Cypher return CASE expression falls outside supported GFQL subset on official IC4 new-topics query

[lmeyerov]

Current benchmark evidence on origin/master 2f79ba1ede0f5285cea986885812d62302016658 from dgx-spark shows interactive-complex-4 / new-topics moved past the older #981 row-lowering bucket after #1023, but still does not execute.

Repro lane:

• artifact: results/runs/dgx-spark-snb-interactive-ic4-origin-master-2f79ba1ed-fresh-r1/
• command: uv run python scripts/run_dgx_spark_suite.py --suite snb-interactive --config configs/suites/snb-interactive-ic4-conformance-sf1.yaml --output-dir results/runs/dgx-spark-snb-interactive-ic4-origin-master-2f79ba1ed-fresh-r1 --graphistry-repo-path /tmp/pygraphistry-origin-master-2f79ba1ed --host-uv

Current direct Cypher failure:

• status: primitive_gap
• error: [unsupported-cypher-query] Cypher expression is outside the currently supported local GFQL subset | field: return.item | value: CASE WHEN 1275350400000 <= post.creationDate ...

Interpretation:

• this looks like the official IC4 return-side CASE ... END expression is now the remaining front-end blocker for this query shape
• older current-master pin 14dafaf6b9 failed earlier under #981, so this is a real post-#1023 frontier change, not a stale repro

Expected:

• direct Cypher should either support the official CASE return expression on this query or fail in a more precise existing tracked lane if one already exists.

[lmeyerov]

Tracking update via meta #1261:

• Child slices created: #1265 (S1), #1266 (S2), #1267 (S3), #1268 (S4), #1269 (S5)
• S1 branch progress: added regression lock test test_issue_1038_ic4_return_side_case_expression_regression_lock in graphistry/tests/compute/gfql/cypher/test_lowering.py
• Local targeted check passes on current origin/master baseline:
  • PYTHONPATH=. python -m pytest graphistry/tests/compute/gfql/cypher/test_lowering.py -k "issue_1038_ic4_return_side_case_expression_regression_lock" -q
  • result: 1 passed

Current local observation: IC4-shaped RETURN-side CASE probe is executing on latest baseline; residual may now be narrower/staler than originally reported. Follow-up decomposition remains in the child slices above.

[lmeyerov]

Conformance/status sync from #1261 decomposition wave:

• Branch/PR: issue-1261-case-return-decomp / #1271
• Landed in branch:
  • S1 lock test (RETURN-side IC4 CASE)
  • S2 parser locks
  • S3 binder/scope locks
  • S4 fail-fast fix for aggregates nested in row CASE expressions (626bf418)

Local check evidence:

• PYTHONPATH=. python -m pytest graphistry/tests/compute/gfql/cypher/test_lowering.py -k "issue_1038_" -q → 2 passed
• cuDF unavailable in current environment (ModuleNotFoundError: cudf) so GPU-path parity remains explicitly pending.

Tracking:

• Meta: #1261
• S5 receipt lane: #1269

[lmeyerov]

Follow-up evidence for #1038 from #1261 slice execution:

• Review-skill workflow completed to convergence for PR #1271 (2 waves, no findings)

  • artifact: plans/issue-1261-case-return-decomposition-program/review/final-report.md

• DGX GPU validation completed on ssh dgx-spark for both RAPIDS versions required in the review guidance:

  • 25.02 run: 4 passed (/tmp/issue1261_rapids2502.log)
  • 26.02 run: 4 passed (/tmp/issue1261_rapids2602.log)

Targeted tests included:

• test_issue_1038_ic4_return_side_case_expression_regression_lock
• test_issue_1038_rejects_aggregate_inside_row_case_expression
• test_parse_ic4_style_return_side_case_expression
• test_binder_ic4_return_case_across_with_distinct_scope

[lmeyerov]

Closing as delivered via #1271 (merged May 4, 2026; commit 74efb35). The #1261 S1-S5 decomposition landed with parser/binder/lowering guardrails, conformance receipts, and changelog updates.