Unify geometry IDs for Immersed Boundaries and ICPP Fixes #1543

[Riship749]

This PR unifies the redundant 2D and 3D geometry integers into single, dimension-agnostic IDs within the pre-processor for both patch_ib and patch_icpp.

Changes:

• Merged Rectangles (3) and Cuboids (9) -> ID 3.
• Merged 2D Airfoils (4) and 3D Airfoils (11) -> ID 4.
• Merged Circles (2), Spheres (8), and Cylinders (10) -> ID 2.

The unified subroutines now dynamically check for 3D constraints (e.g., Z-centroids and Z-lengths) using the p > 0 condition. Legacy compatibility is maintained by routing the old 3D IDs directly to the newly unified subroutines. Obsolete 3D-specific subroutines have been removed to reduce code duplication.

Testing:
Compiled successfully locally and passed the full ./mfc.sh test regression suite.

Fixes #1543

[sbryngelson]

Nice dedup direction — collapsing the near-identical 2D/3D check subroutines is worthwhile. But the unification is applied to the validation layer only, while patch generation (m_icpp_patches.fpp, IB equivalent) still dispatches on the original IDs (2/3/8/9/10/11). That mismatch turns several of the simplified checks into silent validation gaps:

1. 3D IDs no longer rejected in lower-D runs — the original p == 0 prohibits on sphere/cuboid/cylinder/3D-airfoil are gone, so e.g. geometry = 8 in a 2D case passes validation but hits the 3D generator.
2. ICPP circle dropped its n == 0 guard — a circle in a 1D run is now accepted (IB + rectangle checks still keep this guard, so it looks accidental).
3. Cylinder length validation weakened — gating "exactly one length" behind "any length > 0" lets a no-length or negative-length cylinder through.

Details inline. The "compiles + ./mfc.sh test passes" result is expected — these are all gaps in rejecting bad input, which a suite of valid cases won't exercise. Worth adding (or confirming) a case that feeds a 3D ID into a 2D run.

Smaller nits (typos / missing space in messages) also inline.

[codecov]

Codecov Report

❌ Patch coverage is 28.88889% with 32 lines in your changes missing coverage. Please review.
✅ Project coverage is 60.87%. Comparing base (1f02a6f) to head (9313ce9).
⚠️ Report is 6 commits behind head on master.

Files with missing lines	Patch %	Lines
src/pre_process/m_check_ib_patches.fpp	13.33%	8 Missing and 5 partials ⚠️
src/common/m_helper_basic.fpp	28.57%	9 Missing and 1 partial ⚠️
src/pre_process/m_check_patches.fpp	10.00%	7 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1547      +/-   ##
==========================================
+ Coverage   60.57%   60.87%   +0.30%     
==========================================
  Files          73       73              
  Lines       20264    20176      -88     
  Branches     2949     2946       -3     
==========================================
+ Hits        12275    12283       +8     
+ Misses       6000     5891     -109     
- Partials     1989     2002      +13     

☔ 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.

[Riship749]

The generation dispatchers in m_icpp_patches and m_ib_patches are now fully unified and handle the legacy IDs natively based on the p > 0 and n > 0 dimensional blocks. I also added the count and n==0 validator checks you requested.
Codecov might still flag the new @:PROHIBIT lines since the standard test suite doesn't feed bad dimension/length inputs to trigger them.

[Copilot]

Pull request overview

This PR unifies 2D/3D patch geometry IDs for ICPP and immersed boundary (IB) patches in the pre-processor, routing legacy 3D IDs to the unified implementations and reducing duplicated geometry handling.

Changes:

• Unified geometry-ID dispatch for sphere/cylinder/circle (2/8/10), cuboid/rectangle (3/9), and swept patches (4/11) in ICPP patch application.
• Updated patch-geometry validation logic to accept unified/legacy IDs and apply 3D-only constraints when p > 0.
• Removed now-redundant 3D-specific geometry-check subroutines for ICPP and IB.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 4 comments.

File	Description
src/pre_process/m_icpp_patches.fpp	Updates ICPP geometry dispatch to accept unified + legacy IDs and route to unified shape routines.
src/pre_process/m_check_patches.fpp	Updates ICPP geometry validation dispatch/constraints for unified + legacy IDs, with 3D checks gated by p > 0.
src/pre_process/m_check_ib_patches.fpp	Updates IB geometry validation dispatch/constraints for unified + legacy IDs, with 3D checks gated by p > 0.

[sbryngelson]

Thanks for taking this on, @Riship749 — the direction here was right, and I've pushed a follow-up commit (e91aa2b) onto your branch that builds on it. A summary of what changed and why:

The main issue was that the unification only existed in the pre-processor's checkers. The simulation side (m_ib_patches.fpp, m_compute_levelset.fpp, and the moment-of-inertia code in m_ibm.fpp) still dispatches on the old strict 2D/3D IDs, so e.g. a 3D case with patch_ib(1)%geometry = 2 passed all checks but then silently produced no immersed boundary at all — the worst kind of failure for a CFD code, since the run completes and looks plausible.

Rather than teaching every downstream dispatch site about the new IDs, the follow-up collapses unified/legacy IDs to one canonical ID per dimensionality at a single point — f_canonical_geometry, called right after the case file is read in both pre_process and simulation. Everything downstream then works unchanged, which also let me:

• revert the m_icpp_patches.fpp dispatch edits (no longer needed), including the icpp 4/11 merge — those are line/plane sweeps in the patch_icpp namespace (not airfoils, that's patch_ib), they take mutually exclusive parameters, and the checkers made those branches unreachable anyway;
• restore the cylinder strictness that got lost: geometry 10 with no lengths (or a negative one) is a hard error again instead of silently becoming a sphere, keyed on the canonical ID;
• extend the unification to the airfoil (4/11) and STL model (5/12) IB pairs, completing the list in Unify 2D and 3D Patch Geometries #1543;
• add two golden tests declaring a 3D IB sphere and cylinder via geometry 2 — the unified-sphere golden is bit-identical to the legacy geometry-8 case, which is a nice end-to-end proof the canonicalization works.

Net effect vs master is −41 source LOC with the docs tables updated to match. Full test suite passes locally apart from a few chemistry cases that failed for environmental reasons unrelated to this PR; CI will confirm.

[sbryngelson]

@Riship749 heads up — your force-push earlier today removed the follow-up commit I had pushed onto this branch (see the comment above). I've re-applied it on top of your amended commit (now 68891386). Please pull before pushing again (git pull --rebase) so we don't lose it a second time.

On your amendments: the tightened two-step cylinder length check is covered by the re-applied commit (equivalent pair of checks, keyed on the canonical ID so that a cylinder with no lengths is also an error). I deliberately did not keep the line-sweep checker relaxation (allowing p > 0 for geometry 4): the icpp 4/11 pair are line/plane sweeps with mutually exclusive parameters (normal(3) forbidden vs required), so they can't be unified the way the circle/sphere/cylinder family can — the follow-up commit reverts that part and keeps icpp 4 and 11 as separate shapes. The airfoil 4/11 unification from #1543 lives in the patch_ib namespace and is handled by the canonicalization.

asdf

[Copilot]

Pull request overview

Copilot reviewed 10 out of 12 changed files in this pull request and generated 1 comment.