Phase 14: PDF aus Domaenenmodell via WeasyPrint

- src/render/pdf.py kapselt WeasyPrint mit lazy Import; Pango/Cairo-Fehler
  erscheinen erst beim Aufruf, nicht beim Modul-Load
- src/build.py: --pdf rendert jeweils ride.N.M.pdf neben index.html;
  fehlende Systemlibs werden geloggt, Build laeuft weiter
- templates/html/review.html: print-only DOI-Zeile im Header (A6:
  DOI muss auf Seite 1 des PDFs erscheinen, da Sidebar im Print verschwindet)
- static/css/ride.css: print-Stylesheet ausgebaut — @page A4, Chrome
  weg, page-break-after auf Headings, Link-URLs als Klammertext
- .github/workflows/build.yml: apt-install libpango/libpangoft2/
  libharfbuzz-subset; CI ruft jetzt python -m src.build --pdf

Tests:
- 4 neue (PDF-Magic-Bytes, DOI-im-PDF, print-CSS-Contract, DOI-Header-Line);
  WeasyPrint-Tests skippen sauber wenn GTK fehlt (lokales Windows)
- 455 grun + 2 skipped (waren 452)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: chpollin

.github/workflows/build.yml

@@ -64,6 +64,12 @@ jobs:
           cache: pip
           cache-dependency-path: ride-static/requirements.txt
 
+      # WeasyPrint links Pango/Cairo at import time. The Ubuntu runner
+      # ships a partial GTK stack; install the WeasyPrint-listed runtime
+      # libs explicitly so `import weasyprint` succeeds in CI.
+      - name: Install WeasyPrint system libraries
+        run: sudo apt-get update && sudo apt-get install -y libpango-1.0-0 libpangoft2-1.0-0 libharfbuzz-subset0
+
       - name: Install dependencies
         working-directory: ride-static
         run: pip install -r requirements.txt
@@ -119,7 +125,7 @@ jobs:
         working-directory: ride-static
         run: |
           if [ -f src/build.py ] || python -c "import src.build" 2>/dev/null; then
-            python -m src.build --base-url=/${{ github.event.repository.name }}
+            python -m src.build --base-url=/${{ github.event.repository.name }} --pdf
           else
             echo "::warning::src.build is not implemented yet (Phase 8). Emitting placeholder site/."
             mkdir -p site

requirements.txt

@@ -16,7 +16,10 @@ pyyaml>=6.0
 markdown>=3.5
 
 # PDF (Phase 14)
-# weasyprint>=60.0          # uncomment when Phase 14 starts
+# WeasyPrint pulls Pango/Cairo at import time. Local Windows installs
+# additionally need the GTK3 runtime; CI on Ubuntu uses libpango1.0-0
+# + libpangoft2-1.0-0 from apt (see .github/workflows/build.yml).
+weasyprint>=60.0
 
 # Search (Phase 11)
 # pagefind is a binary fetched by the build script, not a Python package.

src/build.py

@@ -270,6 +270,7 @@ def build(
     linkcheck: bool = False,
     matomo_url: str = "",
     matomo_site_id: str = "",
+    pdf: bool = False,
 ) -> int:
     """Run the build. Returns the number of review pages written."""
     if not corpus_dir.exists():
@@ -337,6 +338,11 @@ def build(
     oai_files = _write_oai_pmh_snapshot(tuple(rendered), site, out_root)
     redirect_count = write_redirects(tuple(rendered), out_root, base_url=site.base_url)
 
+    pdf_count = 0
+    pdf_failed: list[tuple[str, str]] = []
+    if pdf:
+        pdf_count, pdf_failed = _render_pdfs(parsed, out_root)
+
     # Phase 13 / Welle 10: validation + link-probe + aggregated build report.
     validation_report = None
     if validate:
@@ -380,13 +386,59 @@ def build(
     if link_report:
         print(f"Linkcheck: {link_report.alive} alive, {link_report.dead} dead "
               f"({link_report.probed} probed)")
+    if pdf:
+        print(f"PDF: {pdf_count} rendered, {len(pdf_failed)} failed")
     print("Wrote api/build-info.json")
 
     _print_asset_summary(asset_reports)
 
     return len(rendered)
 
 
+def _render_pdfs(
+    parsed: list,
+    out_root: Path,
+) -> tuple[int, list[tuple[str, str]]]:
+    """Render every parsed review's HTML to a sibling PDF.
+
+    Returns ``(success_count, failures)`` where ``failures`` is a list
+    of ``(review_id, error_message)`` pairs. The whole pass surfaces
+    cleanly (count = 0) when WeasyPrint or its system libraries
+    cannot be loaded, so a missing GTK install on a developer machine
+    does not block the rest of the build.
+
+    Phase 14 / Welle 11. The HTML is read from disk so the print
+    output reflects exactly what was deployed to the static tree —
+    no separate template, no second render pass.
+    """
+    try:
+        from src.render.pdf import render_review_pdf
+    except (ImportError, OSError) as exc:
+        print(
+            "PDF: WeasyPrint unavailable, skipping. "
+            f"Install instructions: https://doc.courtbouillon.org/weasyprint/  ({exc})",
+            file=sys.stderr,
+        )
+        return 0, []
+
+    count = 0
+    failed: list[tuple[str, str]] = []
+    for path, review in parsed:
+        review_id = review.id or path.stem
+        page_dir = out_root / "issues" / (review.issue or "0") / review_id
+        html_path = page_dir / "index.html"
+        if not html_path.exists():
+            continue  # render pass skipped this review (already in `failed`)
+        pdf_path = page_dir / f"{review_id}.pdf"
+        try:
+            render_review_pdf(html_path, pdf_path)
+            count += 1
+        except Exception as exc:  # noqa: BLE001 — keep building on per-file failure
+            failed.append((review_id, str(exc)))
+            print(f"PDF failed: {review_id}: {exc}", file=sys.stderr)
+    return count, failed
+
+
 def _write_build_info(
     *,
     out_root: Path,
@@ -535,16 +587,13 @@ def main(argv: Optional[list[str]] = None) -> int:
     parser = argparse.ArgumentParser(description="Build the ride.i-d-e.de static site.")
     parser.add_argument("--reviews", type=int, default=None, help="Limit to first N reviews (for iteration)")
     parser.add_argument("--base-url", default="", help="Deploy URL prefix; empty for relative paths")
-    parser.add_argument("--pdf", action="store_true", help="Also run the WeasyPrint PDF pass (Phase 14)")
+    parser.add_argument("--pdf", action="store_true", help="Render a PDF next to every review's HTML via WeasyPrint")
     parser.add_argument("--no-validate", action="store_true", help="Skip the RelaxNG validation pre-check")
     parser.add_argument("--linkcheck", action="store_true", help="Probe external bibliography URLs (slow ~5min, off by default)")
     parser.add_argument("--matomo-url", default="", help="Matomo tracker URL (e.g. https://matomo.example.org/); empty disables tracking")
     parser.add_argument("--matomo-site-id", default="", help="Matomo site id; required when --matomo-url is set")
     args = parser.parse_args(argv)
 
-    if args.pdf:
-        print("--pdf is a Phase 14 placeholder; no PDF rendered yet.", file=sys.stderr)
-
     if bool(args.matomo_url) != bool(args.matomo_site_id):
         parser.error("--matomo-url and --matomo-site-id must be set together")
 
@@ -555,6 +604,7 @@ def main(argv: Optional[list[str]] = None) -> int:
         linkcheck=args.linkcheck,
         matomo_url=args.matomo_url,
         matomo_site_id=args.matomo_site_id,
+        pdf=args.pdf,
     )
     print(f"Wrote {written} review pages to {SITE_DIR}")
     return 0

src/render/pdf.py

@@ -0,0 +1,33 @@
+"""PDF rendering via WeasyPrint — Phase 14.
+
+Per [[requirements#R3 Rezension herunterladen]] and
+[[requirements#A6 PDF-Pfad]] every review ships a PDF next to its HTML
+page. We feed WeasyPrint the already-rendered ``index.html`` so the
+print output reflects the same domain model and templates as the web
+view; the ``@media print`` block in ``ride.css`` strips chrome
+(nav, sidebar, WIP banner) and surfaces a DOI line on the first page.
+
+WeasyPrint pulls Pango/Cairo at import time. The import lives inside
+:func:`render_review_pdf` so missing system libraries surface as a
+per-call ``ImportError`` rather than aborting :mod:`src.build` at
+module load — callers can then decide between skip and hard-fail.
+"""
+from __future__ import annotations
+
+from pathlib import Path
+
+
+def render_review_pdf(html_path: Path, pdf_path: Path) -> None:
+    """Render an already-written review HTML file to PDF on disk.
+
+    Relative asset URLs (figures, stylesheet, fonts) resolve relative
+    to ``html_path``'s directory — the build writes everything next
+    to ``index.html`` so this matches the deployed layout.
+
+    Raises ``ImportError`` when WeasyPrint or its system libraries
+    cannot be loaded. The caller is expected to print a helpful
+    install hint and continue without PDFs rather than crash.
+    """
+    from weasyprint import HTML  # local import — see module docstring
+
+    HTML(filename=str(html_path)).write_pdf(target=str(pdf_path))

static/css/ride.css

@@ -984,13 +984,33 @@ main.ride-shell {
   *, *::before, *::after { animation: none !important; transition: none !important; }
 }
 
-/* ─── 8. Print stylesheet (Phase 14 placeholder) ─────────────────────── */
-
-/* Phase 14 (PDF via WeasyPrint) replaces this block with a full print style.
-   Until then a minimal fallback keeps the HTML-Druck halbwegs lesbar. */
+/* Print-only DOI line in the review header — hidden on screen, shown
+   when the page is rendered to PDF via WeasyPrint (requirements A6:
+   DOI on the first page). The Meta sidebar carries the DOI for web
+   readers; in print, the sidebar is suppressed. */
+.ride-review__doi-print { display: none; }
+
+/* ─── 8. Print stylesheet (Phase 14) ─────────────────────────────────── */
+
+/* Used by the browser's Drucken-Dialog and by WeasyPrint when generating
+   the per-review PDF. Strips chrome, drops the sidebar, surfaces the
+   first-page DOI line. WeasyPrint honours @page, page margins and
+   page-break-* properties — kept simple here; tune per-template if a
+   review's tables overflow. */
 @media print {
-  .ride-sidebar, .ride-search-slot, .ride-skip { display: none; }
-  .ride-page { grid-template-columns: minmax(0, 1fr); }
+  @page { size: A4; margin: 18mm 16mm; }
+  body { font-size: 11pt; }
+  .ride-wip, .ride-header, .ride-nav, .ride-footer,
+  .ride-sidebar, .ride-search-slot, .ride-skip,
+  .ride-cite__btns, .ride-cite-data { display: none; }
+  .ride-page, .ride-review { grid-template-columns: minmax(0, 1fr); }
+  .ride-content, .ride-shell { max-width: 100%; padding: 0; margin: 0; }
+  .ride-review__doi-print {
+    display: block; margin: 0 0 var(--ride-space-4) 0;
+    font-size: 10pt; color: #444;
+  }
   a { color: inherit; text-decoration: none; }
   a[href^="http"]::after { content: " (" attr(href) ")"; font-size: 0.9em; color: #555; }
+  h1, h2, h3 { page-break-after: avoid; }
+  figure, table { page-break-inside: avoid; }
 }

templates/html/review.html

@@ -73,6 +73,16 @@ <h1 class="ride-review__title">{{ review.title }}</h1>
       {%- endif %}
     </header>
 
+    {#- Print-only DOI line per requirements A6 — must appear on the
+        first page of the PDF rendering. Hidden on screen because the
+        Meta sidebar already exposes the DOI to web readers; in the
+        PDF the sidebar is suppressed by ``@media print``. -#}
+    {%- if review.doi %}
+    <p class="ride-review__doi-print">
+      DOI: <a href="https://doi.org/{{ review.doi }}">{{ review.doi }}</a>
+    </p>
+    {%- endif %}
+
     {%- if abstract_section -%}
     <section class="ride-review__abstract" aria-labelledby="abstract-heading">
       <h3 id="abstract-heading">{{ site.strings.abstract | default('Abstract') }}</h3>

tests/test_css_contract.py

@@ -64,6 +64,30 @@ def test_tag_pills_meet_target_size_minimum():
     assert "min-height: 24px" in rule
 
 
+def test_print_stylesheet_hides_chrome_and_shows_doi():
+    """Phase 14 print stylesheet drives the WeasyPrint output.
+
+    Three load-bearing rules:
+      - chrome (nav, sidebar, footer) is suppressed so the PDF is
+        body-only;
+      - .ride-review__doi-print flips from display:none to display:block
+        in print so the DOI lands on page 1 (requirements A6);
+      - @page sets paper size + margins for a predictable PDF.
+    """
+    css = _css()
+    print_block_start = css.find("@media print")
+    assert print_block_start != -1
+    # Slice generously — the @media block contains nested rules with
+    # their own braces, so we just take everything from @media print to
+    # the next top-level marker.
+    block = css[print_block_start:]
+    assert "@page" in block
+    assert ".ride-sidebar" in block
+    # display:block on the DOI line is the load-bearing flip.
+    assert ".ride-review__doi-print" in block
+    assert "display: block" in block
+
+
 def test_reduced_motion_preference_is_honoured():
     """WCAG 2.3.3 (AAA, but project policy): users who set
     ``prefers-reduced-motion: reduce`` get no animations, no

tests/test_render_html.py

@@ -498,6 +498,30 @@ def test_console_banner_omitted_when_build_info_missing():
     assert "console.info(" not in html
 
 
+# ── Print-only DOI line (Phase 14, A6) ───────────────────────────────
+
+
+def test_review_html_carries_print_only_doi_line_when_doi_set():
+    """A6: DOI must surface on the first page of the PDF rendering.
+
+    The Meta sidebar carries the DOI for web readers but is hidden in
+    the print stylesheet. The print-only line in the header guarantees
+    the DOI lands on page 1 of the PDF.
+    """
+    review = _minimal_review(doi="10.18716/ride.a.13.7")
+    html = render_review(review)
+    assert 'class="ride-review__doi-print"' in html
+    assert "10.18716/ride.a.13.7" in html
+    assert 'href="https://doi.org/10.18716/ride.a.13.7"' in html
+
+
+def test_review_html_omits_print_doi_line_when_doi_missing():
+    """No DOI → no orphan label in the PDF (corpus has reviews without DOI)."""
+    review = _minimal_review(doi=None)
+    html = render_review(review)
+    assert 'class="ride-review__doi-print"' not in html
+
+
 # ── Cookieless Matomo (R16) ──────────────────────────────────────────