How can I use sed to replace multiple characters?

Question

I have this script:

replace 3

echo 12345 | sed "s/^\(.\){"$1"}/\1hi/"

I also tried this:

echo 12345 | sed "s/^\(.{"$1"}\)/\1hi/"

In this situation I want the script to add 'hi' after the first 3 characters of '12345' (123hi45). This is a script so the '3' is an argument and can change. I'm really stuck here. Thanks in advance!

Solution

Your command would have worked if you had escaped the curly braces (I've also quoted the $1 that you leave unquoted for unknown reasons):

$ set -- 3
$ echo 12345 | sed "s/^\(.\{$1\}\)/\1hi/"
123hi45

The repetition modifier {n} is an extended regular expression modifier, which in a basic regular expression is written as \{n\}. The sed utility is using basic regular expressions by default.

You would save a few characters by rewriting it as

echo 12345 | sed "s/^.\{$1\}/&hi/"

Personally I would have taken another approach...

---

You want to add the string hi after the 3rd character in 12345 where "the 3rd" is given by the value in $1.

echo 12345 | sed 's/./&hi/'"$1"

When $1 is 3, then the sed expression would look like

s/./&hi/3

This would replace the 3rd match of . (any character) with that same character (this is what & does in the replacement) followed by hi.

Putting a digit, n, at the end of an s command in sed like this makes sed substitute the n:th match of the pattern.

Test running (with a modified input and replacement for readability reasons):

$ set -- 3
$ echo abcde | sed 's/./&<hi>/'"$1"
abc<hi>de

$ set -- 4
$ echo abcde | sed 's/./&<hi>/'"$1"
abcd<hi>e

$ set -- 1
$ echo abcde | sed 's/./&<hi>/'"$1"
a<hi>bcde

Comments

[Stéphane Chazelas]

\{n,m\} was added to BRE before {n,m} was added to ERE. As it broke backward compatibility for ERE, there are some egrep or awk implementations that still don't support it.