How to find the target of a Unix symlink?

[kj]

How does one find the target(s) of a Unix symlink?

I guess one klugey way would be to pick through the output of
"/bin/ls -al":

sub get_target {
my ($link, $seen) = @_;
$seen ||= {};
return $link if $seen->{$link}; # circularity
return $link unless -e $link;
my $ls = (`/bin/ls -al $link`)[0]; # ick!
return $link unless $ls =~ /\s${link} ->\s+(.*?)\s*$/;
$seen->{$link} = 1;
get_target($1, $seen);
}

Is there a more civilized way to do this?

Thanks!

kj

 

[Todd]

How does one find the target(s) of a Unix symlink?

I guess one klugey way would be to pick through the output of
"/bin/ls -al":

sub get_target {
my ($link, $seen) = @_;
$seen ||= {};
return $link if $seen->{$link}; # circularity
return $link unless -e $link;
my $ls = (`/bin/ls -al $link`)[0]; # ick!
return $link unless $ls =~ /\s${link} ->\s+(.*?)\s*$/;
$seen->{$link} = 1;
get_target($1, $seen);
}

Is there a more civilized way to do this?

Thanks!

kj

perldoc -f readlink

my $target = readlink $link_name;

Todd

 

[kj]

How does one find the target(s) of a Unix symlink?

I guess one klugey way would be to pick through the output of
"/bin/ls -al":

sub get_target {
my ($link, $seen) = @_;
$seen ||= {};
return $link if $seen->{$link}; # circularity
return $link unless -e $link;
my $ls = (`/bin/ls -al $link`)[0]; # ick!
return $link unless $ls =~ /\s${link} ->\s+(.*?)\s*$/;
$seen->{$link} = 1;
get_target($1, $seen);
}

Well, the above is clearly a disaster...

Is there a more civilized way to do this?

I found readlink, but my gripe with it is that it doesn't follow
links beyond the first hop. (I.e. if "first" points to "second",
and "second" points to "third", readlink "first" returns "second"
not "third.) stat, on the other hand, does a nice job of finding
the ultimate target of a sequence of links, but one ends up with
a device and an inode, not a filename for the target.

Is there a way to get the filename (or filenames) associated with
a dev+inode combination?

Thanks in advance,

kj

 

[John W. Krahn]

How does one find the target(s) of a Unix symlink?

perldoc -f readlink


John

 

[Darren Dunham]

I found readlink, but my gripe with it is that it doesn't follow
links beyond the first hop. (I.e. if "first" points to "second",
and "second" points to "third", readlink "first" returns "second"
not "third.) stat, on the other hand, does a nice job of finding
the ultimate target of a sequence of links, but one ends up with
a device and an inode, not a filename for the target.

Is there a way to get the filename (or filenames) associated with
a dev+inode combination?

No, not easily. The System rarely (never?) needs to do this, so there
is no index for it.

You would have to 'find' on the filesystem (ususally expensive), trying
to match the inode, and must realize that there may be more than one
filename for it.

 

[Ben Morrow]

I found readlink, but my gripe with it is that it doesn't follow
links beyond the first hop. (I.e. if "first" points to "second",
and "second" points to "third", readlink "first" returns "second"
not "third.)

This is the only way to get all the information. A simple way to follow
all links:

sub readalllinks {
my $file = shift;
while (-l $file) {
$file = readlink $file;
}
}

stat, on the other hand, does a nice job of finding
the ultimate target of a sequence of links, but one ends up with
a device and an inode, not a filename for the target.

Is there a way to get the filename (or filenames) associated with
a dev+inode combination?

No, except in special cases.

Ben

 

[kj]

This is the only way to get all the information. A simple way to follow
all links:

sub readalllinks {
my $file = shift;
while (-l $file) {
$file = readlink $file;
}
}

Hmm. I think this fails if $file's target is outside $file's
directory and is specified using a relative path. I think this
does "the right thing":

use Cwd;
use File::Basename;
sub readalllinks {
my $cwd = my $dir = cwd . '/';
my $file = shift;
my $path = ($file =~ m,^/,) ? $file : "${dir}${file}";
$dir = (fileparse($path))[1];

while (-l $path) {
$file = readlink $path;
$path = ($file =~ m,^/,) ? $file : "${dir}${file}";
$dir = (fileparse($path))[1];
}
(my $basename, $dir) = fileparse $path;
chdir $dir;
$path = cwd . "/$basename";
chdir $cwd;
$path;
}

....but it is eggly.

kj

 

[chris-usenet]

You would have to 'find' on the filesystem (ususally expensive), trying
to match the inode, and must realize that there may be more than one
filename for it.

Or zero matches (the file might have been unlinked but still in use).

Chris

 

[Peter Scott]

How does one find the target(s) of a Unix symlink? [...]
Is there a more civilized way to do this?

File::Spec::Link

 

[Ben Morrow]

Hmm. I think this fails if $file's target is outside $file's
directory and is specified using a relative path.

Yes, of course it does. Whoops

I think this
does "the right thing":

...but it is eggly.

Having seen your use of Cwd, I remembered that this will do just fine:

use Cwd qw/realpath/;

my $file = realpath($file);



But a cleaner rewrite would be

use Cwd qw/cwd/;
use File::Spec::Functions qw/rel2abs catpath splitpath/;

sub readalllinks {
my $rel = shift;
my $dir = cwd;
my $abs;

while ( -l ($abs = rel2abs $rel, $dir) ){
$dir = catpath +(splitpath $abs)[0,1];
$rel = readlink $abs;
}

return $abs;
}

 

[Joe Smith]

Is there a way to get the filename (or filenames) associated with
a dev+inode combination?

To get an idea as to how difficult that is, read this:
ftp://lsof.itap.purdue.edu/pub/tools/unix/lsof/FAQ

-Joe