beta.33 candidate: an Author's text kept as typed, the editor writes stored blocks back unchanged, tier 1 fidelity, tier 2 runtime (from 8a46cd4)

Author: iOSDevSK

.claude-plugin/marketplace.json

@@ -14,7 +14,7 @@
       "source": "./",
       "displayName": "html2wp — HTML to WordPress",
       "description": "BETA. Convert a static HTML site into a standalone WordPress block theme, verified against the original at three viewport widths — and reviewed by you before handover. The conversion runs through a verified Astro 5 build, so the Astro project is a real output too — compared pixel by pixel against the source before anything WordPress-shaped is generated.",
-      "version": "1.0.0-beta.32",
+      "version": "1.0.0-beta.33",
       "author": {
         "name": "BELNEM s.r.o.",
         "email": "hello@html2wp.dev",

.claude-plugin/plugin.json

@@ -2,7 +2,7 @@
   "name": "html2wp",
   "displayName": "html2wp — HTML to WordPress",
   "description": "BETA. Convert a static HTML site — or anything that builds to one (Lovable, Bolt, v0, Next export) — into a standalone WordPress block theme: verified pixel by pixel, pages, blog, forms, menus, SEO and redirects, with no plugin required and verification gates at every stage. Every conversion ends in your own page-by-page review; the gates catch most fidelity drift, not all of it.",
-  "version": "1.0.0-beta.32",
+  "version": "1.0.0-beta.33",
   "author": {
     "name": "BELNEM s.r.o.",
     "email": "hello@html2wp.dev",

.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "html2wp",
-  "version": "1.0.0-beta.32",
+  "version": "1.0.0-beta.33",
   "description": "Convert static HTML or buildable web projects into verified, editable WordPress block themes.",
   "author": {
     "name": "BELNEM s.r.o.",

VERSION

@@ -1 +1 @@
-1.0.0-beta.32
+1.0.0-beta.33

skills/html2wp/VERSION

@@ -1 +1 @@
-1.0.0-beta.32
+1.0.0-beta.33

skills/html2wp/assets/scripts/gutenberg-package.py

@@ -348,6 +348,13 @@ def validate_evidence(root, report, theme_report=None):
                     if not any(row.get('kind') == kind and row.get('id') == expected and row.get('width') == width
                                and (valid_editor_visual(row, region) or valid_empty_part(row, region)) for row in report['editorVisual']):
                         raise ValueError(f'Missing matching editor {region} visual for {expected} at {width}px')
+    # The editor writes back what the import stored (gutenberg-verify-local.py
+    # SERIALIZATION): a passing "serialization" row for every surface with blocks.
+    serialized = {(row.get('kind'), row.get('id')) for row in report.get('serialization') or []
+                  if row.get('passed') is True and row.get('byteIdentical') is True and not row.get('attributeLoss')}
+    for row in report['editor']:
+        if row.get('count', 0) > 0 and row.get('kind') in ('page', 'post', 'templates', 'template-parts') and (row.get('kind'), row.get('id')) not in serialized:
+            raise ValueError(f'Missing passing serialization gate for {row.get("kind")} {row.get("slug") or row.get("id")}: the editor would rewrite its stored blocks')
     for name in ('style.css', 'theme.json', 'templates/index.html', 'functions.php'):
         if not (root / name).is_file():
             raise ValueError('Missing ' + name)

skills/html2wp/assets/scripts/gutenberg-plan.py

@@ -53,11 +53,23 @@ def local_file(root, name):
     return result
 
 
+# The white space HTML collapses (space, tab, line feed, form feed, carriage
+# return). A non-breaking space ( , U+00A0) is text: it keeps a
+# headline's last two words on one line, and Python's \s would eat it.
+HTML_SPACE = re.compile(r'[ \t\n\f\r]+')
+
+
 def collapse(text):
-    """Source text as HTML lays it out: a run of whitespace is one space. An
+    """Source text as HTML lays it out: a run of white space is one space. An
     element's text is edited as rich text, where a source line break (the
     markup's indentation) is a real line break (rich() does the same)."""
-    return re.sub(r'\s+', ' ', text)
+    return HTML_SPACE.sub(' ', text)
+
+
+def line(text):
+    """Source text as one line of content, its ends trimmed (a title, a
+    card's summary); norm() compares, this is what WordPress stores."""
+    return collapse(text or '').strip(' ')
 
 
 class Node:
@@ -132,6 +144,8 @@ def plain(node):
 
 
 def norm(text):
+    """Text for comparing: every run of white space (a non-breaking space
+    too) is one space, so a headline's   still matches its card."""
     return ' '.join((text or '').split())
 
 
@@ -205,7 +219,7 @@ def classify(values, metas, card=False, static=False, tag=''):
         return 'postTitle', None
     if all(m.get('excerpt') and (v == norm(m['excerpt']) or (card and len(v) > 20 and norm(m['excerpt']).startswith(v.rstrip('.… ')))) for v, m in zip(values, metas)):
         return 'postExcerpt', None
-    if all(m.get('categories') and v in m['categories'] for v, m in zip(values, metas)):
+    if all(m.get('categories') and v in [norm(c) for c in m['categories']] for v, m in zip(values, metas)):
         return 'postTerms', None
     if all(m.get('readTime') and v == norm(m['readTime']) for v, m in zip(values, metas)):
         return 'postReadTime', None
@@ -379,7 +393,7 @@ def diff_walk(nodes, metas, overrides, values, mapper, native_title=True, stop=N
 
     def record(bind, fmt, per_instance):
         for store, value in zip(values or [], per_instance):
-            store.setdefault(bind, (norm(value), fmt))
+            store.setdefault(bind, (line(value), fmt))
 
     # A link that carries each post's own address or title (a share bar):
     # the live link carries the post being shown (bind postShare).
@@ -868,7 +882,7 @@ def rich(self, node, preserve=False):
             if not isinstance(child, Node):
                 # The editor's rich text is pre-wrap: source indentation and
                 # line breaks would show and re-wrap lines. HTML collapses them.
-                pieces.append(html.escape(child if preserve else re.sub(r'\s+', ' ', child), quote=False))
+                pieces.append(html.escape(child if preserve else collapse(child), quote=False))
                 continue
             if child.tag not in INLINE_TAGS:
                 return None
@@ -1400,13 +1414,13 @@ def query_block(self, node, targets, cards=None):
                     self.posts[key]['cardImage'] = image
             # The category a card prints is its post's category, when the
             # site uses that word as a category elsewhere too.
-            if values.get('postTerms') and not self.posts[key].get('cardTerms') and values['postTerms'][0].casefold() in self.terms:
+            if values.get('postTerms') and not self.posts[key].get('cardTerms') and norm(values['postTerms'][0]).casefold() in self.terms:
                 self.posts[key]['cardTerms'] = [values['postTerms'][0]]
         # A label only the cards print (no article, category link or topic
         # filter names it) fits a category as well as a badge or a person:
         # it stays bound, and it is not filed as the post's category.
         guessed = [(key, values['postTerms'][0]) for key, values in zip(targets, found)
-                   if values.get('postTerms') and not self.posts[key].get('categories') and values['postTerms'][0].casefold() not in self.terms]
+                   if values.get('postTerms') and not self.posts[key].get('categories') and norm(values['postTerms'][0]).casefold() not in self.terms]
         if guessed:
             self.finding('query-card-terms', 'Cards print a label no page names as a category (no article, category link or topic filter): ' + ', '.join(key + ' "' + term + '"' for key, term in guessed)[:300] + '; bound as the post categories, not recorded as them')
         # The lists that show the posts in their order (prepare reads the
@@ -1420,7 +1434,7 @@ def query_block(self, node, targets, cards=None):
         category = None
         if not lone and len(terms) == 1 and all(values.get('postTerms') for values in found):
             term = next(iter(terms))
-            members = {key for key, meta in self.posts.items() if term in (meta.get('categories') or meta.get('cardTerms') or [])}
+            members = {key for key, meta in self.posts.items() if norm(term) in [norm(c) for c in meta.get('categories') or meta.get('cardTerms') or []]}
             if members == set(targets) and len(members) < len(self.posts):
                 category = term
         for n in walk(cards[0]):
@@ -2504,7 +2518,7 @@ def __init__(self, entries, manifest):
         for key, entry in self.products.items():
             main = next((n for n, part, place in entry['items'] if not part and place == 'main' and isinstance(n, Node)), None)
             category = first_hint(entry['root'], self.shop.get('productCategory'))
-            self.meta[key] = {'name': entry['h1'], 'category': norm(plain(category)) if category is not None else ''}
+            self.meta[key] = {'name': norm(entry['h1']), 'category': norm(plain(category)) if category is not None else ''}
         self.categories = sorted({m['category'] for m in self.meta.values() if m['category']})
 
     def target(self, mapper, node):
@@ -3200,8 +3214,9 @@ def prepare(args):
     counter, lists = [0], []
     # html2wp's own recorder output rides along: the interaction runtime the
     # prerender emitted (never a source application script) and its entrance
-    # animation <style>. Source inline CSS stays a finding: it can be page
-    # specific and its cascade position matters.
+    # animation <style>. A source <style> block is page specific and its
+    # cascade position matters: it becomes a dist asset in that page's own
+    # head order (below).
     runtime_scripts, head_styles, head_boots = set(), {}, {}
     for page in pages:
         source = local_file(dist, page['file']).read_bytes()
@@ -3309,7 +3324,7 @@ def prepare(args):
         entries.append({'key': page['key'], 'page': page, 'kind': page.get('kind'), 'source': source, 'mapper': mapper, 'items': items, 'frame': frame, 'root': body,
                         'body_class': body.attrs.get('class', '') if body is not parser.root else '',
                         'sections': ['source-' + str(i + 1).zfill(4) for i in range(len(items))], 'title': title, 'description': description,
-                        'h1': norm(plain(heading)) if heading is not None else norm(page.get('title') or ''),
+                        'h1': line(plain(heading)) if heading is not None else line(page.get('title') or ''),
                         'urls': list(dict.fromkeys(u for u in urls if u and re.match(r'https?://', u)))})
     # Post metadata drives card/article classification (spec 2 D3/D6).
     metas = {}

skills/html2wp/assets/scripts/gutenberg-test-plan.py

@@ -1049,6 +1049,30 @@ def test_a_search_box_in_a_mail_form_is_its_text_field(self):
         self.assertIn(('unmapped-field', 'file'), items)
         self.assertIn(('unmapped-field', 'hidden'), items)
 
+    def test_a_non_breaking_space_is_text_not_layout(self):
+        #   keeps two words on one line (a headline's last two); HTML
+        # never collapses it, and neither may element or rich text.
+        self.assertEqual(planner.collapse('a\xa0 \n\t b'), 'a\xa0 b')
+        self.assertEqual(planner.line('  Notes on\xa0growth \n'), 'Notes on\xa0growth')
+        self.assertEqual(planner.norm('Notes on\xa0growth'), 'Notes on growth', 'compared, it is a space')
+        _, proposals, _ = self.plan({'index.html': '<body><div class="kicker">Step&nbsp;01</div><p>Keep the last two&nbsp;words</p>'
+                                                   '<h2>Notes on strategy &amp;&nbsp;growth</h2></body>'})
+        kicker, para, heading = proposals['index']['blocks']
+        self.assertEqual(kicker['attributes']['text'], 'Step\xa001')
+        self.assertEqual(para['attributes']['content'], 'Keep the last two\xa0words')
+        self.assertEqual(heading['attributes']['content'], 'Notes on strategy &amp;\xa0growth')
+
+    def test_a_headline_s_non_breaking_space_stays_in_the_post_title(self):
+        # The article's headline joins its last two words; the card does not.
+        # WordPress stores the headline's own title, and the card still binds
+        # it (compared, a non-breaking space is a space).
+        contract, proposals, findings = self.journal_site(headline_nbsp=True)
+        titles = {p['key']: p['title'] for p in planner.read(self.manifest)['pages'] if p['key'] in 'abcd'}
+        self.assertEqual(titles['a'], 'Openers that stop the\xa0scroll')
+        heading = next(b for b in planner.walk_blocks(contract['templates']['home']) if b['attributes'].get('tagName') == 'h2')
+        self.assertEqual(heading['attributes'].get('bind'), 'postTitle')
+        self.assertFalse([f for f in findings['journal'] if f['code'] == 'query-card-title'])
+
     def test_rich_text_collapses_source_whitespace_except_pre(self):
         _, proposals, _ = self.plan({'index.html': self.app(body='<p class="x">Every   program\n      is built</p><pre>a\n   b</pre>')})
         para, pre = proposals['index']['blocks'][1]['innerBlocks'][:2] if proposals['index']['blocks'][1].get('innerBlocks') else proposals['index']['blocks'][1:3]
@@ -1384,7 +1408,7 @@ def test_article_template_post_metadata_and_queries(self):
             self.run_cli('complete', '--task=' + task['id'], '--owner=test')
         self.run_cli('finalize')  # coverage stays complete and ordered
 
-    def journal_site(self, hint=False, short=('c',), article_category=True, separator='', second_page=False, topics=(), authors=False, share=False, pair='plain'):
+    def journal_site(self, hint=False, short=('c',), article_category=True, separator='', second_page=False, topics=(), authors=False, share=False, pair='plain', headline_nbsp=False):
         """Posts as one article sheet (head, image, prose, share bar) beside a
         prev/next pair and a call to action; a listing whose cards print the
         category as bare text before <time>, several posts on one day, a
@@ -1396,7 +1420,8 @@ def journal_site(self, hint=False, short=('c',), article_category=True, separato
         address, and its share bar carries it and its title. pair: 'plain'
         (newer then older, alike), 'sides' (← Previous is the older, Next →
         the newer, each its own class) or 'newest' (every article links the
-        newest other post)."""
+        newest other post). headline_nbsp: each article's headline keeps its
+        last two words together (&nbsp;), which its card does not."""
         posts = [('a', 'Openers that stop the scroll', 'Strategy', 'July 27, 2026'), ('b', 'A label tripled its reach', 'Case Notes', 'July 27, 2026'),
                  ('c', 'Affiliate done honestly', 'Strategy', 'July 26, 2026'), ('d', 'Three signals worth reading', 'Insights', 'July 26, 2026')]
 
@@ -1437,7 +1462,7 @@ def card(post, prefix=''):
                    '<a href="https://www.linkedin.com/sharing/share-offsite/?url=' + url_quote(canonical, safe='') + '">LinkedIn</a>'
                    '<a href="mailto:?subject=' + url_quote(title, safe='') + '&amp;body=' + url_quote(canonical, safe='') + '">Email</a><a class="self" href="' + canonical + '">Link</a>')
             files[key + '.html'] = page(
-                '<article class="sheet"><header class="head"><a class="back" href="journal.html">Blog</a><h1>' + title + '</h1>'
+                '<article class="sheet"><header class="head"><a class="back" href="journal.html">Blog</a><h1>' + ('&nbsp;'.join(title.rsplit(' ', 1)) if headline_nbsp else title) + '</h1>'
                 '<p class="dateline"><span>Published <time>' + date + '</time></span>' + ('<a class="cat" href="category/strategy.html">' + category + '</a>' if article_category else '') + '</p>' + writer(key) + '</header>'
                 '<div class="image"><img src="img/' + key + '.png" alt=""></div><div class="article-body">' + body + '</div>'
                 '<div class="share"><span>Share this</span>' + bar + '</div></article>'

skills/html2wp/assets/scripts/gutenberg-test-verdicts.py

@@ -86,6 +86,30 @@ def test_failures_travel_as_keys_only(self):
         self.assertEqual(gates['G-roundtrip']['failedKeys'], ['front-page', 'new-page'])
         self.assertNotIn('/unknown-route/', json.dumps(gates))
 
+    def test_a_serialization_row_that_does_not_write_back_fails_the_roundtrip(self):
+        # What the editor's first save writes back: a page by its key, a
+        # template or part by its literal template-<slug> / part-<slug>.
+        rows = [{'kind': 'page', 'id': 7, 'slug': 'front-page', 'path': '/', 'passed': True, 'byteIdentical': True, 'attributeLoss': []},
+                {'kind': 'page', 'id': 8, 'slug': 'about', 'path': '/about/', 'passed': False, 'byteIdentical': False, 'attributeLoss': []},
+                {'kind': 'templates', 'id': 'demo//single', 'passed': False, 'byteIdentical': True, 'attributeLoss': [{'key': 'foo'}]},
+                {'kind': 'template-parts', 'id': 'demo//header', 'passed': True, 'byteIdentical': True, 'attributeLoss': []}]
+        gates = self.payload(report(serialization=rows))
+        self.assertEqual((gates['G-roundtrip']['verdict'], gates['G-roundtrip']['failedKeys']), ('failed', ['about', 'template-single']))
+        self.assertEqual(gates['G-editor']['verdict'], 'passed', 'the editor gate reads its own rows')
+        for row in rows:
+            row['passed'] = True
+        self.assertEqual(self.payload(report(serialization=rows))['G-roundtrip'], {'gate': 'G-roundtrip', 'verdict': 'passed', 'pages': 2})
+        # Without the save/reload run a failing row still fails the gate; a passing one proves nothing.
+        rows[3]['passed'] = False
+        bare = report(serialization=rows)
+        for e in bare['editor']:
+            e.pop('roundtrip')
+        bare.pop('newPost'); bare.pop('newPage')
+        gate = self.payload(bare)['G-roundtrip']
+        self.assertEqual((gate['verdict'], gate['failedKeys']), ('failed', ['part-header']))
+        rows[3]['passed'] = True
+        self.assertEqual(self.payload(bare)['G-roundtrip']['verdict'], 'not-run')
+
     def test_absent_sections_are_not_passes(self):
         r = report(editorVisual=[], newPage=None, preview=None)
         r.pop('newPage'); r.pop('preview')