####Functions Needed to Fit Covariate MMSBM#####

#edgelist = an m(m-1) x 3 matrix.  (sender, receiver, tie value)#
#m is network size
 #covariates = an m(m-1) x p matrix (p is the number of covariates) - these can be edgelevel or node-level, but nodelevel will require some dups.
#logitb is a matrix showing the group-group tie probabilities
#BETA is a vector of p parameters
#M=m(m-1) is number of ties
#group is a vector of n group memberships#


library(Rcpp)
sourceCpp("SBM-Ccode.cpp")

loglikeSD=function(logitb, group, edgelist, covariates, beta){
  return(sum(logLikeVectorC(edgelist, logitb, group, covariates, beta)))
       }



logprior.beta<-function(beta, mu, sigmasq){
value=-1/2*log(sigmasq)-1/(2*sigmasq)*(beta-mu)^2
return(value)
}

    f1 <- function(row, k) sapply(1:k, function(it) sum(row==it))

logitprior=function(logitb, prior, diag){
       a=prior$B$diag[1]
      b=prior$B$diag[2]
        val=a*logitb-(a+b)*log(1+exp(logitb))
    return(val)
  }

constrain.logit=function(logitb){
  if(logitb>10){logitb=10}
  if(logitb< -10){logitb=-10}
  return(logitb)
}




SBM.ID.rotation <- function(ID.labels, label.count=max(ID.labels)) {
  #ID.labels=c(3,4,4,1,1,2,5); label.count=max(ID.labels)
  
  #first pin down the keepers.
  assigned <- replaced <- rep(NA, label.count)
  replacements <- rep(NA, length(ID.labels))
  
  for (ii in 1:label.count) {
    if (is.na(replacements[ii])) {
      assigned[ID.labels[ii]] <- ii
      replaced[ii] <- ID.labels[ii]
      replacements[ID.labels==ID.labels[ii]] <- ii
    }
  }
  while (sum(is.na(assigned))>0) {
    #node <- min(which(is.na(replacements)))
    vacancy <- min(which(is.na(assigned)))
    newlabel <- min(which(is.na(replaced)))
    
    assigned[vacancy] <- newlabel
    replaced[newlabel] <- vacancy
    
    replacements[ID.labels==vacancy] <- newlabel
  }
  
  return (assigned)
}


group.reassign=function(groups, assignment){
    newgroups=groups
    for(i in 1:length(assignment)){
        newgroups[which(groups==i)]=assignment[i]}
    return(newgroups)}