[Analytics-Engine] Seed row counts for tables referenced only inside a subquery

[LantaoJin]

Description

Row counts are not seeded for tables referenced only inside a subquery, so those scans keep Calcite's default estimate. This is not MPP related. see #22877 for details.

Related Issues

Resolves #22877

Check List

• Functionality includes testing.
• API changes companion pull request created, if applicable.
• Public documentation issue/PR created, if applicable.

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

🧪 PR contains tests

🔒 No security concerns identified

✅ No TODO sections

🔀 No multiple PR themes

⚡ No major issues detected

[github-actions]

PR Code Suggestions ✨

Explore these optional code suggestions:

Category	Suggestion	Impact
General	Ensure subquery expressions are reliably visited RelNode.accept(RexShuttle) only visits RexNodes at the current node level and does not recurse into inputs, but here inputs are already recursed via getInputs(). However, a subquery nested inside another subquery's RexSubQuery.rel will only be found because collectTableScans recurses into subQuery.rel — which is correct. But note that RelNode.accept(RexShuttle) behavior depends on the node type; some RelNodes (e.g. TableScan) return this without visiting expressions. Consider explicitly extracting RexNodes via a RelShuttle or iterating RelOptUtil.InputFinder-style over each node's expressions to guarantee subqueries in Project/Filter are visited consistently. sandbox/plugins/analytics-engine/src/main/java/org/opensearch/analytics/exec/IndexRowCountFetcher.java [106-112] for (RelNode input : node.getInputs()) { collectTableScans(input, out); } -// A subquery is not an input — it hangs off a RexNode (a Filter condition or a Project expression) -// as a RexSubQuery, and this fetcher runs BEFORE decorrelation turns it into a join. Walking only -// getInputs() therefore misses every table that appears solely inside a subquery, and each of those -// scans then falls back to Calcite's default row count. That default is small, so the estimates -// derived from it collapse: filters and aggregates above such a scan come out at a row or two, which -// makes a large build look tiny and lets plan choices be made on a number with no basis in the data. -node.accept(new RexShuttle() { +RexShuttle subQueryVisitor = new RexShuttle() { @Override public RexNode visitSubQuery(RexSubQuery subQuery) { collectTableScans(subQuery.rel, out); return super.visitSubQuery(subQuery); } -}); +}; +if (node instanceof org.apache.calcite.rel.core.Filter f) { + subQueryVisitor.apply(f.getCondition()); +} else if (node instanceof org.apache.calcite.rel.core.Project p) { + subQueryVisitor.apply(p.getProjects()); +} else if (node instanceof org.apache.calcite.rel.core.Join j) { + subQueryVisitor.apply(j.getCondition()); +} else { + node.accept(subQueryVisitor); +} Suggestion importance[1-10]: 5 __ Why: The concern about RelNode.accept(RexShuttle) behavior across node types is legitimate — some RelNodes may not visit all their RexNodes uniformly. However, the tests in the PR demonstrate that the current approach works for Filter, Project, and scalar subquery cases, so the impact is moderate and somewhat speculative.	Low

[github-actions]

✅ Gradle check result for dc2eaaa: SUCCESS

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 71.60%. Comparing base (f81134c) to head (dc2eaaa).
⚠️ Report is 20 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff              @@
##               main   #22878      +/-   ##
============================================
- Coverage     71.61%   71.60%   -0.02%     
+ Complexity    77315    77308       -7     
============================================
  Files          6170     6170              
  Lines        359671   359671              
  Branches      52450    52450              
============================================
- Hits         257591   257550      -41     
- Misses        81636    81650      +14     
- Partials      20444    20471      +27     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.