How to match the first POSIX `[:cntrl:]` character with older BSD `awk`?

Fravadona 0
Viewed 85

I wanted to match the first ASCII control character of an input string with awk but I stumbled upon what seems to be a bug on older BSDs:

#!/bin/bash

printf 'a \b c\n' |

awk 'match( $0, /[[:cntrl:]]/ ) { print RSTART }'

1

The expected output would be:

3

What could be wrong with the code?

bsd awk
4 Answers

What about using a character class designating control characters in this way with hex numbers:

$ printf $'a \b c\n' | awk 'match( $0, /[\x01-\x1f]/ ) { print RSTART }'
3
userQLI20

I can't test it as I don't have the old version of BSD awk you mentioned but would negating the [:print:] character class work there:

$ printf $'a \b c\n' | awk 'match( $0, /[^[:print:]]/ ) { print RSTART }'
3

I couldn't find a POSIX definition of [:print:] but according to the gawk manual:

[:print:] Printable characters (characters that are not control characters)

so in theory negating it might work.

Ed Morton0

  1. it's nawk that happens to show up in some BSD- descendant distributions, but it has nothing to do with bsd itself

  2. from my personal trial and error attempts, it seems nawk can gsub() a null byte properly into a replacement string, but there's no way to get nawk to print out a null byte, whether it's

-- coming in from /dev/stdin or files,

-- requested via sprintf("%c", 0),

-- or hard-coded into a variable, e.g. myvar = "\0" ;

nawk returns a 1 (true) for this test when all else returns false :

  nawk 'BEGIN { print ("\1"~"\0") }'

  1

nawk in its current form is also very screwy when it comes to invalid data inputs - a simple regex test for existence of a byte results in fatal error :

% nawk 'BEGIN { print length(_="\123\321\456\777") }' 
4

% nawk 'BEGIN { print length(_="\123\321\456\777"), (_~"\321") }'

nawk: multibyte conversion failure at ?...
 source line number 1
 context is
    BEGIN { print length(_="\123\321\456\777"), >>>  (_~"\321") <<<

… a test that gawk in UTF-8 properly handles without any error messages:

LC_ALL='en_US.UTF-8' LANG='en_US.UTF-8' gawk -e '

    BEGIN { _="\123\321\456\777"; print (_~"\321") }' 

1
RARE Kpop Manifesto0

It seems like it is the NUL byte in [:cntrl:] that causes problems; if you put it at the end of the list, it works on BSD and GNU awk (but not with Solaris):

printf 'a \b c\n' |
awk 'match( $0, /[\x01-\x1F\x7F\x00]/ ) { print RSTART }'

3

While it isn't my intent, matching the NUL byte is still impossible, but it'll work on more recent awk. Be careful to not add anything else after \x00 though.

Fravadona0
Related