Introduced parse_ws_pcdata_single flag: only parses whitespace-only PCDATA if it's the only child of the parent node (middle ground between default flags and parse_ws_pcdata)

git-svn-id: http://pugixml.googlecode.com/svn/trunk@825 99668b35-9821-0410-8761-19e4c4f06640

2 files changed, 20 insertions, 5 deletions

diff --git a/src/pugixml.cpp b/src/pugixml.cpp
index 3680fc5..a3c6abd 100644
--- a/src/pugixml.cpp
+++ b/src/pugixml.cpp
@@ -1898,7 +1898,7 @@ namespace
 		
 		// Parser utilities.
 		#define SKIPWS()			{ while (IS_CHARTYPE(*s, ct_space)) ++s; }
-		#define OPTSET(OPT)			( optmsk & OPT )
+		#define OPTSET(OPT)			( optmsk & (OPT) )
 		#define PUSHNODE(TYPE)		{ cursor = append_node(cursor, alloc, TYPE); if (!cursor) THROW_ERROR(status_out_of_memory, s); }
 		#define POPNODE()			{ cursor = cursor->parent; }
 		#define SCANFOR(X)			{ while (*s != 0 && !(X)) ++s; }
@@ -2402,10 +2402,20 @@ namespace
 
 					SKIPWS(); // Eat whitespace if no genuine PCDATA here.
 
-					if ((!OPTSET(parse_ws_pcdata) || mark == s) && (*s == '<' || !*s))
-					{
-						continue;
-					}
+                    if (*s == '<')
+                    {
+                        // We skipped some whitespace characters because otherwise we would take the tag branch instead of PCDATA one
+                        assert(mark != s);
+
+                        if (!OPTSET(parse_ws_pcdata | parse_ws_pcdata_single))
+                        {
+                            continue;
+                        }
+                        else if (OPTSET(parse_ws_pcdata_single))
+                        {
+                            if (s[1] != '/' || cursor->first_child) continue;
+                        }
+                    }
 
 					s = mark;
 							
diff --git a/src/pugixml.hpp b/src/pugixml.hpp
index 1826b45..11bf279 100644
--- a/src/pugixml.hpp
+++ b/src/pugixml.hpp
@@ -164,6 +164,11 @@ namespace pugi
     // This flag determines if document type declaration (node_doctype) is added to the DOM tree. This flag is off by default.
 	const unsigned int parse_doctype = 0x0200;
 
+	// This flag determines if plain character data (node_pcdata) that is the only child of the parent node and that consists only
+    // of whitespace is added to the DOM tree.
+    // This flag is off by default; turning it on may result in slower parsing and more memory consumption.
+	const unsigned int parse_ws_pcdata_single = 0x0400;
+
 	// 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.