The somewhat recent change to the CFG for Java assert statements introduced an accidental cartesian product. The product is between completions and assert statements, both of which are somewhat limited in number, which is why we haven't discovered this before now. It was identified by the join-order metric, since the CP was confusing the optimiser to select a particularly bad join order.

Before:

              444   ~4%    {2} r34 = JOIN Completion::Completion#3da5aa6a WITH num#Completion::NormalCompletion#17a01869 CARTESIAN PRODUCT OUTPUT Rhs.0, Lhs.0        257653644   ~0%    {3}    | JOIN WITH `ControlFlowGraph::ControlFlowGraphImpl::last/3#f72f2d15_201#join_rhs` ON FIRST 1 OUTPUT Rhs.1, Lhs.1, Rhs.2            17760   ~0%    {3}    | JOIN WITH `Statement::AssertStmt.getMessage/0#dispred#911385b1_10#join_rhs` ON FIRST 1 OUTPUT Rhs.1, Lhs.1, Lhs.2            17760   ~0%    {3}    | JOIN WITH num#ControlFlowGraph::ControlFlow::TAssertThrowNode#fd4ea142 ON FIRST 1 OUTPUT Lhs.2, Lhs.1, Rhs.1

After:

[2025-07-18 13:46:42] Evaluated non-recursive predicate _ControlFlowGraph::ControlFlowGraphImpl::last/3#f72f2d15_201#join_rhs_num#Completion::NormalCompleti__#shared@4f47f1qe in 92ms (size: 580301).Evaluated relational algebra for predicate _ControlFlowGraph::ControlFlowGraphImpl::last/3#f72f2d15_201#join_rhs_num#Completion::NormalCompleti__#shared@4f47f1qe with tuple counts:        580301  ~1%    {3} r1 = JOIN `ControlFlowGraph::ControlFlowGraphImpl::last/3#f72f2d15_201#join_rhs` WITH num#Completion::NormalCompletion#17a01869 ON FIRST 1 OUTPUT Lhs.1, Lhs.0, Lhs.2                       return r1...            40   ~3%    {3} r4 = JOIN `_ControlFlowGraph::ControlFlowGraphImpl::last/3#f72f2d15_201#join_rhs_num#Completion::NormalCompleti__#shared` WITH `Statement::AssertStmt.getMessage/0#dispred#911385b1_10#join_rhs` ON FIRST 1 OUTPUT Rhs.1, Lhs.1, Lhs.2            40   ~0%    {3}    | JOIN WITH num#ControlFlowGraph::ControlFlow::TAssertThrowNode#fd4ea142 ON FIRST 1 OUTPUT Lhs.2, Lhs.1, Rhs.1

Java: Fix accidental CP in CFG for asserts.

bc2e7d4

CopilotAI review requested due to automatic review settings

July 18, 2025 12:00

aschackmull requested a review froma team as acode owner

July 18, 2025 12:00

aschackmull added the no-change-note-requiredThis PR does not need a change note label

Jul 18, 2025

CopilotAI reviewed

Jul 18, 2025

View reviewed changes

Copy link

Contributor

CopilotAI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others.Learn more.

Pull Request Overview

This PR fixes an accidental cartesian product in the control flow graph (CFG) implementation for Java assert statements. The issue was causing the query optimizer to select inefficient join orders, significantly impacting performance.

Replaces direct pattern matching withNormalCompletion() with explicit variable binding and constraint
Eliminates cartesian product between completions and assert statements in the query execution plan
Improves query performance by enabling better join order selection