awk cheat sheet: fields, patterns, sums and printf

updated 2026-10-20

tool: awk
versions: GNU awk 5.x, POSIX awk
os: linux, macos, windows

awk reads input line by line, splits each line into fields ($1, $2, … $NF) and runs pattern { action } rules. The examples below all use the same small CSV file.

bash

cat > sales.csv <<'EOF'
region,product,units,price
north,apple,10,0.5
south,pear,4,0.75
north,pear,7,0.75
east,apple,12,0.5
EOF
echo "--- columns 1 and 3:"; awk -F, 'NR > 1 { print $1, $3 }' sales.csv
echo "--- rows where units > 6:"; awk -F, 'NR > 1 && $3 > 6' sales.csv
echo "--- total units:"; awk -F, 'NR > 1 { sum += $3 } END { print sum }' sales.csv
echo "--- revenue per region:"; awk -F, 'NR > 1 { r[$1] += $3 * $4 } END { for (k in r) printf "%-6s %6.2f\n", k, r[k] }' sales.csv | sort

Output

--- columns 1 and 3:
north 10
south 4
north 7
east 12
--- rows where units > 6:
north,apple,10,0.5
north,pear,7,0.75
east,apple,12,0.5
--- total units:
33
--- revenue per region:
east     6.00
north   10.25
south    3.00

Basics

Syntax	Meaning
awk '{ print $1 }' file	first field of every line
awk -F, '...'	comma as field separator (FS)
$0	the whole line
$NF / $(NF-1)	last / second to last field
NR / FNR	line number overall / in the current file
NF	number of fields on the line
-v OFS="\t"	output field separator

Patterns

Pattern	Selects
/error/	lines matching a regex
$3 > 100	lines where field 3 is greater than 100
$1 == "north"	exact field match
NR > 1	everything except the header line
/start/,/end/	ranges of lines between two matches
BEGIN { ... } / END { ... }	run before the first / after the last line

One-liners

Task	Command
Count matching lines	awk '/error/ { n++ } END { print n+0 }' log
Unique values of a column	awk -F, '!seen[$2]++ { print $2 }' file
Lines longer than 80 characters	awk 'length > 80' file
Use a shell variable	awk -v limit="$LIMIT" '$3 > limit' file
Format columns	awk '{ printf "%-20s %8.2f\n", $1, $2 }' file

Sources

errors · sedsed: -e expression #1, char N: unterminated `s' commandGNU sed reports unterminated `s' command when a substitution misses its final delimiter, often because the pattern or replacement contains an unescaped slash.compat · sedsed -i on macOS vs Linux: GNU sed vs BSD sedGNU sed takes -i with an optional attached suffix, BSD sed on macOS needs an argument: sed -i '' on macOS, sed -i on Linux. Other GNU vs BSD sed differences.errors · findfind: paths must precede expressionfind reports paths must precede expression when an unquoted glob such as *.txt is expanded by the shell into several file names. Quote it: find . -name '*.txt'.

Basics

Patterns

One-liners

Sources

Related pages