[SakarPaisen]

What

Closes #5.

compare_text now returns unchanged_lines, counted from the equal opcodes of the alignment it already builds. The unified diff string and the added_lines / removed_lines counts are untouched.

>>> {k: v for k, v in compare_text("hello\nold", "hello\nnew").items() if k != "diff"}
{'added_lines': 1, 'removed_lines': 1, 'unchanged_lines': 1}

Scope

• src/mcp_tools_lab/tools/diff.py — one counter for tag == "equal", using old_end - old_start, plus a docstring line.
• tests/test_diff.py — five new tests.
• docs/tool-guide.md — describes the field and the added_lines + unchanged_lines == lines in after invariant.

Tests added

• Documented example: "hello\nold" vs "hello\nnew" → unchanged_lines=1, added_lines=1, removed_lines=1.
• Identical two-line input → 2; two empty inputs → 0.
• Entirely different inputs → 0.
• "same\r\n" vs "same" → 1, so the existing newline normalization applies to the count too.
• Repeated lines "a\nb\nb\nc" vs "a\nb\nc\nc" → unchanged_lines=3, added_lines=1, removed_lines=1. I checked this against the real SequenceMatcher alignment instead of assuming where the repeated b/c pair up.

Checks

• python -m pytest tests/test_diff.py -q — 7 passed
• python -m pytest -q — 67 passed
• python .github/scripts/check_rules.py — Scanned 11 files under src/ for print() and network imports: OK

Python: 3.13 on Windows 11.

Open item for the maintainers: if the case-insensitive comparison change (#13) adds an ignore_case option later, this count should follow the same comparison mode; the brief for #13 already says so. I left that out of scope here.

(Full disclosure: I used an AI coding assistant for this change, and I reviewed and verified everything submitted here.)