Copilot commented on code in PR #63104:
URL: https://github.com/apache/doris/pull/63104#discussion_r3212908761
##########
be/test/exec/common/schema_util_test.cpp:
##########
@@ -1224,20 +1225,16 @@ TEST_F(SchemaUtilTest,
get_extended_compaction_schema_nested_group_preserves_typ
rowsets, target_schema);
ASSERT_TRUE(status.ok()) << status.to_string();
- EXPECT_EQ(target_schema->num_columns(), 2);
+ // get_extended_compaction_schema rebuilds from the base columns. Real
compaction targets do
+ // not carry pre-existing extracted columns, so NG compaction must rely on
rowset metadata for
+ // regular paths and on get_compaction_typed_columns() for typed paths.
+ EXPECT_EQ(target_schema->num_columns(), 1);
const PathInData typed_path("v1.owner", true);
- const auto typed_column_index = target_schema->field_index(typed_path);
- ASSERT_NE(typed_column_index, -1);
-
- const auto& preserved_typed_column =
target_schema->column(typed_column_index);
- EXPECT_TRUE(preserved_typed_column.path_info_ptr()->get_is_typed());
- EXPECT_EQ(preserved_typed_column.type(),
FieldType::OLAP_FIELD_TYPE_STRING);
+ EXPECT_EQ(target_schema->field_index(typed_path), -1);
const auto* path_set_info = target_schema->try_path_set_info(1);
ASSERT_NE(path_set_info, nullptr);
- ASSERT_TRUE(path_set_info->typed_path_set.contains("owner"));
- EXPECT_EQ(path_set_info->typed_path_set.at("owner").indexes.size(), 1);
-
EXPECT_EQ(path_set_info->typed_path_set.at("owner").indexes[0]->index_name(),
"v1_owner_idx");
+ EXPECT_FALSE(path_set_info->typed_path_set.contains("owner"));
Review Comment:
The updated NestedGroup compaction logic now materializes regular (non-NG)
paths via `collect_regular_types_outside_nested_group()` +
`get_compaction_subcolumns_from_data_types()`, but there is no unit test
covering the positive case (i.e., rowset metadata contains both NG paths and a
sibling regular path like `owner`, and the resulting compaction schema
materializes `v1.owner` and records it in `path_set_info.sub_path_set`). The
current test only asserts that pre-existing extracted columns in the target
schema are ignored. Adding a regression test for the new contract would help
prevent future breakage.
##########
be/src/exec/common/variant_util.cpp:
##########
@@ -1374,27 +1293,37 @@ Status
VariantCompactionUtil::get_extended_compaction_schema(
?
empty_extended_info
: info_it->second;
auto& paths_set_info = uid_to_paths_set_info[column->unique_id()];
- if (ng_root_uids.contains(column->unique_id())) {
- const auto plan =
build_nested_group_compaction_materialization_plan(
- uid_to_existing_ng_subcolumns.contains(column->unique_id())
- ?
uid_to_existing_ng_subcolumns.at(column->unique_id())
- : std::vector<TabletColumnPtr> {},
- extended_info);
- append_nested_group_compaction_columns(target, column, plan,
output_schema,
- paths_set_info);
- LOG(INFO) << "Variant column uid=" << column->unique_id()
- << " keeps nested-group root with regular extracted
columns in compaction "
- "schema";
- continue;
- }
- if (!should_check_variant_path_stats(*column)) {
- VLOG_DEBUG << "skip extended schema compaction for variant uid="
<< column->unique_id()
- << " because the column disables variant path stats";
- continue;
+ const bool use_nested_group_compaction_schema =
ng_root_uids.contains(column->unique_id());
+ if (!use_nested_group_compaction_schema) {
+ if (!should_check_variant_path_stats(*column)) {
+ VLOG_DEBUG << "skip extended schema compaction for variant
uid="
+ << column->unique_id()
+ << " because the column disables variant path
stats";
+ continue;
+ }
+ if (extended_info.has_nested_group) {
+ LOG(INFO) << "Variant column uid=" << column->unique_id()
+ << " has nested group, keep original column in
compaction schema";
+ continue;
+ }
Review Comment:
The `extended_info.has_nested_group` check inside the
`!use_nested_group_compaction_schema` branch appears to be unreachable:
`use_nested_group_compaction_schema` is computed from `ng_root_uids`, and
`ng_root_uids` is built using the same `has_nested_group` predicate. Keeping
this branch makes control flow harder to follow and suggests a state that
cannot occur; consider removing it (or adjust `ng_root_uids` computation if
there really are cases where `has_nested_group` should not imply NG schema).
--
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]
