gavinchou commented on code in PR #66849:
URL: https://github.com/apache/doris/pull/66849#discussion_r3842610286
##########
be/src/storage/schema.cpp:
##########
@@ -99,6 +118,72 @@ void ReadSchema::_init_before_column_ordinals() {
}
}
+void ReadSchema::init_row_binlog_column_mappings(const TabletSchema&
tablet_schema) {
+ DORIS_CHECK_GE(_op_ordinal, 0);
+ DORIS_CHECK_EQ(_before_column_ordinals.size(), _num_block_columns);
+
+ _row_binlog_value_column_pairs.clear();
+ _row_binlog_value_pairs_complete = false;
+
+ std::vector<ColumnId> value_column_ids;
+ value_column_ids.reserve(tablet_schema.num_columns() -
tablet_schema.num_key_columns());
+ for (ColumnId cid = 0; cid < tablet_schema.num_columns(); ++cid) {
+ if (static_cast<int32_t>(cid) == tablet_schema.binlog_tso_col_idx() ||
+ static_cast<int32_t>(cid) == tablet_schema.binlog_lsn_col_idx() ||
+ static_cast<int32_t>(cid) == tablet_schema.binlog_op_col_idx() ||
+ tablet_schema.column(cid).is_key()) {
+ continue;
+ }
+ value_column_ids.push_back(cid);
+ }
+
+ if (value_column_ids.empty() || value_column_ids.size() % 2 != 0) {
+ return;
+ }
+
+ const size_t value_column_count = value_column_ids.size() / 2;
+ for (size_t i = 0; i < value_column_count; ++i) {
+ const auto& after = tablet_schema.column(value_column_ids[i]);
+ const auto& before = tablet_schema.column(value_column_ids[i +
value_column_count]);
+ if (before.name() != binlog::build_before_column_name(after.name()) ||
+ !row_binlog_value_columns_have_same_type(after, before)) {
+ return;
+ }
+ }
+
+ // A valid physical row-binlog layout starts from an identity mapping.
Only AFTER value
+ // columns map to their BEFORE companions; keys, metadata and BEFORE
columns map to themselves.
+ for (ColumnId ordinal = 0; ordinal < _num_block_columns; ++ordinal) {
+ _before_column_ordinals[ordinal] = ordinal;
+ }
+
+ bool complete = true;
+ _row_binlog_value_column_pairs.reserve(value_column_count);
+ for (size_t i = 0; i < value_column_count; ++i) {
+ const auto after_cid = value_column_ids[i];
+ const auto before_cid = value_column_ids[i + value_column_count];
+ const int32_t after_ordinal =
ordinal_by_uid(tablet_schema.column(after_cid).unique_id());
+ const int32_t before_ordinal =
ordinal_by_uid(tablet_schema.column(before_cid).unique_id());
+ if (after_ordinal < 0 || before_ordinal < 0 ||
+ static_cast<size_t>(after_ordinal) >= _num_block_columns ||
+ static_cast<size_t>(before_ordinal) >= _num_block_columns) {
+ complete = false;
+ continue;
+ }
+
+ const auto after = cast_set<ColumnId>(after_ordinal);
+ const auto before = cast_set<ColumnId>(before_ordinal);
+ _before_column_ordinals[after] = before;
+ if (!_read_types[after]->equals(*_read_types[before])) {
+ complete = false;
+ continue;
+ }
+ _row_binlog_value_column_pairs.emplace_back(after, before);
Review Comment:
from csun5285: 这俩做的是一件事,直接在ReadSchema::_init_before_column_ordinals
初始化_row_binlog_value_column_pairs 就行
<img width="1496" height="896" alt="Image"
src="https://github.com/user-attachments/assets/01435339-2b32-4344-a9a5-4c1ebaf6a954"
/>
--
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]