Sed

Usage ​

See the manual locally with info sed.

sed OPTIONS INPUTFILE

Commands ​

Replace: s ​

Any character other than backslash or newline can be used instead of a slash to delimit the pattern and the replacement.

Within the pattern and the replacement, the pattern delimiter itself can be used as a literal character if it is preceded by a backslash.

sed 's/pattern/replacement/g'
sed 's#pattern#replacement#g'    # Using hash instead of slash
sed '0,/pattern/s//replacement/' # Once
sed 's/.*/\u&/'                  # Capitalize

Remove ANSI escape sequences:

sed 's/\x1b\[[0-9;]*m//g'           # Remove color sequences only
sed 's/\x1b\[[0-9;]*[mGKH]//g'      # Remove color and move sequences
sed 's/\x1b\[[0-9;]*[a-zA-Z]//g'    # Remove all escape sequences

Remove carriage return \M:

sed 's/\r//'

Remove trailing newlines:

sed -Ez '$ s/\n+$//'

Find and replace all matches recursively in files:

git grep --files-with-matches foo | xargs -n 1 sed -i 's/foo/boo/g'

Finding patterns ​

Regex option p : If the substitution was made, then print the new pattern space.

sed -nE 's/.*"total":([0-9]+).*/\1/p'
sed -n 's/^Your code has been rated at \([-0-9.]*\)\/.*/\1/p' .pylint

Note: \d does not work on sed. Use [0-9] instead.

For example, to make a linter fail if it contains more than N errors, run:

echo 'Found 4 errors' | sed -nE 's/^Found ([0-9]+) error.*$/\1/p' | xargs test 4 -ge

It is also possible to just print some specific lines, for example from line 10 to 20:

sed -n '10,20p'

Change: c ​

Replaces a line with some text.

sed '1c\#!/data/data/com.termux/files/usr/bin/env fish' script.fish # Changes only the first line
sed '1c\first line\
second line' # Add multiple lines

Insert: i ​

Add text to previous line.

sed '/pattern/i\new text'
sed '0,/pattern/i\new text' # Once
sed '1i\first line\
second line' # Add multiple lines

Append: a ​

Add text to next line.

sed '/pattern/a\new text'
sed '0,/pattern/a\new text' # Once
sed '1a\first line\
second line' # Add multiple lines

Delete: d ​

Delete line.

sed '/pattern/d'
sed '/start_pattern/,/end_pattern/d'                  # Delete lines between two patterns (inclusive)
sed '/start_pattern/,/end_pattern/{/end_pattern/!d}'  # Delete lines between two patterns (exclude)
sed '/start_pattern/,+1{/start_pattern/!d}'           # Delete following line to the match

Read file: r ​

Append text read from file. Do not escape the / character on the file path.

sed -e '/{{ CONTENTS }}/r /tmp/contents.html' -e '/{{ CONTENTS }}/d'

Stop reading input: q ​

Given that we just want to print one line from a file, one can use this command.

sed '5q;d' file

Explanation:

Commands within a script or script-file can be separated by semicolons ; or newlines. Multiple scripts can be specified with -e or -f options.

We have two commands: 5q d.

• 5q: This will only run at the 5th line. sed by default prints the line (-n is not specified). So the 5th line is printed and sed will quit before executing d for that line.
• d: This will run on all lines from 1st to 4th. Which means that even though sed prints all lines by default, it is also deleting all lines by default. Except for the 5th line, which q is run and the script stops immediately.