正则表达式在sed中不匹配

Question

I have used regexs for years, and never come across this problem. On example websites (like http://regexone.com/lesson/1 ) where I can play around with what I'm trying to do, it matches, but in the unix shell using sed , it doesn't match. I discovered this when trying to write logcheck skipping rules.

$ echo 'Hello, world!' | sed '/^\w\w\wlo, wor.*$/d'
$

Works, but

$ echo 'Hello, world!' | sed '/^\w{3}lo, wor.*$/d'
Hello, world!

doesn't. It doesn't see 3 alphanums with the {3} it seems.

I found this out by trying to do reductions on

$ echo "Jul 15 11:31:08 gateway-laptop dbus[3076]: [system] Successfully activated service 'org.freedesktop.PackageKit'"|sed "/^\w{3} [ :0-9]{11} [._[:alnum:]-]+ dbus\[[0-9]+\]: \[system\].*/d"
Jul 15 11:31:08 gateway-laptop dbus[3076]: [system] Successfully activated service 'org.freedesktop.PackageKit'

Which I would have thought should match. Reducing this complexity, this doesn't match

$ echo "Jul 15 11:31:08 gateway-laptop dbus[3076]: [system] Successfully activated service 'org.freedesktop.PackageKit'"|sed "/^\w{3}.*"/d
Jul 15 11:31:08 gateway-laptop dbus[3076]: [system] Successfully activated service 'org.freedesktop.PackageKit'

It looks like its behavior ^\\w{3} should match the first 3 alphanumeric characters in the line, then the .* should match the rest of the line to the EOL.

Answer 1

Escape the braces otherwise sed's default behavior is to match {} or () literally:

echo 'Hello, world!' | gsed '/^\w\{3\}lo, wor.*$/d'

Or else you need -r flag for extended regex capabilities:

echo 'Hello, world!' | gsed -r '/^\w{3}lo, wor.*$/d'