diff --git a/src/json.cc b/src/json.cc
index 98225167..b664ef15 100644
--- a/src/json.cc
+++ b/src/json.cc
@@ -2049,40 +2049,61 @@ std::string json::parser::parseString()
// the result of the parse process
std::string result;
- // iterate with pos_ over the whole string
- for (; pos_ < buffer_.size(); pos_++) {
+ // iterate with pos_ over the whole input until we found the end and return
+ // or we exit via error()
+ for (; pos_ < buffer_.size(); pos_++)
+ {
char currentChar = buffer_[pos_];
- // uneven amount of backslashes means the user wants to escape something
- if (!evenAmountOfBackslashes) {
+ if (!evenAmountOfBackslashes)
+ {
+ // uneven amount of backslashes means the user wants to escape something
+ // so we know there is a case such as '\X' or '\\\X' but we don't
+ // know yet what X is.
+ // at this point in the code, the currentChar has the value of X
// slash, backslash and quote are copied as is
if ( currentChar == '/'
|| currentChar == '\\'
- || currentChar == '"') {
+ || currentChar == '"')
+ {
result += currentChar;
- } else {
- // All other characters are replaced by their respective special character
- if (currentChar == 't') {
- result += '\t';
- } else if (currentChar == 'b') {
- result += '\b';
- } else if (currentChar == 'f') {
- result += '\f';
- } else if (currentChar == 'n') {
- result += '\n';
- } else if (currentChar == 'r') {
- result += '\r';
- } else if (currentChar == 'u') {
- pos_++;
- result += parseUnicodeEscape();
- } else {
- error("expected one of \\,/,b,f,n,r,t behind backslash.");
- }
- // TODO implement \uXXXX
}
- } else {
- if (currentChar == '"') {
+ else
+ {
+ // All other characters are replaced by their respective special character
+ if (currentChar == 't')
+ result += '\t';
+ else if (currentChar == 'b')
+ result += '\b';
+ else if (currentChar == 'f')
+ result += '\f';
+ else if (currentChar == 'n')
+ result += '\n';
+ else if (currentChar == 'r')
+ result += '\r';
+ else if (currentChar == 'u')
+ {
+ // \uXXXX[\uXXXX] is used for escaping unicode, which
+ // has it's own subroutine.
+ result += parseUnicodeEscape();
+ // the parsing process has brought us one step behind the
+ // unicode escape sequence:
+ // \uXXXX
+ // ^
+ // so we need to go one character back or the parser
+ // would skip the character we are currently pointing at
+ // (as the for-loop will drecement pos_ after this iteration).
+ pos_--;
+ }
+ else // user did something like \z and we should report a error
+ error("expected one of \\,/,b,f,n,r,t,u behind backslash.");
+ }
+ }
+ else
+ {
+ if (currentChar == '"')
+ {
// currentChar is a quote, so we found the end of the string
@@ -2093,7 +2114,9 @@ std::string json::parser::parseString()
// bring the result of the parsing process back to the caller
return result;
- } else if (currentChar != '\\') {
+ }
+ else if (currentChar != '\\')
+ {
// all non-backslash characters are added to the end of the result string.
// the only backslashes we want in the result are the ones that are escaped (which happens above).
result += currentChar;
@@ -2121,34 +2144,74 @@ std::string json::parser::parseString()
error("expected '\"'");
}
-std::string json::parser::unicodeToUTF8(unsigned int codepoint) {
- // it's just a ASCII compatible codepoint,
- // so we just interpret the point as a character
- if (codepoint <= 0x7f) {
+
+/*!
+Turns a code point into it's UTF-8 representation.
+You should only pass numbers < 0x10ffff into this function
+(everything else is a invalid code point).
+
+@return the UTF-8 representation of the given codepoint
+
+@pre This method isn't accessing the members of the parser
+
+@post This method isn't accessing the members of the parser
+*/
+std::string json::parser::codepointToUTF8(unsigned int codepoint)
+{
+ // this method contains a lot of bit manipulations to
+ // build the bytes for UTF-8.
+
+ // the '(... >> S) & 0xHH'-patterns are used to retrieve
+ // certain bits from the code points.
+
+ // all static casts in this method have boundary checks
+
+ // we initialize all strings with their final length
+ // (e.g. 1 to 4 bytes) to save the reallocations.
+
+
+ if (codepoint <= 0x7f)
+ {
+ // it's just a ASCII compatible codepoint,
+ // so we just interpret the point as a character
+ // and return ASCII
+
return std::string(1, static_cast<char>(codepoint));
}
+ // if true, we need two bytes to encode this as UTF-8
else if (codepoint <= 0x7ff)
{
- std::string result(2, static_cast<char>(0xc0 | ((codepoint >> 6) & 0x1f)));
- result[1] = static_cast<char>(0x80 | (codepoint & 0x3f));
+ // the 0xC0 enables the two most significant two bits
+ // to make this a two-byte UTF-8 character.
+ std::string result(2, static_cast<char>(0xC0 | ((codepoint >> 6) & 0x1F)));
+ result[1] = static_cast<char>(0x80 | (codepoint & 0x3F));
return result;
}
+ // if true, now we need three bytes to encode this as UTF-8
else if (codepoint <= 0xffff)
{
- std::string result(3, static_cast<char>(0xe0 | ((codepoint >> 12) & 0x0f)));
- result[1] = static_cast<char>(0x80 | ((codepoint >> 6) & 0x3f));
- result[2] = static_cast<char>(0x80 | (codepoint & 0x3f));
+ // the 0xE0 enables the three most significant two bits
+ // to make this a three-byte UTF-8 character.
+ std::string result(3, static_cast<char>(0xE0 | ((codepoint >> 12) & 0x0F)));
+ result[1] = static_cast<char>(0x80 | ((codepoint >> 6) & 0x3F));
+ result[2] = static_cast<char>(0x80 | (codepoint & 0x3F));
return result;
}
- else if (codepoint <= 0x1fffff)
+ // if true, we need maximal four bytes to encode this as UTF-8
+ else if (codepoint <= 0x10ffff)
{
- std::string result(4, static_cast<char>(0xf0 | ((codepoint >> 18) & 0x07)));
- result[1] = static_cast<char>(0x80 | ((codepoint >> 12) & 0x3f));
- result[2] = static_cast<char>(0x80 | ((codepoint >> 6) & 0x3f));
- result[3] = static_cast<char>(0x80 | (codepoint & 0x3f));
+ // the 0xE0 enables the four most significant two bits
+ // to make this a three-byte UTF-8 character.
+ std::string result(4, static_cast<char>(0xF0 | ((codepoint >> 18) & 0x07)));
+ result[1] = static_cast<char>(0x80 | ((codepoint >> 12) & 0x3F));
+ result[2] = static_cast<char>(0x80 | ((codepoint >> 6) & 0x3F));
+ result[3] = static_cast<char>(0x80 | (codepoint & 0x3F));
return result;
- } else {
+ }
+ else
+ {
+ // Can't be tested without direct access to this private method.
std::string errorMessage = "Invalid codepoint: ";
errorMessage += codepoint;
error(errorMessage);
@@ -2156,39 +2219,110 @@ std::string json::parser::unicodeToUTF8(unsigned int codepoint) {
}
/*!
-Parses the JSON style unicode escape sequence (\uXXXX).
+Parses 4 hexadecimal characters as a number.
-@return the utf-8 character the escape sequence escaped
+@return the value of the number the hexadecimal characters represent.
-@pre An opening quote \p " was read in the main parse function @ref parse.
- pos_ is the position after the opening quote.
+@pre pos_ is pointing to the first of the 4 hexadecimal characters.
-@post The character after the closing quote \p " is the current character @ref
- current_. Whitespace is skipped.
+@post pos_ is pointing to the character after the 4 hexadecimal characters.
*/
-std::string json::parser::parseUnicodeEscape() {
+unsigned int json::parser::parse4HexCodepoint()
+{
const auto startPos = pos_;
- if (pos_ + 3 >= buffer_.size()) {
+
+ // check if the remaining buffer is long enough to even hold 4 characters
+ if (pos_ + 3 >= buffer_.size())
+ {
error("Got end of input while parsing unicode escape sequence \\uXXXX");
}
+
+ // make a string that can hold the pair
std::string hexCode(4, ' ');
- for(; pos_ < startPos + 4; pos_++) {
+
+ for(; pos_ < startPos + 4; pos_++)
+ {
+ // no boundary check here as we already checked above
char currentChar = buffer_[pos_];
+
+ // check if we have a hexadecimal character
if ( (currentChar >= '0' && currentChar <= '9')
|| (currentChar >= 'a' && currentChar <= 'f')
- || (currentChar >= 'A' && currentChar <= 'F')) {
+ || (currentChar >= 'A' && currentChar <= 'F'))
+ {
// all is well, we have valid hexadecimal chars
// so we copy that char into our string
hexCode[pos_ - startPos] = currentChar;
- } else {
+ }
+ else
+ {
error("Found non-hexadecimal character in unicode escape sequence!");
}
}
- pos_--;
- // case is safe as 4 hex characters can't present more than 16 bits
- return unicodeToUTF8(static_cast<unsigned int>(std::stoul(hexCode, nullptr, 16)));
+ // the cast is safe as 4 hex characters can't present more than 16 bits
+ // the input to stoul was checked to contain only hexadecimal characters (see above)
+ return static_cast<unsigned int>(std::stoul(hexCode, nullptr, 16));
}
+/*!
+Parses the unicode escape codes as defined in the ECMA-404.
+The escape sequence has two forms:
+1. \uXXXX
+2. \uXXXX\uYYYY
+where X and Y are a hexadecimal character (a-zA-Z0-9).
+
+Form 1 just contains the unicode code point in the hexadecimal number XXXX.
+Form 2 is encoding a UTF-16 surrogate pair. The high surrogate is XXXX, the low surrogate is YYYY.
+
+@return the UTF-8 character this unicode escape sequence escaped.
+
+@pre pos_ is pointing at at the 'u' behind the first backslash.
+
+@post pos_ is pointing at the character behind the last X (or Y in form 2).
+*/
+std::string json::parser::parseUnicodeEscape()
+{
+ // jump to the first hex value
+ pos_++;
+ // parse the hex first hex values
+ unsigned int firstCodepoint = parse4HexCodepoint();
+
+
+ if (firstCodepoint >= 0xD800 && firstCodepoint <= 0xDBFF)
+ {
+ // we found invalid code points, which means we either have a malformed input
+ // or we found a high surrogate.
+ // we can only find out by seeing if the next character also wants to encode
+ // a unicode character (so, we have the \uXXXX\uXXXX case here).
+
+ // jump behind the next \u
+ pos_ += 2;
+ // try to parse the next hex values.
+ // the method does boundary checking for us, so no need to do that here
+ unsigned secondCodepoint = parse4HexCodepoint();
+ // ok, we have a low surrogate, check if it is a valid one
+ if (secondCodepoint >= 0xDC00 && secondCodepoint <= 0xDFFF)
+ {
+ // calculate the final code point from the pair according to the spec
+ unsigned int finalCodePoint =
+ // high surrogate occupies the most significant 22 bits
+ (firstCodepoint << 10)
+ // low surrogate occupies the least significant 15 bits
+ + secondCodepoint
+ // there is still the 0xD800, 0xDC00 and 0x10000 noise in the result
+ // so we have to substract with (0xD800 << 10) + DC00 - 0x10000 = 0x35FDC00
+ - 0x35FDC00;
+
+ // we transform the calculated point into UTF-8
+ return codepointToUTF8(finalCodePoint);
+ }
+ else
+ error("missing low surrogate");
+
+ }
+ // We have Form 1, so we just interpret the XXXX as a code point
+ return codepointToUTF8(firstCodepoint);
+}
/*!
diff --git a/src/json.h b/src/json.h
index 1b5e8fcf..90c5ded8 100644
--- a/src/json.h
+++ b/src/json.h
@@ -419,8 +419,10 @@ class json
/// parse a quoted string
inline std::string parseString();
/// transforms a unicode codepoint to it's UTF-8 presentation
- inline std::string unicodeToUTF8(unsigned int codepoint);
- /// parses a unicode escape sequence
+ inline std::string codepointToUTF8(unsigned int codepoint);
+ /// parses 4 hex characters that represent a unicode codepoint
+ inline unsigned int parse4HexCodepoint();
+ /// parses \uXXXX[\uXXXX] unicode escape characters
inline std::string parseUnicodeEscape();
/// parse a Boolean "true"
inline void parseTrue();
diff --git a/test/json_unit.cc b/test/json_unit.cc
index fb89a2a1..ab679fbf 100644
--- a/test/json_unit.cc
+++ b/test/json_unit.cc
@@ -1652,10 +1652,6 @@ TEST_CASE("Parser")
CHECK(json::parse("\"a\\nz\"") == json("a\nz"));
CHECK(json::parse("\"\\n\"") == json("\n"));
- // escape unicode characters
- CHECK(json::parse("\"\\u002F\"") == json("/"));
- CHECK(json::parse("\"\\u00E4\"") == json(u8"\u00E4"));
-
// escaping senseless stuff
CHECK_THROWS_AS(json::parse("\"\\z\""), std::invalid_argument);
CHECK_THROWS_AS(json::parse("\"\\ \""), std::invalid_argument);
@@ -1665,6 +1661,44 @@ TEST_CASE("Parser")
CHECK_THROWS_AS(json::parse("\""), std::invalid_argument);
}
+ SECTION("unicode_escaping")
+ {
+ // two tests for uppercase and lowercase hex
+
+ // normal forward slash in ASCII range
+ CHECK(json::parse("\"\\u002F\"") == json("/"));
+ CHECK(json::parse("\"\\u002f\"") == json("/"));
+ // german a umlaut
+ CHECK(json::parse("\"\\u00E4\"") == json(u8"\u00E4"));
+ CHECK(json::parse("\"\\u00e4\"") == json(u8"\u00E4"));
+ // weird d
+ CHECK(json::parse("\"\\u0111\"") == json(u8"\u0111"));
+ // unicode arrow left
+ CHECK(json::parse("\"\\u2190\"") == json(u8"\u2190"));
+ // pleasing osiris by testing hieroglyph support
+ CHECK(json::parse("\"\\uD80C\\uDC60\"") == json(u8"\U00013060"));
+ CHECK(json::parse("\"\\ud80C\\udc60\"") == json(u8"\U00013060"));
+
+
+ // no hex numbers behind the \u
+ CHECK_THROWS_AS(json::parse("\"\\uD80v\""), std::invalid_argument);
+ CHECK_THROWS_AS(json::parse("\"\\uD80 A\""), std::invalid_argument);
+ CHECK_THROWS_AS(json::parse("\"\\uD8v\""), std::invalid_argument);
+ CHECK_THROWS_AS(json::parse("\"\\uDv\""), std::invalid_argument);
+ CHECK_THROWS_AS(json::parse("\"\\uv\""), std::invalid_argument);
+ CHECK_THROWS_AS(json::parse("\"\\u\""), std::invalid_argument);
+ CHECK_THROWS_AS(json::parse("\"\\u\\u\""), std::invalid_argument);
+ CHECK_THROWS_AS(json::parse("\"a\\uD80vAz\""), std::invalid_argument);
+ // missing part of a surrogate pair
+ CHECK_THROWS_AS(json::parse("\"bla \\uD80C bla\""), std::invalid_argument);
+ CHECK_THROWS_AS(json::parse("\"\\uD80C bla bla\""), std::invalid_argument);
+ CHECK_THROWS_AS(json::parse("\"bla bla \\uD80C bla bla\""), std::invalid_argument);
+ // senseless surrogate pair
+ CHECK_THROWS_AS(json::parse("\"\\uD80C\\uD80C\""), std::invalid_argument);
+ CHECK_THROWS_AS(json::parse("\"\\uD80C\\u0000\""), std::invalid_argument);
+ CHECK_THROWS_AS(json::parse("\"\\uD80C\\uFFFF\""), std::invalid_argument);
+ }
+
SECTION("boolean")
{
// accept the exact values