fix(tesseract): CASE handling fixes in multi-stage measures

[waralexrom]

Check List

• Tests have been run in packages where changes have been made if available
• Linter has been run for changed code
• Tests for the changes have been added if not covered yet
• Docs have been added / updated if required

[claude]

Claude finished @waralexrom's task in 2m 43s —— View job

---

Code Review: fix(tesseract): CASE handling fixes in multi-stage measures

• Gather context and read changed files
• Review main logic changes in multi_stage_query_planner.rs
• Review test fixtures and test cases
• Review snapshot files
• Post comprehensive review

---

Summary

This PR fixes a correctness bug in how multi-stage CASE/SWITCH measures generate leaf CTEs. The core issue: when the same dependency (e.g., {CUBE.total_amount}) appeared across multiple WHEN branches, the old code would only apply the first branch's filter to the CTE, discarding subsequent branches. This produced incorrect results — only one status's data was aggregated instead of all relevant statuses.

What Changed

Root Cause Fix (multi_stage_query_planner.rs):
The old code iterated per-branch, using a processed_deps HashSet to skip already-seen dependencies. If total_amount appeared in both the "completed" and "pending" branches, only the "completed" filter was applied.

The new approach is a two-pass design:

1. Collection pass: Walks all WHEN and ELSE branches, collecting dependencies into a HashMap that merges the union of switch values each dep needs.
2. Emission pass: Creates one CTE per unique dependency with a combined IN filter (e.g., status IN ('completed', 'pending')).

This correctly handles:

• Same dep in multiple WHEN branches → merged filter values
• ELSE branch with known values → uses complement values
• ELSE branch with unknown universe → unrestricted (no prefilter), safe fallback

Schema Fix (integration_multi_stage.yaml): Changed status_weighted_amount from type: number to type: sum. This is a necessary correction — number type doesn't aggregate in multi-stage context.

Tests: Removed #[ignore] from 3 previously-known-broken tests, added a new test_case_switch_dimension_only test, and added stable ordering to test_case_switch_with_dimension. All 4 snapshot files contain plausible expected results.

Code Quality

The refactored code is cleaner than the original (the //TODO refactor needed comment was rightfully removed). The record closure pattern is a clean way to merge values, and the match on (&mut entry.1, branch_values) handles the three cases (already unrestricted, becoming unrestricted, accumulating values) concisely.

Potential Issues

1. order + HashMap dual-tracking: The order: Vec<String> and deps: HashMap<...> pattern works correctly but introduces a subtle invariant — they must stay in sync. An IndexMap would be a cleaner single data structure, though this is not a correctness concern. (See inline comment.)

2. Linear contains() for dedup: The values.contains(&v) check on line 273 is O(n) per value. This is fine for typical CASE expressions with a handful of branches, but is worth noting.

3. No negative test coverage: There's no test for CASE with CaseSwitchItem::Sql(...) (non-member switch), which takes the early return Ok(false) path. Not a regression from this PR, but worth adding eventually.

Verdict

Looks good. The fix is correct, well-structured, and properly tested. The two-pass approach cleanly solves the dependency merging problem that the old per-branch iteration couldn't handle. The snapshot results (e.g., 375/600/750 for weighted amounts across all statuses) match expected behavior.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 78.78%. Comparing base (187ee17) to head (75443a2).
⚠️ Report is 6 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master   #10764   +/-   ##
=======================================
  Coverage   78.78%   78.78%           
=======================================
  Files         466      466           
  Lines       92061    92061           
  Branches     3383     3383           
=======================================
+ Hits        72527    72532    +5     
+ Misses      19042    19037    -5     
  Partials      492      492           

Flag	Coverage Δ
cube-backend	58.01% <ø> (ø)
cubesql	83.41% <ø> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ovr]

Nit: about usage of Result instead of panic, otherwise LGMT 👍