GeezerVison

Refactor: Resolve High Coupling, God Class, and High Cognitive Complexity

This PR addresses three critical architectural code smells identified via PMD static analysis.
All changes are purely structural refactorings — zero changes to public APIs, method
signatures, or observable behavior.

Issues Addressed

1. High Object Coupling in DiffUtils.java (CBO Score: 29)

Problem: DiffUtils was directly coupled to 29 internal classes simultaneously, acting
as a monolithic control panel for algorithm engines, patch containers, and formatting
utilities. This made future algorithm changes risky and violated separation of concerns.

Fix — Delegation Pattern:

• PatchUtils (new) — owns all patch apply / unpatch / merge logic
• InlineDiffUtils (new) — owns character and word-level inline annotation
• DiffAlgorithmDefaults (new) — factory for default algorithm engine instances
• DiffUtils is now a slim façade that delegates to these specialists

2. God Class in DiffRowGenerator.java (~780 lines, SRP Violation)

Problem: A single 780-line class simultaneously managed diff orchestration, HTML tag
rendering, and asymmetric delta normalization — three unrelated concerns that violated the
Single Responsibility Principle.

Fix — Single-Responsibility Decomposition:

• InlineTagRenderer (new) — isolates HTML span tag injection and newline boundary handling
• DeltaDecompressor (new) — normalizes asymmetric ChangeDelta objects (e.g. 3 deleted
  lines vs 1 inserted) into uniform 1-to-1 row pairs for table rendering
• InlineDiffAnnotatorConfig (new) — immutable value object bundling all annotation
  configuration flags, eliminating parameter bloat
• InlineDiffAnnotator (new) — centralized token-level word/character highlighting engine
• DiffRowGenerator reduced from ~780 lines to ~240 lines; now a pure orchestrator

3. High Cognitive Complexity in UnifiedDiffReader.java (Score: 45 reduced to ~5)

Problem: The parse() method had a Cognitive Complexity score of 45 — three times the
allowed limit of 15 — caused by 4 levels of deeply nested while/if blocks and 16-parameter
vararg rule lists repeated across methods.

Fix — Phase-Driven Assembly Line Pattern:

• Consolidated 16 regex rule varargs into pre-allocated static arrays
• parseHeaderSection() (new) — reads global preamble before any file diff
• parseFileHeader() (new) — consumes file metadata (---, +++, git index lines)
• parseChunkSection() and isChunkFinished() (new) — processes @@ chunk headers
  and line-by-line additions/deletions
• parseTailSection() (new) — handles trailing footer content after all file diffs
• parse() is now a flat, readable ~15-line sequential orchestrator

Files Changed

Design Principles Applied

• Single Responsibility Principle — each new class has exactly one job
• Delegation Pattern — façades forward to specialists instead of doing everything
• Open/Closed Principle — new algorithms or renderers can be added without touching core façades
• Backward Compatibility — all public method signatures fully preserved

Functional Validation

All existing tests pass with zero regressions. The 3 skipped tests are pre-existing
upstream skips entirely unrelated to this refactoring.

Tests run: 11, Failures: 0, Errors: 0  —  GenerateUnifiedDiffTest
Tests run:  1, Failures: 0, Errors: 0  —  ChunkTest
Tests run:  8, Failures: 0, Errors: 0  —  PatchWithAllDiffAlgorithmsTest
Tests run:  2, Failures: 0, Errors: 0  —  PatchWithMyerDiffTest
Tests run:  1, Failures: 0, Errors: 0  —  PatchWithMyerDiffWithLinearSpaceTest
Tests run:  4, Failures: 0, Errors: 0  —  DiffRowGeneratorEqualitiesTest
Tests run: 44, Failures: 0, Errors: 0  —  DiffRowGeneratorTest
Tests run:  4, Failures: 0, Errors: 0  —  StringUtilsTest
Tests run: 32, Failures: 0, Errors: 0  —  UnifiedDiffReaderTest
Tests run:  7, Failures: 0, Errors: 0  —  UnifiedDiffRoundTripTest       (1 pre-existing skip)
Tests run:  1, Failures: 0, Errors: 0  —  UnifiedDiffRoundTripNewLineTest (1 pre-existing skip)
Tests run:  2, Failures: 0, Errors: 0  —  UnifiedDiffWriterTest
Tests run:  2, Failures: 0, Errors: 0  —  HistogramDiffTest (jgit module)

Total — Tests run: 147, Failures: 0, Errors: 0, Skipped: 3
BUILD SUCCESS