Another bug in getSample with numSamples and population chains about bayesiantools

[TankredO]

The error seems to happen here:

file: classMcmcSampler.R
function: getSample.mcmcSampler

      # Sample size
      if(thin == 1 && !is.null(numSamples)){
        # This should already be the thinning step?
        sel <- seq(1,dim(temp)[1], len = (floor(numSamples/length(sampler$chain))))
        temp <- temp[sel,] 
      }
      
      #############
      
      if (!is.null(whichParameters)) temp = temp[,whichParameters]
      
      out[[i]] = coda::mcmc(temp, start = start, end = end, thin = thin)

I don't really understand what this is supposed to do. Should the thinning be applied such that numSamples determines how many samples collected per chain or from all chains together? (seems like the latter)

The problem that happens here - to my undestanding - is that the chain is already thinned out after "if(thin == 1 && !is.null(numSamples)){ .... }". In this case the coda::mcmc() function with thinning=1 is only there to create the mcmc object from the temp matrix. However, if that is the intention we need to adjust the end parameter of coda::mcmc, whenever the temp matrix is already thinned.