Make the manual's EPUB from the same JPEG pictures as its PDF

pandoc embedded every recording whole, so the EPUB came out at 63 MB for
readers that mostly show a GIF's first frame anyway. tools/manual/pdf.py
now makes both downloads from its JPEG copies of the pictures: 7.5 MB.

.gitea/workflows/manual-pages.yml

@@ -62,7 +62,6 @@ jobs:
       # A4, a cover and a contents page with page numbers, a chapter a page —
       # so it says what the page says; tools/manual/pdf.py says why it goes
       # through JPEG copies of the pictures first.
-      # The EPUB comes from the Markdown, which pandoc reads better than a page.
       - name: Make the PDF and the EPUB
         run: |
           set -e
@@ -73,10 +72,8 @@ jobs:
           python3 -m venv /tmp/wp
           /tmp/wp/bin/pip install -q weasyprint
           mkdir -p site
-          /tmp/wp/bin/python tools/manual/pdf.py docs/manual site/darkroom-manual.pdf
+          /tmp/wp/bin/python tools/manual/pdf.py docs/manual \
-          title=$(sed -n 's/^# //p' docs/manual/README.md | head -1)
+            site/darkroom-manual.pdf site/darkroom-manual.epub
-          (cd docs/manual && pandoc README.md -o ../../site/darkroom-manual.epub \
-            --metadata title="$title" --toc --toc-depth=2 --epub-chapter-level=2)
           ls -l site
 
       # One commit, force-pushed: the branch is a build output, and its history

tools/manual/pdf.py

@@ -1,14 +1,17 @@
 #!/usr/bin/env python3
-"""The manual as a PDF: docs/manual/index.html through its print stylesheet.
+"""The manual as a PDF, and as an EPUB: the downloads the page links to.
 
-    python3 tools/manual/pdf.py docs/manual out.pdf
+    python3 tools/manual/pdf.py docs/manual out.pdf [out.epub]
 
-Needs WeasyPrint (and so Pillow): `pip install weasyprint`. The Manual pages
+The PDF is docs/manual/index.html through its print stylesheet; the EPUB is
-workflow runs this and publishes the result beside the page.
+README.md through pandoc, which reads Markdown better than a page. Needs
+WeasyPrint (and so Pillow), and pandoc for the EPUB. The Manual pages
+workflow runs this and publishes both beside the page.
 
 The page's pictures are PNG screenshots and GIF recordings, 16 MB and 46 MB
-of them, and WeasyPrint embeds a picture as it finds it: losslessly, a GIF
+of them, and WeasyPrint and pandoc embed a picture as they find it:
-whole. So the PDF is made from a copy of the page whose pictures are JPEGs
+losslessly, a GIF whole (63 MB of EPUB, for e-readers that mostly do not
+animate). So both are made from copies whose pictures are JPEGs
 (a recording's first frame, which is all a page can show) at the size they
 were recorded. Full chroma, because a 4:2:0 JPEG smears the coloured noise
 the AI denoise close-ups exist to show.
@@ -16,6 +19,7 @@ the AI denoise close-ups exist to show.
 import os
 import re
 import shutil
+import subprocess
 import sys
 import tempfile
 
@@ -31,7 +35,7 @@ def as_jpeg(src, dst):
         im.convert('RGB').save(dst, 'JPEG', quality=QUALITY, subsampling=0, optimize=True)
 
 
-def main(manual_dir, out):
+def main(manual_dir, pdf, epub=None):
     with tempfile.TemporaryDirectory() as tmp:
         os.mkdir(os.path.join(tmp, 'media'))
         renamed = {}
@@ -42,16 +46,27 @@ def main(manual_dir, out):
             jpg = f'{stem}.{ext[1:].lower()}.jpg'   # two pictures may share a stem
             as_jpeg(os.path.join(manual_dir, 'media', name), os.path.join(tmp, 'media', jpg))
             renamed[name] = jpg
-        page = open(os.path.join(manual_dir, 'index.html'), encoding='utf-8').read()
+
-        page = re.sub(r'src="media/([^"]+)"',
+        def copy(source, pattern, out):
-                      lambda m: f'src="media/{renamed.get(m.group(1), m.group(1))}"', page)
+            text = open(os.path.join(manual_dir, source), encoding='utf-8').read()
-        with open(os.path.join(tmp, 'index.html'), 'w', encoding='utf-8') as f:
+            text = re.sub(pattern, lambda m: m.group(1) + renamed.get(m.group(2), m.group(2)), text)
-            f.write(page)
+            with open(os.path.join(tmp, out), 'w', encoding='utf-8') as f:
-        HTML(os.path.join(tmp, 'index.html')).write_pdf(os.path.join(tmp, 'manual.pdf'))
+                f.write(text)
-        shutil.move(os.path.join(tmp, 'manual.pdf'), out)
+            return os.path.join(tmp, out)
+
+        page = copy('index.html', r'(src="media/)([^"]+)', 'index.html')
+        HTML(page).write_pdf(os.path.join(tmp, 'manual.pdf'))
+        shutil.move(os.path.join(tmp, 'manual.pdf'), pdf)
+
+        if epub:
+            md = copy('README.md', r'(\]\(media/)([^)\s]+)', 'README.md')
+            title = next(l[2:].strip() for l in open(md, encoding='utf-8') if l.startswith('# '))
+            subprocess.run(['pandoc', 'README.md', '-o', 'manual.epub', '--metadata', f'title={title}',
+                            '--toc', '--toc-depth=2', '--epub-chapter-level=2'], cwd=tmp, check=True)
+            shutil.move(os.path.join(tmp, 'manual.epub'), epub)
 
 
 if __name__ == '__main__':
-    if len(sys.argv) != 3:
+    if len(sys.argv) not in (3, 4):
         sys.exit(__doc__)
-    main(sys.argv[1], sys.argv[2])
+    main(*sys.argv[1:])