);
+ writer.commit();
+ json->append(column->get_data_at(0).data, column->get_data_at(0).size);
+}
+
+static bool is_column_selected(const FieldSchema& field_schema,
+ const std::set<uint64_t>& column_ids) {
+ return column_ids.empty() || column_ids.find(field_schema.get_column_id())
!= column_ids.end();
+}
+
+static bool has_selected_column(const FieldSchema& field_schema,
+ const std::set<uint64_t>& column_ids) {
+ if (is_column_selected(field_schema, column_ids)) {
+ return true;
+ }
+ return std::any_of(field_schema.children.begin(),
field_schema.children.end(),
+ [&column_ids](const FieldSchema& child) {
+ return has_selected_column(child, column_ids);
+ });
+}
+
+static bool is_direct_variant_leaf_type(const DataTypePtr& data_type) {
+ const auto& type = remove_nullable(data_type);
+ switch (type->get_primitive_type()) {
+ case TYPE_BOOLEAN:
+ case TYPE_TINYINT:
+ case TYPE_SMALLINT:
+ case TYPE_INT:
+ case TYPE_BIGINT:
+ case TYPE_LARGEINT:
+ case TYPE_DECIMALV2:
+ case TYPE_DECIMAL32:
+ case TYPE_DECIMAL64:
+ case TYPE_DECIMAL128I:
+ case TYPE_DECIMAL256:
+ case TYPE_STRING:
+ case TYPE_CHAR:
+ case TYPE_VARCHAR:
+ return true;
+ case TYPE_ARRAY: {
+ const auto* array_type = assert_cast<const DataTypeArray*>(type.get());
+ return is_direct_variant_leaf_type(array_type->get_nested_type());
+ }
+ default:
+ return false;
+ }
+}
+
+static bool can_direct_read_typed_value(const FieldSchema& field_schema, bool
allow_variant_wrapper,
+ const std::set<uint64_t>& column_ids) {
+ if (!has_selected_column(field_schema, column_ids)) {
+ return true;
+ }
+ if (allow_variant_wrapper && is_variant_wrapper_field(field_schema)) {
+ const int value_idx = find_child_idx(field_schema, "value");
+ const int typed_value_idx = find_child_idx(field_schema,
"typed_value");
+ return (value_idx < 0 ||
+ !has_selected_column(field_schema.children[value_idx],
column_ids)) &&
+ typed_value_idx >= 0 &&
+
can_direct_read_typed_value(field_schema.children[typed_value_idx], false,
+ column_ids);
+ }
+
+ const auto& type = remove_nullable(field_schema.data_type);
+ if (type->get_primitive_type() == TYPE_STRUCT) {
+ return std::all_of(field_schema.children.begin(),
field_schema.children.end(),
+ [&column_ids](const FieldSchema& child) {
+ return can_direct_read_typed_value(child, true,
column_ids);
+ });
+ }
+ return is_direct_variant_leaf_type(field_schema.data_type);
+}
+
+static void fill_variant_field_info(FieldWithDataType* value) {
+ FieldInfo info;
+ variant_util::get_field_info(value->field, &info);
+ DCHECK_LE(info.num_dimensions, std::numeric_limits<uint8_t>::max());
+ value->base_scalar_type_id = info.scalar_type_id;
+ value->num_dimensions = static_cast<uint8_t>(info.num_dimensions);
+}
+
+static Status field_to_variant_field(const FieldSchema& field_schema, const
Field& field,
+ FieldWithDataType* value, bool* present) {
+ if (field.is_null()) {
+ *present = false;
+ return Status::OK();
+ }
+ *present = true;
+ const DataTypePtr& type = remove_nullable(field_schema.data_type);
+ switch (type->get_primitive_type()) {
+ case TYPE_BOOLEAN:
+ case TYPE_TINYINT:
+ case TYPE_SMALLINT:
+ case TYPE_INT:
+ case TYPE_BIGINT:
+ case TYPE_LARGEINT:
+ case TYPE_DECIMALV2:
+ case TYPE_DECIMAL32:
+ case TYPE_DECIMAL64:
+ case TYPE_DECIMAL128I:
+ case TYPE_DECIMAL256:
+ case TYPE_STRING:
+ case TYPE_CHAR:
+ case TYPE_VARCHAR:
+ case TYPE_ARRAY:
+ value->field = field;
+ fill_variant_field_info(value);
+ value->precision = type->get_precision();
+ value->scale = type->get_scale();
+ return Status::OK();
+ case TYPE_FLOAT: {
+ const auto float_value = field.get<TYPE_FLOAT>();
+ value->field = std::isfinite(float_value) ? field : Field();
+ fill_variant_field_info(value);
+ return Status::OK();
+ }
+ case TYPE_DOUBLE: {
+ const auto double_value = field.get<TYPE_DOUBLE>();
+ value->field = std::isfinite(double_value) ? field : Field();
+ fill_variant_field_info(value);
+ return Status::OK();
+ }
+ case TYPE_TIMEV2:
+ value->field = Field::create_field<TYPE_BIGINT>(
+ static_cast<int64_t>(std::llround(field.get<TYPE_TIMEV2>())));
+ value->base_scalar_type_id = TYPE_BIGINT;
+ return Status::OK();
+ case TYPE_DATE:
+ value->field =
Field::create_field<TYPE_BIGINT>(variant_date_value(field.get<TYPE_DATE>()));
+ value->base_scalar_type_id = TYPE_BIGINT;
+ return Status::OK();
+ case TYPE_DATETIME:
+ value->field = Field::create_field<TYPE_BIGINT>(
+ variant_datetime_value(field.get<TYPE_DATETIME>()));
+ value->base_scalar_type_id = TYPE_BIGINT;
+ return Status::OK();
+ case TYPE_DATEV2:
+ value->field =
+
Field::create_field<TYPE_BIGINT>(variant_date_value(field.get<TYPE_DATEV2>()));
+ value->base_scalar_type_id = TYPE_BIGINT;
+ return Status::OK();
+ case TYPE_DATETIMEV2:
+ value->field = Field::create_field<TYPE_BIGINT>(
+ variant_datetime_value(field.get<TYPE_DATETIMEV2>()));
+ value->base_scalar_type_id = TYPE_BIGINT;
+ return Status::OK();
+ case TYPE_TIMESTAMPTZ:
+ value->field = Field::create_field<TYPE_BIGINT>(
+ variant_datetime_value(field.get<TYPE_TIMESTAMPTZ>()));
+ value->base_scalar_type_id = TYPE_BIGINT;
+ return Status::OK();
+ case TYPE_VARBINARY:
+ return Status::NotSupported("Parquet VARIANT binary typed_value is not
supported");
+ default:
+ return Status::Corruption("Unsupported Parquet VARIANT typed_value
Doris type {}",
+ type->get_name());
+ }
+}
+
+static Status typed_value_to_json(const FieldSchema& typed_value_field, const
Field& field,
+ const std::string& metadata, std::string*
json, bool* present);
+
+static Status serialize_field_to_json(const DataTypePtr& data_type, const
Field& field,
+ std::string* json) {
+ MutableColumnPtr column = data_type->create_column();
+ column->insert(field);
+
+ auto json_column = ColumnString::create();
+ VectorBufferWriter writer(*json_column);
+ auto serde = data_type->get_serde();
+ DataTypeSerDe::FormatOptions options;
+ RETURN_IF_ERROR(serde->serialize_one_cell_to_json(*column, 0, writer,
options));
+ writer.commit();
+ *json = json_column->get_data_at(0).to_string();
+ return Status::OK();
+}
+
+static Status scalar_typed_value_to_json(const FieldSchema& field_schema,
const Field& field,
+ std::string* json, bool* present) {
+ FieldWithDataType value;
+ RETURN_IF_ERROR(field_to_variant_field(field_schema, field, &value,
present));
+ if (!*present) {
+ return Status::OK();
+ }
+ if (value.field.is_null()) {
+ *json = "null";
+ return Status::OK();
+ }
+
+ DataTypePtr json_type;
+ if (value.base_scalar_type_id != PrimitiveType::INVALID_TYPE) {
+ json_type =
DataTypeFactory::instance().create_data_type(value.base_scalar_type_id, false,
+
value.precision, value.scale);
+ } else {
+ json_type = remove_nullable(field_schema.data_type);
+ }
+ return serialize_field_to_json(json_type, value.field, json);
+}
+
+static Status variant_to_json(const FieldSchema& variant_field, const Field&
field,
+ const std::string* inherited_metadata,
std::string* json,
+ bool* present) {
+ if (field.is_null()) {
+ *present = false;
+ return Status::OK();
+ }
+
+ const auto& fields = field.get<TYPE_STRUCT>();
+ const int metadata_idx = find_child_idx(variant_field, "metadata");
+ const int value_idx = find_child_idx(variant_field, "value");
+ const int typed_value_idx = find_child_idx(variant_field, "typed_value");
+
+ std::string metadata;
+ bool has_metadata = false;
+ if (inherited_metadata != nullptr) {
+ metadata = *inherited_metadata;
+ has_metadata = true;
+ }
+ if (metadata_idx >= 0) {
+ bool metadata_present = false;
+ RETURN_IF_ERROR(get_binary_field(fields[metadata_idx], &metadata,
&metadata_present));
+ has_metadata = metadata_present;
+ }
+
+ if (typed_value_idx >= 0) {
+
RETURN_IF_ERROR(typed_value_to_json(variant_field.children[typed_value_idx],
Review Comment:
This early return drops residual `value` data for nested shredded VARIANT
wrappers whenever `typed_value` is present. `variant_to_json()` is used by
`typed_array_to_json()` and `typed_struct_to_json()` for nested/array wrapper
fields, so a valid layout like `element { value, typed_value { metric int } }`
for an object `{"metric":1,"extra":"x"}` will emit only the typed branch
(`{"metric":1}`) and lose `extra`, which is stored in the wrapper's `value`.
The top-level `variant_to_variant_map()` path merges `value` and `typed_value`,
but this JSON reconstruction path does not. Please merge both branches here (or
route nested/array wrappers through the map merge path) and add coverage for a
nested or array shredded value where `value` and `typed_value` are both present.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
