@lark-apaas/coding-steering 0.1.32-dev.a87aa13 → 0.1.32-dev.f6e7ce8

package/steering/design-html/skills/pptx-style-extract/scripts/verify_layout_assets.py

@@ -0,0 +1,400 @@
+#!/usr/bin/env python3
+"""Verify that generated slides honor every asset bound to their PPTX layout."""
+import argparse
+import hashlib
+import os
+import posixpath
+import re
+import sys
+from collections import Counter
+from html.parser import HTMLParser
+from urllib.parse import urlsplit
+
+from check_v2 import Pack
+
+
+URL_RE = re.compile(r'url\(\s*[\'"]?([^\'")\s]+)', re.I)
+VOID_TAGS = {
+    'area', 'base', 'br', 'col', 'embed', 'hr', 'img', 'input',
+    'link', 'meta', 'param', 'source', 'track', 'wbr',
+}
+
+
+def normalized_path(value):
+    path = urlsplit(value).path.replace('\\', '/')
+    return posixpath.normpath(path).lstrip('./')
+
+
+def bound_asset_ids(value, known_asset_ids):
+    found = Counter()
+    if isinstance(value, dict):
+        for key, child in value.items():
+            if (key in ('asset', 'background')
+                    and isinstance(child, str)
+                    and child in known_asset_ids):
+                found[child] += 1
+            found.update(bound_asset_ids(child, known_asset_ids))
+    elif isinstance(value, list):
+        for child in value:
+            found.update(bound_asset_ids(child, known_asset_ids))
+    return found
+
+
+def layout_asset_contract(pack):
+    known_asset_ids = set(pack.assets)
+    owners = {}
+    for layout_name, (layout, _) in pack.layouts.items():
+        for asset_id, _, _ in layout_asset_instances(
+                layout, known_asset_ids, pack.canvas):
+            owners.setdefault(asset_id, set()).add(layout_name)
+    return owners
+
+
+def asset_urls(pack, asset_prefix, asset_ids):
+    prefix = normalized_path(asset_prefix).rstrip('/')
+    urls = {}
+    for asset_id in asset_ids:
+        entry, _ = pack.assets.get(asset_id, (None, None))
+        if not isinstance(entry, dict):
+            continue
+        path = entry.get('path')
+        if not isinstance(path, str) or not path:
+            continue
+        relative = normalized_path(path)
+        if relative.startswith('assets/'):
+            relative = relative[len('assets/'):]
+        urls.setdefault(posixpath.join(prefix, relative), set()).add(asset_id)
+    return urls
+
+
+def bound_asset_instances(value, known_asset_ids):
+    """Return every positioned fixed-asset instance nested in slots or flow."""
+    instances = []
+    if isinstance(value, dict):
+        box = value.get('box')
+        asset_id = value.get('asset')
+        if (isinstance(box, list) and len(box) == 4
+                and isinstance(asset_id, str)
+                and asset_id in known_asset_ids):
+            instances.append((asset_id, value.get('role') or 'asset', box))
+        for key, child in value.items():
+            if key not in ('asset', 'background'):
+                instances.extend(bound_asset_instances(child, known_asset_ids))
+    elif isinstance(value, list):
+        for child in value:
+            instances.extend(bound_asset_instances(child, known_asset_ids))
+    return instances
+
+
+def layout_asset_instances(layout, known_asset_ids, canvas):
+    """Return every fixed image instance declared by one layout."""
+    if not isinstance(layout, dict):
+        return []
+    instances = []
+    background = layout.get('background')
+    if canvas and isinstance(background, str) and background in known_asset_ids:
+        instances.append(
+            (background, 'background', [0, 0, canvas[0], canvas[1]]))
+    instances.extend(bound_asset_instances(layout, known_asset_ids))
+    return instances
+
+
+def inline_styles(value):
+    styles = {}
+    for declaration in (value or '').split(';'):
+        if ':' not in declaration:
+            continue
+        name, raw = declaration.split(':', 1)
+        styles[name.strip().lower()] = re.sub(
+            r'\s*!important\s*$', '', raw.strip().lower())
+    return styles
+
+
+def css_number(value):
+    match = re.fullmatch(r'(-?(?:\d+(?:\.\d*)?|\.\d+))(?:px)?', value or '')
+    return float(match.group(1)) if match else None
+
+
+def element_box(reference, canvas):
+    styles = reference['styles']
+    if reference['slide_root'] and not any(
+            name in styles for name in ('left', 'top', 'width', 'height')):
+        return [0.0, 0.0, float(canvas[0]), float(canvas[1])]
+    values = [
+        css_number(styles.get(name))
+        for name in ('left', 'top', 'width', 'height')
+    ]
+    if styles.get('position') != 'absolute' or any(
+            value is None for value in values):
+        return None
+    return values
+
+
+def boxes_match(actual, expected, tolerance=1.0):
+    return actual is not None and all(
+        abs(actual_value - expected_value) <= tolerance
+        for actual_value, expected_value in zip(actual, expected)
+    )
+
+
+def file_sha256(path):
+    digest = hashlib.sha256()
+    with open(path, 'rb') as stream:
+        for chunk in iter(lambda: stream.read(1024 * 1024), b''):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def copied_asset_problems(pack, html_path, asset_prefix, asset_ids):
+    """Verify copied fixed assets still contain the source PPTX bytes."""
+    problems = []
+    html_root = os.path.dirname(os.path.abspath(html_path))
+    prefix = normalized_path(asset_prefix).rstrip('/')
+    for asset_id in sorted(asset_ids):
+        entry, _ = pack.assets.get(asset_id, (None, None))
+        path = entry.get('path') if isinstance(entry, dict) else None
+        if not isinstance(path, str) or not path:
+            continue
+        relative = normalized_path(path)
+        if relative.startswith('assets/'):
+            relative = relative[len('assets/'):]
+        source_path = os.path.join(pack.root, 'assets', *relative.split('/'))
+        copied_relative = posixpath.join(prefix, relative)
+        copied_path = os.path.join(html_root, *copied_relative.split('/'))
+        if not os.path.isfile(copied_path):
+            problems.append(
+                '固定素材 %s 未复制到项目: %s' % (asset_id, copied_relative))
+            continue
+        if not os.path.isfile(source_path):
+            problems.append(
+                '风格包中的固定素材 %s 不存在: %s' % (asset_id, path))
+            continue
+        if file_sha256(copied_path) != file_sha256(source_path):
+            problems.append(
+                '固定素材 %s 已被替换或改写，必须使用 PPTX 原文件' % asset_id)
+    return problems
+
+
+def urls_from_attrs(attrs):
+    urls = []
+    for key, value in attrs:
+        if not value:
+            continue
+        if key.lower() in ('src', 'href'):
+            urls.append(value)
+        elif key.lower() == 'srcset':
+            urls.extend(item.strip().split(' ', 1)[0] for item in value.split(','))
+        elif key.lower() == 'style':
+            urls.extend(URL_RE.findall(value))
+    return urls
+
+
+class SlideAssetParser(HTMLParser):
+    def __init__(self):
+        super().__init__()
+        self.slides = []
+        self._stack = []
+        self._style_depth = 0
+        self.outside_urls = []
+
+    def handle_starttag(self, tag, attrs):
+        tag = tag.lower()
+        attrs_map = {key.lower(): value for key, value in attrs}
+        urls = urls_from_attrs(attrs)
+        parent = self._stack[-1] if self._stack else None
+        parent_slide = parent['slide'] if parent else None
+        slide_root = bool(
+            tag == 'section' and parent and parent['tag'] == 'deck-stage')
+        slide = ({
+            'layout': attrs_map.get('data-pptx-layout'),
+            'references': [],
+        } if slide_root else parent_slide)
+        if slide_root:
+            self.slides.append(slide)
+        styles = inline_styles(attrs_map.get('style'))
+        hidden = bool(
+            (parent and parent['hidden'])
+            or 'hidden' in attrs_map
+            or attrs_map.get('aria-hidden', '').lower() == 'true'
+            or styles.get('display') == 'none'
+            or styles.get('visibility') in ('hidden', 'collapse')
+            or styles.get('content-visibility') == 'hidden'
+            or css_number(styles.get('width')) == 0
+            or css_number(styles.get('height')) == 0
+            or (
+                css_number(styles.get('opacity')) is not None
+                and css_number(styles.get('opacity')) <= 0
+            )
+        )
+        if slide is not None and (urls or attrs_map.get('data-pptx-asset')):
+            slide['references'].append({
+                'asset': attrs_map.get('data-pptx-asset'),
+                'hidden': hidden,
+                'slide_root': slide_root,
+                'source': attrs_map.get('src'),
+                'styles': styles,
+                'tag': tag,
+                'urls': urls,
+            })
+        elif urls:
+            self.outside_urls.extend(urls)
+        if tag == 'style':
+            self._style_depth += 1
+        if tag not in VOID_TAGS:
+            self._stack.append({
+                'hidden': hidden,
+                'slide': slide,
+                'tag': tag,
+            })
+
+    def handle_startendtag(self, tag, attrs):
+        self.handle_starttag(tag, attrs)
+        if tag.lower() not in VOID_TAGS:
+            self.handle_endtag(tag)
+
+    def handle_endtag(self, tag):
+        tag = tag.lower()
+        if tag == 'style' and self._style_depth:
+            self._style_depth -= 1
+        if tag not in VOID_TAGS and self._stack:
+            self._stack.pop()
+
+    def handle_data(self, data):
+        if not self._style_depth:
+            return
+        self.outside_urls.extend(URL_RE.findall(data))
+
+
+def validate_layout_assets(pack_dir, html_path, asset_prefix):
+    """Return violations of the asset contract declared by each layout."""
+    pack = Pack(pack_dir)
+    owners = layout_asset_contract(pack)
+    known_urls = asset_urls(pack, asset_prefix, owners)
+    if not known_urls:
+        return []
+    asset_urls_by_id = {
+        asset_id: url
+        for url, asset_ids in known_urls.items()
+        for asset_id in asset_ids
+    }
+
+    with open(html_path, encoding='utf-8') as stream:
+        text = stream.read()
+    parser = SlideAssetParser()
+    parser.feed(text)
+    parser.close()
+
+    problems = copied_asset_problems(pack, html_path, asset_prefix, owners)
+    for url in parser.outside_urls:
+        for asset_id in sorted(known_urls.get(normalized_path(url), ())):
+            problems.append(
+                '模板资产 %s 出现在 slide section 外，无法核验页型归属' % asset_id)
+    for number, slide in enumerate(parser.slides, 1):
+        layout = slide['layout']
+        if not layout:
+            problems.append('第 %d 页缺少 data-pptx-layout，无法核验模板资产归属' % number)
+            continue
+        if layout not in pack.layouts:
+            problems.append('第 %d 页声明了不存在的模板页型: %s' % (number, layout))
+            continue
+        expected = layout_asset_instances(
+            pack.layouts[layout][0], set(pack.assets), pack.canvas)
+        actual = []
+        for reference in slide['references']:
+            referenced_ids = set()
+            normalized_urls = [normalized_path(url) for url in reference['urls']]
+            for url in normalized_urls:
+                referenced_ids.update(known_urls.get(url, ()))
+            asset_id = reference['asset']
+            if not asset_id:
+                for referenced_id in sorted(referenced_ids):
+                    problems.append(
+                        '第 %d 页模板资产 %s 缺少 data-pptx-asset 实例标记'
+                        % (number, referenced_id))
+                continue
+            if asset_id not in pack.assets:
+                problems.append(
+                    '第 %d 页声明了不存在的模板资产: %s' % (number, asset_id))
+                continue
+            expected_url = asset_urls_by_id.get(asset_id)
+            uses_expected_source = (
+                len(normalized_urls) == 1
+                and normalized_urls[0] == expected_url
+                and (
+                    reference['tag'] != 'img'
+                    or (
+                        reference['source']
+                        and normalized_path(reference['source']) == expected_url
+                    )
+                )
+            )
+            if not uses_expected_source:
+                problems.append(
+                    '第 %d 页固定实例 %s 未引用对应的 PPTX 原素材'
+                    % (number, asset_id))
+            if reference['hidden']:
+                problems.append(
+                    '第 %d 页固定实例 %s 不可隐藏' % (number, asset_id))
+            actual.append({
+                'asset': asset_id,
+                'box': element_box(reference, pack.canvas),
+            })
+
+        used_asset_counts = Counter(instance['asset'] for instance in actual)
+        for asset_id in sorted(used_asset_counts):
+            if layout not in owners.get(asset_id, set()):
+                allowed = '、'.join(sorted(owners.get(asset_id) or ())) or '（无）'
+                problems.append(
+                    '第 %d 页页型 %s 不得使用 %s；只允许: %s'
+                    % (number, layout, asset_id, allowed))
+        unmatched = list(actual)
+        for asset_id, _, expected_box in expected:
+            matching_index = next((
+                index for index, instance in enumerate(unmatched)
+                if instance['asset'] == asset_id
+                and boxes_match(instance['box'], expected_box)
+            ), None)
+            if matching_index is not None:
+                unmatched.pop(matching_index)
+                continue
+            same_asset = next((
+                instance for instance in unmatched
+                if instance['asset'] == asset_id
+            ), None)
+            if same_asset:
+                problems.append(
+                    '第 %d 页固定实例 %s 的位置尺寸必须为 %s，当前为 %s'
+                    % (number, asset_id, expected_box, same_asset['box']))
+                unmatched.remove(same_asset)
+            else:
+                problems.append(
+                    '第 %d 页页型 %s 缺少固定实例 %s，位置尺寸应为 %s'
+                    % (number, layout, asset_id, expected_box))
+        for instance in unmatched:
+            if layout in owners.get(instance['asset'], set()):
+                problems.append(
+                    '第 %d 页页型 %s 额外使用了固定实例 %s'
+                    % (number, layout, instance['asset']))
+    return problems
+
+
+def main(argv=None):
+    parser = argparse.ArgumentParser(
+        description='Verify that generated deck HTML honors PPTX layout asset bindings.')
+    parser.add_argument('pack_dir')
+    parser.add_argument('html_path')
+    parser.add_argument('--asset-prefix', required=True)
+    args = parser.parse_args(argv)
+
+    problems = validate_layout_assets(args.pack_dir, args.html_path, args.asset_prefix)
+    if not problems:
+        print('PPTX_LAYOUT_ASSETS: PASS')
+        return 0
+    print('PPTX_LAYOUT_ASSETS: FAIL count=%d' % len(problems))
+    for problem in problems:
+        print('[layoutAssets] %s' % problem)
+    return 1
+
+
+if __name__ == '__main__':
+    sys.exit(main())

package/steering/design-html/skills/pptx-style-extract/scripts/verify_logo_scope.py

@@ -0,0 +1,12 @@
+#!/usr/bin/env python3
+"""Backward-compatible entrypoint for PPTX layout asset verification."""
+import sys
+
+from verify_layout_assets import main, validate_layout_assets
+
+
+validate_logo_scope = validate_layout_assets
+
+
+if __name__ == '__main__':
+    sys.exit(main())

package/steering/design-html/skills/pptx-style-extract/v2-format-spec.md

@@ -133,7 +133,7 @@ layouts:
 
 - **`background` 三形态**：`<asset-id>` / `{<theme>: <asset-id>}` / `{color: <colors-token>}`（`color` 是保留键，主题名禁止叫 color）。`asset` 两形态：`<asset-id>` / `{<theme>: <asset-id>}`。
 - **背景安全扩展**：有真实背景图的 archetype 建议写 `text_safe: [x,y,w,h]`、`avoid: [{box: [x,y,w,h], reason: "..."}]`、`pairing_rule: "..."`。这些是消费约束，不参与封闭枚举；用于避免标题、正文、图表、卡片、表格、时间线及其容器外接矩形覆盖背景视觉主体、强光斑或深色透明区；透明容器也不能跨进禁放区。
-- **流式页型**：内容长度会变化的内容页可用 `flow.regions` 表达纵向区带。`stack` 表达单列顺序，`grid` 表达并列列组，`free` 中的 item 必须带 `box`，用于 logo、页码、页眉和页脚等固定锚点。并列卡片可在 `grid.items` 中使用一层 `{role: group, css, gap, items}`：group 的 `css` 是卡片容器样式，内部 `items` 按顺序排布；不继续嵌套 group。纵向位置与留白由消费模型结合实际内容决定，不把样张的 `y` 坐标当作流式硬约束。
+- **流式页型**：内容长度会变化的内容页可用 `flow.regions` 表达纵向区带。`stack` 表达单列顺序，`grid` 表达并列列组，`free` 中的 item 必须带 `box`，用于 logo、页码、页眉和页脚等固定锚点。并列卡片可在 `grid.items` 中使用一层 `{role: group, css, gap, items}`：group 的 `css` 是卡片容器样式，内部 `items` 按顺序排布；不继续嵌套 group。区带可带自己的 `margin: [左, 右]`，覆盖 `flow` 整块的 `margin`（居中卡片组和贴左标题横向范围本就不同）；不带则继承整块 `margin`。纵向位置与留白由消费模型结合实际内容决定，不把样张的 `y` 坐标当作流式硬约束。
 - **`decor`（可选）**：这一页无文字的图形骨架——图标托底的圆、卡片、分隔线。每条 `{box, geom, css}`：`box` 定位，`css` 是可直接写进 style 的声明串，`geom` 取源形状的 prst（`ellipse` 另加 `border-radius: 50%`）。圆角以每条 `css` 为准，没有 `border-radius` 就按 `0`；不得因 `geom: roundRect` 自行补圆角，因为 OOXML 的 roundRect 可以有零圆角调节点。层级在背景之上、`slots` 之下；带 `asset` 的槽落在 decor 之上是版式本意，不算重叠。
 - **slot 样式契约**：`box` 只承载 `[x,y,w,h]` 几何；可渲染属性统一放进 `css`，并可直接写入 HTML `style`。PPTX `bodyPr.insets_px` 转成 `box-sizing: border-box; padding: ...`，字号/字重/颜色/水平与垂直对齐/行高/字距/旋转分别转成标准 CSS。禁止在 slot 中输出 `size` / `weight` / `color` / `align` / `valign` / `insets_px` 等旧字段。
 - **文本角色判断**：脚本把实例页及其引用版式中的现有文本槽、几何和 CSS 完整写入草案；`text_roles` 只供模型把这些槽判断为 `title | subtitle | header | footer | body`，不控制槽位去留。判断不清时用 `body`，不归纳模板中不存在的标题、页眉或页脚。