ENH: Ingest ITKPrincipalComponentsAnalysis into Modules/Numerics

[hjmjohnson]

Ingest the standalone remote module
InsightSoftwareConsortium/ITKPrincipalComponentsAnalysis
into Modules/Numerics/PrincipalComponentsAnalysis/. PCA on scalar,
vector, and mesh-vertex data is now in-tree under
EXCLUDE_FROM_DEFAULT.

Built directly on upstream/main (post-#6236, so pixi 0.68 line
continuations apply); the v4 ingestion guidelines defined in #6204
(sanitize-history pass, deny-pattern pass, whitelist-only paths,
ghostflow-clean per-commit subjects) were applied locally to produce
this PR's content. Closes the consolidation work of issue #6060 for
this module.

Ingest results

Aspect	Value
Whitelist	Utilities/Maintenance/RemoteModuleIngest/whitelists/PrincipalComponentsAnalysis.list
Final tree merge count	16 (incl. ingest merge)
Whitelisted files	50 (include/, test/, wrapping/, CMakeLists.txt, itk-module.cmake, LICENSE)
Content-links at HEAD	All .cid (upstream already used ExternalData), 0 .md5, 0 .shaNNN, 0 raw binaries
examples/ and doc/ directories	Auto-excluded; remain in archived upstream
License	Apache-2.0
git blame	Walks back to original upstream contributors (Marc Bowers, Laurent Younes, and others)

Local validation

• pre-commit run --all-files: clean
• cmake -DModule_PrincipalComponentsAnalysis:BOOL=ON .: configures clean
• Compile check: new module's test source files build clean
  (PrincipalComponentsAnalysisHeaderTest1.cxx.o,
  itkVectorKernelPCATest.cxx.o).
• Test execution will fetch the existing .cid blobs through the
  standard ITKTestingData mirror at CI time.

Commits at PR tip (top → bottom)

1. ENH: Enable Module_PrincipalComponentsAnalysis in configure-ci
2. COMP: Remove PrincipalComponentsAnalysis .remote.cmake (in-tree)
3. DOC: Add PrincipalComponentsAnalysis README pointing at upstream
4. ENH: Ingest ITKPrincipalComponentsAnalysis into Modules/Numerics
   (merge commit --no-ff --allow-unrelated-histories to
   sanitized filter-repo'd upstream history; preserves 15 upstream
   merges and git blame to original authors)
5. ENH: Add v4 whitelist for PrincipalComponentsAnalysis ingest

Phase B (post-merge) — upstream archival

After this PR merges, run Phase B from a fresh worktree:

Utilities/Maintenance/RemoteModuleIngest/archive-remote-module.sh \
    PrincipalComponentsAnalysis

Refs: #6060

[dzenanz]

Some changes are needed.

[greptile-apps]

Greptile Summary

This PR ingests the standalone remote module ITKPrincipalComponentsAnalysis into Modules/Numerics/PrincipalComponentsAnalysis/, removing the corresponding Modules/Remote/*.remote.cmake file and enabling it in CI. The module provides itk::VectorFieldPCA for PCA on scalar, vector, and mesh-vertex data, along with a GaussianDistanceKernel helper, and is marked EXCLUDE_FROM_DEFAULT.

• The header itkVectorFieldPCA.h defines both GaussianDistanceKernel and VectorFieldPCA; GaussianDistanceKernel has uninitialized double member fields that can trigger undefined behavior if Evaluate() is called before SetKernelSigma().
• The implementation contains a structural dead-code pattern (redundant if (m_PointSet) after an unconditional throw) and the m_VertexCount member is never assigned.
• The test file carries a duplicate #include and an unused debugOut stream variable inherited from the upstream.

Confidence Score: 4/5

The module is marked EXCLUDE_FROM_DEFAULT so it won't affect existing builds; the main concern is an uninitialized-member issue in GaussianDistanceKernel that would only affect callers who invoke Evaluate() before SetKernelSigma().

The overall ingest is well-structured and follows the v4 guidelines. The one substantive concern is that GaussianDistanceKernel's default constructor leaves m_KernelSigma and m_OneOverMinusTwoSigmaSqr uninitialized — calling Evaluate() on a freshly-constructed kernel before SetKernelSigma() produces undefined behavior. All other findings are editorial cleanups inherited from the upstream and do not affect correctness for normal use.

Modules/Numerics/PrincipalComponentsAnalysis/include/itkVectorFieldPCA.h deserves a second look for the uninitialized members in GaussianDistanceKernel.

Important Files Changed

Filename	Overview
Modules/Numerics/PrincipalComponentsAnalysis/include/itkVectorFieldPCA.h	Header for GaussianDistanceKernel and VectorFieldPCA; GaussianDistanceKernel has uninitialized double members and m_VertexCount is declared but never written.
Modules/Numerics/PrincipalComponentsAnalysis/include/itkVectorFieldPCA.hxx	Template implementation of VectorFieldPCA; contains a redundant always-true null check after an unconditional throw.
Modules/Numerics/PrincipalComponentsAnalysis/test/itkVectorKernelPCATest.cxx	Test for kernel PCA; has a duplicate vnl_vector.h include and an unused debugOut std::ofstream variable.
Modules/Numerics/PrincipalComponentsAnalysis/itk-module.cmake	Module descriptor; correctly lists ITKCommon, ITKMesh, ITKIOMesh, ITKIOImageBase as DEPENDS and sets EXCLUDE_FROM_DEFAULT.
Modules/Numerics/PrincipalComponentsAnalysis/test/CMakeLists.txt	Test CMakeLists correctly registers itkVectorKernelPCATest with all 41 ExternalData .cid inputs via DATA{}.
Modules/Remote/PrincipalComponentsAnalysis.remote.cmake	Remote module descriptor correctly removed now that the module is in-tree.
pyproject.toml	Adds -DModule_PrincipalComponentsAnalysis:BOOL=ON to the configure-ci task so the new module is built in CI.
Modules/Numerics/PrincipalComponentsAnalysis/wrapping/itkGaussianDistanceKernel.wrap	Wraps GaussianDistanceKernel for all real types via ITK's standard itk_wrap_class/itk_wrap_template pattern; looks correct.

Class Diagram

%%{init: {'theme': 'neutral'}}%%
classDiagram
    class KernelFunctionBase~TRealValueType~ {
        <<abstract>>
        +Evaluate(u: TRealValueType) TRealValueType
    }
    class GaussianDistanceKernel~TRealValueType~ {
        -m_KernelSigma: double
        -m_OneOverMinusTwoSigmaSqr: double
        +SetKernelSigma(s: double)
        +Evaluate(u: TRealValueType) TRealValueType
    }
    class Object {
        <<ITK>>
    }
    class VectorFieldPCA~TVectorFieldElementType, TPCType~ {
        -m_PCAEigenValues: VectorType
        -m_BasisVectors: BasisSetTypePointer
        -m_VectorFieldSet: VectorFieldSetTypePointer
        -m_PointSet: InputPointSetPointer
        -m_KernelFunction: KernelFunctionPointer
        -m_ComponentCount: unsigned int
        -m_VertexCount: unsigned int
        +Compute()
        +GetAveVectorField() MatrixType
        +GetPCAEigenValues() VectorType
        +GetBasisVectors() BasisSetType
    }
    KernelFunctionBase <|-- GaussianDistanceKernel
    Object <|-- VectorFieldPCA
    VectorFieldPCA --> KernelFunctionBase : uses (optional)

Loading

_{Reviews (1): Last reviewed commit: "COMP: Address PrincipalComponentsAnalysi..." | Re-trigger Greptile}