fix: update mathml

app/services/converter.py

@@ -296,29 +296,77 @@ class Converter:
     def _latex_to_mathml_cached(latex_formula: str) -> str:
         """Cached conversion of LaTeX formula to MathML.
 
+        Uses Pandoc for conversion to ensure Word compatibility.
+        Pandoc generates standard MathML that Word can properly import.
+
         Uses LRU cache to avoid recomputing for repeated formulas.
         """
         try:
-            # Use latex2mathml library for conversion (fast, pure Python)
-            return latex_to_mathml(latex_formula)
-        except Exception as e:
-            # Fallback: try with Pandoc (slower, but more robust)
+            # Use Pandoc for Word-compatible MathML (primary method)
+            mathml_html = pypandoc.convert_text(
+                f"${latex_formula}$",
+                "html",
+                format="markdown+tex_math_dollars",
+                extra_args=["--mathml"],
+            )
+            # Extract just the <math> element from the HTML
+            match = Converter._RE_MATH_ELEMENT.search(mathml_html)
+            if match:
+                mathml = match.group(0)
+                # Post-process for Word compatibility
+                return Converter._postprocess_mathml_for_word(mathml)
+            
+            # If no match, return as-is
+            return mathml_html.rstrip("\n")
+            
+        except Exception as pandoc_error:
+            # Fallback: try latex2mathml (less Word-compatible)
             try:
-                mathml_html = pypandoc.convert_text(
-                    f"${latex_formula}$",
-                    "html",
-                    format="markdown+tex_math_dollars",
-                    extra_args=["--mathml"],
-                )
-                # Extract just the <math> element from the HTML
-                match = Converter._RE_MATH_ELEMENT.search(mathml_html)
-                if match:
-                    return match.group(0)
-                return mathml_html.rstrip("\n")
-            except Exception as pandoc_error:
+                mathml = latex_to_mathml(latex_formula)
+                return Converter._postprocess_mathml_for_word(mathml)
+            except Exception as e:
                 raise RuntimeError(
-                    f"MathML conversion failed: {e}. Pandoc fallback also failed: {pandoc_error}"
+                    f"MathML conversion failed: {pandoc_error}. latex2mathml fallback also failed: {e}"
                 ) from e
+    
+    @staticmethod
+    def _postprocess_mathml_for_word(mathml: str) -> str:
+        """Post-process MathML to improve Word compatibility.
+        
+        Applies transformations to make MathML more compatible with Word:
+        - Change display="inline" to display="block" for better rendering
+        - Decode Unicode entities to actual characters (Word prefers this)
+        - Clean up unnecessary attributes
+        
+        Args:
+            mathml: MathML string.
+            
+        Returns:
+            Word-compatible MathML string.
+        """
+        # Change display to block for better Word rendering
+        mathml = mathml.replace('display="inline"', 'display="block"')
+        
+        # If no display attribute, add it
+        if 'display=' not in mathml and '<math' in mathml:
+            mathml = mathml.replace('<math', '<math display="block"', 1)
+        
+        # Decode common Unicode entities to actual characters (Word prefers this)
+        unicode_map = {
+            '&#x0002B;': '+',
+            '&#x02026;': '⋯',
+            '&#x022EE;': '⋮',
+            '&#x0003D;': '=',
+            '&#x0007C;': '|',
+            '&#x0002C;': ',',
+            '&#x00028;': '(',
+            '&#x00029;': ')',
+        }
+        
+        for entity, char in unicode_map.items():
+            mathml = mathml.replace(entity, char)
+        
+        return mathml
 
     def _latex_to_mathml(self, latex_formula: str) -> str:
         """Convert LaTeX formula to standard MathML.