Add initial support for parse_embed_pcdata

When this flag is true, PCDATA value is saved to the parent element instead of
allocating a new node.

This prevents some documents from round-tripping since it loses information,
but can provide a significant memory reduction and parsing speedup for some
documents.

2 files changed, 18 insertions, 4 deletions

diff --git a/src/pugixml.cpp b/src/pugixml.cpp
index 35c0d8e..de87dcf 100644
--- a/src/pugixml.cpp
+++ b/src/pugixml.cpp
@@ -3360,13 +3360,22 @@ PUGI__NS_BEGIN
 							
 					if (cursor->parent || PUGI__OPTSET(parse_fragment))
 					{
-						PUGI__PUSHNODE(node_pcdata); // Append a new node on the tree.
-						cursor->value = s; // Save the offset.
+						if (!PUGI__OPTSET(parse_embed_pcdata))
+						{
+							PUGI__PUSHNODE(node_pcdata); // Append a new node on the tree.
+
+							cursor->value = s; // Save the offset.
+
+							PUGI__POPNODE(); // Pop since this is a standalone.
+						}
+						else
+						{
+							if (cursor->parent && !cursor->value)
+								cursor->value = s; // Save the offset.
+						}
 
 						s = strconv_pcdata(s);
 								
-						PUGI__POPNODE(); // Pop since this is a standalone.
-						
 						if (!*s) break;
 					}
 					else
diff --git a/src/pugixml.hpp b/src/pugixml.hpp
index 540e6ba..4ed6f55 100644
--- a/src/pugixml.hpp
+++ b/src/pugixml.hpp
@@ -158,6 +158,11 @@ namespace pugi
 	// is a valid document. This flag is off by default.
 	const unsigned int parse_fragment = 0x1000;
 
+	// This flag determines if plain character data is be stored in the parent element's value. This significantly changes the structure of
+	// the document and does not allow some documents to round-trip; this flag is only recommended for parsing documents with a lot of
+	// PCDATA nodes in a very memory-constrained environment. This flag is off by default.
+	const unsigned int parse_embed_pcdata = 0x2000;
+
 	// The default parsing mode.
 	// Elements, PCDATA and CDATA sections are added to the DOM tree, character/reference entities are expanded,
 	// End-of-Line characters are normalized, attribute values are normalized using CDATA normalization rules.