Use MathML attributes for PDFs read in Adobe Acrobat (#17984)

This fixes #17980

Summary of the issue:
For PDF, the code was not grabbing MathML attributes needed for speech.

Description of user facing changes
The speech for math, and rarely the braille (e.g., bevelled fractions in Nemeth), was not always correct due to attributes not being picked up and the defaults being used.

Description of development approach
Unfortunately, the PDF interface does not allow grabbing all the attributes. Instead, one must ask for each attribute individually. Most attributes don't affect speech or braille, so it is not necessary to get them. I looked at what MathCAT used and added those.

Author: NSoiffer

source/NVDAObjects/IAccessible/adobeAcrobat.py

@@ -119,61 +119,96 @@ def _isEqual(self, other):
 			return self.accID == other.accID
 		return super(AcrobatNode, self)._isEqual(other)
 
-	def _getNodeMathMl(self, node):
+	@staticmethod
+	def getMathMLAttributes(node: IPDDomElement, attrList: list) -> str:
+		"""Get the MathML attributes in 'attrList' for a 'node' (MathML element)."""
+		attrValues = ""
+		for attr in attrList:
+			# "NSO" comes from the PDF spec
+			val = node.GetAttribute(attr, "NSO")
+			if val:
+				attrValues += f' {attr}="{val}"'
+		return attrValues
+
+	def _getNodeMathMl(self, node: IPDDomElement) -> str:
+		"""Traverse the MathML tree and return an XML string representing the math"""
+
 		tag = node.GetTagName()
-		yield "<%s" % tag
-		# Output relevant attributes.
-		if tag == "mfenced":
-			for attr in "open", "close", "separators":
-				val = node.GetAttribute(attr, "XML-1.00")
-				if val:
-					yield ' %s="%s"' % (attr, val)
-		yield ">"
+		answer = f"<{tag}"
+		# Output relevant attributes
+		id = node.GetID()
+		if id:
+			answer += f' id="{id}"'
+		# The PDF interface lacks a way to get all the attributes, so we have to get specific ones
+		# The attributes below affect accessibility
+		answer += AcrobatNode.getMathMLAttributes(node, ["intent", "arg"])
+		match tag:
+			case "mi" | "mn" | "mo" | "mtext":
+				answer += AcrobatNode.getMathMLAttributes(node, ["mathvariant"])
+			case "mfenced":
+				answer += AcrobatNode.getMathMLAttributes(node, ["open", "close", "separators"])
+			case "menclose":
+				answer += AcrobatNode.getMathMLAttributes(node, ["notation", "notationtype"])
+			case "annotation-xml" | "annotation":
+				answer += AcrobatNode.getMathMLAttributes(node, ["encoding"])
+			case "ms":
+				answer += AcrobatNode.getMathMLAttributes(node, ["open", "close"])
+			case _:
+				pass
+		answer += ">"
 		val = node.GetValue()
 		if val:
-			yield val
+			answer += val
 		else:
 			for childNum in range(node.GetChildCount()):
 				try:
 					subNode = node.GetChild(childNum).QueryInterface(IPDDomElement)
 				except COMError:
 					continue
 				for sub in self._getNodeMathMl(subNode):
-					yield sub
-		yield "</%s>" % tag
+					answer += sub
+		return answer + f"</{tag}>"
 
 	def _get_mathMl(self) -> str:
 		"""Return the MathML associated with a Formula tag"""
+		# There are two ways that MathML can be represented in a PDF:
+		# 1. As a series of nested tags, each with a MathML element as the value.
+		# 2. As a Formula tag with MathML as the value (comes from MathML in an Associated File)
 		if self.pdDomNode is None:
 			log.debugWarning("_get_mathMl: self.pdDomNode is None!")
 			raise LookupError
+
+		# see if it is MathML tagging is used
+		for childNum in range(self.pdDomNode.GetChildCount()):
+			try:
+				child = self.pdDomNode.GetChild(childNum).QueryInterface(IPDDomElement)
+			except COMError:
+				log.debugWarning(f"COMError trying to get {childNum=}")
+				continue
+			if log.isEnabledFor(log.DEBUG):
+				log.debug(f"\t(PDF) get_mathMl: tag={child.GetTagName()}")
+			if child.GetTagName() == "math":
+				answer = "".join(self._getNodeMathMl(child))
+				log.debug(f"_get_mathMl (PDF): found tagged MathML = {answer}")
+				return answer
+
 		mathMl = self.pdDomNode.GetValue()
 		if log.isEnabledFor(log.DEBUG):
 			log.debug(
 				(
-					f"_get_mathMl: math recognized: {mathMl.startswith('<math')}, "
+					f"_get_mathMl (PDF): math recognized: {mathMl.startswith('<math')}, "
 					f"child count={self.pdDomNode.GetChildCount()},"
-					f"\n  name='{self.pdDomNode.GetName()}', value='{mathMl}'"
+					f"\n  name='{self.pdDomNode.GetName()}', value found from AF ='{mathMl}'"
 				),
 			)
 		# this test and the replacement doesn't work if someone uses a namespace tag (which they shouldn't, but..)
 		if mathMl.startswith("<math"):
 			return mathMl.replace('xmlns:mml="http://www.w3.org/1998/Math/MathML"', "")
-		# Alternative for tagging: all the sub expressions are tagged -- gather up the MathML
-		for childNum in range(self.pdDomNode.GetChildCount()):
-			try:
-				child = self.pdDomNode.GetChild(childNum).QueryInterface(IPDDomElement)
-			except COMError:
-				log.debugWarning(f"COMError trying to get childNum={childNum}")
-				continue
-			if log.isEnabledFor(log.DEBUG):
-				log.debug(f"\tget_mathMl: tag={child.GetTagName()}")
-			if child.GetTagName() == "math":
-				return "".join(self._getNodeMathMl(child))
-		# fall back to return the contents, which is hopefully alt text
-		if log.isEnabledFor(log.DEBUG):
-			log.debug("_get_mathMl: didn't find MathML -- returning value as mtext")
-		return f"<math><mtext>{self.pdDomNode.GetValue()}</mtext></math>"
+
+		# not MathML -- fall back to return the contents, which is hopefully alt text, inside an <mtext>
+		answer = f"<math><mtext>{mathMl}</mtext></math>"
+		log.debug(f"_get_mathMl: didn't find MathML -- returning value as mtext: {answer}")
+		return answer
 
 
 class RootNode(AcrobatNode):

user_docs/en/changes.md

@@ -598,6 +598,7 @@ There are many minor bug fixes for applications, such as Thunderbird, Adobe Read
 
 ### Bug Fixes
 
+* Fixed math attributes being read in Adobe Reader, this resulted in poor or wrong speech and braille. (#17980)
 * Windows 11 fixes:
   * NVDA will once again announce hardware keyboard input suggestions. (#16283, @josephsl)
   * In Version 24H2 (2024 Update and Windows Server 2025), mouse and touch interaction can be used in quick settings. (#16348, @josephsl)