From 6071a2d98c13b6c1ff643d67e41df6df98bd694c Mon Sep 17 00:00:00 2001
From: Quentin <1771956812@qq.com>
Date: Tue, 29 Sep 2026 15:33:33 +0800
Subject: [PATCH] =?UTF-8?q?feat(symbol):=20=E6=AD=A3=E6=96=87=E6=A0=87?=
 =?UTF-8?q?=E5=87=86=20Unicode=20=E6=95=B0=E5=AD=A6=E7=AC=A6=E5=8F=B7=20?=
 =?UTF-8?q?=E2=86=92=20LaTeX=EF=BC=88=E4=B8=8E=E5=85=AC=E5=BC=8F=E5=90=8C?=
 =?UTF-8?q?=E4=B8=80=E6=9D=A1=E6=B8=B2=E6=9F=93=E8=B7=AF=E5=BE=84=EF=BC=89?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

背景：公式（OMML/MathType/EQ）已转 LaTeX，但正文里的裸 Unicode 符号
（⊆ ∈ ∪ ≤ × → ∅ ∞ ∑ √ …）docx4j 只原样输出，靠系统字体渲染，字形/字号
与公式不一致；且 EqToLatexUtil 未覆盖的符号会以裸字符留在公式 LaTeX 内部。

实现 UnicodeSymbolLatex：
- 53 条映射，取自 jby omml_mathml2latex/data/symbols.py，保证与公式内转换一致
  （⊆→\subseteq、∈→\in、≤→\leqslant、×→\times、→→\rightarrow、∅→\emptyset …）
- 文本节点内：符号 → \( … \)（MathJax 渲染）
- 已在 \(…\)/\[…\] 内的片段：只换命令不加包裹（修 LaTeX 内部裸字符）
- 跳过 <script>/<style>/<textarea> 与所有标签属性
- 开关 wordparse.convert-unicode-symbols（默认 true）

效果：46 份教辅重新生成后，正文裸 Unicode 符号 0 处，公式片段 5394 处
---
 .../com/wordparse/config/WordParseConfig.java |   4 +
 .../wordparse/core/UnicodeSymbolLatex.java    | 161 ++++++++++++++++++
 .../com/wordparse/core/WordParseUtil.java     |   4 +
 .../main/resources/application.yml.example    |   1 +
 4 files changed, 170 insertions(+)
 create mode 100644 java-svc/src/main/java/com/wordparse/core/UnicodeSymbolLatex.java

diff --git a/java-svc/src/main/java/com/wordparse/config/WordParseConfig.java b/java-svc/src/main/java/com/wordparse/config/WordParseConfig.java
index 8eb1998..92dca1b 100644
--- a/java-svc/src/main/java/com/wordparse/config/WordParseConfig.java
+++ b/java-svc/src/main/java/com/wordparse/config/WordParseConfig.java
@@ -15,6 +15,8 @@ public class WordParseConfig {
     private String pythonCommand = "python3";
     /** 私有区符号图的基址（默认我们自己的 RustFS，不依赖 jby CDN） */
     private String symbolBaseUrl = "https://fs.code-lab.cn/temp/wordparse-symbols/";
+    /** 是否把正文里的标准 Unicode 数学符号转成 LaTeX（默认开） */
+    private Boolean convertUnicodeSymbols = Boolean.TRUE;
 
     public String getTmpPath() { return tmpPath; }
     public Double getImageScalingFactor() { return imageScalingFactor; }
@@ -23,6 +25,8 @@ public class WordParseConfig {
     public String getPythonCommand() { return pythonCommand; }
     public String getSymbolBaseUrl() { return symbolBaseUrl; }
     public void setSymbolBaseUrl(String symbolBaseUrl) { this.symbolBaseUrl = symbolBaseUrl; }
+    public Boolean getConvertUnicodeSymbols() { return convertUnicodeSymbols; }
+    public void setConvertUnicodeSymbols(Boolean v) { this.convertUnicodeSymbols = v; }
 
     public void setTmpPath(String v) { this.tmpPath = v; }
     public void setImageScalingFactor(Double v) { this.imageScalingFactor = v; }
diff --git a/java-svc/src/main/java/com/wordparse/core/UnicodeSymbolLatex.java b/java-svc/src/main/java/com/wordparse/core/UnicodeSymbolLatex.java
new file mode 100644
index 0000000..7d74bbc
--- /dev/null
+++ b/java-svc/src/main/java/com/wordparse/core/UnicodeSymbolLatex.java
@@ -0,0 +1,161 @@
+package com.wordparse.core;
+
+import java.util.LinkedHashMap;
+import java.util.Map;
+import java.util.Set;
+import java.util.regex.Matcher;
+import java.util.regex.Pattern;
+
+/**
+ * 正文里的标准 Unicode 数学符号 → LaTeX（供前端 MathJax 渲染）。
+ *
+ * 为什么需要：公式（OMML/MathType/EQ）已转成 LaTeX，但正文里的裸 Unicode 符号
+ * （⊆ ∈ ∪ ≤ × …）docx4j 只会原样输出，靠系统字体渲染，字形/字号与公式不一致。
+ * 这里把正文符号也包成 \(...\)，与公式走同一条渲染路径。
+ *
+ * 映射表取自 jby 的 omml_mathml2latex/data/symbols.py，保证与公式内转换一致。
+ * 跳过：<script>/<style> 内容、标签属性、以及已处于 \(...\) / \[...\] 内的文本。
+ */
+public final class UnicodeSymbolLatex {
+
+    private static final Map<String, String> MAP;
+
+    static {
+        Map<String, String> m = new LinkedHashMap<>();
+        m.put("±", "\\pm");
+        m.put("×", "\\times");
+        m.put("÷", "\\div");
+        m.put("←", "\\leftarrow");
+        m.put("↑", "\\uparrow");
+        m.put("→", "\\rightarrow");
+        m.put("↓", "\\downarrow");
+        m.put("⇌", "\\rightleftharpoons");
+        m.put("⇒", "\\Longrightarrow");
+        m.put("⇔", "\\Longleftrightarrow");
+        m.put("∅", "\\emptyset");
+        m.put("∈", "\\in");
+        m.put("∉", "\\notin");
+        m.put("∏", "\\prod");
+        m.put("∑", "\\sum");
+        m.put("√", "\\sqrt{}");
+        m.put("∝", "\\propto");
+        m.put("∞", "\\infty");
+        m.put("∠", "\\angle");
+        m.put("∥", "\\parallel");
+        m.put("∩", "\\cap");
+        m.put("∪", "\\cup");
+        m.put("∫", "\\int");
+        m.put("∴", "\\therefore");
+        m.put("∵", "\\because");
+        m.put("∽", "\\backsim");
+        m.put("≈", "\\approx");
+        m.put("≌", "\\cong");
+        m.put("≠", "\\neq");
+        m.put("≡", "\\equiv");
+        m.put("≤", "\\leqslant");
+        m.put("≥", "\\geqslant");
+        m.put("⊂", "\\subset");
+        m.put("⊃", "\\supset");
+        m.put("⊆", "\\subseteq");
+        m.put("⊇", "\\supseteq");
+        m.put("⊥", "\\perp");
+        m.put("⊙", "\\odot");
+        m.put("△", "\\triangle");
+        m.put("∼", "\\sim");
+        m.put("≅", "\\cong");
+        m.put("⇐", "\\Longleftarrow");
+        m.put("⟹", "\\Longrightarrow");
+        m.put("⟸", "\\Longleftarrow");
+        m.put("⟺", "\\Longleftrightarrow");
+        m.put("∀", "\\forall");
+        m.put("∃", "\\exists");
+        m.put("∂", "\\partial");
+        m.put("∇", "\\nabla");
+        m.put("ℝ", "\\mathbb{R}");
+        m.put("ℕ", "\\mathbb{N}");
+        m.put("ℤ", "\\mathbb{Z}");
+        m.put("ℚ", "\\mathbb{Q}");
+        MAP = Map.copyOf(m);
+    }
+
+    private static final Pattern MATH_SPAN =
+            Pattern.compile("\\\\\\(.*?\\\\\\)|\\\\\\[.*?\\\\\\]", Pattern.DOTALL);
+    private static final Pattern TOKEN = Pattern.compile("<[^>]*>|[^<]+");
+    private static final Set<String> SKIP_TAGS = Set.of("script", "style", "textarea");
+
+    private UnicodeSymbolLatex() {}
+
+    public static boolean containsSymbol(String text) {
+        for (String key : MAP.keySet()) {
+            if (text.contains(key)) {
+                return true;
+            }
+        }
+        return false;
+    }
+
+    /** 转换整段 HTML：只处理文本节点，跳过属性/脚本/已数学化的片段 */
+    public static String convert(String html) {
+        Matcher tokens = TOKEN.matcher(html);
+        StringBuilder out = new StringBuilder(html.length() + 256);
+        String skipUntil = null;
+        while (tokens.find()) {
+            String t = tokens.group();
+            if (t.startsWith("<")) {
+                out.append(t);
+                String lower = t.toLowerCase();
+                if (skipUntil == null) {
+                    for (String tag : SKIP_TAGS) {
+                        if (lower.startsWith("<" + tag) && !lower.endsWith("/>")) {
+                            skipUntil = "</" + tag;
+                        }
+                    }
+                } else if (lower.startsWith(skipUntil)) {
+                    skipUntil = null;
+                }
+                continue;
+            }
+            out.append(skipUntil != null ? t : convertText(t));
+        }
+        return out.toString();
+    }
+
+    /** 单个文本片段：保护已有 \(...\) 区间，其余部分替换符号 */
+    private static String convertText(String text) {
+        if (!containsSymbol(text)) {
+            return text;
+        }
+        StringBuilder sb = new StringBuilder(text.length() + 64);
+        int last = 0;
+        Matcher m = MATH_SPAN.matcher(text);
+        while (m.find()) {
+            sb.append(replaceSymbols(text.substring(last, m.start())));
+            // 数学区内部同样规范化：裸 ∈ / ⊆ 等 → LaTeX 命令（EqToLatexUtil 未覆盖的字符）
+            sb.append(replaceSymbolsBare(m.group()));
+            last = m.end();
+        }
+        sb.append(replaceSymbols(text.substring(last)));
+        return sb.toString();
+    }
+
+    private static String replaceSymbols(String s) {
+        String r = s;
+        for (Map.Entry<String, String> e : MAP.entrySet()) {
+            if (r.contains(e.getKey())) {
+                r = r.replace(e.getKey(), "\\(" + e.getValue() + "\\)");
+            }
+        }
+        return r;
+    }
+
+    /** 数学区内部：只换命令，不加 \( \) 包裹 */
+    private static String replaceSymbolsBare(String s) {
+        String r = s;
+        for (Map.Entry<String, String> e : MAP.entrySet()) {
+            if (r.contains(e.getKey())) {
+                r = r.replace(e.getKey(), e.getValue());
+            }
+        }
+        return r;
+    }
+}
diff --git a/java-svc/src/main/java/com/wordparse/core/WordParseUtil.java b/java-svc/src/main/java/com/wordparse/core/WordParseUtil.java
index 465d0db..5318062 100644
--- a/java-svc/src/main/java/com/wordparse/core/WordParseUtil.java
+++ b/java-svc/src/main/java/com/wordparse/core/WordParseUtil.java
@@ -1243,6 +1243,10 @@ public class WordParseUtil {
         }
         // ⇌（U+E225 U+E226 成对出现）直接转 LaTeX
         result = result.replace("\ue225\ue226", "\\(\\rightleftharpoons\\)");
+        // 正文里的标准 Unicode 数学符号（⊆ ∈ ∪ ≤ × …）→ LaTeX，与公式走同一条渲染路径
+        if (Boolean.TRUE.equals(config.getConvertUnicodeSymbols())) {
+            result = UnicodeSymbolLatex.convert(result);
+        }
         writeHtml(result);
         return result;
     }
diff --git a/java-svc/src/main/resources/application.yml.example b/java-svc/src/main/resources/application.yml.example
index a1ff2f0..20c2aae 100644
--- a/java-svc/src/main/resources/application.yml.example
+++ b/java-svc/src/main/resources/application.yml.example
@@ -15,6 +15,7 @@ wordparse:
   transform-scripts-path: /opt/transform
   python-command: python3
   symbol-base-url: https://fs.code-lab.cn/temp/wordparse-symbols/
+  convert-unicode-symbols: true
   oss:
     endpoint: https://fs.code-lab.cn
     access-key: ${WORDPARSE_OSS_AK:}