Loop: use gbm.auto trycatch to only move to next number if it run successfully

[SimonDedman]

[SimonDedman]

loop runs whole gbm.auto. in gbm.auto it checks the outcome of assign(gbm.step.SD) which I can't do in loop. check the saved outputs?

either numbered folder only contains grv.csv ( which shouldn't exist, now removed) i.e. empty folder, or metrics are bad (can't see how to test, gbm.auto fails with error more than it succeeds with bad results.

Not sure if this will work given gbm.auto fails with error, stopping the loop, but starter for 10:

userdefined <- 10
maxattempts <- 10

for (i in 1:userdefined) {
  success <- FALSE  # Initialize success flag
  attempts <- 0      # Initialize attempt counter
  
  setwd(paste0("./", i)) # move to it
  
  while (!success) {
    attempts <- attempts + 1
    
    # Simulate your code that might fail (replace with your actual code)
    gbm.auto(grids = grids, # run i'th gbm.auto
             samples = samples,
             expvar = expvar,
             resvar = resvar,
             randomvar = randomvar,
             tc = tc,
             lr = lr,
             bf = bf,
             n.trees = n.trees,
             ZI = ZI,
             fam1 = fam1,
             fam2 = fam2,
             simp = simp,
             gridslat = gridslat,
             gridslon = gridslon,
             multiplot = multiplot,
             cols = cols,
             linesfiles = linesfiles,
             smooth = smooth,
             savedir = "./", # current directory created & moved to above
             savegbm = savegbm,
             loadgbm = loadgbm,
             varint = varint,
             map = map,
             shape = shape,
             RSB = RSB,
             BnW = BnW,
             alerts = alerts,
             pngtype = pngtype,
             gaus = gaus,
             MLEvaluate = MLEvaluate,
             ...) # accept other gbm.auto values than these basics
    
    # test outcome: if folder empty or metrics shit (can't see how to test this) then success FALSE else TRUE
    success <- ifelse(is.character(list.files()) & length(list.files()) == 0, FALSE, TRUE)
    
    # result <- tryCatch({
    #   # Your code here
    #   if (runif(1) < 0.3) {
    #     stop("Simulated failure") # Simulate a failure 30% of the time
    #   } else {
    #     paste("Success on attempt", attempts, "for i =", i) # Simulate success
    #   }
    # }, error = function(e) {
    #   print(paste("Attempt", attempts, "failed for i =", i, ":", e$message))
    #   NULL # Return NULL on error, so success remains FALSE
    # })
    # 
    # if (!is.null(result)) {
    #   print(result)
    #   success <- TRUE # Set success to TRUE if the code ran without error
    # } # close if (!is.null(result)
    
    # Optional: Add a maximum number of retries to prevent infinite loops
    if (attempts > maxattempts) {
      print(paste("Maximum retries reached for loop ", i, ". Skipping."))
      break # Exit the while loop if max retries are exceeded
    } # close if (attempts > maxattempts)
    
  } # close while
  setwd("../") # move back up to root folder
}

[SimonDedman]

Report.csv for bad results:

Best Gaussian BRT Model combo: Gaus_BRT.tc5.lr0.5.bf0.9 trees: 1900 Training Data Correlation: 1 CV Mean Deviance: 0 CV Deviance SE: 0 CV D squared: 1 CV Mean Correlation: 1 CV Correlation SE: 0 CV RMSE: 0 Deviance% explained relative to null, training: 100 Deviance% explained relative to null, CV: 100