Alex Rivera | Logout

Sed expression doesn't allow optional grouped string

Asked 2011-05-27T18:26:16.713
77

I'm trying to use the following regex in a sed script but it doesn't work:

sed -n '/\(www\.\)\?teste/p'

The regex above doesn't seem to work. sed doesn't seem to apply the ? to the grouped www\..

It works if you use the -E parameter that switches sed to use the Extended Regex, so the syntax becomes:

sed -En '/(www\.)?teste/p'

This works fine but I want to run this script on a machine that doesn't support the -E operator. I'm pretty sure that this is possible and I'm doing something very stupid.

regex sed
Edit
Report

1 Answer

133

Standard sed only understands POSIX Basic Regular Expressions (BRE), not Extended Regular Expressions (ERE), and the ? is a metacharacter in EREs, but not in BREs.

Your version of sed might support EREs if you turn them on. With GNU sed, the relevant options are -r and --regexp-extended, described as "use extended regular expressions in the script".

However, if your sed does not support it - quite plausible - then you are stuck. Either import a version of sed that does support them, or redesign your processing. Maybe you should use awk instead.

2014-02-21

I don't know why I didn't mention that even though sed does not support the shorthand ? or \? notation, it does support counted ranges with \{n,m\}, so you can simulate ? with \{0,1\}:

sed -n '/\(www\.\)\{0,1\}teste/p' << EOF
http://www.tested.com/
http://tested.com/
http://www.teased.com/
EOF

which produces:

http://www.tested.com/
http://tested.com/

Tested on Mac OS X 10.9.1 Mavericks with the standard BSD sed and with GNU sed 4.2.2.

answered 2011-05-27T21:09:11.623

Your Answer