Short Course on R Tools

Mehdi Maadooliat and Hossein Haghbin

Marquette University
SCoRT - Summer 2025

Motivating Example - What is Data Science?

R vs Python

Motivating Example - Functional Singular Spectrum Analysis

Outline

• Session 1. OOP in R
• Session 2. Web Apps with Shiny
• Session 3. Rcpp – High Performance with C++ in R
• Session 4. Python in R
• Session 5. Build Your Own R Package
• Session 6. CRAN Submission
• Session 7. GitHub Essentials

🧠 Session 1: OOP in R

Motivating Example - OOP

Motivating Example - OOP (Cont.)

OOP in R

• 🧩 OOP Paradigms in R:
  • Functional OOP: plot(lm(...))
  • Encapsulated OOP: learner$train()
  • Modern tools: mlr3, R6, S7

• 🏗 OOP Systems:
  • S3: Lightweight & informal — uses UseMethod()
  • S4: Formal and type-safe — uses setClass() and @
  • R6: Mutable, used in apps like shiny
  • S7: Newest and unified — multiple dispatch

🛠️ OOP in Practice

🔁 Comparison of Systems

Sys.	Style	Used In	✅ Pros	⚠️ Cons
S3	Functional	Base R	Simple, extensible	No schema, fragile
S4	Formal	Bioconductor	Type-safe, robust	Verbose, complex
R6	Encapsulated	Shiny, Keras	Fast, mutable	Weak type safety
S7	Unified OOP	Modern dev	Combines S3 & S4 benefits	Still evolving

Session 2:🌐 Web Apps with Shiny

FSSA - Shiny

🔗 Open in a browser

Web Apps with Shiny

• ✨ Shiny = Interactive web apps in pure R (no HTML/JS!)

• 📦 Components:
  • ui: Layout, inputs/outputs
  • server: Logic, reactivity

• 🔌 Inputs & Outputs:
  • Inputs: sliderInput(), textInput()
  • Outputs: renderPlot(), renderText(), renderTable()

• 🧪 Use Cases:
  • Dashboards, simulations, teaching tools

Shiny Concepts & Deployment in Workshop

• 🚀 Deployment Options:
  • shinyapps.io, Posit Connect, university servers
  • shinylive: deploy as static site (experimental)

• ⚡Reactivity System:
  • reactive(): memoized reactive expressions
  • isolate(): break reactive chains
  • observeEvent(): trigger actions without outputs

• 🧠 Hands-on:
  • Explore shinyapps gallery for inspiration

⚡ Session 3: Rcpp –
High Performance with C++ in R

Context & Motivation

• Diagonal averaging is key in many algorithms (e.g. SSA reconstruction).
  
• Given any matrix \(G \in \mathbb{R}^{L\times K}\), we want \[ \scriptsize{y_t = \frac{1}{\#\{(i,j):\,i+j-1 = t\}} \sum_{i+j-1 = t} G_{i,j}, \quad t = 1,\dots, L+K-1.} \]
• Pure‑R implementation uses two nested loops over \(i\) and \(j\).

Pure R Code

hankelize_R <- function(X){
  L <- nrow(X)
  K <- ncol(X)
  N <- L + K - 1
  result <- numeric(N)
  count <- numeric(N)
 
  for(i in 1:L){
    for(j in 1:K){
      result[i + j - 1] <- result[i + j - 1] + X[i,j]
      count[i + j - 1] <- count[i + j - 1] + 1
    }
  }
 
  return(result / count)
}

Rcpp Code

library(Rcpp)

cppFunction('
NumericVector hankelize_rcpp(NumericMatrix X) {
  int L = X.nrow();
  int K = X.ncol();
  int N = L + K - 1;
 
  NumericVector result(N);
  NumericVector count(N);
 
  for(int i = 0; i < L; i++){
    for(int j = 0; j < K; j++){
      result[i + j] += X(i,j);
      count[i + j] += 1;
    }
  }
 
  for(int i = 0; i < N; i++){
    result[i] /= count[i];
  }
 
  return result;
}')

Benchmark

library(microbenchmark)

# Large random matrix (e.g., 100 x 1000)
set.seed(123)
X_large <- matrix(runif(1e5), nrow = 100, ncol = 1000)

bench <- microbenchmark(
  R_version = hankelize_R(X_large),
  Rcpp_version = hankelize_rcpp(X_large),
  times = 10
)

print(bench)
># Unit: microseconds
>#          expr     min      lq     mean   median      uq     max neval
>#     R_version 21225.1 21550.8 26604.53 25180.35 30969.3 34455.1    10
>#  Rcpp_version   176.9   177.8   299.87   194.50   253.7  1161.3    10

• Dramatic speedup: ~89× faster.
• Pattern can be applied to many \(O(n^2)\) tasks in time series, spatial stats, and beyond.

🔬 Rcpp in Action

• 🚀 Rcpp bridges R and C++ for high-performance computing.
• 🏁 Use cases:
  • Speed up loops, recursion, and matrix operations
  • Access C++ libraries and STL
  • Integrate C++ in R packages

🐍 4. Python in R

🔁 Example: Use Python from R

🐍 Native Python

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.neural_network import MLPClassifier
iris = load_iris()
X, y = iris.data, iris.target
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
mlp = MLPClassifier(hidden_layer_sizes=(10,))
mlp.fit(X_train, y_train)
y_pred = mlp.predict(X_test)

📦 R with reticulate

library(reticulate)
sklearn <- import("sklearn")
model_selection <- sklearn$model_selection
neural_network <- sklearn$neural_network
X <- as.matrix(iris[, 1:4])         
y <- as.integer(iris$Species) - 1     
train_test <- model_selection$train_test_split(X, y, test_size = 0.2, random_state = 42)
X_train <- train_test[[1]]
X_test <- train_test[[2]]
y_train <- train_test[[3]]
y_test <- train_test[[4]]
mlp <- neural_network$MLPClassifier(hidden_layer_sizes = tuple(10L))
mlp$fit(X_train, y_train)
y_pred <- mlp$predict(X_test)

💡 Why Call Python in R?

• 🔄 Avoid rewriting code from one language to another
• 🔁 Call shared models or preprocessing functions written in Python
• 📦 Reuse Python packages or APIs within your R workflow
• 📤 Publish reproducible reports that include both R and Python code

What is reticulate?

• R package for interoperability with Python
• Maintains a shared Python session accessible from R
• Automatic conversion of many R types to Python and vice versa
• Supports inline Python code in R Markdown and R scripts

🎥 Demo: Python from R in RStudio

📦 5. Build Your Own R Package

💡 Motivations to Package Your Code

🧰 Why build a package?

• ♻️ Reuse and organize reusable code chunks
• 📖 Provide documentation and examples
• 🧪 Add unit tests to ensure reliability
• 🌐 Distribute:
  • via CRAN for wide access
  • via GitHub for collaborative development
  • or keep it private for internal use

🏗 Package Structure Overview

mypkg/
├── DESCRIPTION         # 📋 Package metadata
├── NAMESPACE           # 🔁 Exported functions & imports
├── R/                  # 📂 Your R functions
└── man/                # 📚 Auto-generated documentation

💡 Other folders you may add:

• 📂 tests/ → Unit tests (testthat)
• 📂 vignettes/ → Long-form docs
• 📂 data/ → Internal or example datasets

🎥 Demo: Rfssa Package

🛠 Package States in R

R packages transition through five development states:

• 🗂️ Source: your raw package folder
• 📦 Bundled: compressed .tar.gz for sharing
• 🧱 Binary: platform-specific precompiled version
• 📚 Installed: available in your R library
• 🧠 In-Memory: actively loaded via library()

🔁 Transitioning Between States

6. 📤 CRAN Submitting

🚀 Submitting to CRAN

• 📦 CRAN = Comprehensive R Archive Network
  
• 🎯 Goal: Make your R package public, discoverable, and installable
• ✔ Must pass R CMD check with:
  • ❌ No ERRORs
  • ⚠️ No WARNINGs
  • 📝 Minimal and justifiable NOTEs
• 📤 Submit with: devtools::release()

📩 Confirmation Email (After Submission)

📬 Submission Confirmed Email

⏳ Pending Manual Inspection Email

✅ Package Initially Accepted Email

🎉 Good news!

📦 Package Built on CRAN Email

🧱 CRAN has built binaries for major platforms (Windows, …).

🌐 CRAN Page Created

🌍 Your package is now live! Searchable and installable via:

install.packages("Rfssa")

7. 📤 GitHub Essentials

📬 Example: Rfssa Package Github

🌍 Why Use Git & GitHub?

• 🔁 Version Control: Track and manage changes in code
• 🤝 Collaboration: Work with others without overwriting files
• 🧪 Reproducibility: Restore older versions when bugs occur
• 🔄 Open Source: Share your package with the world

🧰 Git = Local tracking tool
☁️ GitHub = Online hosting and teamwork platform

🔁 Daily Git Workflow in RStudio

🎯 A typical work cycle for package developers:

Step	Description
✏️ Edit	Make changes to .R or .Rmd files
✅ Stage	Select files to track in Git tab
💬 Commit	Save with a message (e.g., “Fix bug”)
🚀 Push	Send to GitHub

🎥 Demo: Rfssa Github Package

🙏 Thank you!

Questions & Discussion