Handle multi-byte unicode characters in json parsing (#1023)

* handle multi-byte unicode characters in json parsing * Properly check high surrogate start and end Co-Authored-By: Tymolc <[email protected]>
microsoft · Jan 28, 2019 · 5a885dd · 5a885dd
1 parent 8f0393d
commit 5a885dd
Show file tree

Hide file tree

Showing 2 changed files with 76 additions and 21 deletions.
diff --git a/Release/src/json/json_parsing.cpp b/Release/src/json/json_parsing.cpp
@@ -139,6 +139,7 @@ class JSON_Parser
 
     virtual bool CompleteComment(Token& token);
     virtual bool CompleteStringLiteral(Token& token);
+    int convert_unicode_to_code_point();
     bool handle_unescape_char(Token& token);
 
 private:
@@ -652,7 +653,15 @@ bool JSON_StringParser<CharType>::CompleteComment(typename JSON_Parser<CharType>
     return true;
 }
 
-void convert_append_unicode_code_unit(JSON_Parser<wchar_t>::Token& token, utf16char value)
+void convert_append_unicode_code_unit(JSON_Parser<utf16char>::Token& token, utf16string value)
+{
+    token.string_val.append(value);
+}
+void convert_append_unicode_code_unit(JSON_Parser<char>::Token& token, utf16string value)
+{
+    token.string_val.append(::utility::conversions::utf16_to_utf8(value));
+}
+void convert_append_unicode_code_unit(JSON_Parser<utf16char>::Token& token, utf16char value)
 {
     token.string_val.push_back(value);
 }
@@ -662,6 +671,37 @@ void convert_append_unicode_code_unit(JSON_Parser<char>::Token& token, utf16char
     token.string_val.append(::utility::conversions::utf16_to_utf8(utf16));
 }
 
+template<typename CharType>
+int JSON_Parser<CharType>::convert_unicode_to_code_point()
+{
+    // A four-hexdigit Unicode character.
+    // Transform into a 16 bit code point.
+    int decoded = 0;
+    for (int i = 0; i < 4; ++i)
+    {
+        auto ch = NextCharacter();
+        int ch_int = static_cast<int>(ch);
+        if (ch_int < 0 || ch_int > 127) return -1;
+#ifdef _WIN32
+        const int isxdigitResult = _isxdigit_l(ch_int, utility::details::scoped_c_thread_locale::c_locale());
+#else
+        const int isxdigitResult = isxdigit(ch_int);
+#endif
+        if (!isxdigitResult) return -1;
+
+        int val = _hexval[static_cast<size_t>(ch_int)];
+
+        _ASSERTE(val != -1);
+
+        // Add the input char to the decoded number
+        decoded |= (val << (4 * (3 - i)));
+    }
+    return decoded;
+}
+
+#define H_SURROGATE_START 0xD800
+#define H_SURROGATE_END 0xDBFF
+
 template<typename CharType>
 inline bool JSON_Parser<CharType>::handle_unescape_char(Token& token)
 {
@@ -682,26 +722,31 @@ inline bool JSON_Parser<CharType>::handle_unescape_char(Token& token)
         case 't': token.string_val.push_back('\t'); return true;
         case 'u':
         {
-            // A four-hexdigit Unicode character.
-            // Transform into a 16 bit code point.
-            int decoded = 0;
-            for (int i = 0; i < 4; ++i)
+            int decoded = convert_unicode_to_code_point();
+            if (decoded == -1)
             {
-                ch = NextCharacter();
-                int ch_int = static_cast<int>(ch);
-                if (ch_int < 0 || ch_int > 127) return false;
-#ifdef _WIN32
-                const int isxdigitResult = _isxdigit_l(ch_int, utility::details::scoped_c_thread_locale::c_locale());
-#else
-                const int isxdigitResult = isxdigit(ch_int);
-#endif
-                if (!isxdigitResult) return false;
+                return false;
+            }
+
+            // handle multi-block characters that start with a high-surrogate
+            if (decoded >= H_SURROGATE_START && decoded <= H_SURROGATE_END)
+            {
+                // skip escape character '\u'
+                if (NextCharacter() != '\\' || NextCharacter() != 'u')
+                {
+                    return false;
+                }
+                int decoded2 = convert_unicode_to_code_point();
+
+                if (decoded2 == -1)
+                {
+                    return false;
+                }
 
-                int val = _hexval[static_cast<size_t>(ch_int)];
-                _ASSERTE(val != -1);
+                utf16string compoundUTF16 = {static_cast<utf16char>(decoded), static_cast<utf16char>(decoded2)};
+                convert_append_unicode_code_unit(token, compoundUTF16);
 
-                // Add the input char to the decoded number
-                decoded |= (val << (4 * (3 - i)));
+                return true;
             }
 
             // Construct the character based on the decoded number
@@ -1015,9 +1060,13 @@ std::unique_ptr<web::json::details::_Value> JSON_Parser<CharType>::_ParseValue(
 {
     switch (tkn.kind)
     {
-        case JSON_Parser<CharType>::Token::TKN_OpenBrace: { return _ParseObject(tkn);
+        case JSON_Parser<CharType>::Token::TKN_OpenBrace:
+        {
+            return _ParseObject(tkn);
         }
-        case JSON_Parser<CharType>::Token::TKN_OpenBracket: { return _ParseArray(tkn);
+        case JSON_Parser<CharType>::Token::TKN_OpenBracket:
+        {
+            return _ParseArray(tkn);
         }
         case JSON_Parser<CharType>::Token::TKN_StringLiteral:
         {

diff --git a/Release/tests/functional/json/parsing_tests.cpp b/Release/tests/functional/json/parsing_tests.cpp
@@ -159,7 +159,7 @@ SUITE(parsing_tests)
             input.append(2, ch);
             json::value val = json::value::parse(input);
             VERIFY_IS_TRUE(val.is_object());
-        VERIFY_ARE_EQUAL(U("2"), val[U("1"]).serialize());
+            VERIFY_ARE_EQUAL(U("2"), val[U("1")].serialize());
         }
     }
 
@@ -213,6 +213,12 @@ SUITE(parsing_tests)
         const auto euro = to_string_t("\xE2\x82\xAC");
         VERIFY_ARE_EQUAL(euro, str.as_string());
 
+        // UTF-16 character with surrogate pair
+        str = json::value::parse(U("\"\\ud83d\\ude00\""));
+        // Grinning Face emoji as a hexadecimal UTF-8
+        const auto emoji = to_string_t("\xF0\x9F\x98\x80");
+        VERIFY_ARE_EQUAL(emoji, str.as_string());
+
         VERIFY_PARSING_THROW(json::value::parse(U("\"\\u0klB\"")));
     }