几年前是用的0.8版本的，python与C++下使用情况一致。

这次别人用时发现不正确，即遇到了这种情况load a python trained xgboost in c++ api, the predict result is empty - Stack Overflow 即

auto ret = XGBoosterPredict(booster_, data, 0, 0, &len, &out);
  if (ret < 0) {
    LOG(ERROR) << "xgboost inference error !";
    return -1;
  }
  std::cout << "len=" << len << std::endl;

即类似这样，函数返回0，而且out_len输出也是0！！！找了很久，发现原来anaconda升级了xgboost，xgboost已经是V1.4.0版本。而我C++下依旧用V0.8就会导致这种情况。而且无任何报错。

后来重新编译xgboostV1.4.0版本，生成xgboostV1.4.0的动态库。

但测试时如果依旧用以前的代码，会报错：

bst_ulong const *outshape;
ret=XGBoosterPredictFromDMatrix(h_booster,h_test,
		R"({"type": 0,"training": False,"iteration_begin": 0,"iteration_end": 0,"strict_shape": true})",
		&outshape,&out_len,&f);
Error:symbol lookup error: /home/jumper/projects/XRT_PC_Stream_Mathprocess_Simulation/Debug/XRT_PC_Stream_Mathprocess_Simulation: undefined symbol: XGBoosterPredictFromDMatrix

XGBoost Predict flag -1, info: [13:48:34] /home/jumper/workspace/xgboost1.4/xgboost/src/common/json.cc:458: Unknown construct, around character position: 23
    ining": False",i
    ~~~~~~~^~~~~~~~~
Stack trace:
  [bt] (0) /usr/lib/libxgboost.so(dmlc::LogMessageFatal::~LogMessageFatal()+0x61) [0x7ffff21ee571]
  [bt] (1) /usr/lib/libxgboost.so(xgboost::JsonReader::Error(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) const+0xd30) [0x7ffff222a500]
  [bt] (2) /usr/lib/libxgboost.so(xgboost::JsonReader::Parse()+0x112) [0x7ffff222afa2]
  [bt] (3) /usr/lib/libxgboost.so(xgboost::JsonReader::ParseObject()+0x42a) [0x7ffff222ed6a]
  [bt] (4) /usr/lib/libxgboost.so(xgboost::JsonReader::Parse()+0x1ac) [0x7ffff222b03c]
  [bt] (5) /usr/lib/libxgboost.so(xgboost::JsonReader::Load()+0x1d) [0x7ffff222b0dd]
  [bt] (6) /usr/lib/libxgboost.so(xgboost::Json::Load(xgboost::StringView)+0x42) [0x7ffff222b142]
  [bt] (7) /usr/lib/libxgboost.so(XGBoosterPredictFromDMatrix+0x6d) [0x7ffff21e206d]
  [bt] (8) /home/jumper/projects/XRT_PC_Stream_Mathprocess_Simulation/Debug/XRT_PC_Stream_Mathprocess_Simulation() [0x40aaa9]

找了很久，原来加入了json格式，试了很久

//solution:
						BoosterHandle h_booster;
						int ret=XGBoosterCreate(0,0,&h_booster);
						printf("XGBoost initialization flag: %d\n",ret);
						ret=XGBoosterLoadModel(h_booster,"/home/jumper/projects/xgbmodel/xgb_32.model");
						printf("XGBoost Load model flag: %d\n\n",ret);

						float f_testdata[featuresize]={0.0f};
						for(int feaid=0;feaid!=featuresize;feaid++)
						{
							if(feaid<histsize)
							{
								f_testdata[feaid]=channel1hist.ptr<float>(feaid)[0];
//								cout<<feaid<<":"<<f_testdata[feaid]<<endl;
							}
							else
							{
								f_testdata[feaid]=channel2hist.at<float>(feaid-histsize,0);
//								cout<<feaid<<":"<<f_testdata[feaid]<<endl;
							}
						}
						DMatrixHandle h_test;
						ret=XGDMatrixCreateFromMat((float *)f_testdata,1,featuresize,-1,&h_test);
						printf("XGBoost creates features matrix flag %d, info: %s\n",ret,XGBGetLastError());

						bst_ulong out_len;
						const float *f;
						bst_ulong const *outshape;

						//1:wrong
//						ret=XGBoosterPredictFromDMatrix(h_booster,h_test,
//								R"({"type": 0,"training": False,"iteration_begin": 0,"iteration_end": 0,"strict_shape": true})",
//								&outshape,&out_len,&f);

						//2:wrong
//						ret=XGBoosterPredictFromDMatrix(h_booster,h_test,
//							R"(
//							{
//								"type": 0,
//								"training": False,
//								"iteration_begin": 0,
//								"iteration_end": 0,
//								"strict_shape": true
//							}
//							)",
//							&outshape,&out_len,&f);

						//3:wrong
//						const char* strxgb = "{\"type\": 0,\"training\": False,\"iteration_begin\": 0,\"iteration_end\": 0,\"strict_shape\": true}";

						//4:correct
						const char* strxgb ="{ \"training\": false, \"type\": 0,\"iteration_begin\": 0,\"iteration_end\": 0,\"strict_shape\": false}";

						ret=XGBoosterPredictFromDMatrix(h_booster,h_test,strxgb,&outshape,&out_len,&f);

						printf("XGBoost Predict flag %d, info: %s\n",ret,XGBGetLastError());
						cout<<out_len<<" "<<f[0]<<" "<<f[1]<<endl;

						XGDMatrixFree(h_test);
						XGBoosterFree(h_booster);

至此，C++终于和python下一致了。