sed

sed [options] 'command' [file...]

sed [options] -e 'command1' -e 'command2' [file...]

The following files are used as examples.

data.txt

Hello World foo foo
This is a foo test
Another line here

config.txt

# Database settings
host=localhost
# Port number
port=3306

s/before/after/ replaces the first match on each line. Adding g replaces all matches.

sed 's/foo/bar/' data.txt
Hello World bar foo
This is a bar test
Another line here

The following example demonstrates this:

sed 's/foo/bar/g' data.txt
Hello World bar bar
This is a bar test
Another line here

/pattern/d deletes lines matching the pattern. The example below removes comment lines (lines starting with #).

sed '/^#/d' config.txt
host=localhost
port=3306

/^$/d removes blank lines.

printf "line1\n\nline2\n\nline3\n" | sed '/^$/d'
line1
line2
line3

Combining -n and the p command extracts only specific lines.

sed -n '2,4p' config.txt
host=localhost
# Port number
port=3306

-e runs multiple substitutions at once.

sed -e 's/foo/bar/g' -e 's/World/Earth/' data.txt
Hello Earth bar bar
This is a bar test
Another line here

-i edits the file directly (in-place substitution). -i.bak creates a backup before editing.

sed -i 's/old/new/g' file.txt

The following example demonstrates this:

sed -i.bak 's/old/new/g' file.txt

Removes leading and trailing whitespace (trim).

echo "  hello world  " | sed 's/^[[:space:]]*//; s/[[:space:]]*$//'
hello world

Deletes the first line (useful for skipping CSV headers, for example).

echo -e "name,age\nKiryu Kazuma,37\nMajima Goro,39" | sed '1d'
Kiryu Kazuma,37
Majima Goro,39

sed supports POSIX extended regular expressions. With GNU sed, use the -E option to enable ERE syntax without escape characters.

Deletes lines consisting only of digits (BRE).

printf "abc\n123\nfoo456\n789\n" | sed '/^[0-9]*$/d'
abc
foo456

Use -E for extended regular expressions (ERE). No backslash needed for + and ?.

printf "color\ncolour\n" | sed -E 's/colou?r/hue/g'
hue
hue

Use capture groups (\1) to preserve part of the match. Converts "YYYY-MM-DD" to "DD/MM/YYYY".

echo "2025-03-28" | sed -E 's/([0-9]{4})-([0-9]{2})-([0-9]{2})/\3\/\2\/\1/'
28/03/2025

macOS ships with BSD sed, while Linux ships with GNU sed. They differ in option syntax and regex handling, so a script that works on one platform may fail on the other. This is especially important when your local machine runs macOS but your CI/CD or servers run Linux.

The backup-extension form works on both platforms.

sed -i.bak 's/localhost/127.0.0.1/g' config.txt

config.txt is rewritten; the original is saved as config.txt.bak.

For extended regex, use -E — it is supported by both GNU sed and BSD sed (GNU sed also accepts -r, but BSD sed does not).

sed -E 's/colou?r/hue/g' data.txt
hue
hue

Newline and tab insertion differs between the two, so using printf or $'\n' is the most portable approach.

sed "s/:/$(printf '\n')/g" data.txt

To use GNU sed on macOS, install it via Homebrew. After installation, use the gsed command.

brew install gnu-sed
gsed -i 's/old/new/g' file.txt

You can also detect the environment in a script and switch automatically.

if sed --version > /dev/null 2>&1; then
    SED="sed"
else
    SED="gsed"
fi
$SED -i 's/old/new/g' file.txt

GNU sed supports --version but BSD sed does not, which makes this a reliable way to detect the environment.

Common mistake 1: slash as delimiter

Using / as the delimiter makes path substitution awkward and can cause syntax errors. You can use any character as the delimiter — |, @, #, etc. — so choose a different one when the substitution involves slashes.

sed 's|/var/log|/var/www|g' file.txt

Common mistake 2: using p without -n

Using p without -n causes each line to be printed twice — once by the normal output and once by p. To extract specific lines only, always pair -n with p.

sed 'p' data.txt
Hello World foo foo
Hello World foo foo
This is a foo test
This is a foo test
Another line here
Another line here

The following example demonstrates this:

sed -n 'p' data.txt
Hello World foo foo
This is a foo test
Another line here

The following example demonstrates this:

sed -n '2,4p' data.txt
This is a foo test
Another line here

When using sed -i for in-place editing, creating a backup first or working within a git-managed repository helps protect your files. On macOS, BSD sed is used instead of GNU sed, so some option syntax differs (for example, macOS requires an empty string: sed -i "").

For more complex text processing, use awk. For simple searches, see also grep.