Potential Bug: innerRegionBoxScalingFactor silently ignored for common open/closed configurations (no warning, mesh unchanged)`

[AlejandroMunozManterola]

Summary

adapterOptions.innerRegionBoxScalingFactor is read from the input JSON and stored on ShapesClassification, but it only affects mesh construction in one specific branch: open problem + no open boundary in the STEP (synthetic rectangle + disk). In other common setups the factor is ignored with no warning, and identical factor values produce identical .msh files.
This caused confusion when trying to scale the “inner region” for a realistic unshielded bundle.

Expected behavior (user expectation)

• Changing innerRegionBoxScalingFactor (e.g. 1.1 vs 1.3) should change the near-region vacuum construction and/or mesh near the bundle, or
• Tulip should warn that the option has no effect for the current geometry/classification.

Actual behavior

• Factor is only used in ShapesClassification::buildOpenVacuumDomain() when isOpenBoundaryDefined() is false, inside:
  nearBoxSize = bbMaxLen * innerRegionBoxScalingFactor_
• If isOpenCase == false, buildClosedVacuumDomain() runs → factor unused.
• If isOpenCase == true but OpenBoundary_* exists in STEP (mapped via open layer), vacuum is cut(openVacuum, nonVacuumSurfaces) → factor unused.
  No log message when the user sets the factor in an ignored path.

Code references

cases_for_study.zip

• AdapterOptions.h — documents factor for “bundles with no open boundary defined” (easy to miss other cases).
• ShapesClassification::buildVacuumDomain() — open vs closed branch.
• ShapesClassification::buildOpenVacuumDomain() — lines ~460–476 (STEP open) vs ~477–535 (synthetic, factor used).

Reproduction

A) Closed classification (factor sweep has no effect)

1. Use a bundle STEP without OpenBoundary_0 / inner_region (e.g. trimmed no_open_no_inner).
2. Same layers, only change innerRegionBoxScalingFactor: 1.1, 1.2, 1.3.
3. Re-run adapter + solver (fresh .msh each time).
   Result: identical .msh size/hash; identical FEM+MP. Classification uses buildClosedVacuumDomain().

B) Paper-like open boundary in STEP (factor in JSON has no effect)

1. Use inner_region STEP with OpenBoundary_0 and open layer in JSON (Tabla II case).
2. Add "innerRegionBoxScalingFactor": 1.2 in adapterOptions vs default 1.3.
   Result: mesh/solution unchanged vs base; vacuum built from CAD open volume, not synthetic near box.

Impact

• Users may run parameter sweeps that do nothing.
• Name collides with STEP layer inner_region and output field innerRegionBox, which come from other code paths.
• Post-processing innerRegionBox in .tulip.out.json is derived from the mesh (Model::getInnerRegionBoundingBox), not from innerRegionBoxScalingFactor.

Suggested improvements

1. Warn (or error in strict mode) when innerRegionBoxScalingFactor is set but the active branch does not use it; log: isOpenCase, isOpenBoundaryDefined(), branch taken.
2. Document a small decision table in docs/tulip_data_format.md under adapterOptions.
3. Optional: clarify naming (syntheticNearRegionScale?) or extend behavior if scaling CAD inner_region / near mesh is intended (separate design discussion).

Environment

• Tulip adapter + driver, realistic multi-wire bundle with dielectrics (31 conductors). See files attached.