I want to read a JSON document with RapidJSON. My JSON document contains a multi-dimensional matrix and has the following form:
{
"shape": [1, 120, 42],
"type": "float32",
"data": [0.123, 0.234, ..., 0.345]
}
I would like to write a ReadArray
function that can read the arrays with numbers (e.g., shape and data in this example) into a std::vector
. The type of the vector should be int for the shape and appropriate (float/int) for the data.
What I have tried is:
template <typename ValueType>
void ReadArray(rapidjson::Value& jsonArray, std::vector<ValueType> output) {
for (auto it = jsonTensor.Begin(); it != jsonTensor.End(); it++) {
output.emplace_back(it->Get()); // error: No matching member function for call to 'Get'
}
}
I will call this as follows:
std::string dataPath("/path/to/data.json"); // data.json content see above
std::ifstream dataInputStream(dataPath);
rapidjson::IStreamWrapper dataInputStreamWrapper(dataInputStream);
rapidjson::Document jsonTensor;
jsonTensor.ParseStream(dataInputStreamWrapper);
auto jsonShape = jsonTensor["shape"].GetArray();
std::vector<int64_t> shape;
ReadArray(jsonShape, shape);
auto jsonData = jsonTensor["data"].GetArray();
std::vector<float> data;
ReadArray(jsonData, data);
At the location where I try to get the value from the JSON array, I get the error "No matching member function for call to 'Get'" [1]. Using rapidjson::GenericValue
instead of rapidjson::Value
is also not working for the same reason. rapidjson::GenericArray<false, ValueType>
instead raises "No matching function for call to 'ReadTensor'" in the location where I run ReadArray
[2].
How could I solve this issue? How could I convert the rapidjson::Value
to an array type or should I try something completely else here?
[1] Error message when using Value
:
/home/user/test/main.cpp: In function ‘void ReadArray(rapidjson::Value&, std::vector<ValueType>)’:
/home/user/test/main.cpp:17:37: error: no matching function for call to ‘rapidjson::GenericValue<rapidjson::UTF8<> >::Get()’
output.emplace_back(it->Get());
^
In file included from /home/user/test/main.cpp:7:0:
/home/user/.local/include/rapidjson/document.h:1912:7: note: candidate: template<class T> T rapidjson::GenericValue<Encoding, Allocator>::Get() const [with T = T; Encoding = rapidjson::UTF8<>; Allocator = rapidjson::MemoryPoolAllocator<rapidjson::CrtAllocator>]
T Get() const { return internal::TypeHelper<ValueType, T>::Get(*this); }
^
/home/user/.local/include/rapidjson/document.h:1912:7: note: template argument deduction/substitution failed:
/home/user/test/main.cpp:17:37: note: couldn't deduce template parameter ‘T’
output.emplace_back(it->Get());
^
In file included from /home/user/test/main.cpp:7:0:
/home/user/.local/include/rapidjson/document.h:1915:7: note: candidate: template<class T> T rapidjson::GenericValue<Encoding, Allocator>::Get() [with T = T; Encoding = rapidjson::UTF8<>; Allocator = rapidjson::MemoryPoolAllocator<rapidjson::CrtAllocator>]
T Get() { return internal::TypeHelper<ValueType, T>::Get(*this); }
^
/home/user/.local/include/rapidjson/document.h:1915:7: note: template argument deduction/substitution failed:
/home/user/test/main.cpp:17:37: note: couldn't deduce template parameter ‘T’
output.emplace_back(it->Get());
^
[2] Error message when using GenericArray
:
/home/user/test/main.cpp: In function ‘int main(int, char**)’:
/home/user/test/main.cpp:89:31: error: no matching function for call to ‘ReadArray(rapidjson::GenericArray<false, rapidjson::GenericValue<rapidjson::UTF8<> > >&, std::vector<long int>&)’
ReadArray(jsonShape, shape);
^
/home/user/test/main.cpp:15:6: note: candidate: template<class ValueType> void ReadArray(rapidjson::GenericArray<false, ValueType>&, std::vector<ValueType>)
void ReadArray(rapidjson::GenericArray<false, ValueType>& jsonTensor, std::vector<ValueType> output) {
^
/home/user/test/main.cpp:15:6: note: template argument deduction/substitution failed:
/home/user/test/main.cpp:89:31: note: deduced conflicting types for parameter ‘ValueType’ (‘rapidjson::GenericValue<rapidjson::UTF8<> >’ and ‘long int’)
ReadArray(jsonShape, shape);
^
Thanks to @273K, I could get it to work as follows:
template <typename ValueType>
void ReadArray(rapidjson::Value jsonArray, std::vector<ValueType>& output) {
for (auto it = jsonArray.Begin(); it != jsonArray.End(); it++) {
auto value = it->template Get<ValueType>();
output.emplace_back(value);
}
}
int main(int argc, char* argv[]) {
std::ifstream dataInputStream(dataPath);
rapidjson::IStreamWrapper dataInputStreamWrapper(dataInputStream);
rapidjson::Document jsonTensor;
jsonTensor.ParseStream(dataInputStreamWrapper);
auto jsonShape = jsonTensor["shape"].GetArray();
std::vector<int64_t> shape;
ReadArray(jsonShape, shape);
}
The missing link was to use it->template Get<ValueType>()
instead of it->Get()
.