How to use regex as field separator in awk?

Question

I'm trying to use regex as a field seperator in awk. From my reading this seems possible but I can't get the syntax right.

rpm -qa | awk '{ 'FS == [0-9]' ; print $1 }'
awk: cmd. line:1: { FS
awk: cmd. line:1:     ^ unexpected newline or end of string

Thoughts? The goal if not obviouse is to get a list of software without version number.

Kusalananda · Accepted Answer · 2018-12-21T09:45:16.787

30

You have mucked up your quotes and syntax. To set the input field separator, the easiest way to do it is with the -F option on the command line:

awk -F '[0-9]' '{ print $1 }'

or

awk -F '[[:digit:]]' '{ print $1 }'

This would use any digit as the input field separator, and then output the first field from each line.

The [0-9] and [[:digit:]] expressions are not quite the same, depending on your locale. See "Difference between [0-9], [[:digit:]] and \d".

One could also set FS in the awk program itself. This is usually done in a BEGIN block as it's a one-time initialisation:

awk 'BEGIN { FS = "[0-9]" } { print $1 }'

Note that single quotes can't be used in a single-quoted string in the shell, and that awk strings always use double quotes.

edited Dec 21 '18 at 09:45

answered Oct 07 '11 at 18:04

Kusalananda

Is it possible to access the FS and see the string it matched? – Roland Apr 06 '20 at 11:07
For a more natural separator, I think `FS="( | )+"` works (the first one is a tab, and 2nd one is a space) – Sridhar Sarnobat Jul 31 '22 at 18:29
@SridharSarnobat That's not relevant to the current question (which deals with `FS` set to something that matches digits) but would be more naturally written as `FS='[[:blank:]]+'`, which is very close to what `awk` would use if you don't set `FS` at all. – Kusalananda Jul 31 '22 at 18:32
Thanks, that's better. I can't think of a case when I'd use a number as a field separator but maybe I'm not creative enough :) – Sridhar Sarnobat Jul 31 '22 at 18:33
@SridharSarnobat I don't really know what this user's data looks like. – Kusalananda Jul 31 '22 at 18:35

score 15 · Answer 2 · edited Aug 31 '19 at 19:25

15

+1 for Kusalananda's answer. Alternately, the FS variable can be set in the BEGIN block:

awk 'BEGIN {FS="[0-9]"} {print $1}'

Changing FS in a action block won't take effect until the next line is read

$ printf "%s\n" "abc123 def456" "ghi789 jkl0" | awk '{FS="[0-9]"; print $1}'
abc123
ghi

The other errors in the question:

edited Aug 31 '19 at 19:25

Kusalananda

answered Oct 07 '11 at 18:20

glenn jackman

1

"Changing FS in a action block won't take effect until the next line is read" I've been looking all over for that info. – Samizdis Jun 26 '17 at 15:50
1

plus: can't use single quotes for string value in awk, even if you pass them from shell correctly – dave_thompson_085 Jun 27 '18 at 09:24
Is it possible to access the FS and see the string it matched? – Roland Apr 06 '20 at 11:07
`FS` is a variable so you do anything like any other variable (e.g `print FS`). To get the parts that match FS, with POSIX awk, I think you cannot. With GNU awk you could write `n = split($0, fields, FS, separators)` where `fields` and `separators` are arrays. – glenn jackman Apr 06 '20 at 12:55

2 Answers2