Add Support for Rendering Various Inline and Block-level Mathematical Formulas

2025-11-27 10:29:27 +08:00
parent 4e882560da
commit 5e9da9cfbf
3 changed files with 195 additions and 21 deletions
--- a/ReportEngine/renderers/html_renderer.py
+++ b/ReportEngine/renderers/html_renderer.py
@@ -1262,7 +1262,8 @@ class HTMLRenderer:

    def _render_math(self, block: Dict[str, Any]) -> str:
        """渲染数学公式，占位符交给外部MathJax或后处理"""
-        latex = self._escape_html(block.get("latex", ""))
+        latex_raw = block.get("latex", "")
+        latex = self._escape_html(self._normalize_latex_string(latex_raw))
        math_id = self._escape_attr(block.get("mathId", "")) if block.get("mathId") else ""
        id_attr = f' data-math-id="{math_id}"' if math_id else ""
        return f'<div class="math-block"{id_attr}>$$ {latex} $$</div>'
@@ -1988,6 +1989,66 @@ class HTMLRenderer:

        return text_value, marks

+    @staticmethod
+    def _normalize_latex_string(raw: Any) -> str:
+        """去除外层数学定界符，兼容 $...$、$$...$$、\\(\\)、\\[\\] 等格式"""
+        if not isinstance(raw, str):
+            return ""
+        latex = raw.strip()
+        patterns = [
+            r'^\$\$(.*)\$\$$',
+            r'^\$(.*)\$$',
+            r'^\\\[(.*)\\\]$',
+            r'^\\\((.*)\\\)$',
+        ]
+        for pat in patterns:
+            m = re.match(pat, latex, re.DOTALL)
+            if m:
+                latex = m.group(1).strip()
+                break
+        return latex
+
+    def _render_text_with_inline_math(self, text: Any, math_id: str | list | None = None) -> str | None:
+        """
+        识别纯文本中的数学定界符并渲染为math-inline/math-block，提升兼容性。
+
+        - 支持 $...$、$$...$$、\\(\\)、\\[\\]。
+        - 若未检测到公式，返回None。
+        """
+        if not isinstance(text, str) or not text:
+            return None
+
+        pattern = re.compile(r'(\$\$(.+?)\$\$|\$(.+?)\$|\\\((.+?)\\\)|\\\[(.+?)\\\])', re.S)
+        cursor = 0
+        parts: List[str] = []
+        idx = 0
+        id_iter = iter(math_id) if isinstance(math_id, list) else None
+        for m in pattern.finditer(text):
+            start, end = m.span()
+            if start > cursor:
+                parts.append(self._escape_html(text[cursor:start]))
+            raw = next(g for g in m.groups()[1:] if g is not None)
+            latex = self._normalize_latex_string(raw)
+            idx += 1
+            # 若已有math_id，直接使用，避免与SVG注入ID不一致；否则按局部序号生成
+            if id_iter:
+                mid = next(id_iter, f"auto-math-{idx}")
+            else:
+                mid = math_id or f"auto-math-{idx}"
+            id_attr = f' data-math-id="{self._escape_attr(mid)}"'
+            is_display = m.group(1).startswith('$$') or m.group(1).startswith('\\[')
+            if is_display:
+                parts.append(f'<div class="math-block"{id_attr}>$$ {self._escape_html(latex)} $$</div>')
+            else:
+                parts.append(f'<span class="math-inline"{id_attr}>\\( {self._escape_html(latex)} \\)</span>')
+            cursor = end
+
+        if cursor == 0:
+            return None
+        if cursor < len(text):
+            parts.append(self._escape_html(text[cursor:]))
+        return "".join(parts)
+
    @staticmethod
    def _coerce_inline_payload(payload: Dict[str, Any]) -> Dict[str, Any] | None:
        """尽力将字符串里的内联节点恢复为dict，修复渲染遗漏"""
@@ -2013,12 +2074,19 @@ class HTMLRenderer:
        text_value, marks = self._normalize_inline_payload(run)
        math_mark = next((mark for mark in marks if mark.get("type") == "math"), None)
        if math_mark:
-            latex = math_mark.get("value")
+            latex = self._normalize_latex_string(math_mark.get("value"))
            if not isinstance(latex, str) or not latex.strip():
-                latex = text_value
+                latex = self._normalize_latex_string(text_value)
            math_id = self._escape_attr(run.get("mathId", "")) if run.get("mathId") else ""
            id_attr = f' data-math-id="{math_id}"' if math_id else ""
            return f'<span class="math-inline"{id_attr}>\\( {self._escape_html(latex)} \\)</span>'
+
+        # 尝试从纯文本中提取数学公式（即便没有math mark）
+        math_id_hint = run.get("mathIds") or run.get("mathId")
+        mathified = self._render_text_with_inline_math(text_value, math_id_hint)
+        if mathified is not None:
+            return mathified
+
        text = self._escape_html(text_value)
        styles: List[str] = []
        prefix: List[str] = []
--- a/ReportEngine/renderers/math_to_svg.py
+++ b/ReportEngine/renderers/math_to_svg.py
@@ -4,6 +4,7 @@ LaTeX 数学公式转 SVG 渲染器
 """

 import io
+import re
 from typing import Optional
 import matplotlib
 import matplotlib.pyplot as plt
@@ -40,8 +41,22 @@ class MathToSVG:
            SVG 字符串，如果转换失败则返回 None
        """
        try:
-            # 清理 LaTeX 字符串
-            latex = latex.strip()
+            # 清理 LaTeX 字符串，去除外层定界符，兼容 $...$ / $$...$$ / \\( \\) / \\[ \\]
+            latex = (latex or "").strip()
+            patterns = [
+                r'^\$\$(.*)\$\$$',
+                r'^\$(.*)\$$',
+                r'^\\\[(.*)\\\]$',
+                r'^\\\((.*)\\\)$',
+            ]
+            for pat in patterns:
+                m = re.match(pat, latex, re.DOTALL)
+                if m:
+                    latex = m.group(1).strip()
+                    break
+            # 清理控制字符并做常见兼容
+            latex = re.sub(r'[\x00-\x1f\x7f]', '', latex)
+            latex = latex.replace(r'\\tfrac', r'\\frac').replace(r'\\dfrac', r'\\frac')
            if not latex:
                logger.warning("空的 LaTeX 公式")
                return None
--- a/ReportEngine/renderers/pdf_renderer.py
+++ b/ReportEngine/renderers/pdf_renderer.py
@@ -10,6 +10,7 @@ import copy
 import os
 import sys
 import io
+import re
 from pathlib import Path
 from typing import Any, Dict
 from datetime import datetime
@@ -544,23 +545,62 @@ class PDFRenderer:
                    continue
                marks = run.get('marks') or []
                math_mark = next((m for m in marks if m.get('type') == 'math'), None)
-                if not math_mark:
+
+                if math_mark:
+                    # 仅单个math mark
+                    raw = math_mark.get('value') or run.get('text') or ''
+                    latex = self._normalize_latex(raw)
+                    is_display = bool(re.match(r'^\s*(\$\$|\\\[)', str(raw)))
+                    if not latex:
+                        continue
+                    block_counter[0] += 1
+                    math_id = run.get('mathId') or f"math-inline-{block_counter[0]}"
+                    run['mathId'] = math_id
+                    try:
+                        svg_content = (
+                            self.math_converter.convert_display_to_svg(latex)
+                            if is_display else
+                            self.math_converter.convert_inline_to_svg(latex)
+                        )
+                        if svg_content:
+                            svg_map[math_id] = svg_content
+                            logger.debug(f"公式 {math_id} 转换为SVG成功")
+                        else:
+                            logger.warning(f"公式 {math_id} 转换为SVG失败: {latex[:50]}...")
+                    except Exception as exc:
+                        logger.error(f"转换内联公式 {latex[:50]}... 时出错: {exc}")
                    continue
-                latex = (math_mark.get('value') or run.get('text') or '').strip()
-                if not latex:
+
+                # 无math mark，尝试解析文本中的多个公式
+                text_val = run.get('text')
+                if not isinstance(text_val, str):
                    continue
-                block_counter[0] += 1
-                math_id = f"math-inline-{block_counter[0]}"
-                try:
-                    svg_content = self.math_converter.convert_inline_to_svg(latex)
-                    if svg_content:
-                        svg_map[math_id] = svg_content
-                        run['mathId'] = math_id
-                        logger.debug(f"公式 {math_id} 转换为SVG成功")
-                    else:
-                        logger.warning(f"公式 {math_id} 转换为SVG失败: {latex[:50]}...")
-                except Exception as exc:
-                    logger.error(f"转换内联公式 {latex[:50]}... 时出错: {exc}")
+                segments = self._find_all_math_in_text(text_val)
+                if not segments:
+                    continue
+                ids_for_html: list[str] = []
+                for idx, (latex, is_display) in enumerate(segments, start=1):
+                    if not latex:
+                        continue
+                    block_counter[0] += 1
+                    math_id = f"auto-math-{block_counter[0]}"
+                    ids_for_html.append(math_id)
+                    try:
+                        svg_content = (
+                            self.math_converter.convert_display_to_svg(latex)
+                            if is_display else
+                            self.math_converter.convert_inline_to_svg(latex)
+                        )
+                        if svg_content:
+                            svg_map[math_id] = svg_content
+                            logger.debug(f"公式 {math_id} 转换为SVG成功")
+                        else:
+                            logger.warning(f"公式 {math_id} 转换为SVG失败: {latex[:50]}...")
+                    except Exception as exc:
+                        logger.error(f"转换内联公式 {latex[:50]}... 时出错: {exc}")
+                if ids_for_html:
+                    # 将ID列表写回run，便于HTML渲染时使用相同ID（顺序对应segments）
+                    run['mathIds'] = ids_for_html

        for block in blocks:
            if not isinstance(block, dict):
@@ -570,7 +610,7 @@ class PDFRenderer:

            # 处理math类型
            if block_type == 'math':
-                latex = block.get('latex', '').strip()
+                latex = self._normalize_latex(block.get('latex', ''))
                if latex:
                    block_counter[0] += 1
                    math_id = f"math-block-{block_counter[0]}"
@@ -679,6 +719,57 @@ class PDFRenderer:

        return html

+    @staticmethod
+    def _normalize_latex(raw: Any) -> str:
+        """去除外层数学定界符，兼容 $...$、$$...$$、\\(\\)、\\[\\] 等格式"""
+        if not isinstance(raw, str):
+            return ""
+        latex = raw.strip()
+        patterns = [
+            r'^\$\$(.*)\$\$$',
+            r'^\$(.*)\$$',
+            r'^\\\[(.*)\\\]$',
+            r'^\\\((.*)\\\)$',
+        ]
+        for pat in patterns:
+            m = re.match(pat, latex, re.DOTALL)
+            if m:
+                latex = m.group(1).strip()
+                break
+        # 清理控制字符、防止mathtext解析失败
+        latex = re.sub(r'[\x00-\x1f\x7f]', '', latex)
+        # 常见兼容：\tfrac/\dfrac -> \frac
+        latex = latex.replace(r'\tfrac', r'\frac').replace(r'\dfrac', r'\frac')
+        return latex
+
+    @staticmethod
+    def _find_first_math_in_text(text: Any) -> tuple[str, bool] | None:
+        """从纯文本中提取首个数学片段，返回(内容, 是否display)"""
+        if not isinstance(text, str):
+            return None
+        pattern = re.compile(r'\$\$(.+?)\$\$|\$(.+?)\$|\\\((.+?)\\\)|\\\[(.+?)\\\]', re.S)
+        m = pattern.search(text)
+        if not m:
+            return None
+        raw = next(g for g in m.groups() if g is not None)
+        latex = raw.strip()
+        is_display = bool(m.group(1) or m.group(4))  # $$ or \[ \]
+        return latex, is_display
+
+    @staticmethod
+    def _find_all_math_in_text(text: Any) -> list[tuple[str, bool]]:
+        """从纯文本中提取所有数学片段，返回[(内容, 是否display)]"""
+        if not isinstance(text, str):
+            return []
+        pattern = re.compile(r'\$\$(.+?)\$\$|\$(.+?)\$|\\\((.+?)\\\)|\\\[(.+?)\\\]', re.S)
+        results = []
+        for m in pattern.finditer(text):
+            raw = next(g for g in m.groups() if g is not None)
+            latex = raw.strip()
+            is_display = bool(m.group(1) or m.group(4))
+            results.append((latex, is_display))
+        return results
+
    def _inject_wordcloud_images(self, html: str, img_map: Dict[str, str]) -> str:
        """
        将词云PNG data URI注入HTML，替换对应canvas