R-autofun/check-spsann-arguments.R

In samuel-rosa/spsann: Optimization of Spatial Samples via Simulated Annealing

# Check spsann arguments
# 
# COMMAND
# eval(.check_spsann_arguments())
#
# SUMMARY
# 1. Check the mandatory OPTIM arguments 'points' and 'candi';
# 2. If all is fine, check the mandatory OPTIM argument 'iterations';
# 3. If all is fine, check the mandatory OPTIM argument 'acceptance';
# 4. If all is fine, check the mandatory OPTIM argument 'stopping';
# 5. Check if the optional OPTIM arguments 'weights', 'nadir' and 'utopia' are 
#    missing;
# 6. If the optional OPTIM arguments are not missing, check the mandatory MOOP
#    arguments 'weights', 'nadir' and 'utopia';
# 7. Check if acceptance probabilities should be tracked.
#
# NOTES
# 1. This code chunk is used only with the family of 'optim...()' functions.
# 2. A good reading on argument checking is the article at
#    http://www.r-bloggers.com/a-warning-about-warning/
# 
res <- NULL

# Check missing arguments
aa <- c("points", "candi")
bb <- c(missing(points), missing(candi))
if (any(bb)) {
  i <- which(bb == TRUE)
  res <- c("missing argument: '", aa[i], "'\n", sep = "")
} else {
  
  # Argument 'candi'
  aa <- any(apply(candi, 2, is.numeric) == FALSE)
  bb <- ncol(candi) != 2
  cc <- any(c(c("x", "y") != colnames(candi)) == TRUE)
  if (aa || bb || cc) {
    res <- c("'candi' must have two named numeric columns: 'x' and 'y'")
  } else {
    
    # Argument 'schedule'
    if (!is.list(schedule) || length(schedule) != 11) {
      res <- c("'schedule' must be a list with 11 sub-arguments")
    }
  }
}

# Multi-objective optimization
aa <- objective %in% c("ACDC", "CLHS", "SPAN")
# aa <- all(c(!is.null(weights), !is.null(utopia), !is.null(nadir)))
MOOP <- ifelse(aa, TRUE, FALSE)
# if (aa) {
#   MOOP <- length(which(weights != 0))
#   MOOP <- ifelse(MOOP > 1, TRUE, FALSE)
#   #COST <- ifelse(weights$COST == 0, FALSE, TRUE)
# }

# Check consistency of argument 'weights'
if (MOOP) {
  
  if (missing(weights) || !is.list(weights) || is.null(names(weights)) || sum(unlist(weights)) != 1) {
    res <- "'weights' must be a list with named sub-arguments that sum to 1"
    
  } else if (length(weights) < 2) {
    res <- "'weights' must contain more than on sub-arguments"
  }
}

# Arguments 'nadir' and 'utopia'
if (MOOP && objective != "CLHS") {
  aa <- !is.list(utopia)
  bb <- !length(utopia) == 1
  cc <- is.null(names(utopia))
  if (aa || bb || cc) {
    res <- c("'utopia' must be a list with one named sub-arguments")
  } else {
    
    aa <- !is.list(nadir)
    if (aa) {
      res <- c("'nadir' must be a list with named sub-arguments")
    }
    aa <- names(nadir)
    if (length(aa) >= 2) {
      if (length(aa) > 2) {
        res <- c("you must choose a single 'nadir' option")
      }
    } else {
      if (aa == "sim" || aa == "seeds") { 
        res <- c("the number of simulations and their seeds must be set")
      }
      if (aa == "abs") {
        res <- c("sorry but 'nadir' cannot be calculated")
      }
    }  
  }
}

# Check consistency of argument 'track'
if (track || plotit) {
  
  # If plotting is required, then objective function values MUST be tracked
  if (plotit) {
    track <- TRUE
  }
  
  # Are we dealing with a single- or multi-objective optimization problem?
  if (MOOP && !missing(weights)) {
    
    # If multi-objective optimization problem, then the number of columns of the data.frame used to store the 
    # objective function values is set by the lenght of the weights argument plus one column to store the
    # weighted sum of the single objective function values.
    # 2019-03-10: if the used does not set the weights, then optimCLHS sets list(O1 = 1/3, O2 = 1/3, O3 = 1/3)
    # by default. A four-columns data.frane is created to store the objetive function values. However, if the 
    # user provides only continuos variables but not the weights, then optimCLHS has no way of guessing that it
    # should have created a three-columns data.frame. Thus, from now on, the user MUST set the weights to avoid
    # this issue. Besides, it is important for the user to always set the weights to make sure that s/he is 
    # aware of what s/he is doing.
    energies <- as.data.frame(matrix(NA_real_, nrow = 1, ncol = length(weights) + 1))
    colnames(energies) <- c("obj", sort(names(weights)))
    
  } else {
    energies <- data.frame(obj = NA_real_)
  }
}

# Output and cleanup
if (!is.null(res)) stop (res, call. = FALSE)
# rm(aa, bb, cc, dd, res)
rm(aa, bb, cc, res)