feat: epub2md – EPUB nach kapitelweisem Markdown

Kommandozeilenwerkzeug, das DRM-freie EPUB-2/3-Dateien offline in
kapitelweise Markdown-Dateien umwandelt. Erzeugt werden außerdem Medien,
metadata.yaml, manifest.json, conversion-report.json und Diagnose-Dateien.

- Spine-basierte Kapitelbildung mit --chapter-level, epub:type-Abschnitten,
  Fortsetzungsdateien und Erkennung von Layout-Überschriften
- Code mit exakter Einrückung, sicheren Fences und konservativer
  Spracherkennung
- MathML->TeX (konservativ), TeX aus alt-Texten, SVG-/Bildformeln
  bleiben Bilder
- Fußnoten-Modi end-of-chapter, end-of-book, inline und keep-links
- Interne Links/Anker werden umgeschrieben, Linkcheck über die Ausgabe
- Sicher: kein Entpacken, Schutz vor Path Traversal und XXE,
  DRM-Erkennung, atomares Schreiben, --force nur für frühere Ausgaben
- 82 pytest-Tests mit synthetischen EPUBs und Invariantentests für Beispiele/

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tests/test_math.py

@@ -0,0 +1,99 @@
+import pytest
+from lxml import etree
+
+from conftest import PNG, SVG, Book, Doc
+from epub2md.mathml import Unsupported, mathml_to_tex
+
+M = 'xmlns="http://www.w3.org/1998/Math/MathML"'
+
+
+def tex(xml: str) -> str:
+    return mathml_to_tex(etree.fromstring(f"<math {M}>{xml}</math>"))
+
+
+@pytest.mark.parametrize(
+    ("xml", "expected"),
+    [
+        ("<mfrac><mi>a</mi><mi>b</mi></mfrac>", r"\frac{a}{b}"),
+        ("<msup><mi>x</mi><mn>2</mn></msup>", "x^{2}"),
+        ("<msub><mi>a</mi><mi>i</mi></msub>", "a_{i}"),
+        ("<msubsup><mi>x</mi><mi>i</mi><mn>2</mn></msubsup>", "x_{i}^{2}"),
+        ("<msqrt><mi>x</mi></msqrt>", r"\sqrt{x}"),
+        ("<mroot><mi>x</mi><mn>3</mn></mroot>", r"\sqrt[3]{x}"),
+        ("<munderover><mo>∑</mo><mrow><mi>i</mi><mo>=</mo><mn>1</mn></mrow><mi>n</mi></munderover><msub><mi>a</mi><mi>i</mi></msub>",
+         r"\sum_{i = 1}^{n} a_{i}"),
+        ("<mi>α</mi><mo>≤</mo><mi>β</mi>", r"\alpha \leq \beta"),
+        ("<mover><mi>x</mi><mo>^</mo></mover>", r"\hat{x}"),
+        ("<mi mathvariant='double-struck'>R</mi>", r"\mathbb{R}"),
+        ("<mtext>für alle</mtext>", r"\text{für alle}"),
+        ("<mrow><mo>(</mo><mtable><mtr><mtd><mn>1</mn></mtd><mtd><mn>0</mn></mtd></mtr>"
+         "<mtr><mtd><mn>0</mn></mtd><mtd><mn>1</mn></mtd></mtr></mtable><mo>)</mo></mrow>",
+         r"\left( \begin{matrix} 1 & 0 \\ 0 & 1 \end{matrix} \right)"),
+    ],
+)
+def test_mathml_to_tex(xml, expected):
+    assert tex(xml) == expected
+
+
+def test_tex_annotation_is_preferred():
+    xml = ('<semantics><mi>x</mi><annotation encoding="application/x-tex">\\mathbf{x}</annotation></semantics>')
+    assert tex(xml) == r"\mathbf{x}"
+
+
+@pytest.mark.parametrize("xml", ["<menclose><mi>x</mi></menclose>", "<mmultiscripts><mi>x</mi></mmultiscripts>",
+                                 "<mtable><mlabeledtr><mtd>1</mtd></mlabeledtr></mtable>"])
+def test_unsupported_mathml_raises(xml):
+    with pytest.raises(Unsupported):
+        tex(xml)
+
+
+BODY = (
+    "<h1>Formeln</h1>"
+    f'<p>Inline <math {M}><mfrac><mi>a</mi><mi>b</mi></mfrac></math> im Satz.</p>'
+    f'<div class="equation"><math {M} display="block"><msup><mi>x</mi><mn>2</mn></msup></math> <span>(3.7)</span></div>'
+    f'<p>Unbekannt: <math {M}><menclose notation="box"><mi>z</mi></menclose></math></p>'
+    '<p>SVG: <span class="math"><img src="f.svg" alt="Formel 1"/></span></p>'
+    '<p>Bild ohne TeX: <span class="equation"><img src="eq.png" alt="Summe"/></span></p>'
+    '<p>Bild mit TeX: <span class="InlineEquation"><img src="eq2.png" alt="$$a &amp;amp; b \\leq c$$"/></span></p>'
+)
+
+
+def math_book() -> Book:
+    return Book(docs=[Doc("m.xhtml", BODY)], media={"f.svg": SVG, "eq.png": PNG, "eq2.png": PNG})
+
+
+def test_math_auto(convert):
+    run = convert(math_book())
+    text = run.chapter("001")
+    assert r"Inline $\frac{a}{b}$ im Satz." in text
+    assert "$$\nx^{2}\n$$" in text
+    assert "(3.7)" in text  # equation number visible
+    assert "<math" in text and "menclose" in text  # unsupported MathML kept, not lost
+    assert "![Formel 1](../media/images/" in text  # SVG formula stays SVG
+    assert "![Summe](../media/images/" in text
+    assert r"$a & b \leq c$" in text  # TeX from alt, double escaping repaired
+    codes = [w["code"] for w in run.report["warnings"]]
+    assert "mathml-kept" in codes and "formula-image" in codes
+    svgs = [m for m in run.report["media"]["files"] if m["file"].endswith(".svg")]
+    assert svgs and svgs[0]["media_type"] == "image/svg+xml"
+
+
+def test_math_tex_mode_uses_backslash_delimiters(convert):
+    text = convert(math_book(), "--math", "tex").chapter("001")
+    assert r"\(\frac{a}{b}\)" in text
+    assert "\\[\nx^{2}\n\\]" in text
+    assert "$" not in text.replace("\\$", "")
+
+
+def test_math_mathml_mode_keeps_mathml(convert):
+    text = convert(math_book(), "--math", "mathml").chapter("001")
+    assert '<math xmlns="http://www.w3.org/1998/Math/MathML"><mfrac><mi>a</mi><mi>b</mi></mfrac></math>' in text
+    assert "frac{" not in text
+
+
+def test_math_image_mode(convert):
+    run = convert(math_book(), "--math", "image")
+    text = run.chapter("001")
+    assert text.count("![") == 3  # svg + both formula images
+    assert "<math" in text  # pure MathML without image is not lost
+    assert any(w["code"] == "mathml-no-image" for w in run.report["warnings"])