One of the great strengths of Schnek is the ability to extend the setup file parser with user defined functions. As we already introduced in a previous example adding all the functions from the cmath library is done with a single line of code. It turns out that adding your own custom functions is just as simple.

Let’s start again creating a simulation block with two variables

class SimulationBlock : public Block {
private:
double value;
int n;
protected:
void initParameters(BlockParameters &parameters) {
}

public:
void writeValues() {
std::cout << "value = " << value << std::endl;
std::cout << "n = " << n << std::endl;
}
};


The variable value is a double and n is an integer. As before we have two functions. initParameters() registers the variables with the parser and writeValues simply writes out the values of the two variables.

The code for configuring and starting up the parser is also almost the same as we have seen before.

int main() {
BlockClasses blocks;

blocks.registerBlock("sim").setClass<SimulationBlock>();

std::ifstream in("example_setup_functions.setup");

Parser P("my_simulation", "sim", blocks);

registerCMath(P.getFunctionRegistry());

// ADDITIONAL FUNCTIONS WILL BE REGISTERED HERE

try {
pBlock application = P.parse(in);

SimulationBlock &mysim = dynamic_cast<SimulationBlock&>(*application);
mysim.evaluateParameters();
mysim.writeValues();
}
catch (ParserError &e) {
std::cerr << "Parse error in " << e.getFilename() << " at line " << e.getLine() << ": "<< e.message << std::endl;
exit(-1);
}
catch (EvaluationException &e) {
std::cerr << "Evaluation Error: " << e.getMessage() << std::endl;
exit(-1);
}

return 0;
}


Note here the comment after the call to registerCMath(). This comment indicates the place where we will insert the code for registering our custom functions. Before we do that, we need to define a function. Let’s create a function that calculates the normal distribution.

double normal(double x) {
return exp(-x*x)/sqrt(2.0*M_PI);
}


To register this function with the parser we need to add a single line to the code. This line is added at the location of the comment after the call to registerCMath().

P.getFunctionRegistry().registerFunction("normal",normal);


The string "normal" specifies the name of the function in the setup file. This can be anything you like. The second argument is the function in C++ itself. In fact, without the parentheses, normal specifies the function pointer.

Note:This needs to be a proper function and can’t be a class member function or a function object.

This is all there is to it.

Warning! There is one caveat when using user defined functions. The Schnek routines for evaluating expressions are intelligent. When using a DependencyUpdater multiple times to evaluate an expression with varying independent parameters, they optimise the evaluation to minimise repeated evaluation of constant expressions. While doing this they assume that functions are deterministic, i.e. that a function will always return the same value when given the same arguments. This poses a problem when adding functions such as random number generators. The random number generator will be called once and the result will be re-used for all future evaluations of the expression. To avoid this problem, these type of functions should be made depenent on a dummy parameter so that they are not optimised away.

The code for this example can be found here.