Morris experiments may set parameters out of bounds

[ScriBanana]

Hi,

Describe the bug
Morris experiment sometimes sets parameters to values outside their assigned bounds. In particular, above 1.0 for a float with bounds 0.0 and 1.0. The parameter value in "Parameters to explore" gives the maximum value instead of the erroneous value taken by the parameter, as can be verified in the console or an assert statement, for example.

I have had no such issue with other exploration methods (Sobol or OFAT).

To Reproduce

model MorrisTest
global {
	bool endSim <- false;
	float a;
	float b;
	reflex roll {
		assert a <= 1.0;
		assert b <= 1.0;
		endSim <- true;
	}
}
experiment explo type: batch autorun: true until: endSim {
	parameter "a" var: a min: 0.0 max: 1.0;
	parameter "b" var: b min: 0.0 max: 1.0;
	method morris outputs: [] sample: 100 levels: 4 report: "morris.txt" results: "morris_raw.csv";
}

And check the value of the failed parameter in the console.

[ScriBanana]

I tried to reproduce with the example from the library (Exploration.gaml) with the following modifications with no success.

experiment batch_abstract type:batch virtual:true until:(time > end_cycle) {
	init {is_batch <- true;}
	parameter "Predator energy transfer:" var: predator_energy_transfer min: 0.0 max: 1.0;
	parameter "Predator energy reproduce:" var: predator_energy_reproduce min: 0.0 max: 1.0;
}
experiment Morris parent: batch_abstract type: batch until:( time > 2 ) {
	init {
		assert predator_energy_transfer <= 1.0;
		assert predator_energy_reproduce <= 1.0;
	}
	method morris outputs:[] sample:100 levels:4 report:"Results/morris.txt" results:"Results/morris_raw.csv";
}

[chapuisk]

Hey @ScriBanana, I dig a bit in the sampling method and "mistakes were made". The problem should be solve as soon as the PR #369 has been reviewed and accepted.

[ScriBanana]

Hi @chapuisk,
Nice, thanks ! I will try it out if I find the time.