I have not tried 1.13.2 FMUs. I get my fmus from the XC repo: https://github.com/modelica/fmi-cross-check
But it seems they have removed all OM FMUs (except arm)
I am asking because I have had some issues on my side with a lisp cffi connection tool I built that seems to work with 1.12.0 version (me-fmu) of OM [with simulation running :-) ] but now completely fails with the last version of OM. Please give me time to understand your package (I am not good at all in C++ lol) and I will return to you ... hopefully. :-) Thank you very much.
is there any solution? thanks!
Hello! I'm trying to convert a Python script I made with PyFMI into C++ with FMI4cpp.
The program is this one (https://gitlab.com/JackHack96/fmu2opcua/-/blob/master/src/FmuSimulation.py).
The actual problem is that in Python with just one line of code I can pass a list made of <value_reference, actual_value> and write the values to the respective ports (line 40 of the Python script).
What would be the quickest way to replicate this in C++ with FMI4cpp?
The program is this one (https://gitlab.com/JackHack96/fmu2opcua/-/blob/master/src/FmuSimulation.py).
The actual problem is that in Python with just one line of code I can pass a list made of <value_reference, actual_value> and write the values to the respective ports (line 40 of the Python script).
What would be the quickest way to replicate this in C++ with FMI4cpp?
I was looking at
std::variant, I tought that I could make an std::unordered_map<fmi2ValueReference, std::variant> with all the possibile variable types, but I'm not entirely sure this is the correct approach (I still have to learn a lot about C++ though I've been studying it for 1 year now)
the map approach would not work as a single
fmi2ValueReference could represent each type in the variant (eg. valueref of 1 is valid for both int/real/boolean/string)
at the same time.
void FMU::set_inputs(const std::unordered_map<fmi2ValueReference, std::any> &inputs) {
std::unique_lock<std::mutex> lock(input_mutex);
for (const auto &i : inputs) {
if (i.second.type() == typeid(fmi2String)) {
if (!fmu_instance->write_string(i.first, std::any_cast<fmi2String>(i.second))) {
std::cerr << "Error! step() returned with status: " << to_string(fmu_instance->last_status()) << std::endl;
break;
}
} else if (i.second.type() == typeid(fmi2Boolean)) {
if (!fmu_instance->write_boolean(i.first, std::any_cast<fmi2Boolean>(i.second))) {
std::cerr << "Error! step() returned with status: " << to_string(fmu_instance->last_status()) << std::endl;
break;
}
} else if (i.second.type() == typeid(fmi2Integer)) {
if (!fmu_instance->write_integer(i.first, std::any_cast<fmi2Integer>(i.second))) {
std::cerr << "Error! step() returned with status: " << to_string(fmu_instance->last_status()) << std::endl;
break;
}
} else if (i.second.type() == typeid(fmi2Real)) {
if (!fmu_instance->write_real(i.first, std::any_cast<fmi2Real>(i.second))) {
std::cerr << "Error! step() returned with status: " << to_string(fmu_instance->last_status()) << std::endl;
break;
}
}
}
}
I actually went another route. I mean, what I wanted to achieve was the ability to set port values with a map (eg. set port 1 to true, set port 2 to 2.5 and so on)
So I discovered another C++17 things called
std::any which seems to fit
In theory, this way I can create a map of <ID,value> without having to do
write_real/write_boolean and so on for each one of the ports
Hi! I'm currently using FMI4cpp 0.8.0 on linux (gcc 6.3), and I'm getting a bunch of "undefined reference to `vtable for fmi4cpp::fmi2::fmu'" errors. After some google/stackoverflow reading, it seems that the issue is the lack of virtual constructors for the classes. MSVC has no issue with this. Is this something that is known/intentional? I just want to check before I create any issue :)
Hey Lars, we are in a situation where we need the raw DLL handle in order to implement a callback for an FMU, something that is currently not supported by the standard. We have patched fmi4cpp locally to make this work. Is this something you would be interested in to have in the official fmi4cpp? I can imagine other cases where this might be necessary, for example licensing checks.
