DOCX Comparison

The system SHALL provide a CorrelationStatus enum with the following values: Nil, Normal, Unknown, Inserted, Deleted, Equal, Group, MovedSource, MovedDestination, FormatChanged.

The system SHALL detect "continuation patterns" in legal numbering where a paragraph at ilvl > 0 continues a flat sequence rather than creating a nested hierarchy. When detected, the system SHALL use the effective level (level 0) properties instead of the declared level.

A continuation pattern exists when:
1. The paragraph is the first at this level in the current sequence, AND
2. The level's start value equals the parent level's counter + 1

The system SHALL calculate footnote display numbers sequentially based on document order, NOT using raw XML w:id attribute values. The w:id is a reference identifier linking footnoteReference to footnote definitions; display numbers are determined by the order footnotes appear in the document flow.

The system SHALL provide a detectMovesInAtomList() function that identifies relocated content after LCS comparison. The algorithm:
1. Groups consecutive atoms by correlationStatus into blocks (Deleted blocks, Inserted blocks)
2. Extracts text from each block by joining content element values
3. Filters blocks by minimum word count (configurable, default: 3)
4. Calculates Jaccard word similarity between deleted and inserted blocks
5. Converts matching pairs (above threshold) to MovedSource and MovedDestination

The system SHALL provide configurable settings for move detection:
- detectMoves: Enable/disable move detection (default: true)
- moveSimilarityThreshold: Jaccard threshold for move matching (default: 0.8)
- moveMinimumWordCount: Minimum words for move consideration (default: 3)
- caseInsensitive: Case-insensitive similarity matching (default: false)

The system SHALL generate native Word move tracking markup when moves are detected:

For moved source (content moved FROM):
- w:moveFromRangeStart with w:id, w:name, w:author, w:date
- w:moveFrom containing the moved content
- w:moveFromRangeEnd with matching w:id

For moved destination (content moved TO):
- w:moveToRangeStart with w:id, w:name, w:author, w:date
- w:moveTo containing the moved content
- w:moveToRangeEnd with matching w:id

The system SHALL provide a FormatChangeInfo interface with:
- oldRunProperties: The w:rPr element from the original document (may be null)
- newRunProperties: The w:rPr element from the modified document (may be null)
- changedProperties: Array of friendly property names that differ (e.g., "bold", "italic")

The system SHALL provide a detectFormatChangesInAtomList() function that identifies formatting differences in Equal atoms after LCS comparison. The algorithm:
1. Iterates through atoms with correlationStatus === Equal
2. Skips atoms without comparisonUnitAtomBefore reference
3. Extracts w:rPr from ancestor w:r element for both original and modified atoms
4. Normalizes w:rPr elements (removes existing w:rPrChange, sorts children)
5. Compares normalized properties for equality
6. Converts non-equal atoms to FormatChanged status with formatChange info

The system SHALL provide configurable settings for format change detection:
- detectFormatChanges: Enable/disable format change detection (default: true)

The system SHALL generate native Word format change tracking markup (w:rPrChange) when format changes are detected.

For format-changed content:
- The current w:rPr contains the NEW properties
- w:rPrChange is added as a child of w:rPr containing the OLD properties
- w:rPrChange includes w:id, w:author, and w:date attributes

The system SHALL include format changes in GetRevisions() output with type FormatChanged, extracting revision information from w:rPrChange elements.