@godcreator/word-report 0.4.0 → 0.5.0

package/engine/src/docreport/compose.py

@@ -7,7 +7,8 @@ Word objects and deferred MathType/field placeholders.
 from pathlib import Path
 import re
 
-from docreport.word import utf16_length, find_text
+from docreport.plain_math import INLINE_MATH_STYLE
+from docreport.word import utf16_length, find_text, MarkerFinder
 
 
 def mm(value):
@@ -179,23 +180,21 @@ class Composer:
         if self._finished:
             raise RuntimeError('Composer.finish() may only be called once.')
         doc = self.doc
+        fields = MarkerFinder(doc)
         for token, code, bookmark in self._pending_fields:
-            rng = find_text(doc, token)
+            rng = fields.find(token)
             field = doc.Fields.Add(rng, -1, code, False)
             field.Code.Font.Bold, field.Code.Font.Italic = 0, 0
             if bookmark:
                 doc.Bookmarks.Add(bookmark, field.Result)
+        numbers = MarkerFinder(doc)
         for begin, end, bookmark in self._pending_numbers:
-            rng = doc.Content.Duplicate
-            if not rng.Find.Execute(FindText=end, Wrap=0):
-                raise RuntimeError(f'Number boundary missing: {end}')
-            end_position = rng.Start
-            rng.Delete()
-            rng = doc.Content.Duplicate
-            if not rng.Find.Execute(FindText=begin, Wrap=0):
-                raise RuntimeError(f'Number boundary missing: {begin}')
-            start_position = rng.Start
-            rng.Delete()
+            opening = numbers.find(begin)
+            closing = numbers.find(end)
+            end_position = closing.Start
+            closing.Delete()
+            start_position = opening.Start
+            opening.Delete()
             doc.Bookmarks.Add(bookmark, doc.Range(
                 start_position, end_position - utf16_length(begin)))
         for table in doc.Tables:
@@ -227,6 +226,7 @@ class ContentComposer(Composer):
         self.heading_lists = {}
         self.appendix_numbering = None
         self.appendix_count = 0
+        self._inline_math_style = None
 
     def point(self):
         rng = self.doc.Content.Duplicate if self.container is None else self.container.Range.Duplicate
@@ -286,7 +286,13 @@ class ContentComposer(Composer):
             if kind == 'text':
                 value = node['text']
             elif kind == 'math':
-                value = self.math_token(node, True)
+                if 'plainText' not in node:
+                    value = self.math_token(node, True)
+                else:
+                    for run in node['plainText']:
+                        spans.append((utf16_length(text), utf16_length(run['text']), ['math'] + run.get('marks', [])))
+                        text += run['text']
+                    continue
             elif kind == 'ref':
                 value = self.ref_token(node)
             elif kind == 'break':
@@ -309,14 +315,39 @@ class ContentComposer(Composer):
     def rich_paragraph(self, nodes, style=None, prefix='', suffix=''):
         text, spans = self.inline(nodes)
         rng = self.paragraph(prefix + text + suffix, style)
+        self.apply_marks(rng, utf16_length(prefix), spans)
+        return rng
+
+    def apply_marks(self, rng, offset, spans):
         for start, length, marks in spans:
-            run = self.part(rng, utf16_length(prefix) + start, length)
+            if not marks:
+                continue
+            run = self.part(rng, offset + start, length)
             for mark in marks:
                 if mark == 'code':
                     run.Font.Name = 'Consolas'
+                elif mark == 'math':
+                    run.Style = self.inline_math_style()
                 else:
                     setattr(run.Font, {'strong': 'Bold', 'em': 'Italic', 'sup': 'Superscript', 'sub': 'Subscript'}[mark], True)
-        return rng
+
+    def inline_math_style(self):
+        """Character style of inline formulas written as text; created in the document, the DOTX stays unchanged."""
+        if self._inline_math_style is None:
+            style = self.doc.Styles.Add(INLINE_MATH_STYLE, 2)
+            style.Font.Name = 'Times New Roman'
+            self._inline_math_style = style
+        return self._inline_math_style
+
+    def single_paragraph_cell(self, cell, node):
+        """Write a one-paragraph cell in one assignment; the general path inserts then trims a paragraph."""
+        text, spans = self.inline(node['content'])
+        rng = cell.Range
+        rng.Text = text
+        rng = self.part(cell.Range, 0, utf16_length(text))
+        rng.Style = self.doc.Styles.Item(self.layout.get(node['id'], {}).get('style') or self.style)
+        self.apply_marks(rng, 0, spans)
+        self.mark(node['id'], rng, '_b')
 
     def chapter_sequence(self, kind, bookmark):
         """「章号-序号」 for an equation, figure or table in the current top-level chapter."""
@@ -489,11 +520,14 @@ class ContentComposer(Composer):
         saved = self.container, self.cell, self.style, self.width
         for r, c, rs, cs, cell in placements:
             self.container = cells[r, c].Cells.Item(1)
-            # Merge can concatenate the original empty cell paragraphs.
-            self.container.Range.Text = ''
             self.cell = True
             self.style = 'Table Header' if r < node['headerRows'] else 'Table Text'
             self.width = sum(widths[c:c + cs]) - 8
+            if len(cell['blocks']) == 1 and cell['blocks'][0]['type'] == 'paragraph':
+                self.single_paragraph_cell(self.container, cell['blocks'][0])
+                continue
+            # Merge can concatenate the original empty cell paragraphs.
+            self.container.Range.Text = ''
             self.blocks(cell['blocks'], depth)
             self.trim_container()
         self.container, self.cell, self.style, self.width = saved

package/engine/src/docreport/mathtype.py

@@ -10,7 +10,7 @@ import re
 import struct
 from hashlib import sha256
 
-from .word import find_text
+from .word import MarkerFinder
 
 DEFAULT_DLL = Path(r"C:\Program Files (x86)\MathType\MathPage\64\MathPage.wll")
 
@@ -138,12 +138,13 @@ def fill_and_reopen(draft, destination, pending, record, dll=DEFAULT_DLL):
         word.Visible, word.DisplayAlerts, word.AutomationSecurity = False, 0, 3
         missing = comtypes.automation.VARIANT.missing
         doc = word.Documents.Open(str(draft), False, False, False)
+        markers = MarkerFinder(doc)
         for item in pending:
             equation = item["equation"]
             tex = mathtype_tex(equation["latex"])
             if tex != equation["latex"]:
                 record("tex-adaptation", id=equation["id"], sourceTex=equation["latex"], mathTypeTex=tex)
-            target = find_text(doc, item["token"])
+            target = markers.find(item["token"])
             target.Text = ""
             shape = doc.InlineShapes.AddOLEObject("Equation.DSMT4", missing, False, False,
                                                  missing, missing, missing, target)

package/engine/src/docreport/plain_math.py

@@ -0,0 +1,53 @@
+"""Inline formulas written as Word text: their character style and the saved-package fix-up.
+
+The report marks each such formula with <WordPlainMath> and supplies the runs; the engine
+does not decide which TeX can become text.
+"""
+
+
+INLINE_MATH_STYLE = 'DR Inline Math'
+
+
+def clear_east_asian_hints(path):
+    """Drop w:hint="eastAsia" from runs in the inline-math character style.
+
+    Word adds the hint when it inserts φ, ≈ or ° into a Chinese paragraph, which renders
+    them in the East Asian font; COM exposes no setter for it, so the saved package is edited.
+    """
+    import os
+    import re
+    import tempfile
+    import zipfile
+
+    with zipfile.ZipFile(path) as source:
+        styles = source.read('word/styles.xml').decode('utf8')
+        match = re.search(r'<w:style\b[^>]*w:styleId="([^"]+)"[^>]*>(?:(?!</w:style>).)*?<w:name w:val="'
+                          + re.escape(INLINE_MATH_STYLE) + '"', styles, re.S)
+        if not match:
+            return 0
+        marker = f'<w:rStyle w:val="{match[1]}"/>'
+        cleared = 0
+
+        def clear(run):
+            nonlocal cleared
+            text = run[0]
+            if marker not in text or 'w:hint="eastAsia"' not in text:
+                return text
+            cleared += 1
+            text = text.replace(' w:hint="eastAsia"', '')
+            return text.replace('<w:rFonts/>', '')
+
+        fd, temporary = tempfile.mkstemp(suffix='.docx', dir=os.path.dirname(path))
+        os.close(fd)
+        try:
+            with zipfile.ZipFile(temporary, 'w') as target:
+                for item in source.infolist():
+                    data = source.read(item)
+                    if item.filename.startswith('word/') and item.filename.endswith('.xml') and marker.encode() in data:
+                        data = re.sub(r'<w:r(?:\s[^>]*)?>.*?</w:r>', clear, data.decode('utf8'), flags=re.S).encode('utf8')
+                    target.writestr(item, data, compress_type=item.compress_type)
+        except BaseException:
+            os.unlink(temporary)
+            raise
+    os.replace(temporary, path)
+    return cleared

package/engine/src/docreport/verify.py

@@ -59,6 +59,39 @@ def abstract_state(doc, content):
     return results
 
 
+WORD_NS = '{http://schemas.openxmlformats.org/wordprocessingml/2006/main}'
+# Word writes these into a field's result when a REF/SEQ target or switch is broken; readers then see them as body text.
+FIELD_ERROR_PREFIXES = ('错误!不能识别的开关参数', '错误!未找到引用源', '错误!找不到引用源', '错误!未定义书签',
+                        'error!unknownswitchargument', 'error!referencesourcenotfound', 'error!bookmarknotdefined')
+
+
+def field_results(root):
+    """Yield (field code, displayed result) for every simple and complex field in one XML part."""
+    open_fields = []  # [code, result, past separator] per nested complex field
+    for node in root.iter():
+        if node.tag == WORD_NS + 'fldSimple':
+            yield node.get(WORD_NS + 'instr', ''), ''.join(t.text or '' for t in node.iter(WORD_NS + 't'))
+        elif node.tag == WORD_NS + 'fldChar':
+            kind = node.get(WORD_NS + 'fldCharType')
+            if kind == 'begin':
+                open_fields.append(['', '', False])
+            elif kind == 'separate' and open_fields:
+                open_fields[-1][2] = True
+            elif kind == 'end' and open_fields:
+                code, result, _ = open_fields.pop()
+                yield code, result
+        elif open_fields and node.tag == WORD_NS + 'instrText' and not open_fields[-1][2]:
+            open_fields[-1][0] += node.text or ''
+        elif open_fields and node.tag == WORD_NS + 't' and open_fields[-1][2]:
+            open_fields[-1][1] += node.text or ''
+
+
+def reject_field_error_results(root):
+    for code, result in field_results(root):
+        if re.sub(r'\s+', '', result).lower().replace('！', '!').startswith(FIELD_ERROR_PREFIXES):
+            raise RuntimeError(f'Word field result error: {code.strip()}: {result}')
+
+
 def package_state(path, expected_equations):
     with ZipFile(path) as archive:
         objects, omml = [], []
@@ -66,6 +99,7 @@ def package_state(path, expected_equations):
             if not part.startswith('word/') or not part.endswith('.xml'):
                 continue
             root = ET.fromstring(archive.read(part))
+            reject_field_error_results(root)
             found = [o for o in root.findall(".//{urn:schemas-microsoft-com:office:office}OLEObject")
                      if o.get("ProgID") == "Equation.DSMT4"]
             objects.extend(found)
@@ -103,6 +137,11 @@ def content_state(path, content, directory):
                 if compact(node['text']) not in available:
                     raise RuntimeError(f"Source text missing after save: {node['text'][:100]}")
                 checked += 1
+            elif node.get('type') == 'math' and 'plainText' in node:
+                written = ''.join(run['text'] for run in node['plainText'])
+                if compact(written) not in available:
+                    raise RuntimeError(f"Inline formula text missing after save: {written[:100]}")
+                checked += 1
             elif node.get('type') == 'abstract':
                 keywords = '关键词：' + '；'.join(node['keywords'])
                 if compact(node.get('title', '摘要')) not in available or compact(keywords) not in available:

package/engine/src/docreport/word.py

@@ -72,6 +72,29 @@ def find_text(doc, text):
     raise ValueError(f"Generation marker missing: {text}")
 
 
+class MarkerFinder:
+    """Find generation markers written in body order without rescanning from the top each time.
+
+    The body story is searched from the previous body hit; a marker not found there
+    (notes, headers, or out of order) falls back to the full search of every story.
+    """
+
+    MAIN_STORY = 1
+
+    def __init__(self, doc):
+        self.doc = doc
+        self.position = 0
+
+    def find(self, text):
+        target = self.doc.Range(self.position, self.doc.Content.End)
+        target.Find.ClearFormatting()
+        if not target.Find.Execute(text, True, False, False, False, False, True, 0):
+            target = find_text(self.doc, text)
+        if target.StoryType == self.MAIN_STORY:
+            self.position = target.Start
+        return target
+
+
 def refresh(doc):
     """Resolve sequence fields before cross-references, then refresh TOC pages."""
     for _ in range(2):

package/package.json

@@ -1,11 +1,11 @@
 {
   "name": "@godcreator/word-report",
-  "version": "0.4.0",
+  "version": "0.5.0",
   "author": {
     "name": "godcreator"
   },
   "type": "module",
-  "description": "报告：JSON 报告文档的格式、校验与分章组装（.）；读报告目录、检查、导出 Word 的命令行 report-docx（./node）；报告页与报告列表（./site），两页与图片的 loader（./site/server）；engine/ 是驱动本机 Word 的 Python 导出引擎与 .dotx 模板；template/ 是建站器复制进站点 app/features/word-report/ 的接线文件（两条页面路由、图片路由、侧栏入口、样式）。",
+  "description": "报告：JSON 报告文档的格式、校验与分章组装，MDX 写法编译成它（.）；读报告目录、检查、导出 Word、JSON 报告转 MDX 的命令行 report-docx（./node）；报告页与报告列表（./site），两页与图片的 loader（./site/server）；engine/ 是驱动本机 Word 的 Python 导出引擎与 .dotx 模板；template/ 是建站器复制进站点 app/features/word-report/ 的接线文件（两条页面路由、图片路由、侧栏入口、样式）。",
   "license": "MIT",
   "exports": {
     ".": {
@@ -43,7 +43,14 @@
   ],
   "dependencies": {
     "@godcreator/comments": "^0.1.3",
-    "sharp": "^0.35.5"
+    "remark-frontmatter": "^5.0.0",
+    "remark-gfm": "^4.0.1",
+    "remark-math": "^6.0.0",
+    "remark-mdx": "^3.1.0",
+    "remark-parse": "^11.0.0",
+    "sharp": "^0.35.5",
+    "unified": "^11.0.5",
+    "yaml": "^2.9.1"
   },
   "peerDependencies": {
     "@fumadocs/base-ui": "^16.17.1",

package/dist/report-files-CgWzLxH-.mjs

@@ -1,72 +0,0 @@
-import { a as parseReportJson, n as safeRelativePath, o as ReportContentError, r as validateDocument } from "./assemble-D25NLmLx.mjs";
-import { readFile, readdir, realpath, stat } from "node:fs/promises";
-import path from "node:path";
-//#region src/node/report-files.ts
-/** 站点里的报告：站点根下 reports/<报告 id>/content.json，章节与图片在同一个报告目录里 */
-const reportsDirName = "reports";
-/** 报告 id 就是报告目录名，出现在网址里：小写字母、数字与 -，以字母或数字开头 */
-const REPORT_ID_PATTERN = /^[a-z0-9][a-z0-9-]{0,63}$/;
-/** 报告的入口，相对站点目录 */
-function reportEntryOf(reportId) {
-	if (!REPORT_ID_PATTERN.test(reportId)) throw new ReportContentError(`报告 id 只能是小写字母、数字与 -，以字母或数字开头：${JSON.stringify(reportId)}`);
-	return `${reportsDirName}/${reportId}/content.json`;
-}
-/** 站点里的报告 id：reports/ 下有 content.json 的目录，按名字排；目录名不合 REPORT_ID_PATTERN 的直接报错，不悄悄跳过 */
-async function reportIdsOf(siteDir) {
-	const reportsDir = path.join(siteDir, reportsDirName);
-	let entries;
-	try {
-		entries = await readdir(reportsDir, { withFileTypes: true });
-	} catch (error) {
-		if (error.code === "ENOENT") return [];
-		throw error;
-	}
-	const ids = [];
-	for (const entry of entries) {
-		if (!entry.isDirectory()) continue;
-		if (!await stat(path.join(reportsDir, entry.name, "content.json")).then((info) => info.isFile(), (error) => {
-			if (error.code === "ENOENT") return false;
-			throw error;
-		})) continue;
-		reportEntryOf(entry.name);
-		ids.push(entry.name);
-	}
-	return ids.sort();
-}
-/**
-* 读一份报告组装要的全部 JSON 文件（入口与它 include 到的章节），键是相对站点目录的路径，交给 assembleDocument。
-* 不读图片、不解码：网页显示用；检查与导出用 readReport，它还核对图片与快照
-*/
-async function readReportFiles(siteDir, entry) {
-	const root = await realpath(siteDir);
-	const files = {};
-	const load = async (relative) => {
-		if (Object.hasOwn(files, relative)) return;
-		const absolute = path.join(root, relative);
-		let real;
-		try {
-			real = await realpath(absolute);
-		} catch {
-			throw new ReportContentError("Missing or unreadable file", relative);
-		}
-		const inside = path.relative(root, real);
-		if (inside === ".." || inside.startsWith(".." + path.sep) || path.isAbsolute(inside)) throw new ReportContentError("File escapes report root", relative);
-		const text = await readFile(real, "utf8");
-		files[relative] = text;
-		const document = parseReportJson(text, relative);
-		validateDocument(document, relative);
-		for (const [index, block] of document.blocks.entries()) {
-			if (block.type !== "include") continue;
-			try {
-				safeRelativePath(block.src);
-			} catch (error) {
-				throw new ReportContentError(error.message, relative, `/blocks/${index}/src`);
-			}
-			await load(path.posix.join(path.posix.dirname(relative), block.src));
-		}
-	};
-	await load(entry);
-	return files;
-}
-//#endregion
-export { reportsDirName as a, reportIdsOf as i, readReportFiles as n, reportEntryOf as r, REPORT_ID_PATTERN as t };