Differential evaluation crossover_gap = 0

[dklong-csu]

If crossover_gap in DifferentialEvaluationMetropolisHastings is equal to 0 -- i.e. if the DE-MC algorithm is always used instead of also having normal metropolis steps -- then the sampler attempts the operation generation % crossover_gap which ends up being generation % 0 and you get a divide by zero error.

[bangerth]

I think that the variable is misnamed -- it's the interval, not the gap. If so, then it needs to be a positive number, and we should check that with an assertion.

[dklong-csu]

Makes sense, and that aligns with the comment above that comparison. Having crossover_gap=1 would result in "perform crossover every 1 iteration".

[bangerth]

We should change the name of the variable, as well as update the documentation. I'll get around to it eventually, but at least now you know what you should choose.

[bangerth]

I've also special cased the current behavior to avoid the division by zero, see #188 .