From 88e3fe9e2b0e7c0dcd93bd95fe811f92ca6fd1a9 Mon Sep 17 00:00:00 2001 From: shen Date: Fri, 30 Oct 2026 11:42:18 +0800 Subject: [PATCH] fix: preserve per-property inheritance and semantic emphasis --- docs/architecture.md | 2 +- docs/user_manual.md | 4 +- mdhelper/docx_semantic.py | 41 ++++++++------ mdhelper/style_sync.py | 18 +++++- tests/test_style_regressions.py | 97 +++++++++++++++++++++++++++++++++ tests/test_word_named_styles.py | 3 +- tests/word_styles_smoke.ps1 | 19 +++++++ 7 files changed, 163 insertions(+), 21 deletions(-) create mode 100644 tests/test_style_regressions.py diff --git a/docs/architecture.md b/docs/architecture.md index 6250771..9687b3c 100644 --- a/docs/architecture.md +++ b/docs/architecture.md @@ -31,7 +31,7 @@ `style_panel.py` 仅在用户打开时创建,按文档版本缓存语法分析,隐藏时停止刷新;`style_dialog.py` 的实际阅读预览也只在切换到该页时创建。启动路径不创建这些浏览器,不重建主窗口。 -`docx_style.py` 负责格式字段与 OOXML 属性的转换及属性顺序;`docx_semantic.py` 识别各引擎输出的语义、生成命名段落/字符/表格样式,并移除会遮盖样式的重复直接格式。保留编号、公式、关系和书签;组合字符与加粗段落上下文采用命名变体。`tests/word_styles_smoke.ps1` 使用本机 Word 的独立隐藏实例,验证打开文件、样式修改传播及继承,关闭时不保存测试修改。 +`docx_style.py` 负责格式字段与 OOXML 属性的转换及属性顺序;`docx_semantic.py` 识别各引擎输出的语义、生成命名段落/字符/表格样式,并按属性保留继承。保留编号、公式、关系和书签;除用于防止 Word 反向切换的明确字符字形(加粗、斜体、删除线)外,不重复写入直接字体、字号、颜色和段落格式。组合字符采用命名样式。`tests/word_styles_smoke.ps1` 使用本机 Word 的独立隐藏实例,验证打开文件、样式修改传播及继承,关闭时不保存测试修改。 `preferences.py` 定义 `ReaderPreferences` 和 `ApplicationConfiguration`。本地设置与 JSON 导入共用字段、取值检查和旧版迁移规则;`ConfigurationStore` 负责持久化。页面设置保留旧版 PDF 页边距的读取方式。窗口几何状态和草稿仍由各自的界面、文档组件管理。内置文档样式集中在 `style_presets.py`。 diff --git a/docs/user_manual.md b/docs/user_manual.md index ff43542..9573a2f 100644 --- a/docs/user_manual.md +++ b/docs/user_manual.md @@ -173,7 +173,9 @@ Markdown 支持: 三个 Word 导出引擎均创建 `MDHelper 正文`、`MDHelper 标题 1`、`MDHelper 加粗` 等命名样式。导出后,在 Word 的样式面板中修改相应样式,即可统一调整引用该样式的内容;正文的字体和字号也会传递到未单独设置字体的加粗、强调和链接文字。表格边框及内边距通过 Word 表格样式修改。 -Word 每段文字只能引用一个字符样式,因此加粗链接等组合会生成 `MDHelper 超链接+加粗` 等组合样式;组合优先以超链接为基准,其余效果保存在组合样式中。标题中加粗等情况会生成带段落名称的变体,以适配 Word 的加粗叠加规则。组合和变体都可以在 Word 样式面板统一修改,但不会同时继承多个字符样式。嵌套列表使用基于对应列表样式的层级样式,层级缩进可单独修改。 +Word 每段文字只能引用一个字符样式,因此加粗链接等组合会生成 `MDHelper 超链接+加粗` 等组合样式;组合不会同时继承多个字符样式。嵌套列表使用基于对应列表样式的层级样式,层级缩进可单独修改。 + +Word 的字符样式会把加粗、斜体和删除线作为与段落格式叠加的开关。导出采用强调保护:明确字符样式中的这些字形设置会固定在对应文字上,避免修改正文或标题样式时反向关闭强调。字体、字号、颜色、下划线和段落仍由命名样式统一控制;受保护的加粗、斜体和删除线不能仅通过 Word 字符样式关闭,可在 MDHelper 调整后重新导出,或在 Word 中清除对应文字的直接字形格式再使用其原生样式规则。 纸张宽度是全局阅读设置,内置样式不再写死正文宽度,会跟随纸张宽度变化。 diff --git a/mdhelper/docx_semantic.py b/mdhelper/docx_semantic.py index d74f8db..87c8398 100644 --- a/mdhelper/docx_semantic.py +++ b/mdhelper/docx_semantic.py @@ -44,15 +44,26 @@ def _delta(current, base, authored=None): 'keepNext': ('keep_with_next',), 'keepLines': ('keep_lines_together',), 'pageBreakBefore': ('page_break_before',), 'widowControl': ('widow_control',), 'contextualSpacing': ('contextual_spacing',)} for node in list(current): - if any(field in authored for field in fields.get(ET.QName(node).localname, ())): continue old = base.find(node.tag) - if old is not None and node.attrib == old.attrib and node.text == old.text and not len(node) and not len(old): - current.remove(node) - elif old is not None and node.tag in {W + 'rFonts', W + 'spacing', W + 'ind'}: + compound = { + 'rFonts': {'ascii': ('english_font_family',), 'hAnsi': ('english_font_family',), + 'cs': ('english_font_family',), 'eastAsia': ('chinese_font_family',)}, + 'spacing': {**{attr: (f'space_{side}_value', f'space_{side}_unit') + for side in ('before', 'after') for attr in (side, side + 'Lines', side + 'Autospacing')}, + 'line': ('line_height_value', 'line_spacing_mode'), 'lineRule': ('line_height_value', 'line_spacing_mode')}, + 'ind': {**{attr: (f'{side}_indent_value', f'{side}_indent_unit') + for side in ('left', 'right') for attr in (side, side + 'Chars')}, + **{attr: ('indent_mode', 'indent_value') for attr in ('firstLine', 'firstLineChars', 'hanging', 'hangingChars')}}, + }.get(ET.QName(node).localname) + if old is not None and compound and not (node.tag == W + 'spacing' and current.tag == W + 'rPr'): # These properties merge attribute by attribute in Word. for key in list(node.attrib): - if old.get(key) == node.get(key): del node.attrib[key] + explicit = any(field in authored for field in compound.get(ET.QName(key).localname, ())) + if not explicit and old.get(key) == node.get(key): del node.attrib[key] if not node.attrib and not len(node): current.remove(node) + elif not any(field in authored for field in fields.get(ET.QName(node).localname, ())): + if old is not None and node.attrib == old.attrib and node.text == old.text and not len(node) and not len(old): + current.remove(node) class WordStyles: @@ -158,17 +169,6 @@ class WordStyles: _run_properties(props, merged) _run_properties(old, self.character_rule(keys[0])) _delta(props, old) - # Word XORs style-level emphasis with the paragraph's emphasis. A named - # context variant compensates for that (e.g. **text** in a bold heading), - # keeping all formatting editable in the Styles panel, not on each run. - compensate = [field for field in ('bold', 'italic', 'strike') if field in merged and self.effective[paragraph_target].get(field)] - if compensate: - context_id = identifier + 'In' + style_id(paragraph_target)[8:] - if context_id not in self.created: - style = self.new(context_id, 'character', 'MDHelper ' + '+'.join(style_name(k, self.rules) for k in keys) + '(' + style_name(paragraph_target, self.rules) + ')', identifier) - for field in compensate: - for tag in {'bold': ('b', 'bCs'), 'italic': ('i', 'iCs'), 'strike': ('strike',)}[field]: val(child(style, 'rPr'), tag, int(not merged[field])) - identifier = context_id return identifier def syntax_style(self, run_id, target): @@ -311,6 +311,15 @@ class WordStyles: remove_fields(rpr, RUN_FIELDS) if identifier: val(rpr, 'rStyle', identifier) else: remove_fields(rpr, ('rStyle',)) + # Word character-style emphasis XORs with paragraph emphasis. Protect + # only explicit semantic emphasis; fonts, sizes, colors and spacing stay + # style-driven. This is an intentional export contract, not a style copy. + merged = {} + for key in ('link', 'strong', 'emphasis', 'strike', 'inline_code'): + if key in keys: merged.update(self.character_rule(key)) + for field, tags in (('bold', ('b', 'bCs')), ('italic', ('i', 'iCs')), ('strike', ('strike',))): + if field in merged: + for tag in tags: val(rpr, tag, int(merged[field])) def apply(self, document): for paragraph in document.iter(W + 'p'): diff --git a/mdhelper/style_sync.py b/mdhelper/style_sync.py index ac746ab..448ba93 100644 --- a/mdhelper/style_sync.py +++ b/mdhelper/style_sync.py @@ -33,7 +33,7 @@ PROPERTIES = { "fonts": {"font-family", "--mdhelper-font-chinese", "--mdhelper-font-english"}, "size": {"font-size"}, "bold": {"font-weight"}, "italic": {"font-style"}, "color": {"color"}, "tracking": {"letter-spacing"}, "align": {"text-align"}, - 'decoration': {'text-decoration', 'text-decoration-line'}, 'background': {'background-color'}, + 'decoration': {'text-decoration', 'text-decoration-line', '--mdhelper-underline', '--mdhelper-strike'}, 'background': {'background-color'}, 'border_color': {'border-color'}, 'border_width': {'border-width', 'border-style'}, 'padding': {'padding'}, 'marker': {'list-style-type'}, 'level_indent': {'--mdhelper-list-level-indent'}, "line": {"line-height", "--mdhelper-line-spacing-mode"}, "indent": {"text-indent"}, @@ -62,6 +62,7 @@ ALIASES.update({'strong': 'strong', 'em': 'emphasis', 's': 'strike', 'del': 'str ALIASES[_normalize_selector_text(SELECTORS["paragraph"])] = "paragraph" FONT_VARIABLES = {"chinese_font_family": "--mdhelper-font-chinese", "english_font_family": "--mdhelper-font-english"} FONT_EXPRESSION = "var(--mdhelper-font-english, var(--mdhelper-font-chinese, var(--reader-font-family))), var(--mdhelper-font-chinese, sans-serif)" +DECORATION_EXPRESSION = 'var(--mdhelper-text-decoration, none)' @dataclass @@ -301,6 +302,10 @@ class _CoreRuleReader: return self.rule def decoration(self, value): + if re.sub(r'\s+', '', value) == re.sub(r'\s+', '', DECORATION_EXPRESSION): + for key in ('underline', 'strike'): + self.choice('--mdhelper-' + key, key, {'true': True, 'false': False}) + return parts = value.split() if not parts or not set(parts) <= {'none', 'underline', 'line-through'}: raise ValueError self.rule.update(underline='underline' in parts, strike='line-through' in parts) @@ -426,7 +431,10 @@ def analyze_css(source: str, font_size_px: float = 16) -> Analysis: def _core_declarations(rule): - declarations = [d for d in css_declarations(rule) if not d.startswith(("font-family:", "padding-left:"))] + declarations = [d for d in css_declarations(rule) if not d.startswith(("font-family:", "padding-left:", 'text-decoration-line:'))] + if 'underline' in rule or 'strike' in rule: + declarations.extend(f'--mdhelper-{key}: {str(rule[key]).lower()};' for key in ('underline', 'strike') if key in rule) + declarations.append(f'text-decoration-line: {DECORATION_EXPRESSION};') if any(rule.get(key) for key in FONT_VARIABLES): for key, variable in FONT_VARIABLES.items(): if rule.get(key): @@ -541,6 +549,12 @@ def render_css(source: str, rules: dict, font_family: str = '', font_size_px: fl contextual = {key: {"contextual_spacing": True} for key, rule in rules.items() if isinstance(rule, dict) and rule.get("contextual_spacing")} helpers = build_document_style_css(contextual) helpers += '\n' + build_document_style_css(inherited) + # Independent flags retain their inherited counterpart. Literal CSS such as + # text-decoration-line:none still wins over this opt-in variable expression. + for key, rule in effective.items(): + if any(field in rules.get(key, {}) or field in inherited.get(key, {}) for field in ('underline', 'strike')): + decoration = ' '.join(token for field, token in (('underline', 'underline'), ('strike', 'line-through')) if rule.get(field)) or 'none' + helpers += '\n' + SELECTORS[key] + ' { --mdhelper-text-decoration: ' + decoration + '; }' paragraph = rules.get("paragraph", {}) if paragraph.get("font_size"): helpers += f"\n.reader {{ --reader-math-font-size: {fmt(paragraph['font_size'])}{paragraph.get('font_size_unit', 'px')}; }}" diff --git a/tests/test_style_regressions.py b/tests/test_style_regressions.py new file mode 100644 index 0000000..0f1be5d --- /dev/null +++ b/tests/test_style_regressions.py @@ -0,0 +1,97 @@ +import json +import time +import zipfile +from lxml import etree as ET +import pytest +from PySide6.QtCore import Qt +from PySide6.QtWebEngineCore import QWebEnginePage +from mdhelper.docx_exporter import export_markdown_to_docx +from mdhelper.docx_style import W, NS, val +from mdhelper.style_dialog import DocumentStyleEditorDialog +from mdhelper.style_sync import normalize_payload, render_css +from mdhelper.markdown_renderer import MarkdownRenderer +from docx_assertions import effective_document + +PARTIAL_RULES = { + 'h1': {'chinese_font_family': 'SimSun', 'english_font_family': 'Arial', 'font_size': 12, 'font_size_unit': 'pt', + 'space_after_value': 8, 'space_after_unit': 'pt', 'line_height_value': 1.5, 'line_spacing_mode': 'multiple', + 'left_indent_value': 24, 'left_indent_unit': 'pt'}, + 'h2': {'english_font_family': 'Courier New', 'space_before_value': 12, 'space_before_unit': 'pt', + 'right_indent_value': 8, 'right_indent_unit': 'pt'}, + '_metadata': {'h2': {'based_on': 'h1'}}, +} + + +def parts(path): + with zipfile.ZipFile(path) as z: + return ET.fromstring(z.read('word/document.xml')), ET.fromstring(z.read('word/styles.xml')) + + +def browser(qapp, tmp_path, payload, source, expression): + page = QWebEnginePage(); loaded = []; result = [] + def wait(predicate): + end = time.monotonic() + 12 + while not predicate() and time.monotonic() < end: qapp.processEvents(); time.sleep(.01) + assert predicate(), 'WebEngine timed out' + try: + page.loadFinished.connect(loaded.append) + page.setHtml(MarkdownRenderer(tmp_path).render_source(source, tmp_path / 'test.md', + document_style_css=render_css(payload['css'], payload['rules']))) + wait(lambda: loaded) + page.runJavaScript('JSON.stringify(' + expression + ')', result.append) + wait(lambda: result) + return json.loads(result[0]) + finally: page.deleteLater(); qapp.processEvents() + + +@pytest.mark.parametrize('engine', ['builtin', 'python-docx', 'pandoc']) +def test_word_inherits_other_attributes_when_one_attribute_is_explicit(tmp_path, engine): + path = export_markdown_to_docx('# Parent\n\n## Child', tmp_path / 'partial.docx', engine=engine, style_rules=PARTIAL_RULES) + doc, styles = parts(path) + base = styles.find("w:style[@w:styleId='MDHelperH1']", NS) + child = styles.find("w:style[@w:styleId='MDHelperH2']", NS) + assert child.find('w:rPr/w:rFonts', NS).get(W + 'eastAsia') is None + assert child.find('w:rPr/w:rFonts', NS).get(W + 'ascii') == 'Courier New' + assert set(child.find('w:pPr/w:spacing', NS).attrib) == {W + 'before'} + assert set(child.find('w:pPr/w:ind', NS).attrib) == {W + 'right'} + base.find('w:rPr/w:rFonts', NS).set(W + 'eastAsia', 'KaiTi') + base.find('w:pPr/w:spacing', NS).set(W + 'after', '480') + rendered = effective_document(doc, styles) + paragraph = next(p for p in rendered.iter(W + 'p') if 'Child' in ''.join(p.itertext())) + assert paragraph.find('.//w:rFonts', NS).get(W + 'eastAsia') == 'KaiTi' + assert paragraph.find('w:pPr/w:spacing', NS).get(W + 'after') == '480' + + +@pytest.mark.parametrize('engine', ['builtin', 'python-docx', 'pandoc']) +def test_word_semantic_emphasis_survives_paragraph_style_changes(tmp_path, engine): + path = export_markdown_to_docx('Plain **StrongText** *ItalicText* ~~Deleted~~\n\n# Title **HeadingStrong**', tmp_path / 'emphasis.docx', engine=engine) + doc, styles = parts(path) + base = styles.find("w:style[@w:styleId='MDHelperParagraph']/w:rPr", NS) + for tag in ('b', 'i', 'strike'): val(base, tag, 1) + val(styles.find("w:style[@w:styleId='MDHelperH1']/w:rPr", NS), 'b', 0) + rendered = effective_document(doc, styles) + for text, tag in (('StrongText', 'b'), ('HeadingStrong', 'b'), ('ItalicText', 'i'), ('Deleted', 'strike')): + run = next(r for r in rendered.iter(W + 'r') if text in ''.join(r.itertext())) + assert run.find('w:rPr/w:' + tag, NS).get(W + 'val') == '1' + for run in doc.iter(W + 'r'): + assert run.find('w:rPr/w:sz', NS) is None + assert run.find('w:rPr/w:rFonts', NS) is None + assert run.find('w:rPr/w:color', NS) is None + + +@pytest.mark.parametrize('key,changed,source,selector', [('strike', 'underline', '~~Text~~', 's'), ('link', 'strike', '[Text](#text)', 'p a')]) +def test_graphical_decoration_preserves_inherited_counterpart(qt_application, tmp_path, key, changed, source, selector): + dialog = DocumentStyleEditorDialog({}) + try: + dialog.select_target(key) + dialog.bool_boxes[changed].setCheckState(Qt.CheckState.Checked) + payload = dialog.style_payload() + assert payload['rules'][key] == {changed: True} + assert normalize_payload(payload)['rules'][key] == {changed: True} + actual = browser(qt_application, tmp_path, payload, source, f'getComputedStyle(document.querySelector("{selector}")).textDecorationLine') + assert set(actual.split()) == {'underline', 'line-through'} + # Explicitly switching the other property off is still respected. + other = 'strike' if changed == 'underline' else 'underline' + dialog.bool_boxes[other].setCheckState(Qt.CheckState.Unchecked) + assert dialog.style_payload()['rules'][key][other] is False + finally: dialog.close() diff --git a/tests/test_word_named_styles.py b/tests/test_word_named_styles.py index 01e06b8..6eb2629 100644 --- a/tests/test_word_named_styles.py +++ b/tests/test_word_named_styles.py @@ -29,7 +29,8 @@ def test_shared_named_styles_without_direct_format_overrides(tmp_path, engine): assert ppr.find(W + 'pStyle').get(W + 'val') in by_id for run in doc.iter(W + 'r'): rpr = run.find(W + 'rPr') - assert all(rpr.find(W + field) is None for field in RUN_FIELDS) + protected = {'b', 'bCs', 'i', 'iCs', 'strike'} + assert all(rpr.find(W + field) is None for field in set(RUN_FIELDS) - protected) ref = rpr.find(W + 'rStyle') if ref is not None: assert ref.get(W + 'val') in by_id paragraph_ids = {p.find('w:pPr/w:pStyle', NS).get(W + 'val') for p in doc.iter(W + 'p')} diff --git a/tests/word_styles_smoke.ps1 b/tests/word_styles_smoke.ps1 index ab7facb..77655b7 100644 --- a/tests/word_styles_smoke.ps1 +++ b/tests/word_styles_smoke.ps1 @@ -11,10 +11,12 @@ sys.path.insert(0, str(Path('tests').resolve())) from test_word_named_styles import SOURCE from mdhelper.docx_exporter import export_markdown_to_docx from mdhelper.style_catalog import style_name +from test_style_regressions import PARTIAL_RULES folder = Path(tempfile.mkdtemp(prefix='MDHelper-Word-Styles-')) rules = {'h2': {'bold': False}, '_metadata': {'h2': {'based_on': 'h1'}}} for engine in ('builtin', 'python-docx', 'pandoc'): export_markdown_to_docx(SOURCE + '\n\n## Child title', folder / (engine + '.docx'), engine=engine, style_rules=rules) + export_markdown_to_docx('# Parent\n\n## Child', folder / (engine + '-partial.docx'), engine=engine, style_rules=PARTIAL_RULES) (folder / 'names.json').write_text(json.dumps({key: 'MDHelper ' + style_name(key) for key in ('paragraph', 'strong', 'link', 'h1')}), encoding='utf-8') print(folder) '@ | & $Python -B -).Trim() @@ -44,8 +46,25 @@ print(folder) if ($range.Font.Bold -ne 0) { throw "$engine explicit child-style bold=False did not override its base." } $doc.Styles.Item($names.h1).Font.Size = 30 if ($range.Font.Size -ne 30) { throw "$engine base-style change does not reach child heading." } + $range = $doc.Content.Duplicate; [void]$range.Find.Execute('strong second') + $doc.Styles.Item($names.paragraph).Font.Bold = -1 + if ($range.Font.Bold -ne -1) { throw "$engine paragraph bold inverted semantic emphasis." } + $range = $doc.Content.Duplicate; [void]$range.Find.Execute('bold title') + $doc.Styles.Item($names.h1).Font.Bold = 0 + if ($range.Font.Bold -ne -1) { throw "$engine heading edit removed semantic emphasis." } Write-Output "$engine passed: Word opened document, heading emphasis, body and character edits, mixed links, base-style inheritance." } finally { $doc.Close(0); [void][Runtime.InteropServices.Marshal]::FinalReleaseComObject($doc) } + $doc = $word.Documents.Open((Join-Path $testDirectory "$engine-partial.docx"), $false, $true) + try { + $range = $doc.Content.Duplicate; [void]$range.Find.Execute('Child') + $base = $doc.Styles.Item($names.h1) + $base.Font.NameFarEast = 'KaiTi'; $base.ParagraphFormat.SpaceAfter = 24 + $base.ParagraphFormat.LineSpacingRule = 0; $base.ParagraphFormat.LeftIndent = 40 + if ($range.Font.NameFarEast -ne $base.Font.NameFarEast -or $range.ParagraphFormat.SpaceAfter -ne 24 -or $range.ParagraphFormat.LineSpacingRule -ne 0 -or $range.ParagraphFormat.LeftIndent -ne 40) { + throw "$engine partial overrides froze unrelated inherited properties." + } + Write-Output "$engine passed: partial overrides preserve inheritance of fonts, spacing and indents." + } finally { $doc.Close(0); [void][Runtime.InteropServices.Marshal]::FinalReleaseComObject($doc) } } } finally { if ($word) { $word.Quit(); [void][Runtime.InteropServices.Marshal]::FinalReleaseComObject($word) }