sed

What it does

sed (stream editor) reads input line by line into a "pattern space," applies a script of editing commands to each line, and writes the result to standard output. It's built for non-interactive text transformation — substituting text, deleting lines, printing ranges — inside scripts and pipelines, without opening a full editor. The most common use is the substitute command, s/pattern/replacement/, but sed also supports deletion, insertion, and line-range selection.

Beginner examples

• sed 's/foo/bar/' file.txt — replace the first "foo" on each line with "bar"
• sed 's/foo/bar/g' file.txt — replace every "foo" on each line
• sed -n '5p' file.txt — print only line 5
• sed '/error/d' file.txt — remove every line containing "error"

sed 's/localhost/127.0.0.1/' hosts.conf

Advanced examples

• Edit a file in place with a backup: sed -i.bak 's/DEBUG/INFO/g' app.conf
• Print a range of lines like a lightweight head/tail: sed -n '10,20p' access.log
• Chain multiple edits with -e: sed -e 's/foo/bar/' -e 's/baz/qux/' file.txt
• Use extended regex for grouping and alternation: sed -E 's/(GET|POST) /\1: /' access.log
• Insert a line before every match: sed '/^\[section\]/i\# generated by config tool' file.conf

sed -n '/^Error/,/^$/p' build.log

Common mistakes

• Forgetting the g flag and being surprised that only the first match per line changed.
• Using sed -i on macOS without an argument and getting sed: 1: "file.txt": extra characters at the end of s command — BSD sed needs sed -i '' '...' or sed -i.bak '...'.
• Not escaping / inside a pattern or replacement, e.g. matching a path — switch the delimiter instead: sed 's#/usr/local#/opt#'.
• Assuming sed edits the file directly by default; without -i it only writes the transformed output to stdout, leaving the original file untouched.

Tips

• Change the delimiter (s#pattern#replacement#) when your pattern or replacement contains slashes, to avoid a wall of backslash escapes.
• Use & in the replacement to refer to the whole matched text: sed 's/error/[&]/' wraps every match in brackets.
• Combine -n with p to use sed as a precise line-range printer instead of head -n X | tail -n Y.

Best practices

• Always test a sed substitution without -i first (or use -i.bak) so you can review the diff before overwriting a file you can't easily recover.
• Prefer -E/-r for anything beyond a trivial substitution — basic regex escaping (\(, \|, \+) gets unreadable fast.
• In portable scripts targeting both GNU and BSD sed, avoid -i entirely where possible; write to a temp file and mv it into place instead.

Real-world use cases

• Bulk-replacing a hostname or port across many config files during a migration: sed -i 's/old-host/new-host/g' *.conf.
• Stripping comments and blank lines from a config before validating it: sed '/^\s*#/d;/^\s*$/d' nginx.conf.
• Extracting a block of log output between two markers for a bug report: sed -n '/BEGIN TRACE/,/END TRACE/p' app.log.

Common interview questions

• What's the difference between sed 's/a/b/' and sed 's/a/b/g'? Without g, only the first match per line is replaced; g replaces every match on the line.
• How do you use sed to print a range of lines from a file? sed -n '10,20p' file — -n suppresses default printing, and 10,20p prints only lines 10 through 20.
• Why does sed -i behave differently on macOS than on Linux? macOS ships BSD sed, which requires an explicit (possibly empty) backup-suffix argument after -i, while GNU sed on Linux treats the suffix as optional and directly attached to -i.