Part III · Shell and commands

28.4 sed

alsosedreplace textsearch and replacefind and replacesubstitute

TL;DR

sed edits text line by line as it flows through. Its most used command is s/old/new/.
Without g, only the first match on each line changes; s/old/new/g changes them all.
-i edits the file in place. GNU sed (Linux) takes -i alone; macOS needs -i ''.

sed 's/http:/https:/g' links.txt
sed -i.bak 's/DEBUG=1/DEBUG=0/' .env    # keeps a copy in .env.bak
sed -n '10,20p' big.log                 # print lines 10 to 20

Pitfalls

Run the command without -i first and read the output: -i has no undo, unless you give it a suffix like -i.bak.
When the text contains /, use another separator: s|/usr/local|/opt|.
In the replacement, & means "what matched": write \& for a literal ampersand.
As with grep, + and | need -E to have their usual meaning.
sed -i on a symbolic link replaces the link with a regular file, unless you add --follow-symlinks.

Deep dive 2 questions

How does sed work?

It reads one line, runs the script on it, prints the result, then moves to the next line. -n turns off the automatic printing, so only p prints.

A command can be limited to some lines with an address in front: /^#/d deletes comment lines, 5,10s/a/b/ only replaces on lines 5 to 10, and /start/,/end/ covers everything between two matches.

How do I reuse part of the match?

With -E, parentheses capture, and \1, \2 bring the captures back in the replacement. This swaps "user@host" into "host: user":

sed -E 's/([a-z]+)@([a-z.]+)/\2: \1/' contacts.txt

Commands: sed
See also: 28.2 grep and regex
28.3 awk