Meta: #1038 CASE-return residual decomposition program

[lmeyerov]

Summary

Program meta for decomposing and executing #1038 (Cypher CASE return-expression support gap) into mergeable slices.

Parent/Related

• Parent execution DAG: #1259
• Benchmark queue meta: #992
• Source issue: #1038

Goal

Convert the current broad residual into small, test-first implementation lanes that can be run in parallel where safe.

Child slices

• S1: Repro + fixture lock -> #1265
  • pin current IC4-shaped query and expected behavior
  • lock in test_lowering.py
• S2: Parser/AST surface verification -> #1266
  • confirm CASE forms parsed and represented consistently
  • add parser-level regression locks where needed
• S3: Binder/type/scope handling for CASE branches -> #1267
  • validate branch typing/nullability and alias scope
  • add strict failures for unsupported branch constructs
• S4: Lowering/runtime CASE evaluation route -> #1268
  • implement admitted CASE-return shape end-to-end
  • preserve existing behavior outside scoped shape
• S5: Conformance + benchmark receipt -> #1269
  • pandas + cuDF execution checks
  • post receipt back to #1038 and #992

Land order

1. S1 first (must-have gate)
2. S2/S3 parallel if independent
3. S4 after S2+S3
4. S5 closeout

Merged delivery

• PR #1271 merged on May 4, 2026 (squash): test(cypher): lock #1038 IC4 return-side CASE behavior #1271

Close criteria

• #1038 has concrete child issues for each slice
• at least first implementation slice merged or active PR linked
• benchmark/conformance receipt posted

[lmeyerov]

Program complete: all child slices (#1265-#1269) are now merged via PR #1271 and closed with receipts; checklist synced to done.