Merged changes from netmail to allow disabling of escaping invalid ch…

…aracters, and to add WJWOpenMemDocument

include/wjwriter.h

@@ -42,6 +42,19 @@ typedef struct {
 	*/
 	int					base;
 
+	/*
+		If set to TRUE (enabled by default) then any character that can not be
+		encoded as a UTF8 character will be written as \xXX where XX are a hex
+		representation of the value.
+
+		This can be disabled since it is non-standard JSON and will cause parse
+		errors with some parsers.
+
+		If disabled then any invalid characters will not be written to the
+		document.
+	*/
+	XplBool				escapeInvalidChars;
+
 	struct {
 		void			*data;
 		WJWriteCallback	cb;
@@ -115,6 +128,12 @@ EXPORT XplBool			WJWRawValue(char *name, char *value, XplBool done, WJWriter doc
 EXPORT size_t			WJWFileCallback(char *buffer, size_t length, void *data);
 #define WJWOpenFILEDocument(pretty, file) _WJWOpenDocument((pretty), WJWFileCallback, (file), 0)
 
+/*
+	An alternative method of opening a JSON document for writing, which will
+	provide a proper callback for allocating and writing to a memory allocation.
+*/
+EXPORT WJWriter			WJWOpenMemDocument(XplBool pretty, char **mem);
+
 #ifdef __cplusplus
 }
 #endif

src/wjwriter/wjwriter.c

@@ -267,10 +267,11 @@ EXPORT WJWriter _WJWOpenDocument(XplBool pretty, WJWriteCallback callback, void
 		The first value after opening a document should not be preceded by a
 		comma.  skipcomma will be reset after reading that first value.
 	*/
-	doc->public.pretty		= pretty;
-	doc->public.base		= 10;
-	doc->skipcomma			= TRUE;
-	doc->skipbreak			= TRUE;
+	doc->public.pretty				= pretty;
+	doc->public.escapeInvalidChars	= TRUE;
+	doc->public.base				= 10;
+	doc->skipcomma					= TRUE;
+	doc->skipbreak					= TRUE;
 
 	return((WJWriter) doc);
 }
@@ -336,6 +337,39 @@ EXPORT size_t WJWFileCallback(char *buffer, size_t length, void *data)
 	return(0);
 }
 
+static size_t WJWMemCallback(char *buffer, size_t length, void *data)
+{
+	char		**mem = data;
+	size_t		l;
+
+	if (mem) {
+		if (!*mem) {
+			*mem = MemMallocWait(length + 1);
+			memcpy(*mem, buffer, length);
+			(*mem)[length] = '\0';
+		} else {
+			l = strlen(*mem);
+
+			*mem = MemReallocWait(*mem, l + length + 1);
+			memcpy(*mem + l, buffer, length);
+			(*mem)[l + length] = '\0';
+		}
+
+		return(length);
+	}
+
+	return(0);
+}
+
+EXPORT WJWriter WJWOpenMemDocument(XplBool pretty, char **mem)
+{
+	if (!mem) {
+		return(NULL);
+	}
+
+	return(_WJWOpenDocument(pretty, WJWMemCallback, mem, 0));
+}
+
 /*
 	Verify that str points to a valid UTF8 character, and return the length of
 	that character in bytes.  If the value is not a full valid UTF8 character
@@ -440,8 +474,9 @@ static XplBool WJWriteString(char *value, size_t length, XplBool done, WJIWriter
 					e += (l - 1);
 				} else if (l == 1) {
 					/*
-						*e is valid UTF8 but is not a printable character, and will be escaped before
-						being sent, using the JSON-standard "\u00xx" form
+						*e is valid UTF8 but is not a printable character, and
+						will be escaped before being sent, using the
+						JSON-standard "\u00xx" form
 					*/
 					char	unicodeHex[sizeof("\\u0000")];
 
@@ -454,15 +489,18 @@ static XplBool WJWriteString(char *value, size_t length, XplBool done, WJIWriter
 				} else if (l < 0) {
 					/*
 						*e is not valid UTF8 data, and must be escaped before
-						being sent. But JSON-standard does not give us a mechanism
-						so we chose "\xhh" format because of its almost universal comprehension.
+						being sent. But JSON-standard does not give us a
+						mechanism so we chose "\xhh" format because of its
+						almost universal comprehension.
 					*/
 					char	nonUnicodeHex[sizeof("\\x00")];
 
 					WJWrite(doc, v, e - v);
 
-					sprintf(nonUnicodeHex, "\\x%02x", (unsigned char) *e);
-					WJWrite(doc, nonUnicodeHex, sizeof(nonUnicodeHex)-1);
+					if (doc->public.escapeInvalidChars) {
+						sprintf(nonUnicodeHex, "\\x%02x", (unsigned char) *e);
+						WJWrite(doc, nonUnicodeHex, sizeof(nonUnicodeHex)-1);
+					}
 
 					v = e + 1;
 				}