eyepopslikeamosquito has asked for the wisdom of the Perl Monks concerning the following question:

I'd like to be able to mimic the way the Unix shell tokenizes a command line into what a Perl program sees as @ARGV. This little test program hopefully makes clear what I'd like to achieve: 

# Mimic the way the Unix shell breaks up cmdline into @ARGV.
# Before running this script, create a separate test script, g.pl, con
+sisting of:
#   for my $a (@ARGV) { print "a=:$a:\n" }

use strict;
use warnings;

### $qq matches a single or double quoted string.
my $qq = qr/'[^'\\]*(?:\\.[^'\\]*)*'|"[^"\\]*(?:\\.[^"\\]*)*"/;

my $cmd = qq{perl g.pl arg1 'arg 2' "arg 3" "arg \\"4" arg5};
print "cmd=:$cmd:\n\n";
my @args = $cmd =~ /$qq|\S+/g;
for my $t (@args) {
    if (substr($t, 0, 1) eq "'" || substr($t, 0, 1) eq '"') {
        $t = substr($t, 1, -1);
        $t =~ s/\\(['"])/$1/g;
    }
}
for my $a (@args) { print "a=:$a:\n" }
print "\n";
system($cmd);
print "\n";
system { $^X } @args;

[download]
Running this program shows that @args above matches what is seen as @ARGV in g.pl. However, this feels quite tricky and I've got that old re-inventing the wheel feeling. Is there a better way to do this?

Replies are listed 'Best First'.
Re: Mimicking Unix shell quoting in Perl
by Anonymous Monk on Aug 20, 2007 at 08:04 UTC
[reply]
      Looking at Shell::Parser leaves me empty inside. It looks like a very lightweight wrapper around lexical analysis. It doesn't appear to do any glob syntax expansion like eyepopslikeamosquito wanted.

      Not only does bash globbing syntax handle foo*.jpg but it deals with foo?.jpg and foo[1-5][xyz].jpg and even foo.{jpg,jpeg} options too. It's a pretty complicated feature, even if you skip all the $VAR interpolation stuff.

      Speaking of which, eyepopslikeamosquito, is it important for your purposes to do things exactly like the shell? Can you assume you're on a platform that actually has the shell, or must you work smoothly on Win/DOS too? You could just shell out a string to xargs, but this is a significant taint-checking exercise too.

      --
      [ e d @ h a l l e y . c c ]

[reply]
[d/l]
[select]
        The Windows shell doesn't do glob. So even a module that doesn't handle globbing can still be much like a "real shell". It just depends on what you use as a role model. :)
[reply]
Re: Mimicking Unix shell quoting in Perl
by brx (Pilgrim) on Aug 20, 2007 at 15:39 UTC
    Text::Balanced can probably help you 
    use strict;
use warnings;
use Text::Balanced qw(extract_delimited);
my $cmd = qq{perl g.pl arg1 'arg 2' "arg 3" "arg \\"4" arg5};
my @tokens;
while(length($cmd)){
    my $ext = extract_delimited($cmd);
    if (not defined $ext) {
        $cmd=~s/^\s*(\S+)\s*//;
        $ext=$1;
    } else {
        ####no delimiters
        # $ext=substr($ext,1,-1);
    }
    push @tokens,$ext;
}
print join "\n",@tokens;

    [download]
[reply]
[d/l]
Re: Mimicking Unix shell quoting in Perl
by djerius (Beadle) on Aug 20, 2007 at 14:48 UTC
[reply]

Back to Seekers of Perl Wisdom