BUG: Backport BresenhamLine integer-only algorithm and double precision fix

[hjmjohnson]

Backport two BresenhamLine bug fixes from main to release-5.4. Part of #6051.

1. 59e3c069 — Integer-only N-dimensional Bresenham for BuildLine(Index, Index), avoiding floating-point direction conversion entirely
2. b7248b1c — Use double precision (not float) for ending index computation in BuildLine(Direction, length)

Cherry-pick adaptation

Required one API adaptation: itk::Math::Absolute() does not exist on release-5.4 (added on main as part of Math modernization). Replaced with itk::Math::abs() which is the release-5.4 equivalent.

Local build verified: compiles cleanly with the Path and Voronoi modules enabled.

[greptile-apps]

Greptile Summary

This PR backports two BresenhamLine bug fixes from main to release-5.4: (1) a full integer-only N-dimensional Bresenham algorithm for BuildLine(Index, Index) that avoids floating-point direction conversion, and (2) a double-precision variable for the ending-index computation in BuildLine(Direction, length). The itk::Math::Absolute → itk::Math::abs substitution for release-5.4 API compatibility is correct.

Confidence Score: 5/5

Safe to merge; both algorithms are correct and the single finding is a minor precision opportunity, not a defect

Both bug fixes are algorithmically sound — the integer Bresenham correctly guarantees exact start/end points, and the double-precision variable genuinely improves the multiplication step. The only finding (P2) is that the division on line 44 still runs in float because Direction's components are float, so one further cast could be applied, but the code as-is is strictly better than before and introduces no regressions.

No files require special attention

Important Files Changed

Filename	Overview
Modules/Core/Common/include/itkBresenhamLine.hxx	Two backported bug fixes: integer-only Bresenham for BuildLine(Index,Index) and double-precision variable for BuildLine(Direction,length); logic is correct but the float→double precision improvement in the division on line 44 is incomplete due to LType using float components

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["BuildLine(Direction, length)"] --> B["Normalize direction"]
    B --> C["Find dominant axis (max |component|)"]
    C --> D["Compute euclideanLineLen = (length-1) / |Direction[mainAxis]| as double"]
    D --> E["Compute LastIndex: double × float → double, RoundHalfIntegerUp"]
    E --> F["Call BuildLine(StartIndex, LastIndex)"]
    F --> G["Convert IndexArray → OffsetArray"]

    H["BuildLine(p0, p1)"] --> I["Compute absDeltas and step (integer only)"]
    I --> J["Find mainAxis (max |delta|)"]
    J --> K["numPixels = maxAbsDelta + 1"]
    K --> L{"For s = 0..numPixels-1"}
    L --> M["Push currentIndex"]
    M --> N["Advance main axis by step[mainAxis]"]
    N --> O["Each secondary axis: accumulate 2×absDelta; if ≥ maxAbsDelta → step, subtract 2×maxAbsDelta"]
    O --> L
    L -->|done| P["Return IndexArray guaranteed to end at p1"]

Loading

_{Reviews (1): Last reviewed commit: "BUG: BresenhamLine - use double precisio..." | Re-trigger Greptile}

[thewtex]

@hjmjohnson thanks!