R Cheat Sheet: Vectors, Data Frames, Strings and Files

Short answer: In R, everything is a vector: c(3, 1, 2) builds one, operations apply element by element, and indexes start at 1. Assign with <-, subset with x[x > 1] or df[df$cpu > 50, ], loop less and use sapply, vapply or Map instead, chain steps with the native pipe |>, and write short functions as \(x) x + 1. Read data with read.csv(), catch errors with tryCatch(), and summarise with summary(), table() and aggregate().

The examples in this R cheat sheet were run with R 4.3.3 on 9 October 2026, using base R only. The current release is R 4.6.1; functions marked 4.4+ or later were checked against the R NEWS file.

Vectors and basic types

Syntax	What it does	Example
x <- c(3, 1, 2)	Create a vector (= also assigns)	ports <- c(80, 443)
5L, "a", TRUE, NA, NULL	Integer, character, logical, missing value, nothing	class(5L) gives "integer"
1:5, seq(1, 10, by = 3)	Sequences	seq_len(n) is safe when n is 0
rep(x, times = 2), rep(x, each = 2)	Repeat	rep(1:2, each = 2) gives 1 1 2 2
x * 2, x + y	Element-wise arithmetic; shorter vectors recycle	c(3, 1, 2) * 2
%/%, %%, ^	Integer division, modulo, power	7 %/% 2 is 3
sum, mean, max, range, cumsum	Summaries; na.rm = TRUE skips NA	mean(c(1, NA, 3), na.rm = TRUE)
sort, order, rev, unique	order returns positions, for sorting other things	df[order(-df$cpu), ]
%in%, any, all, which	Membership and tests (%notin% is in base from 4.6)	which(x > 1)
ifelse(test, yes, no)	Vectorised if	ifelse(x > 1, "big", "small")
is.na(), as.integer(), as.character()	Test and convert	as.numeric("abc") gives NA with a warning
factor(x, levels = …)	Categorical data with fixed levels	table(f) counts per level
matrix(1:6, nrow = 2)	Matrix (filled by column); %*% multiplies	m[2, 3]

Never compare floating-point numbers with ==: 0.1 + 0.2 == 0.3 is FALSE. Use isTRUE(all.equal(a, b)).

Indexing and subsetting

Syntax	Returns	Example
x[2], x[c(1, 3)]	Elements by position (1-based)	x[length(x)] is the last
x[-1]	Everything except position 1	x[-(1:2)]
x[x > 1]	Elements where a logical test is TRUE	lat[lat > 500]
v["b"]	By name	v <- c(a = 1, b = 2)
l[["key"]] vs l["key"]	The element itself vs a one-element list	cfg[["host"]]
l$key	Shorthand for [["key"]]	df$cpu
df[rows, cols]	Data frame rows and columns; leave one blank for all	df[df$cpu > 50, c("host", "cpu")]
head(x, n), tail(x, n)	First or last n	head(df, 3)

Data frames

Function	What it does	Example
data.frame()	Build a table from equal-length vectors	data.frame(host = c("web01", "db01"), cpu = c(35, 55))
str(df), summary(df)	Structure and per-column summaries	First thing to run on new data
nrow, ncol, names	Size and column names	names(df)
df$new <- …	Add or replace a column	df$high <- df$cpu > 60
subset(df, cond, select = …)	Filter rows and pick columns	subset(df, cpu > 50, select = c(host, cpu))
aggregate(y ~ g, data, FUN)	Group and summarise	aggregate(cpu ~ role, data = df, FUN = mean)
tapply(x, g, f)	Apply a function per group	with(df, tapply(cpu, role, max))
merge(a, b, by = "k")	Join two data frames	merge(df, owners, by = "role")
rbind, cbind	Stack rows or columns	rbind(df1, df2)
split(x, g)	List of pieces per group	split(df$cpu, df$role)
transform(df, …)	Add computed columns	transform(df, cpu_frac = cpu / 100)

df |> subset(role == "web") |> nrow()
mtcars |> lm(mpg ~ wt, data = _) |> coef()   # _ placeholder must be a named argument

Lists, functions and the apply family

Syntax	What it does	Example
list(host = "web01", ports = c(80, 443))	Container for mixed types, can nest	l$meta$os
l$key <- NULL	Remove an element	cfg$password <- NULL
f <- function(x, y = 2, ...) { … }	Function with default and extra arguments; the last value is returned	greet <- function(name, greeting = "Hello")
\(x) x + 1	Short lambda (R 4.1+)	sapply(1:3, \(i) i^2)
<<-	Assign in an enclosing environment (closures)	n <<- n + 1
missing(arg)	Was an argument supplied?	if (missing(b)) "no b"
match.arg(x, choices)	Validate against allowed values, with partial matching	match.arg("med", c("low", "medium"))
lapply / sapply	Apply over a list; return a list / simplify	lapply(files, readLines)
vapply(x, f, integer(1))	Like sapply, with a checked result type	vapply(c("a", "bb"), nchar, integer(1))
Map, mapply	Iterate over several vectors in parallel	Map(\(a, b) a + b, 1:2, 3:4)
Reduce, Filter	Fold and filter	Filter(\(n) n %% 2 == 0, 1:6)
apply(m, 1, f)	Rows (1) or columns (2) of a matrix	apply(m, 2, max)
do.call(f, list)	Call with arguments from a list	do.call(rbind, pieces)
x %||% y (4.4+)	y when x is NULL	opts$timeout %||% 30

Control flow

Syntax	Example
if (…) { … } else if (…) { … } else { … }	if (code >= 500) "server" else "ok"
for (h in hosts) { … }	for (i in seq_along(x)) …
while (cond) { … }, repeat { … }	repeat needs a break
next / break	Skip an iteration / leave the loop
switch(x, a = 1, b = 2, default)	switch("db", web = 80, db = 3306, 0) gives 3306

if needs a single TRUE or FALSE. Use ifelse() or any()/all() when the condition is a vector, otherwise R 4.2+ raises an error.

Strings and regular expressions

Function	What it does	Example
paste(a, b, sep = "-"), paste0()	Join; collapse joins a vector into one string	paste(hosts, collapse = ",")
sprintf()	printf-style formatting, vectorised	sprintf("%-8s %5d", host, port)
format(x, big.mark = ",")	Number formatting	format(1234567.891, big.mark = ",", nsmall = 2)
nchar, toupper, trimws, substr	Length, case, trim, substring	substr(s, 1, 5)
strsplit(s, ",")[[1]]	Split (returns a list)	strsplit("a,b,c", ",")[[1]]
startsWith, endsWith	Prefix and suffix tests	endsWith(f, ".log")
grepl(pat, x)	TRUE/FALSE per element	grepl("^web", hosts)
grep(pat, x, value = TRUE)	Matching elements (grepv() in 4.5+ does the same)	grep("db", hosts, value = TRUE)
sub / gsub	Replace first / all; \\1 for groups	sub("^(\\w+)\\..*$", "\\1", fqdn)
regmatches(x, gregexpr(pat, x))	Extract all matches	regmatches(s, gregexpr("[0-9]+", s))[[1]]
fixed = TRUE, perl = TRUE	Literal match / PCRE (lookarounds)	grepl(".", x, fixed = TRUE)
r"(C:\path)"	Raw string, no escaping (4.0+)	Windows paths, regexes

R strings need double backslashes in patterns: "\\d+" is the regex \d+. A raw string such as r"(\d+)" avoids that.

Files, dates and running R from the shell

Task	Code
Read a CSV	df <- read.csv("hosts.csv") (stringsAsFactors is FALSE by default since 4.0)
Write a CSV	write.csv(df, "out.csv", row.names = FALSE)
Lines of text	readLines(path) / writeLines(lines, path)
Append	cat("line\n", file = path, append = TRUE)
Save any R object	saveRDS(df, "df.rds") then readRDS("df.rds")
Files	file.exists, list.files(dir, pattern = "\\.csv$"), unlink(path), tempfile()
Environment	Sys.getenv("HOME")
Dates	as.Date("2026-10-09") + 30, format(d, "%d %b %Y")
Parse a log time	as.POSIXct("09/Oct/2026:10:00:00", format = "%d/%b/%Y:%H:%M:%S", tz = "UTC")
Time difference	difftime(t2, t1, units = "mins")
Run a command	system2("uptime", stdout = TRUE)

For server admins: Rscript runs R non-interactively, so it fits in cron jobs and pipelines: Rscript report.R, or one-liners such as Rscript -e 'summary(read.csv("latency.csv")$ms)'. Install the interpreter from the distribution (dnf install R from EPEL on AlmaLinux, apt install r-base on Debian and Ubuntu).

Errors and conditions

res <- tryCatch({
  stop("bad port")
}, error = function(e) {
  paste("caught:", conditionMessage(e))
}, finally = {
  cat("cleanup\n")
})
print(res)   # "caught: bad port"

Function	What it does
stop(msg), warning(msg), message(msg)	Signal an error, a warning or a message (messages go to stderr)
tryCatch(expr, error = , warning = , finally = )	Handle conditions; a caught warning stops the expression
withCallingHandlers() + invokeRestart("muffleWarning")	Log a warning and carry on
try(expr, silent = TRUE)	Returns an object of class try-error on failure
stopifnot("port must be numeric" = is.numeric(p))	Assertions with custom messages
on.exit(close(con), add = TRUE)	Cleanup when the function exits
errorCondition(msg, class = "configError")	Custom condition class to catch by name

Quick statistics and plots

Function	What it does	Example
summary(x)	Min, quartiles, mean, max	summary(lat)
quantile(x, c(0.5, 0.95))	Percentiles	p50 and p95 latency
sd, var, median, cor	Spread and correlation	cor(mtcars$mpg, mtcars$wt)
table(x), prop.table()	Counts and proportions	sort(table(codes), decreasing = TRUE)
cut(x, breaks)	Bin numbers into ranges	table(cut(lat, c(0, 100, 500, Inf)))
lm(y ~ x1 + x2, data)	Linear model; coef() and summary() on the result	lm(mpg ~ wt + hp, data = mtcars)
set.seed(1), sample, rnorm	Reproducible random numbers	sample(1:10, 3)
hist, plot, boxplot	Base graphics	Wrap with png("out.png") … dev.off() on a headless server

lat <- c(120, 95, 300, 110, 2000, 130)
quantile(lat, c(0.5, 0.95))
#  50%  95%
#  125 1575

Official documentation (checked 9 October 2026): An Introduction to R · R Language Definition · R NEWS (release notes) · R manuals · The R Project

R Cheat Sheet at a glance

Related guides: Python Cheat Sheet: Syntax, Data Types and File Handling · Excel Cheat Sheet: Formulas, Dynamic Arrays and Shortcuts · SQL Cheat Sheet: MySQL, MariaDB and PostgreSQL Syntax.

Reading a lot of our guides? Go ad-free with srvScripts Pro for $6 a month. It also pays for new guides. Cancel any time.