Dealing with backreference in a sed command inside a shell script

Question

I'm trying to make work a sed command inside of a shell script that takes variables while interacting with the user, but I can not make it work because of the backreference I use in it.... some help...?

Script:

echo "Origin language?\n"
read ol
echo "\nTarget language?\n"
read tl
echo "\ntmx file name?\n"
read file

then at the sed command I have tried these ways:

1.

sed -r "s/lang=\"$ol\">(.*)$/\1/g" "$file" > ${file}_${ol}.txt
sed -r "s/lang=\"$tl\">(.*)$/\1/g" "$file" > ${file}_${tl}.txt

2.

thisol = "<tuv xml:lang=\"$ol\">(.*)$"
thistl = "<tuv xml:lang=\"$tl\">(.*)$"

sed -r "s/$thisol/\1/g" "$file" > ${file}_${ol}.txt
sed -r "s/$thistg/\1/g" "$file" > ${file}_${tl}.txt

3.

thisol = "<tuv xml:lang=\"$ol\">(.*)$"
thistl = "<tuv xml:lang=\"$tl\">(.*)$"
that = "\\\1"

sed -r "s/$thisol/$that/g" "$file" > ${file}_${ol}.txt
sed -r "s/$thistg/$that/g" "$file" > ${file}_${tl}.txt

You need to escape parentheses of the backreference if you use double quotes. Such as `$.*$`. — tvm, Oct 16 '17 at 18:38
That was one of the things I usuccessfuly tried, I tried again though, but no success again.... — Andrés Chandía, Oct 16 '17 at 18:42
Learn what quotes do: `sed -E 's/lang='"$ol"'>(.*)$/\1/g' "$file" > "${file}_${ol}.txt"`. — Ed Morton, Oct 16 '17 at 23:44

score 0 · Accepted Answer · answered Oct 17 '17 at 15:44

Based on the comment made by @EdMorton I could find the answer to this question, I had to modify a little bit the suggestion made by him, but at the end I realized that this was not what I needed, I should have done it with grep, any way here is the correct code to accomplish what is implied on the question, and the final grep command I used (which I learned from here)

sed command:

sed -r '/<tuv xml:lang='\"$ol\"'>/{N;/<seg>/{N;s/<tuv xml:lang='\"$ol\"'>\n\s+<seg>(.*)<\/seg>/\1/}}' "$file.tmx" > "${file}_${ol}.txt"

where:

/<tuv xml:lang='\"$ol\"'>/ the RegEx from the starting line

{N;/<seg>/ the RegEx from the finishing line

{N;s/<tuv xml:lang='\"$ol\"'>\n\s+<seg>(.*)<\/seg> the searched complete RegEx

/\1/ the backreference to the (.*) element

Grep command:

grep -Pzo "<tuv xml:lang=\"$ol\">\n\s+<seg>(.*)<\/seg>" "$file.tmx" > "${file}_${ol}.txt"

Dealing with backreference in a sed command inside a shell script

1 Answers1