The default preset keeps `alpha` and `gamma`, changes lowercase `beta` to uppercase `BETA`, and inserts `delta`. Exact mode treats case as real text. The result intentionally describes the apparent replacement as one removal followed by one addition, then reports the independent insertion. This walkthrough connects the visible records, line numbers, counts and similarity to the exact source versions.
Start with terminated three- and four-line texts
Paste `alpha`, `beta`, `gamma` as three lines on the left, each including its line ending. On the right paste `alpha`, `BETA`, `delta`, `gamma`, also with a final newline. Choose Exact and keep Ignore case off. TextDiff reports three original lines and four revised lines. Because both texts end with LF, no no-newline marker appears. Transport CRLF would normalize to the same tokens and would not alter this example.
Read five ordered records
Record one retains original line 1 as revised line 1. Record two removes original line 2, `beta`. Record three adds revised line 2, `BETA`; record four adds revised line 3, `delta`. The last record retains original line 3 as revised line 4. This order is a shortest insertion/removal script. Red and green adjacency helps a person see a replacement, but TextDiff does not merge those atoms or claim knowledge of the author’s editing action.
Verify counts and case-insensitive variation
The exact result has two retained, one removed and two added records. Similarity is `2 × 2 / (3 + 4)`, or about 57.1%. Turn Ignore case on and compare again: `beta` and `BETA` now share a key while both spellings remain visible. The result becomes three retained, zero removed and one added, with similarity `6 / 7`, about 85.7%. The scores answer different configured questions, so neither should be presented as an absolute quality measure.