Projekty R i zarządzanie obszarem roboczym

Obszar roboczy R: .RData i .Rhistory

# .RData: binary snapshot of all objects in the global environment
# .Rhistory: text file of every command you typed

# See what would be saved right now:
cat('Objects that would be saved:', length(ls()), '\n')
cat('Their names:', paste(ls(), collapse = ', '))

# Workspace saving is controlled by:
# Tools -> Global Options -> General -> Save workspace (NEVER recommended)
# Or at startup: R --no-save --no-restore

Dlaczego unikać automatycznego zapisywania .RData?

# The recommended R startup setting (in RStudio):
# Tools -> Global Options -> General
#   Restore .RData at startup: UNCHECKED
#   Save workspace on exit: NEVER

# This forces you to write scripts that are fully self-contained.
# A fresh session that runs cleanly = reproducible analysis.

# Equivalent command-line flags:
# R --no-save --no-restore

cat('Always start fresh!')

rm() — czyszczenie obszaru roboczego

# Create some objects
a <- 1
b <- 'hello'
df <- data.frame(x = 1:3)

cat('Before rm:', length(ls()), 'objects\n')

# Remove specific objects
rm(a, b)
cat('After rm(a,b):', length(ls()), 'objects\n')

# Remove ALL objects in global environment
rm(list = ls())
cat('After rm(list=ls()):', length(ls()), 'objects')

options() — globalne ustawienia sesji

# Default behavior:
cat(0.000123456, '\n')    # may show 1.23456e-04

# Suppress scientific notation (positive scipen = prefer fixed)
options(scipen = 999)
cat(0.000123456, '\n')    # now shows 0.000123456

# Control printed digits
options(digits = 4)
cat(pi, '\n')              # 3.142

# Check current options:
cat('scipen:', getOption('scipen'), '\n')
cat('digits:', getOption('digits'))

.Renviron — zmienne środowiskowe

# .Renviron file (in ~ or project root):
# DATABASE_URL=postgres://user:pass@host/db
# OPENAI_API_KEY=sk-abc123
# DATA_ROOT=/data/shared

# Read in R:
db_url <- Sys.getenv('DATABASE_URL')
api_key <- Sys.getenv('OPENAI_API_KEY', unset = 'KEY_NOT_SET')

cat('DB URL set:', nchar(db_url) > 0, '\n')
cat('API key:', api_key)

# To edit .Renviron:
# usethis::edit_r_environ()   <- if usethis is installed
# file.edit('~/.Renviron')    <- base R

saveRDS() — zapisywanie pojedynczego obiektu

# Create an object to save
result_list <- list(
  model   = 'linear_regression',
  r2      = 0.87,
  rmse    = 12.4,
  created = Sys.time()
)

# Save to disk
tmp_rds <- tempfile(fileext = '.rds')
saveRDS(result_list, file = tmp_rds)
cat('Saved to:', tmp_rds, '\n')

# Load it back (even in a fresh session)
loaded <- readRDS(tmp_rds)
cat('R-squared:', loaded$r2)

readRDS() — przywracanie pojedynczego obiektu

# Suppose model_results.rds was saved earlier
tmp <- tempfile(fileext = '.rds')
saveRDS(list(coef = c(1.2, 3.4), n = 100), tmp)

# Load with a meaningful name:
model_output <- readRDS(tmp)
cat('Coefficients:', model_output$coef, '\n')
cat('Sample size:', model_output$n, '\n')

# Key advantage over load():
# With load(), the name is baked into the file.
# With readRDS(), YOU choose the name on load.
cat('readRDS gives you naming control')

save() i load() — wiele obiektów

# Save multiple objects at once
scores <- c(85, 92, 78)
labels <- c('Alice', 'Bob', 'Carol')

tmp_rdata <- tempfile(fileext = '.RData')
save(scores, labels, file = tmp_rdata)
cat('Saved to:', tmp_rdata, '\n')

# Remove originals to simulate a new session
rm(scores, labels)
cat('Removed. Objects now:', length(ls()), '\n')

# Load restores original names automatically
load(tmp_rdata)
cat('Restored scores:', scores)

saveRDS a save — kiedy używać której funkcji

# saveRDS: one object, you name it on load
# saveRDS(my_model, 'outputs/model_v2.rds')
# m <- readRDS('outputs/model_v2.rds')  <- your name choice

# save: multiple objects, names fixed in the file
# save(model, validation_df, file = 'outputs/session.RData')
# load('outputs/session.RData')  <- restores 'model' and 'validation_df'

# Practical rule:
# saveRDS for caching single results
# save for snapshots shared between collaborators

cat('saveRDS is usually the better choice')

Sprawdzanie i przywracanie opcji options()

# Capture current options before changing
old_options <- options(scipen = 999, digits = 3)

# Use on.exit to restore them no matter what
on.exit(options(old_options), add = TRUE)

# Now do your work
cat(pi, '\n')           # 3.14
cat(1234567.89, '\n')   # 1234568 (no sci notation)

# When function or script exits, options are restored automatically
cat('Options will restore on exit')

Projekty RStudio do zarządzania obszarem roboczym

# When you open a .Rproj file, RStudio automatically:
# 1. Sets working directory to the project root
# 2. Loads the project-specific .Rhistory
# 3. Optionally restores project .RData

# Best practice for new projects:
# File -> New Project -> New Directory -> New Project

# Your project structure then looks like:
# my_project/
#   my_project.Rproj
#   .Rprofile        (project-level options)
#   .Renviron        (project-level env vars)
#   R/               (scripts)
#   data/            (raw data)
#   output/          (results)

cat('RStudio Projects = clean workspaces')

Szybki test

Zarządzanie obszarem roboczym — najważniejsze informacje

# Typical script header for clean, reproducible analysis:
rm(list = ls())                         # clean workspace
options(scipen = 999, digits = 4)       # formatting

# Load secrets from .Renviron:
db_url <- Sys.getenv('DATABASE_URL')

# Load cached data if available:
cache_file <- file.path('cache', 'processed_data.rds')
if (file.exists(cache_file)) {
  df <- readRDS(cache_file)
  cat('Loaded from cache\n')
} else {
  cat('Cache not found, run data_prep.R first\n')
}

cat('Workspace ready')