Quantifiers in a regular expression used with awk behave unexpected

Question

I want to process this list: (Of course this is just an excerpt.)

    1   S3 -> PC-8-Set
    2   S3 -> PC-850-Set
    3   S3 -> ANSI-Set
    4   S3 -> 7-Bit-NRC
    5   PC-8-Set   -> S3
    6   PC-850-Set -> S3
    7   ANSI-Set   -> S3

This is what I did:

awk -F '[[:blank:]]+' '{printf ("%s ", $2)}' list

This is what I got:

1 2 3 4 5 6 7

Now I thought the quantifier + is equivalent to {1,}, but when I changed the line to

awk -F '[[:blank:]]{1,}' '{printf ("%s ", $2)}' list

I got just blanks and the whole line was read to $1.

Can someone explain this behaviour please? I'm thankful for every answer!

Answer 1

Try

awk --re-interval -F '[[:blank:]]{1,}' '{printf ("%s ", $2)}' list

--re-interval

Allow interval expressions (see Regexp Operators) in regexps. This is now gawk's default behavior. Nevertheless, this option remains both for backward compatibility, and for use in combination with the --traditional option.

Answer 2

You are using a Gawk which is from before this November 2010 commit, found by git bisect.

http://git.savannah.gnu.org/cgit/gawk.git/commit/?id=40b3741f63c19e38077d57f4ce4737916ec5073e

The change indeed hinges on the defaulting behavior with respect to intervals, which become on by default (as POSIX requires them to be).

It looks like the --re-interval option becomes relegated only for use with --traditional; i.e. that if --traditional is enabled, then support for {m,n} goes away, but can be selectively brought back with --re-interval.

In your version, {m,n} is unrecognized by default, with or without --traditional. This is true up to this commit:

commit 00ef0423acd97cb964a2bae54c93a03a8ab50e5e
Author: Arnold D. Robbins <arnold@******>
Date:   Fri Jul 16 14:55:10 2010 +0300

    Move to 3.1.8.

and you're behind that still, on 3.1.5.