sed Cheat Sheet: Commands & Examples

Q: How do I replace a string in a file with sed?

Use sed -i 's/old/new/g' file . -i edits the file in place and g replaces every occurrence on each line, not just the first.

Q: How do I make a backup when using sed -i?

Give -i a suffix with no space: sed -i.bak 's/a/b/' file keeps the original as file.bak .

Q: How do I use a slash in a sed replacement?

Pick a different delimiter, for example sed 's|/usr/local|/opt|g' . Any character after s becomes the delimiter.

Q: How do I delete a line containing a word with sed?

Use sed '/word/d' file to print the result, or sed -i '/word/d' file to change the file.

Q: Why does sed -i fail on macOS?

BSD sed on macOS requires an argument after -i , so use sed -i '' 's/a/b/' file there. GNU sed on Linux takes -i alone.

Most used

sed 's/old/new/g' file              # replace all, print result
sed -i 's/old/new/g' file           # replace all, edit file in place
sed -i.bak 's/old/new/g' file       # same, keep file.bak
sed '/pattern/d' file               # delete lines matching pattern
sed -n '10,20p' file                # print only lines 10-20
sed -E 's/([0-9]+)px/\1rem/g' file  # extended regex with capture group

sed reads input line by line, applies a script to each line and prints the result. Nothing changes on disk unless you use -i. Syntax: sed [options] 'script' [file...]. This page covers GNU sed, the version on every mainstream Linux distro; check with sed --version. Combine it with grep to find files and awk when you need fields and arithmetic.

Substitution

Command	What it does
`sed 's/a/b/' f`	Replace the first a on each line
`sed 's/a/b/g' f`	Replace every a on each line
`sed 's/a/b/2' f`	Replace only the 2nd occurrence on each line
`sed 's/a/b/2g' f`	Replace the 2nd and all later occurrences
`sed 's/a/b/I' f`	Case-insensitive match (GNU extension)
`sed 's\|/old/path\|/new/path\|g' f`	Use `\|` as delimiter to avoid escaping slashes
`sed 's/.*/"&"/' f`	`&` inserts the whole match (wraps each line in quotes)
`sed 's/$key$=$.*$/\2=\1/' f`	BRE capture groups ``, reused as `\1 \2`
`sed -E 's/(key)=(.*)/\2=\1/' f`	Same with extended regex, no escaping of parentheses
`sed 's/[a-z]*/\U&/' f`	Uppercase the match (`\U`, `\L`, `\u`, `\l`, `\E` are GNU)
`sed -n 's/a/b/p' f`	Print only lines where a substitution happened
`sed 's/a/b/w changed.txt' f`	Also write changed lines to a file
`sed -e 's/a/b/' -e 's/c/d/' f`	Run several commands (or separate them with `;`)

Addresses: choosing which lines

Any command can be prefixed with an address. Without one it applies to every line.

Command	What it does
`sed '5s/a/b/' f`	Only line 5
`sed '$s/a/b/' f`	Only the last line
`sed '10,20s/a/b/' f`	Lines 10 to 20
`sed '/start/,/end/s/a/b/' f`	From a line matching start to the next matching end
`sed '/^#/!s/a/b/' f`	`!` negates: every line that is not a comment
`sed '1~2s/a/b/' f`	Every 2nd line starting at 1 (odd lines, GNU)
`sed '/start/,+3d' f`	Matching line plus the next 3 (GNU)
`sed '0,/foo/s//bar/' f`	Replace only the first foo in the whole file (GNU)

Deleting and printing lines

Command	What it does
`sed '3d' f`	Delete line 3
`sed '$d' f`	Delete the last line
`sed '/^$/d' f`	Delete empty lines
`sed '/^\s*#/d' f`	Delete comment lines (including indented ones)
`sed '/^\s*$/d' f`	Delete blank and whitespace-only lines
`sed '1,5d' f`	Delete lines 1 to 5
`sed -n '5p' f`	Print only line 5 (`-n` turns off automatic printing)
`sed -n '/start/,/end/p' f`	Print a block between two patterns
`sed -n '$p' f`	Print the last line
`sed '20q' f`	Print the first 20 lines then quit (like `head -20`, stops reading early)
`sed -n '/pattern/=' f`	Print line numbers of matching lines

Inserting, appending and changing lines

Command	What it does
`sed '1i\# header' f`	Insert a line before line 1
`sed '$a\last line' f`	Append a line after the last line
`sed '/pattern/a\new line' f`	Append after every matching line
`sed '/pattern/i\new line' f`	Insert before every matching line
`sed '/^Port /c\Port 2222' f`	Replace the whole matching line
`sed '/pattern/r other.txt' f`	Insert the contents of a file after matching lines
`sed 'G' f`	Double-space a file (append the empty hold space)
`sed '=' f \| sed 'N;s/\n/ /'`	Number every line

Options

Command	What it does
`sed -i ...`	Edit files in place (GNU: no argument needed)
`sed -i.bak ...`	Edit in place and keep a backup with that suffix
`sed -n ...`	Do not print lines automatically; use `p` to print
`sed -E ...`	Extended regex (`-r` is the older GNU spelling)
`sed -f script.sed f`	Read the sed script from a file
`sed -s ...`	Treat multiple files separately (so `$` means each file's last line)
`sed -z ...`	Lines are separated by NUL instead of newline (lets you match across newlines)
`sed --debug ...`	Show how sed parses and runs the script (GNU 4.6+)

Common recipes

Change a setting in a config file and keep a backup (here, the SSH port; see the SSH cheat sheet before restarting sshd):

sudo sed -i.bak -E 's/^#?Port .*/Port 2222/' /etc/ssh/sshd_config
diff /etc/ssh/sshd_config.bak /etc/ssh/sshd_config

Preview a change first, then apply it:

sed 's/http:/https:/g' site.conf | diff site.conf -
sed -i 's/http:/https:/g' site.conf

Replace across many files found by find:

find . -name '*.md' -type f -exec sed -i 's/linuxtoday\.com/linuxtoday.net/g' {} +

Strip trailing whitespace and Windows line endings:

sed -i 's/[[:space:]]*$//' file.txt
sed -i 's/\r$//' script.sh     # fixes "/bin/bash^M: bad interpreter"

Use a shell variable in the replacement (double quotes, and a delimiter that cannot appear in the value):

new_dir="/srv/www"
sed -i "s|^root .*|root $new_dir;|" site.conf

Print a section of a log between two timestamps:

sed -n '/^2026-10-16 09:00/,/^2026-10-16 09:30/p' app.log

Join all lines into one, comma-separated:

sed -z 's/\n/,/g; s/,$/\n/' list.txt

Uncomment a line:

sudo sed -i 's/^#\s*\(PermitRootLogin\)/\1/' /etc/ssh/sshd_config

Gotchas

GNU vs BSD sed -i. GNU sed (Linux) takes -i or -i.bak. BSD sed (macOS, FreeBSD) requires an argument: sed -i '' 's/a/b/' f. On GNU, sed -i '' ... treats '' as the script and fails. A portable form is sed -i.bak ... && rm f.bak.
-i replaces the file. sed writes a new file and renames it, so hard links are broken, a symlink is replaced by a regular file unless you add --follow-symlinks, and ownership can change if you run it as another user.
Don't combine flags as -ie. GNU sed reads -ie as "in place with backup suffix e" and leaves filee files behind. Write -i -e or -Ei.
Escape the regex, not just the delimiter. . matches any character, so s/1.2.3.4/x/ also matches 1a2b3c4. Use 1\.2\.3\.4. In the replacement, & and \ are special.
sed works line by line. A pattern containing \n won't match across lines unless you use -z or the N command.
\s, \+, \? and I are GNU extensions. They work on Linux but not in strict POSIX or BusyBox sed in every case; use [[:space:]] and -E for portability.
Single quotes stop variable expansion. sed 's/x/$HOME/' inserts the literal text $HOME. Use double quotes when you need a variable.

Reference: the GNU sed manual. For editing in scripts, see the bash scripting tutorial.

Frequently asked questions

How do I replace a string in a file with sed?

Use sed -i 's/old/new/g' file. -i edits the file in place and g replaces every occurrence on each line, not just the first.

How do I make a backup when using sed -i?

Give -i a suffix with no space: sed -i.bak 's/a/b/' file keeps the original as file.bak.

How do I use a slash in a sed replacement?

Pick a different delimiter, for example sed 's|/usr/local|/opt|g'. Any character after s becomes the delimiter.

How do I delete a line containing a word with sed?

Use sed '/word/d' file to print the result, or sed -i '/word/d' file to change the file.

Why does sed -i fail on macOS?

BSD sed on macOS requires an argument after -i, so use sed -i '' 's/a/b/' file there. GNU sed on Linux takes -i alone.

Spotted a mistake or a command that no longer works? Send a correction — pages are re-checked when reports come in.