reticulate

reticulate

R interface to Python.

Setup

library(reticulate)

# Use specific Python
use_python("/usr/bin/python3")
use_virtualenv("myenv")
use_condaenv("myenv")

# Check configuration
py_config()
py_available()

Import Modules

# Import Python modules
np <- import("numpy")
pd <- import("pandas")
sklearn <- import("sklearn")

# Import with conversion disabled
np <- import("numpy", convert = FALSE)

# Import submodules
preprocessing <- import("sklearn.preprocessing")

Call Python

# Run Python code
py_run_string("x = 1 + 1")
py_run_file("script.py")

# Access Python objects
py$x
py$my_function(arg1, arg2)

# Execute in main module
py_run_string("
import pandas as pd
df = pd.DataFrame({'a': [1,2,3]})
")
py$df

Data Conversion

# R to Python
py_df <- r_to_py(mtcars)

# Python to R
r_df <- py_to_r(py$df)

# Automatic conversion (default)
np$array(c(1, 2, 3))  # Returns R vector

# Disable conversion
np <- import("numpy", convert = FALSE)
arr <- np$array(c(1, 2, 3))  # Returns Python object
py_to_r(arr)  # Explicit conversion

NumPy Integration

np <- import("numpy")

# Create arrays
arr <- np$array(matrix(1:9, 3, 3))
arr <- np$zeros(c(3L, 3L))
arr <- np$ones(c(3L, 3L))

# Array operations
np$sum(arr)
np$mean(arr)
np$dot(arr, arr)

# Convert to R
as.matrix(arr)

Pandas Integration

pd <- import("pandas")

# Create DataFrame
py_df <- pd$DataFrame(list(
  a = 1:3,
  b = c("x", "y", "z")
))

# Read files
py_df <- pd$read_csv("data.csv")
py_df <- pd$read_excel("data.xlsx")

# Convert to R
r_df <- py_to_r(py_df)

# R data frame to pandas
py_df <- r_to_py(mtcars)

Scikit-learn

sklearn <- import("sklearn")
preprocessing <- import("sklearn.preprocessing")
model_selection <- import("sklearn.model_selection")

# Preprocessing
scaler <- preprocessing$StandardScaler()
X_scaled <- scaler$fit_transform(X)

# Train/test split
split <- model_selection$train_test_split(X, y, test_size = 0.2)
X_train <- split[[1]]
X_test <- split[[2]]

# Models
linear_model <- import("sklearn.linear_model")
model <- linear_model$LogisticRegression()
model$fit(X_train, y_train)
predictions <- model$predict(X_test)