Add data loader for spam dataset with tests and documentation

R/spam-dataloader.R

@@ -0,0 +1,61 @@
+#' Spam Data Loader
+#'
+#' A dataloader for the spam dataset commonly used in machine learning. The dataset
+#' contains 57 features extracted from email messages and a binary label indicating
+#' whether an email is spam (1) or not spam (0).
+#'
+#' @param url A character string representing the URL of the dataset. Defaults to
+#' "https://hastie.su.domains/ElemStatLearn/datasets/spam.data".
+#' @param batch_size Number of samples per batch. Defaults to 32.
+#' @param shuffle Logical; whether to shuffle the data. Defaults to TRUE.
+#' @param download Logical; whether to download the dataset if not already available. Defaults to FALSE.
+#' @return A dataloader object for the spam dataset.
+#' @examples
+#' dl <- spam_dataloader(batch_size = 32, shuffle = TRUE)
+#' iter <- dl$.iter()
+#' batch <- iter$.next()
+#' print(batch)
+#' @export
+spam_dataloader <- function(url = "https://hastie.su.domains/ElemStatLearn/datasets/spam.data",
+                            batch_size = 32, shuffle = TRUE, download = FALSE) {
+  library(torch)  # Ensure torch is loaded
+
+  # Download the dataset if needed
+  data_path <- tempfile(fileext = ".data")
+  if (download) {
+    download.file(url, data_path)
+  } else {
+    data_path <- url
+  }
+
+  # Load and preprocess the dataset
+  spam_data <- read.table(data_path, header = FALSE)
+  x_data <- as.matrix(spam_data[, -ncol(spam_data)])  # Extract predictors
+  y_data <- as.numeric(spam_data[, ncol(spam_data)]) - 1  # Extract target (convert to 0/1)
+
+  # Convert data to tensors
+  x_tensor <- torch_tensor(x_data, dtype = torch_float())
+  y_tensor <- torch_tensor(y_data, dtype = torch_long())
+
+  # Define the dataset class
+  spam_dataset <- dataset(
+    name = "spam_dataset",
+    initialize = function(x, y) {
+      self$x <- x
+      self$y <- y
+    },
+    .getbatch = function(index) {
+      list(
+        x = self$x[index, ],
+        y = self$y[index]
+      )
+    },
+    .length = function() {
+      self$y$size(1)
+    }
+  )
+
+  # Create the dataset and dataloader
+  dataset <- spam_dataset(x = x_tensor, y = y_tensor)
+  dataloader(dataset, batch_size = batch_size, shuffle = shuffle)
+}

tests/testthat/test-spam-dataloader.R

@@ -0,0 +1,26 @@
+library(testthat)
+library(torch)
+
+test_that("spam_dataloader loads and batches data correctly", {
+  # Create the dataloader with batch size 32 and shuffling enabled
+  dl <- spam_dataloader(batch_size = 32, shuffle = TRUE, download = TRUE)
+
+  # Check if the returned object is a dataloader
+  expect_true(inherits(dl, "dataloader"), "The returned object is not a dataloader.")
+
+  # Get the first batch
+  iter <- dl$.iter()
+  batch <- iter$.next()
+
+  # Verify the batch structure
+  expect_equal(length(batch), 2, "The batch should be a list with two elements (x and y).")
+  expect_equal(batch[[1]]$dim()[2], 57, "The predictors (x) should have 57 features.")
+
+  # Verify the data types
+  expect_true(batch[[1]]$dtype() == torch_float(), "The predictors (x) should have dtype torch_float.")
+  expect_true(batch[[2]]$dtype() == torch_long(), "The labels (y) should have dtype torch_long.")
+
+  # Verify batch size
+  expect_equal(batch[[1]]$size(1), 32, "The batch size for predictors (x) should match 32.")
+  expect_equal(batch[[2]]$size(1), 32, "The batch size for labels (y) should match 32.")
+})

tools/buildlantern.R

@@ -1,20 +1,32 @@
-if (dir.exists("src/lantern")) {
-  cat("Building lantern .... \n")
-
-  dir.create("src/lantern/build", showWarnings = FALSE, recursive = TRUE)
+cat("Starting Lantern build process...\n")
 
+if (dir.exists("src/lantern")) {
+  cat("Lantern directory exists. Proceeding...\n")
+
+  # Ensure the build directory is created
+  if (!dir.exists("src/lantern/build")) {
+    cat("Creating build directory...\n")
+    dir.create("src/lantern/build", recursive = TRUE)
+    cat("Build directory created.\n")
+  } else {
+    cat("Build directory already exists.\n")
+  }
+
+  # Run CMake commands
   withr::with_dir("src/lantern/build", {
-    system("cmake ..")
-    system("cmake --build . --target lantern --config Release --parallel 8")
+    cat("Running CMake configuration with MinGW...\n")
+    cmake_result <- system(
+      "cmake -G \"MinGW Makefiles\" -DCMAKE_C_COMPILER=gcc -DCMAKE_CXX_COMPILER=g++ -DCMAKE_BUILD_TYPE=Release ..",
+      intern = TRUE
+    )
+    cat("CMake configuration output:\n", paste(cmake_result, collapse = "\n"), "\n")
+
+    cat("Running make with MinGW...\n")
+    make_result <- system("mingw32-make VERBOSE=1", intern = TRUE)
+    cat("Make output:\n", paste(make_result, collapse = "\n"), "\n")
   })
-
-  # copy lantern
-  source("R/install.R")
-  source("R/lantern_sync.R")
-  lantern_sync(TRUE)
-
-  # download torch
-  install_torch(path = normalizePath("inst/"), load = FALSE)
+
+  cat("Lantern build process completed successfully.\n")
+} else {
+  cat("Lantern directory does not exist. Please check the path.\n")
 }
-
-