Why does setPredictor prevent empty leaf-nodes?

[EoghanONeill]

If there are empty leaf-nodes, the likelihood for the Metropolis-Hastings step can still be calculated, and the means for the empty leaf-node parameter draws are equal to zero (the prior mean). Is there a reason this is considered to be invalid?

Is it possible to add an option that allows for empty leaf-nodes?

Also I can find updatePredictor in https://github.com/vdorie/dbarts/blob/main/src/dbarts/bartFit.cpp , but I cannot find it in https://cran.r-project.org/web/packages/dbarts/dbarts.pdf . Which is the correct function to use when setting new values for the predictors in each MCMC iteration, setPredictor or updatePredictor?