Repeating a regex pattern

Question

First, I don't know if this is actually possible but what I want to do is repeat a regex pattern. The pattern I'm using is:

sed 's/[^-\t]*\t[^-\t]*\t\([^-\t]*\).*/\1/' films.txt

An input of

250.    7.9    Shutter Island (2010)    110,675

Will return:

Shutter Island (2010)

I'm matching all none tabs, (250.) then tab, then all none tabs (7.9) then tab. Next I backrefrence the film title then matching all remaining chars (110,675).

It works fine, but im learning regex and this looks ugly, the regex [^-\t]*\t is repeated just after itself, is there anyway to repeat this like you can a character like a{2,2}?

I've tried ([^-\t]*\t){2,2} (and variations) but I'm guessing that is trying to match [^-\t]*\t\t?

Also if there is any way to make my above code shorter and cleaner any help would be greatly appreciated.

Answer 1

This works for me:

sed 's/\([^\t]*\t\)\{2\}\([^\t]*\).*/\2/' films.txt

If your sed supports -r you can get rid of most of the escaping:

sed -r 's/([^\t]*\t){2}([^\t]*).*/\2/' films.txt

Change the first 2 to select different fields (0-3).

This will also work:

sed 's/[^\t]\+/\n&/3;s/.*\n//;s/\t.*//' films.txt

Change the 3 to select different fields (1-4).

Answer 2

To use repeating curly brackets and grouping brackets with sed properly, you may have to escape it with backslashes like

sed 's/\([^-\t]*\t\)\{3\}.*/\1/' films.txt

Yes, this command will work properly with your example.

If you feel annoyed to, you can choose to put -r option which enables regex extended mode and forget about backslash escapes on brackets.

sed -r 's/([^-\t]*\t){3}.*/\1/' films.txt

Found that this is almost the same as Dennis Williamson's answer, but I'm leaving it because it's shorter expression to do the same.