How can I duplicate a line and search and replace on the duplicate with sed?

Question

I'm trying to duplicate lines in a text file that contain certain special characters, but in the duplicate, substitute the special character with "regular" ASCII characters. The concrete use-case are accented characters.

Input:

éva
test
frédéric

Desired output:

éva
eva
test
frédéric
frederic

For now I can duplicate the lines containing the é character, but I'm not sure how to search and replace in the capture group.

Here is what I've got so far:

echo 'éva\ntest\nfrédéric' | sed 's/\(.*é.*\)/&\n\1/'

Can I do that with sed? If not, I'll be glad to work with awk...

Answer 1

You can match on é and then apply multiple commands:

sed '/é/{p;s/é/e/g;}'

For any line containing é, this prints the current pattern space, then replaces all és with e (and prints the pattern space again).

The AWK equivalent is

awk '/é/{print; gsub(/é/, "e")}1'

sed’s s command can re-use the address pattern:

sed '/é/{p;s//e/g;}'

and if your replacements are all single-character replacements, the y command is more efficient:

sed '/é/{p;y/é/e/;}'

Answer 2

$ awk '1; gsub(/é/,"e")' file
éva
eva
test
frédéric
frederic

The above uses:

1. The idiomatic true condition 1 to cause awk to perform the default action of printing the current line, then:
2. gsub() to replace any és with es and if that found/replaced any és then it's positive return used in the condition context again causes awk to perform the default action of printing the current (now modified) line.

Note that by using the return code from gsub() to tell us if any és were found it saves us from redundantly having to specify the same regexp /é/ twice in the command.

Answer 3

Yet another sed option - inspired by @EdMorton's awk answer

sed -n 'p;s/é/e/gp' file

Answer 4

Another option, similar to @Stephen Kitt's:

$ sed '/é/p;s/é/e/g'
éva
eva
test
frédéric
frederic

---

• /é/p select the lines that have an é character and print.
• s/é/e/g print the previous lines with the substitution.