THRIFT-2411 - C++: Fixed support for UTF-16 encoding in JSON protocol

Client: C++
Patch: Phongphan Phuttha <phongphan@acm.org>

Support unicode-encoded character including UTF-16 surrogate pair which
allow extended character that is not in Basic Multilingual Plane.

This closes #648
This commit is contained in:
Konrad Grochowski 2015-10-16 11:22:10 +02:00
parent 86da51d2ae
commit a84e139215
3 changed files with 116 additions and 8 deletions

View File

@ -26,6 +26,7 @@
#include <boost/math/special_functions/fpclassify.hpp> #include <boost/math/special_functions/fpclassify.hpp>
#include <boost/lexical_cast.hpp> #include <boost/lexical_cast.hpp>
#include <boost/locale.hpp>
#include <thrift/protocol/TBase64Utils.h> #include <thrift/protocol/TBase64Utils.h>
#include <thrift/transport/TTransportException.h> #include <thrift/transport/TTransportException.h>
@ -285,6 +286,16 @@ static bool isJSONNumeric(uint8_t ch) {
return false; return false;
} }
// Return true if the code unit is high surrogate
static bool isHighSurrogate(uint16_t val) {
return val >= 0xD800 && val <= 0xDBFF;
}
// Return true if the code unit is low surrogate
static bool isLowSurrogate(uint16_t val) {
return val >= 0xDC00 && val <= 0xDFFF;
}
/** /**
* Class to serve as base JSON context and as base class for other context * Class to serve as base JSON context and as base class for other context
* implementations * implementations
@ -709,14 +720,17 @@ uint32_t TJSONProtocol::readJSONSyntaxChar(uint8_t ch) {
} }
// Decodes the four hex parts of a JSON escaped string character and returns // Decodes the four hex parts of a JSON escaped string character and returns
// the character via out. The first two characters must be "00". // the UTF-16 code unit via out.
uint32_t TJSONProtocol::readJSONEscapeChar(uint8_t* out) { uint32_t TJSONProtocol::readJSONEscapeChar(uint16_t* out) {
uint8_t b[2]; uint8_t b[4];
readJSONSyntaxChar(kJSONZeroChar);
readJSONSyntaxChar(kJSONZeroChar);
b[0] = reader_.read(); b[0] = reader_.read();
b[1] = reader_.read(); b[1] = reader_.read();
*out = (hexVal(b[0]) << 4) + hexVal(b[1]); b[2] = reader_.read();
b[3] = reader_.read();
*out = (hexVal(b[0]) << 12)
+ (hexVal(b[1]) << 8) + (hexVal(b[2]) << 4) + hexVal(b[3]);
return 4; return 4;
} }
@ -724,6 +738,7 @@ uint32_t TJSONProtocol::readJSONEscapeChar(uint8_t* out) {
uint32_t TJSONProtocol::readJSONString(std::string& str, bool skipContext) { uint32_t TJSONProtocol::readJSONString(std::string& str, bool skipContext) {
uint32_t result = (skipContext ? 0 : context_->read(reader_)); uint32_t result = (skipContext ? 0 : context_->read(reader_));
result += readJSONSyntaxChar(kJSONStringDelimiter); result += readJSONSyntaxChar(kJSONStringDelimiter);
std::vector<uint16_t> codeunits;
uint8_t ch; uint8_t ch;
str.clear(); str.clear();
while (true) { while (true) {
@ -736,7 +751,22 @@ uint32_t TJSONProtocol::readJSONString(std::string& str, bool skipContext) {
ch = reader_.read(); ch = reader_.read();
++result; ++result;
if (ch == kJSONEscapeChar) { if (ch == kJSONEscapeChar) {
result += readJSONEscapeChar(&ch); uint16_t cp;
result += readJSONEscapeChar(&cp);
if (isHighSurrogate(cp)) {
codeunits.push_back(cp);
} else {
if (isLowSurrogate(cp)
&& codeunits.empty()) {
throw TProtocolException(TProtocolException::INVALID_DATA,
"Missing UTF-16 high surrogate pair.");
}
codeunits.push_back(cp);
codeunits.push_back(0);
str += boost::locale::conv::utf_to_utf<char>(codeunits.data());
codeunits.clear();
}
continue;
} else { } else {
size_t pos = kEscapeChars.find(ch); size_t pos = kEscapeChars.find(ch);
if (pos == std::string::npos) { if (pos == std::string::npos) {
@ -747,8 +777,17 @@ uint32_t TJSONProtocol::readJSONString(std::string& str, bool skipContext) {
ch = kEscapeCharVals[pos]; ch = kEscapeCharVals[pos];
} }
} }
if (!codeunits.empty()) {
throw TProtocolException(TProtocolException::INVALID_DATA,
"Missing UTF-16 low surrogate pair.");
}
str += ch; str += ch;
} }
if (!codeunits.empty()) {
throw TProtocolException(TProtocolException::INVALID_DATA,
"Missing UTF-16 low surrogate pair.");
}
return result; return result;
} }

View File

@ -127,7 +127,7 @@ private:
uint32_t readJSONSyntaxChar(uint8_t ch); uint32_t readJSONSyntaxChar(uint8_t ch);
uint32_t readJSONEscapeChar(uint8_t* out); uint32_t readJSONEscapeChar(uint16_t* out);
uint32_t readJSONString(std::string& str, bool skipContext = false); uint32_t readJSONString(std::string& str, bool skipContext = false);

View File

@ -19,6 +19,8 @@
#define _USE_MATH_DEFINES #define _USE_MATH_DEFINES
#include <cmath> #include <cmath>
#include <iomanip>
#include <sstream>
#include <thrift/transport/TBufferTransports.h> #include <thrift/transport/TBufferTransports.h>
#include <thrift/protocol/TJSONProtocol.h> #include <thrift/protocol/TJSONProtocol.h>
#include "gen-cpp/DebugProtoTest_types.h" #include "gen-cpp/DebugProtoTest_types.h"
@ -269,3 +271,70 @@ BOOST_AUTO_TEST_CASE(test_json_proto_8) {
BOOST_CHECK_THROW(ooe2.read(proto.get()), BOOST_CHECK_THROW(ooe2.read(proto.get()),
apache::thrift::protocol::TProtocolException); apache::thrift::protocol::TProtocolException);
} }
static std::string toHexSequence(const std::string& str) {
std::stringstream ss;
ss << std::hex << std::setfill('0');
for (std::size_t i = 0; i < str.size(); i++) {
ss << "\\x" << int(uint8_t(str[i]));
}
return ss.str();
}
BOOST_AUTO_TEST_CASE(test_json_unicode_escaped) {
const char json_string[] =
"{\"1\":{\"tf\":1},\"2\":{\"tf\":0},\"3\":{\"i8\":127},\"4\":{\"i16\":27000},"
"\"5\":{\"i32\":16},\"6\":{\"i64\":6000000000},\"7\":{\"dbl\":3.1415926"
"53589793},\"8\":{\"str\":\"JSON THIS!\"},\"9\":{\"str\":\"\\u0e01 \\ud835\\udd3e\"},"
"\"10\":{\"tf\":0},\"11\":{\"str\":\"000000\"},\"12\":{\"lst\""
":[\"i8\",3,1,2,3]},\"13\":{\"lst\":[\"i16\",3,1,2,3]},\"14\":{\"lst\":[\"i64"
"\",3,1,2,3]}}";
const char* expected_zomg_unicode = "\xe0\xb8\x81 \xf0\x9d\x94\xbe";
boost::shared_ptr<TMemoryBuffer> buffer(new TMemoryBuffer(
(uint8_t*)(json_string), sizeof(json_string)));
boost::shared_ptr<TJSONProtocol> proto(new TJSONProtocol(buffer));
OneOfEach ooe2;
ooe2.read(proto.get());
BOOST_CHECK_MESSAGE(!ooe2.zomg_unicode.compare(expected_zomg_unicode),
"Expected:\n" << toHexSequence(expected_zomg_unicode) << "\nGotten:\n"
<< toHexSequence(ooe2.zomg_unicode));
}
BOOST_AUTO_TEST_CASE(test_json_unicode_escaped_missing_low_surrogate) {
const char json_string[] =
"{\"1\":{\"tf\":1},\"2\":{\"tf\":0},\"3\":{\"i8\":127},\"4\":{\"i16\":27000},"
"\"5\":{\"i32\":16},\"6\":{\"i64\":6000000000},\"7\":{\"dbl\":3.1415926"
"53589793},\"8\":{\"str\":\"JSON THIS!\"},\"9\":{\"str\":\"\\ud835\"},"
"\"10\":{\"tf\":0},\"11\":{\"str\":\"000000\"},\"12\":{\"lst\""
":[\"i8\",3,1,2,3]},\"13\":{\"lst\":[\"i16\",3,1,2,3]},\"14\":{\"lst\":[\"i64"
"\",3,1,2,3]}}";
boost::shared_ptr<TMemoryBuffer> buffer(new TMemoryBuffer(
(uint8_t*)(json_string), sizeof(json_string)));
boost::shared_ptr<TJSONProtocol> proto(new TJSONProtocol(buffer));
OneOfEach ooe2;
BOOST_CHECK_THROW(ooe2.read(proto.get()),
apache::thrift::protocol::TProtocolException);
}
BOOST_AUTO_TEST_CASE(test_json_unicode_escaped_missing_hi_surrogate) {
const char json_string[] =
"{\"1\":{\"tf\":1},\"2\":{\"tf\":0},\"3\":{\"i8\":127},\"4\":{\"i16\":27000},"
"\"5\":{\"i32\":16},\"6\":{\"i64\":6000000000},\"7\":{\"dbl\":3.1415926"
"53589793},\"8\":{\"str\":\"JSON THIS!\"},\"9\":{\"str\":\"\\udd3e\"},"
"\"10\":{\"tf\":0},\"11\":{\"str\":\"000000\"},\"12\":{\"lst\""
":[\"i8\",3,1,2,3]},\"13\":{\"lst\":[\"i16\",3,1,2,3]},\"14\":{\"lst\":[\"i64"
"\",3,1,2,3]}}";
boost::shared_ptr<TMemoryBuffer> buffer(new TMemoryBuffer(
(uint8_t*)(json_string), sizeof(json_string)));
boost::shared_ptr<TJSONProtocol> proto(new TJSONProtocol(buffer));
OneOfEach ooe2;
BOOST_CHECK_THROW(ooe2.read(proto.get()),
apache::thrift::protocol::TProtocolException);
}