Replace a string pattern by a sequence using sed

Question

I want to replace a string in a text file with a sequence of numbers such as 1 2 3 4 5 6 7 preferably using sed.

Here is what I have tried (without the search string) and seems to work:

for i in {1..5};do echo $(seq 1 $i); done

The output:

1
1 2
1 2 3
1 2 3 4
1 2 3 4 5

However, when I try this:

for i in {1..5};do sed -i "s/abc/$(seq 1 $i)/g" test; done

I get the following error:

sed: -e expression #1, char 8: unterminated `s' command
sed: -e expression #1, char 8: unterminated `s' command
sed: -e expression #1, char 8: unterminated `s' command
sed: -e expression #1, char 8: unterminated `s' command

I have tried enclosing the replacement string in brackets as well, but it does not seem to work.

The first loop

for i in {1..5}

will be replaced by a glob for filenames within which I want to replace some parameters.

Answer 1

TL; DR:

There can be some ways to work around that, I think the cleanest for your sample case is:

for i in {1..5};do sed -i "s/abc/$(seq -s ' ' 1 $i)/g" test; done

That is: we request seq to use one single space to separate each number.

Explanation:

The reason why you got those errors is because the output from:

seq 1 5

is not:

1 2 3 4 5

but rather:

1
2
3
4
5

And since you enclosed the $(seq 1 $i) Command Substitution within double-quotes, the usual Word Splitting is not performed, hence all the newlines (except the last) output by the seq utility are left as is.

Therefore the actual sed command has become:

sed -i "s/abc/1
2
3
4
5
/g" test

which is not accepted because, as far as sed is concerned, you provided it with a s/abc/1 command, i.e. lacking the required final /.

An echo $(seq 1 5) does print 1 2 3 4 5 because the $(seq 1 5) is not within double-quotes, therefore Word Splitting occurs as per shell's default behavior.

We might indeed fix your sample case also as in:

for i in {1..5};do sed -i "s/abc/$(echo $(seq 1 $i))/g" test; done

that is, by doing a Command Substitution (your original seq 1 $i) inside another Command Substitution (the echo one).

This works even within the double-quotes because the $(echo ...) Command Substitution is run in a subshell which is not in the "I-am-inside-a-double-quote" state, so Word Splitting on the output from seq will be performed there.

Answer 2

You can pipe the output to a temporary file, then move the temporary file over the original one. Also, you may have meant to operate on one line at a time, in which case you should drop the g:

for i in {1..5}; do
    sed -e "s/abc/$(seq 1 $i)/" test > test.tmp
    mv test.tmp test
done