sed

Updated

September 4, 2026

Overview

sed (stream editor) transforms text line by line — substitute, delete, print, and simple inserts — without opening an interactive editor. GNU sed is standard on Ubuntu/Debian. Prefer sed for mechanical rewrites; use awk for field logic and an editor for multi-line structural edits.

Syntax

sed [options] 'script' [file...]
sed [options] -e 'script' -e 'script' [file...]
sed [options] -f script.sed [file...]
… | sed 'script'

Common Options

Option Description
-n Quiet; only explicit p prints
-E / -r Extended regex (ERE)
-i[SUFFIX] In-place edit (optional backup suffix)
-e Add a script expression
-f file Read script from file
-z NUL-separated “lines” (GNU; useful with find -print0)

Common script commands

Cmd Meaning
s/pat/repl/flags Substitute (g global, i ignore case, p print, 2 Nth only)
d Delete line
p Print line
a\ / i\ / c\ Append / insert / change (multi-line form in scripts)
q / Q Quit (GNU Q without printing pattern space)
y/src/dst/ Transliterate like tr
= Print line number

Safety

  • sed -i rewrites files — always keep backups (-i.bak) until the expression is trusted.
  • Prefer single quotes around scripts so the shell does not expand $ and `.
  • Never run sed -i on binary files or unknown encodings.
  • On macOS, -i requires a mandatory suffix argument — scripts written for GNU sed may break there.

Key Use Cases

  1. Bulk string replace in config/logs
  2. Strip comments/blank lines
  3. Extract a capture group (print-only substitute)
  4. Normalize line endings / trailing whitespace
  5. Quick in-place flips for feature flags

Examples with Explanations

Substitute first match per line

sed 's/foo/bar/' file.txt

Only the first foo on each line becomes bar.

Global substitute

sed 's/foo/bar/g' file.txt

Every non-overlapping match on the line is replaced.

Extended regex

sed -E 's/[0-9]{1,3}(\.[0-9]{1,3}){3}/IP/g' access.log

With -E, {} and () work without backslashes (GNU ERE).

Delete matching lines

sed '/^#/d; /^$/d' config.txt

Drops comment lines and empty lines — useful for “effective config” views.

In-place with backup

sed -i.bak 's/Enable=false/Enable=true/' app.conf
diff -u app.conf.bak app.conf

Inspect the diff before deleting the .bak.

Line addresses and ranges

sed -n '10,20p' file.txt          # print lines 10–20
sed '1,5d' file.txt               # delete first five
sed '/^## Start/,/^## End/d' doc.md

Addresses can be numbers, $ (last line), or /regex/.

Multiple expressions

sed -e 's/\r$//' -e 's/[[:space:]]\+$//' file.txt

Strip CR from CRLF files and trailing whitespace.

Use match in replacement

sed -E 's/(user=)[^ ]+/\1REDACTED/g' app.log

Parentheses capture; \1 reinserts the prefix. & is the whole match.

Change delimiter when paths have slashes

sed 's|/usr/local/bin|/opt/bin|g' env.sh

Any character after s can be the delimiter (s|||, s###).

Hold space: join two lines (advanced)

sed -n '1h;1!H;${x;s/\n/,/g;p}' list.txt

GNU sed can join lines via hold space; for heavy multi-line work prefer awk or perl.

Dry-run pattern for risky edits

sed 's/prod/staging/g' settings.yml | diff -u settings.yml -
# then:
sed -i.bak 's/prod/staging/g' settings.yml

Always preview against the original before -i.

Understanding Output

By default sed prints every line (after script processing) to stdout. Exit status is 0 on success; GNU sed can return non-zero for invalid options or I/O errors. In-place mode still writes a temporary file then renames — a full disk mid-edit can fail; keep backups.

Notes & Pitfalls

  • Default regex is BRE; +, ?, | need escapes unless you use -E.
  • s/a/b/ only changes the first match; forget g often.
  • & in the replacement is the whole match; write \& for a literal ampersand.
  • sed is line-oriented — multi-line HTML/XML edits are painful.
  • Character classes: prefer [[:space:]] over \s for portable scripts.

Additional Resources

  • man sed
  • GNU sed manual