property_tree/xml_parser is conceptually flawed, because it can't parse XMLs with the mix of texts and child trees

[yurivict]

This XML:

<?xml version="1.0" encoding="UTF-8"?>
<tag>
  Text1
  <tag1>a</tag1>
  Text2
  <tag2>b</tag2>
  Text3
</tag>

parse result has all texts (Text1, Text2, Text3) combined into a single string that is available as the output of the data() function:

data="
  Text1
  
  Text2
  
  Text3
"

and boost leaves no way to determine that <tag1> and <tag2> are in the middle of this text. tag1 and tag2 are available as separate keys, but this loses information about their location.

A larger practical example:
The line <param len="4">const <ptype>GLfloat</ptype> *<name>v</name></param> in https://raw.githubusercontent.com/KhronosGroup/OpenGL-Registry/master/xml/gl.xml can't be parsed using property_tree/xml_parser because the <ptype> tag is between two text elements.