This is an automated email from the ASF dual-hosted git repository.
Gabriel39 pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/doris.git
The following commit(s) were added to refs/heads/master by this push:
new ebb13c31799 [fix](be) Avoid abort on absent schema tree columns
(#67164)
ebb13c31799 is described below
commit ebb13c31799de7c764681ced6288d690c5962790
Author: Socrates <[email protected]>
AuthorDate: Thu Aug 27 15:23:36 2026 +0800
[fix](be) Avoid abort on absent schema tree columns (#67164)
### What problem does this PR solve?
Issue Number: None
Related PR: #66835
Problem Summary:
Table-format predicate probes can reference a column that is not
registered in the schema tree. `StructNode::children_column_exists()`
assumed every probe had an entry, so an absent key triggered a DCHECK in
debug builds or `std::out_of_range` in release builds and terminated the
BE process.
This change treats an unregistered child as unavailable for file-level
processing while preserving `has_children_column()` for callers that
need to distinguish an FE/BE schema-contract mismatch from a
file-missing column. The unit test covers present, known-missing, and
unregistered children.
### Release note
Prevent a BE abort when table-format processing probes a column absent
from the schema tree.
### Check List (For Author)
- Test:
- No local BE build or unit test was run per the requested validation
boundary.
- `build-support/clang-format.sh` passed with clang-format 16.
- `build-support/check-format.sh` passed with clang-format 16.
- `build-support/check-build-hygiene.sh` passed.
- `git diff --check` passed.
- Behavior changed: Yes. An unregistered schema-tree child now reports
unavailable instead of terminating the BE process.
- Does this need documentation: No
---
be/src/format/table/table_schema_change_helper.h | 13 ++++++-------
be/test/format/table/table_schema_change_helper_test.cpp | 11 +++++++++++
2 files changed, 17 insertions(+), 7 deletions(-)
diff --git a/be/src/format/table/table_schema_change_helper.h
b/be/src/format/table/table_schema_change_helper.h
index 24cd989c7f3..8d800044cbd 100644
--- a/be/src/format/table/table_schema_change_helper.h
+++ b/be/src/format/table/table_schema_change_helper.h
@@ -77,11 +77,10 @@ public:
return nullptr;
}
- // Presence-only check (does NOT DCHECK). Distinct from
children_column_exists, which asserts
- // the key exists and then reports the file-side exists flag. Callers
use this to reject a
- // projected column that is absent from the table-side schema tree (an
FE/BE schema-contract
- // mismatch) BEFORE calling children_column_exists, turning a would-be
process abort into a
- // graceful per-query error.
+ // Presence-only check. Distinct from children_column_exists, which
reports the file-side
+ // exists flag and also returns false for an unregistered key. Callers
use this to distinguish
+ // a projected column absent from the table-side schema tree (an FE/BE
schema-contract mismatch)
+ // from a registered column that is absent from the data file.
virtual bool has_children_column(std::string table_column_name) const {
throw std::logic_error(
"has_children_column should not be called on base
TableInfoNode");
@@ -188,8 +187,8 @@ public:
}
bool children_column_exists(std::string table_column_name) const
override {
- DCHECK(children.contains(table_column_name));
- return children.at(table_column_name).exists;
+ auto child = children.find(table_column_name);
+ return child != children.end() && child->second.exists;
}
const schema::external::TField* get_missing_column_field(
diff --git a/be/test/format/table/table_schema_change_helper_test.cpp
b/be/test/format/table/table_schema_change_helper_test.cpp
index 9d42a49c1eb..68cfcf43358 100644
--- a/be/test/format/table/table_schema_change_helper_test.cpp
+++ b/be/test/format/table/table_schema_change_helper_test.cpp
@@ -117,6 +117,17 @@ TEST(PartitionColumnFillerTest,
FillNullableIntPartitionValue) {
}
}
+TEST(MockTableSchemaChangeHelper, UnknownStructChildDoesNotExist) {
+ TableSchemaChangeHelper::StructNode root;
+ root.add_children("file_column", "file_column",
+ std::make_shared<TableSchemaChangeHelper::ScalarNode>());
+ root.add_not_exist_children("missing_file_column");
+
+ EXPECT_TRUE(root.children_column_exists("file_column"));
+ EXPECT_FALSE(root.children_column_exists("missing_file_column"));
+ EXPECT_FALSE(root.children_column_exists("partition_column"));
+}
+
TEST(MockTableSchemaChangeHelper, OrcNameNoSchemaChange) {
std::vector<DataTypePtr> data_types;
std::vector<std::string> column_names;
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
