workshop_setup.R

#!/usr/bin/env Rscript
# ============================================================================
# Workshop Setup Script: R + TensorFlow + Scientific Computing
# ============================================================================
# This script prepares the environment for the workshop by:
#   1. Installing required R packages
#   2. Configuring Python for use with reticulate in Positron
#   3. Installing Python packages to system Python
#   4. Verifying the installation
#
# Designed for Positron on Posit Workbench
# Usage: source("workshop_setup.R")
# ============================================================================

cat("\n")
cat("======================================================================\n")
cat("  Workshop: R + TensorFlow + Scientific Computing\n")
cat("  Setup Script - Installing Dependencies\n")
cat("======================================================================\n\n")

# ============================================================================
# STEP 1: Install Core R Packages
# ============================================================================

cat("STEP 1: Checking and installing R packages...\n")
cat("----------------------------------------------------------------------\n")

# Core packages needed for the workshop
cran_pkgs <- c(
  # Python integration
  "reticulate",

  # Deep learning
  "keras",
  "tensorflow",
  "torch",

  # Data manipulation and visualization
  "tidyverse",
  "ggplot2",

  # MLOps and deployment
  "vetiver",
  "pins",
  "plumber",
  "rsconnect",

  # API consumption
  "httr",
  "jsonlite",

  # Utilities
  "config",
  "withr",
  "stringr",
  "glue",

  # For demos
  "nycflights13",
  "rmarkdown",
  "knitr",

  # Shiny (for apps in part 1)
  "shiny",

  # Development tools
  "devtools",
  "remotes"
)

# Check which packages need to be installed
to_install <- setdiff(cran_pkgs, installed.packages()[, "Package"])

if (length(to_install) > 0) {
  cat(sprintf("Installing %d CRAN packages: %s\n",
              length(to_install),
              paste(to_install, collapse=", ")))
  # Use system default repos (configured by IT/Posit Package Manager)
  install.packages(to_install)
  cat("✓ CRAN packages installed\n\n")
} else {
  cat("✓ All CRAN packages already installed\n\n")
}

# ============================================================================
# STEP 2: Configure Python for Positron
# ============================================================================

cat("STEP 2: Configuring Python for Positron/reticulate...\n")
cat("----------------------------------------------------------------------\n")

# Find system Python installation
python_paths <- c(
  "/opt/python/default/bin/python3",
  "/opt/python/3.12.3/bin/python3",
  "/usr/bin/python3"
)

python_exe <- NULL
for (py_path in python_paths) {
  if (file.exists(py_path)) {
    python_exe <- py_path
    break
  }
}

if (is.null(python_exe)) {
  stop("ERROR: Could not find Python installation")
}

cat(sprintf("Found Python: %s\n", python_exe))

# Create .Rprofile to configure Python for reticulate
# This is needed in Positron to prevent reticulate from creating its own Python env
rprofile_content <- sprintf('# Auto-generated by workshop_setup.R
# Configure Python for reticulate in Positron
Sys.setenv(RETICULATE_PYTHON = "%s")
', python_exe)

# Create .Rprofile in workshop root
writeLines(rprofile_content, ".Rprofile")
cat("✓ Created .Rprofile in workshop root\n")

# Create .Rprofile in each workshop directory and subdirectory
workshop_dirs <- c(
  "part1",
  "part1/reticulated-shiny-app-py2",
  "part1/reticulated-shiny-rmd-py2",
  "part1/Shiny-Python",
  "part2",
  "part2/keras_image_recognition",
  "part_3",
  "part_3/immunotherapy",
  "part_4_torch",
  "part_4_torch/part1",
  "part_4_torch/part2",
  "part_4_torch/part3/immunotherapy",
  "part_5_vetiver",
  "part_5_vetiver/tensorflow_keras",
  "part_5_vetiver/torch"
)

for (dir in workshop_dirs) {
  if (dir.exists(dir)) {
    rprofile_path <- file.path(dir, ".Rprofile")
    writeLines(rprofile_content, rprofile_path)
    cat(sprintf("✓ Created .Rprofile in %s/\n", dir))
  }
}

cat("  This ensures reticulate uses system Python with all packages\n")
cat("  Works from any workshop directory\n")

# Set for current session
Sys.setenv(RETICULATE_PYTHON = python_exe)
cat("✓ Python configured for current session\n\n")

# ============================================================================
# STEP 3: Install GitHub Packages
# ============================================================================

cat("STEP 3: Installing packages from GitHub...\n")
cat("----------------------------------------------------------------------\n")

# ggseqlogo - for sequence logos (used in part 3)
if (!"ggseqlogo" %in% installed.packages()[, "Package"]) {
  cat("Installing ggseqlogo from GitHub...\n")
  remotes::install_github("omarwagih/ggseqlogo", quiet = TRUE)
  cat("✓ ggseqlogo installed\n")
} else {
  cat("✓ ggseqlogo already installed\n")
}

# PepTools - peptide tools (used in part 3)
if (!"PepTools" %in% installed.packages()[, "Package"]) {
  cat("Installing PepTools from GitHub...\n")
  remotes::install_github("leonjessen/PepTools", quiet = TRUE)
  cat("✓ PepTools installed\n")
} else {
  cat("✓ PepTools already installed\n")
}

cat("\n")

# ============================================================================
# STEP 4: Install Python Packages
# ============================================================================

cat("STEP 4: Installing Python packages...\n")
cat("----------------------------------------------------------------------\n")

library(reticulate)

# Get the system Python that reticulate will use
py_exe <- Sys.which("python3")
if (py_exe == "") {
  py_exe <- Sys.which("python")
}

cat(sprintf("Python executable: %s\n", py_exe))

# Check if requirements.txt exists
if (file.exists("requirements.txt")) {
  cat("\nFound requirements.txt - installing Python packages from it...\n")
  cat("(This may take several minutes)\n\n")

  # Install from requirements.txt
  install_cmd <- sprintf("%s -m pip install --user -r requirements.txt", py_exe)
  cat("Running: pip install --user -r requirements.txt\n\n")

  result <- system(install_cmd, intern = FALSE)

  if (result == 0) {
    cat("\n✓ Python packages from requirements.txt installed\n\n")
  } else {
    cat("\n⚠ Some packages may have failed (check output above)\n")
    cat("  This might be ok if packages are already installed\n\n")
  }
} else {
  cat("\n⚠ requirements.txt not found\n")
  cat("  Installing core packages manually...\n\n")

  # Fallback: install core packages
  py_packages <- c("pandas", "numpy", "matplotlib", "seaborn", "tensorflow", "keras")

  for (pkg in py_packages) {
    cat(sprintf("  Installing %s...\n", pkg))
    install_cmd <- sprintf("%s -m pip install --user --quiet %s", py_exe, pkg)
    system(install_cmd)
  }

  cat("\n✓ Core Python packages installed\n\n")
}

# ============================================================================
# STEP 5: Verify Installation
# ============================================================================

cat("STEP 5: Verifying installation...\n")
cat("======================================================================\n\n")

# Check R packages
cat("R Packages Status:\n")
cat("------------------\n")
critical_packages <- c("reticulate", "keras", "tensorflow", "torch", "tidyverse",
                       "vetiver", "pins", "plumber", "httr",
                       "ggseqlogo", "PepTools", "nycflights13")
all_ok <- TRUE
for (pkg in critical_packages) {
  has_pkg <- requireNamespace(pkg, quietly = TRUE)
  status <- if (has_pkg) "✓ INSTALLED" else "✗ MISSING"
  cat(sprintf("  %-15s %s\n", pkg, status))
  if (!has_pkg) all_ok <- FALSE
}
cat("\n")

# Check Python packages
cat("Python Packages Status:\n")
cat("-----------------------\n")
cat("(Restart R session if packages don't show up)\n\n")

# These checks might fail in current session, that's ok
for (pkg in c("pandas", "numpy", "matplotlib", "tensorflow")) {
  status <- tryCatch({
    py_module_available(pkg)
    "✓ available"
  }, error = function(e) {
    "? check after restart"
  })
  cat(sprintf("  %-15s %s\n", pkg, status))
}

cat("\n")

# ============================================================================
# Final Instructions
# ============================================================================

cat("======================================================================\n")
if (all_ok) {
  cat("✓ Setup complete!\n")
} else {
  cat("⚠ Setup complete with some warnings\n")
}
cat("======================================================================\n\n")

cat("IMPORTANT - RESTART R NOW:\n")
cat("--------------------------\n")
cat("In Positron: Session > Restart R Session\n")
cat("Or command: .rs.restartR()\n")
cat("This loads the .Rprofile Python configuration\n\n")

cat("After restarting R:\n")
cat("-------------------\n")
cat("1. The workshop materials are organized in five parts:\n")
cat("   - part1/       : R & Python Integration with reticulate\n")
cat("   - part2/       : Deep Learning with Keras\n")
cat("   - part_3/      : Cancer Immunotherapy Application (Keras)\n")
cat("   - part_4_torch/: Same as parts 1-3 but using Torch R\n")
cat("   - part_5_vetiver/: Modern MLOps with Vetiver (Keras & Torch)\n\n")
cat("2. Start with part1/demo-notebook.Rmd\n\n")
cat("3. Run notebook chunks IN ORDER from top to bottom\n")
cat("   IMPORTANT: Always run the setup chunk first!\n\n")
cat("4. For Shiny apps (runtime: shiny), use:\n")
cat("   rmarkdown::run(\"path/to/app.Rmd\")\n")
cat("   (Do NOT try to render Shiny apps)\n\n")

cat("Tips:\n")
cat("-----\n")
cat("- Python variables are accessible in R via py$variable_name\n")
cat("- R variables are accessible in Python via r.variable_name\n")
cat("- If you get 'module not found' errors, restart R to reload Python config\n")
cat("- The .Rprofile ensures reticulate uses system Python with all packages\n")
cat("- In Positron, reticulate and native Python work side-by-side\n\n")

cat("Setup script completed.\n\n")

invisible(TRUE)