#Training set dependent sample

library(depstats)
library(progress)
library(fst)
library(dplyr)

sample_sizes <- c(50, 100, 200, 400)


#only generate samples for the remaining ones
sample_sizes_remain <- as.character(sample_sizes) %>% setdiff(list.dirs(path = "dependent_samples", full.names = FALSE))

for (sample_size in sample_sizes_remain) {
  cat('working on sample size', sample_size, '...\n')

  n <- as.integer(sample_size)

  set.seed(n)

  #generate dependent samples
  X <- c() #will be 20000 samples
  ############################################################
  #Linear
  X1 <- depstats::depgen(600, n, 'normnoise(randlin(n), runif(1, 0, 1))',
               randrotate = FALSE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(randlin(n), \'0.5 * x\')',
               randrotate = FALSE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(randlin(n), \'0.5 - abs(0.5 * x)\')',
               randrotate = FALSE)
  X <- rbind(X, X1, X2, X3)
  ############################################################
  #Diamond
  X1 <- depstats::depgen(600, n, 'normnoise(diam.cld(n, runif(1, pi / 6, pi / 3)), runif(1, 0, 0.25))',
               randrotate = FALSE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(diam.cld(n, runif(1, pi / 6, pi / 3)), \'0.2 * x\')',
               randrotate = FALSE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(diam.cld(n, runif(1, pi / 6, pi / 3)), \'0.2 - abs(0.2 * x)\')',
               randrotate = FALSE)
  X <- rbind(X, X1, X2, X3)
  ############################################################
  #Triangle
  X1 <- depstats::depgen(600, n, 'normnoise(tri.cld(n, runif(1, 0, 1), runif(1, 0, 1)), runif(1, 0, 0.5))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(tri.cld(n, runif(1, 0, 1), runif(1, 0, 1)), \'0.3 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(tri.cld(n, runif(1, 0, 1), runif(1, 0, 1)), \'0.3 - abs(0.3 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #Crescent
  X1 <- depstats::depgen(600, n, 'normnoise(cre.cld(n,runif(1, 0.5, 1)), runif(1, 0, 0.5))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(cre.cld(n, runif(1, 0.5, 1)), \'0.3 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(cre.cld(n, runif(1, 0.5, 1)), \'0.3 - abs(0.3 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #Point Clouds
  pts1 <- matrix(c( 1, 0,
                   -1, 0,
                    0, 1), ncol = 2, byrow = TRUE)
  pts2 <- matrix(c( 1,  0,
                   -1,  0,
                    0,  0,
                    0,  1,
                    0, -1), ncol = 2, byrow = TRUE)
  pts3 <- matrix(c( 1,  1,
                   -1,  1,
                    0,  0,
                    1, -1,
                   -1, -1), ncol = 2, byrow = TRUE)
  pts <- list(pts1, pts2, pts3)
  X1 <- depstats::depgen(1000, n, 'normnoise(pt.cld(n, pts[sample(c(1, 2, 3), 1)][[1]]), runif(1, 0.05, 0.5), TRUE)',
               randrotate = TRUE)
  X <- rbind(X, X1)
  #Exp
  X1 <- depstats::depgen(600, n, 'normnoise(exp_cv(n, runif(1, 2, 4)), runif(1, 0, 0.5))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(exp_cv(n, runif(1, 2, 4)), \'0.3 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(exp_cv(n, runif(1, 1.5, 3)), \'0.3 - abs(0.3 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #circular
  c1 <- matrix(c(0, 0), ncol = 2, byrow = TRUE)
  c2 <- matrix(c( 1, 0,
                 -1, 0,
                  0, 1), ncol = 2, byrow = TRUE)
  c3 <- matrix(c( 1,  1,
                 -1,  1,
                  1, -1,
                 -1, -1), ncol = 2, byrow = TRUE)
  cs <- list(c1, c2, c3)
  X1 <- depstats::depgen(600, n, 'normnoise(circ.cv(n, cs[sample(c(1, 2, 3), 1)][[1]]), runif(1, 0, 0.5))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(circ.cv(n, cs[sample(c(1, 2, 3), 1)][[1]]), \'0.3 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(circ.cv(n, cs[sample(c(1, 2, 3), 1)][[1]]), \'0.3 - abs(0.3 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #Cross
  X1 <- depstats::depgen(600, n, 'normnoise(rotate(cross.cv(n), runif(1, -pi / 6, pi / 6)), runif(1, 0, 0.5))',
               randrotate = FALSE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(rotate(cross.cv(n), runif(1, -pi / 6, pi / 6)), \'0.3 * x\')',
               randrotate = FALSE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(rotate(cross.cv(n), runif(1, -pi / 6, pi / 6)), \'0.3 - abs(0.3 * x)\')',
               randrotate = FALSE)
  X <- rbind(X, X1, X2, X3)
  #wedge
  X1 <- depstats::depgen(600, n, 'normnoise(wedge.cv(n), runif(1, 0, 1))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(wedge.cv(n), \'0.5 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(wedge.cv(n), \'0.5 - abs(0.5 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #cubic
  X1 <- depstats::depgen(600, n, 'normnoise(cubic.cv(n), runif(1, 0, 0.25))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(cubic.cv(n), \'0.2 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(cubic.cv(n), \'0.2 - abs(0.2 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #w-shape
  X1 <- depstats::depgen(600, n, 'normnoise(w.cv(n), runif(1, 0, 0.3))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(w.cv(n), \'0.2 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(w.cv(n), \'0.2 - abs(0.2 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #parabola
  X1 <- depstats::depgen(600, n, 'normnoise(parabola.cv(n), runif(1, 0, 0.3))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(parabola.cv(n), \'0.2 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(parabola.cv(n), \'0.2 - abs(0.2 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #2-parabola
  X1 <- depstats::depgen(600, n, 'normnoise(twoparabola.cv(n), runif(1, 0, 0.4))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(twoparabola.cv(n), \'0.3 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(twoparabola.cv(n), \'0.3 - abs(0.3 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #sine
  X1 <- depstats::depgen(600, n, 'normnoise(sine.cv(n, runif(1, 1, 4)), runif(1, 0, 1))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(sine.cv(n, runif(1, 1, 4)), \'0.5 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(sine.cv(n, runif(1, 1, 4)), \'0.5 - abs(0.5 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #doppler
  X1 <- depstats::depgen(600, n, 'normnoise(doppler.cv(n, runif(1, 2, 4)), runif(1, 0, 1))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(doppler.cv(n, runif(1, 2, 4)), \'0.5 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(doppler.cv(n, runif(1, 2, 4)), \'0.5 - abs(0.5 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #heavy sine
  X1 <- depstats::depgen(600, n, 'normnoise(heavysine.cv(n), runif(1, 0, 1))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(heavysine.cv(n), \'0.5 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(heavysine.cv(n), \'0.5 - abs(0.5 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #heart
  X1 <- depstats::depgen(600, n, 'normnoise(heart.cv(n), runif(1, 0, 0.5))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(heart.cv(n), \'0.3 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(heart.cv(n), \'0.3 - abs(0.3 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #spiral
  X1 <- depstats::depgen(600, n, 'normnoise(spiral.cv(n,runif(1, 2.5, 3.5)), runif(1, 0, 0.3))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(spiral.cv(n, runif(1, 2.5, 3.5)), \'0.2 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(spiral.cv(n, runif(1, 2.5, 3.5)), \'0.2 - abs(0.2 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #taegeuk
  X1 <- depstats::depgen(600, n, 'normnoise(taegeuk.cv(n), runif(1, 0, 0.3))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(taegeuk.cv(n), \'0.2 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(taegeuk.cv(n), \'0.2 - abs(0.2 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)
  #samtaegeuk
  X1 <- depstats::depgen(600, n, 'normnoise(samtaegeuk.cv(n), runif(1, 0, 0.3))',
               randrotate = TRUE)
  X2 <- depstats::depgen(200, n, 'varnormnoise(samtaegeuk.cv(n), \'0.2 * x\')',
               randrotate = TRUE)
  X3 <- depstats::depgen(200, n, 'varnormnoise(samtaegeuk.cv(n), \'0.2 - abs(0.2 * x)\')',
               randrotate = TRUE)
  X <- rbind(X, X1, X2, X3)

  cat('working on dependence measures ...\n')
  Xind <- depstats::sampleapply(X, n, 1) %>% as.data.frame()

  cat('working on images ...\n')
  Ximage <- depstats::sampleapply(X, n, 0, grid = 25) %>% as.data.frame()

  cat('writing files ...\n')
  dir.create(paste0('./dependent_samples/', sample_size)) #create folder

  write.fst(Xind, paste0('./dependent_samples/', sample_size, '/', 'dep.fst'))
  write.fst(Ximage, paste0('./dependent_samples/', sample_size, '/', 'dep_image.fst'))
}