Sed extract 2nd word

I was wondering why it didn't work when I do:

echo "d_suites/k_val/tests/asm/logs/kf_on_stage1 FAILED 0:00:22 Jul 22 22:33 " | 
    sed 's/[ \t]*\([^ \t]+\)[ \t]+\([^ \t]+\).*/\2/'

but this one (change + to * ) works:

echo "d_suites/k_val/tests/asm/logs/kf_on_stage1 FAILED 0:00开发者_开发百科:22 Jul 22 22:33 " |
     sed 's/[ \t]*\([^ \t]*\)[ \t]*\([^ \t]*\).*/\2/'

Any help will be appreciated.

---

Sed does not support the + wildcard by default.

$ echo "aaabbbccc" | sed "s/a+/XXX/g"
aaabbbccc

You can enable it with the -r flag (on GNU sed) or -E flag (on Mac OS X and, I suspect, *BSD sed) because these options enable the use of extended regular expressions (in opposition to basic regular expressions):

$ echo "aaabbbccc" | sed -E "s/a+/XXX/g"
XXXbbbccc

If you use GNU sed, it supports the + as a repeater in the basic regex mode if you escape it with a backslash:

$ echo "aaabbbccc" | sed "s/a\+/XXX/g"

---

The first one doesn't work because you need to escape your +'s, like this:

echo "d_suites/k_val/tests/asm/logs/kf_on_stage1 FAILED 0:00:22 Jul 22 22:33 " | 
    sed 's/[ \t]*\([^ \t]\+\)[ \t]\+\([^ \t]\+\).*/\2/'

Edit

For more info about why, read this very informative comment.

---

If you're not completely married to sed, awk would be more readable:

echo "..." | awk '{print $2}'

---

Edited - now referring to [^ \t]

[^ \t]+ is 1-n tabs, requiring there to be a non-tab.
[^ \t]* is 0-n tabs, not requiring there to be a non-tab.

You don't have a non-tab in your input string there