Re: Top 10 reasons to start using ack

[diotalevi]

I wrote diotalevi's grep. It's one distinguishing feature is automatically looking inside tarballs and zip files. I find that's really common for me. It does a dirt simple look at the file extension and opens .tgz/.zip/.whatever using the streaming abilities of archive extraction programs. The code isn't sexy and it isn't even good but it does do a pretty good enough job. It'd be possible to do a much better job given a few minutes and the interest. I'd use ack immediately if it got this feature.

sub open_file_harder {
    my ($filename) = @_;
    return if not defined $filename;

    if ( my ($extension) = $filename =~ /(\.[^.]+)\z/mx ) {
        my @readers = (
            [ qr/\.t(?:ar\.)?gz\z/ => qw( gzcat ),    $filename ],
            [ qr/\.zip\z/,         => qw( unzip -p ), $filename ],
            [ qr/\.Z\z/            => qw( zcat ),     $filename ],
            [ qr/\.gz\z/           => qw( gzcat ),    $filename ],
            [ qr/\.bz2\z/          => qw( bzcat ),    $filename ],
        );
        for my $reader (@readers) {
            my ( $pattern, @command ) = @{$reader};
            if ( $extension =~ $pattern ) {
                open3( undef, my $fh, undef, @command );
                return $fh;
            }
        }
    }

    open my $fh, '<', $filename
        or die "Couldn't open $filename: $!";
    return $fh;
}

[petdance]

A couple of thoughts:

* I wouldn't have it on by default. It would only be via the -A,--archive switch, for example.

* How would I show the resultant filename? If the file is in a tarball, what do I show for the filename?

[diotalevi]

I'd be sane to treat these containers like directories. Directories are just another kind of container. Given a tarball baz.tgz in /foo/bar containing the files a and b/c you'd tell the user about matches in the paths /foo/bar/baz.tgz/a and /foo/bar/baz.tgz/b/c. I think Windows "Compressed Folders" work like this.

[petdance]

I could see using "foo.gz" as the filename for a standalone file, using your streaming method.

For tarballs, though, I would have to extract to a temporary directory and delve through that, and that means write privileges that I don't want to assume.