tensortree: TidyTensor

# context("reading fasta file, fasta_encode_tensor")
#
# test_that("fasta_encode_tensor() works properly", {
#   t <- fasta_encode_tensor(system.file("extdata", "seqs.fasta", package = "tidytensor"))
#   expect_equal(t[1,1,1:4], tt(c(0, 0, 0, 1))) # first base of first seq is a T
#   expect_equal(t[1,4,1:4], tt(c(1, 0, 0, 0))) # fourth base of first seq is an A
#
#   expect_equal(t[2,1,1:4], tt(c(1, 0, 0, 0))) # first base of first seq is a A
#   expect_equal(t[2,4,1:4], tt(c(0, 1, 0, 0))) # second base of first seq is an C
#
#   expect_equal(dim(t), c(97, 140, 4))     # correct shape?
#   expect_equal(dimnames(t)[[1]][1], "EP535473")
#   expect_equal(dimnames(t)[[1]][2], "EP7180000000023_Z")
#
#
#   t2 <- fasta_encode_tensor(system.file("extdata", "seqs.fasta", package = "tidytensor"), start = 4, end = 6)
#   expect_equal(t2[1,1,1:4], tt(c(0, 0, 0, 1))) # first base of first seq is a T
#   expect_equal(t2[1,4,1:4], tt(c(0, 0, 1, 0))) # fourth base of first seq is a G
#
#   t3 <- fasta_encode_tensor(system.file("extdata", "seqs.fasta", package = "tidytensor"), ids = c("EP7180000000023_Z", "EP483640", "EP535473", "EP7180000026554"))
#   expect_equal(t3[1,1,1:4], tt(c(1, 0, 0, 0))) # first base of first seq is a A
#   expect_equal(t3[1,4,1:4], tt(c(0, 1, 0, 0))) # fourth base of first seq is a C
#   expect_equal(t3[2,1,1:4], tt(c(0, 1, 0, 0))) # first base of second seq is a C
#   expect_equal(t3[2,4,1:4], tt(c(0, 1, 0, 0))) # fourth base of second seq is a C
#
#   expect_equal(dimnames(t3)[[1]][1], "EP7180000000023_Z")
#   expect_equal(dimnames(t3)[[1]][2], "EP483640")
#
# })
#
#
# test_that("fasta_train_batch() works properly", {
#   t <- fasta_train_batch(system.file("extdata", "seqs.fasta", package = "tidytensor"),
#                            ids = c("EP7180000000023_Z", "EP483640", "EP535473", "EP7180000026554"),
#                            targets = c("one", "two", "three", "two"),
#                            keras_categorical_targets = TRUE)
#   t_tensor <- t[[1]]
#   t_target <- t[[2]]
#   expect_equal(t_tensor[1,1,1:4], tt(c(1, 0, 0, 0))) # first base of first seq is a A
#   expect_equal(t_tensor[1,4,1:4], tt(c(0, 1, 0, 0))) # fourth base of first seq is a C
#   expect_equal(t_tensor[2,1,1:4], tt(c(0, 1, 0, 0))) # first base of second seq is a C
#   expect_equal(t_tensor[2,4,1:4], tt(c(0, 1, 0, 0))) # fourth base of second seq is a C
#
#   expect_equal(dimnames(t_tensor)[[1]][1], "EP7180000000023_Z")
#   expect_equal(dimnames(t_tensor)[[1]][2], "EP483640")
#
#   # ordering of labels is "one", "three", "two" (default factor ordering)
#   expect_equal(t_target[1,], tt(c(1, 0, 0))) # first target should be encoded as c(1, 0)
#   expect_equal(t_target[2,], tt(c(0, 0, 1))) # second target should be encoded as c(0, 1)
#
#   expect_equal(dimnames(t_target)[[1]][1], "EP7180000000023_Z") # first dim, seq names
#   expect_equal(dimnames(t_target)[[2]][1], "one") # second dim, target labels
#   expect_equal(dimnames(t_target)[[2]][2], "three") # second dim, target labels
#   expect_equal(dimnames(t_target)[[2]][3], "two") # second dim, target labels
#
#   # check factor ordering?
#   t <- fasta_train_batch(system.file("extdata", "seqs.fasta", package = "tidytensor"),
#                          ids = c("EP7180000000023_Z", "EP483640", "EP535473", "EP7180000026554"),
#                          targets = factor(c("one", "two", "three", "two"), levels = c("one", "two", "three")),
#                          keras_categorical_targets = TRUE)
#   t_tensor <- t[[1]]
#   t_target <- t[[2]]
#   expect_equal(dimnames(t_target)[[2]][1], "one") # second dim, target labels
#   expect_equal(dimnames(t_target)[[2]][2], "two") # second dim, target labels
#   expect_equal(dimnames(t_target)[[2]][3], "three") # second dim, target labels
#
# })
#
#
# test_that("flow_sequences_from_fasta() works properly", {
#   gen <- flow_sequences_from_fasta(system.file("extdata", "seqs.fasta", package = "tidytensor"),
#                          ids = c("EP7180000000023_Z", "EP483640", "EP535473", "EP7180000026554"),
#                          targets = c("one", "two", "three", "two"),
#                          keras_categorical_targets = TRUE,
#                          batch_size = 2)
#   t1 <- gen()
#   t_tensor <- t1[[1]]
#   t_target <- t1[[2]]
#   expect_equal(t_tensor[1,1,1:4], tt(c(1, 0, 0, 0))) # first base of first seq is a A
#   expect_equal(t_tensor[1,4,1:4], tt(c(0, 1, 0, 0))) # fourth base of first seq is a C
#   expect_equal(t_tensor[2,1,1:4], tt(c(0, 1, 0, 0))) # first base of second seq is a C
#   expect_equal(t_tensor[2,4,1:4], tt(c(0, 1, 0, 0))) # fourth base of second seq is a C
#
#   expect_equal(dimnames(t_tensor)[[1]][1], "EP7180000000023_Z")
#   expect_equal(dimnames(t_tensor)[[1]][2], "EP483640")
#
#   # ordering of labels is "one", "three", "two" (default factor ordering)
#   expect_equal(t_target[1,], tt(c(1, 0, 0))) # first target should be encoded as c(1, 0)
#   expect_equal(t_target[2,], tt(c(0, 0, 1))) # second target should be encoded as c(0, 1)
#
#   expect_equal(dimnames(t_target)[[1]][1], "EP7180000000023_Z") # first dim, seq names
#   expect_equal(dimnames(t_target)[[2]][1], "one") # second dim, target labels
#   expect_equal(dimnames(t_target)[[2]][2], "three") # second dim, target labels
#   expect_equal(dimnames(t_target)[[2]][3], "two") # second dim, target labels
#
#
#   t2 <- gen()
#   t_tensor <- t2[[1]]
#   t_target <- t2[[2]]
#   print(t_tensor, bottom = "2d", show_names = T)
#   expect_equal(t_tensor[1,1,1:4], tt(c(0, 0, 0, 1))) # first base of first seq is a T
#   expect_equal(t_tensor[1,4,1:4], tt(c(1, 0, 0, 0))) # fourth base of first seq is a A
#   expect_equal(t_tensor[2,1,1:4], tt(c(0, 1, 0, 0))) # first base of second seq is a C
#   expect_equal(t_tensor[2,4,1:4], tt(c(0, 0, 0, 1))) # fourth base of second seq is a T
#
#   expect_equal(dimnames(t_tensor)[[1]][1], "EP535473")
#   expect_equal(dimnames(t_tensor)[[1]][2], "EP7180000026554")
#
#   # ordering of labels is "one", "three", "two" (default factor ordering)
#   expect_equal(t_target[1,], tt(c(0, 1, 0))) # first target should be encoded as "three" (c(0, 1, 0))
#   expect_equal(t_target[2,], tt(c(0, 0, 1))) # second target should be encoded as "two" (c(0, 0, 1))
#
#   expect_equal(dimnames(t_target)[[1]][1], "EP535473") # first dim, seq names
#   expect_equal(dimnames(t_target)[[2]][1], "one") # second dim, target labels
#   expect_equal(dimnames(t_target)[[2]][2], "three") # second dim, target labels
#   expect_equal(dimnames(t_target)[[2]][3], "two") # second dim, target labels
#
#
#   gen <- flow_sequences_from_fasta(system.file("extdata", "seqs.fasta", package = "tidytensor"),
#                                    ids = c("EP7180000000023_Z", "EP483640", "EP535473", "EP7180000026554"),
#                                    targets = c(-1, 2, 3, 0),
#                                    keras_categorical_targets = FALSE,
#                                    batch_size = 2)
#   t3 <- gen()
#   t_tensor <- t3[[1]]
#   t_target <- t3[[2]]
#   expect_equal(t_target[1], tt(-1))
#   expect_equal(t_target[2], tt(2))
#
#   t4 <- gen()
#   t_tensor <- t4[[1]]
#   t_target <- t4[[2]]
#   expect_equal(t_target[1], tt(3))
#   expect_equal(t_target[2], tt(0))
# })
#
#
#