26.1 Antalya port - fixes for s3Cluster distributed calls#1395
26.1 Antalya port - fixes for s3Cluster distributed calls#1395ianton-ru wants to merge 8 commits intoantalya-26.1from
Conversation
…_global_join_fixes 25.8 Antalya: Fix joins with Iceberg tables
…for_cluster_requests Revert support for 'IN' in cluster requests
…_cluster_request Fix IN with Iceberg table
ianton-ru
added
antalya
port-antalya
|
@codex review |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 7df85aef41
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| auto join_node = query_node.getJoinTree(); | ||
| query_tree_distributed = join_node->as<JoinNode>()->getLeftTableExpression()->clone(); |
There was a problem hiding this comment.
When object_storage_cluster_join_mode='local', this branch assumes info.has_join means the root QueryNode join tree is a JoinNode, but getQueryTreeInfo() currently marks has_join if it finds any JOIN anywhere in query_tree (including scalar/IN subqueries). In queries where the top-level FROM is just s3Cluster(...) and only a nested subquery contains a JOIN, query_node.getJoinTree() is not a JoinNode, so as<JoinNode>() throws and the query fails. Please restrict JOIN detection/rewrite to the top-level join tree (or validate node type before casting).
Useful? React with 👍 / 👎.
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3829e1a458
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| auto modified_query_tree = query_tree->clone(); | ||
|
|
||
| SearcherVisitor left_table_expression_searcher({QueryTreeNodeType::TABLE, QueryTreeNodeType::TABLE_FUNCTION}, context); | ||
| left_table_expression_searcher.visit(modified_query_tree); |
There was a problem hiding this comment.
The source-table search is run on the entire modified_query_tree and stops at the first TABLE/TABLE_FUNCTION node, so queries with scalar/CTE subqueries can match an unrelated table before the main FROM source. When object_storage_cluster_join_mode='local', the later projection and WHERE rewrite then operates on the wrong node, which can generate invalid remote subqueries or strip the wrong predicates. Limit this search to the main join tree (for example query_node.getJoinTree()) instead of the full query tree.
Useful? React with 👍 / 👎.
| query_node.resolveProjectionColumns(columns); | ||
| auto column_nodes_to_select = std::make_shared<ListNode>(); | ||
| column_nodes_to_select->getNodes().reserve(columns.size()); |
There was a problem hiding this comment.
The rewrite always replaces the SELECT list with collected source columns, but that set can be empty for valid JOIN shapes that do not reference source columns directly (e.g. SELECT count() with CROSS JOIN, or JOIN ... ON 1). In that case the serialized remote query becomes an empty projection (SELECT FROM ...) and fails to parse on remote shards. Add a fallback projection (such as a constant) or skip projection replacement when columns is empty.
Useful? React with 👍 / 👎.
3 participants
Changelog category (leave one):
Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):
Frontports for Antalya 26.1
CI/CD Options
Exclude tests:
Regression jobs to run: