Getting a List of Files Via Glob

[Zoogie]

I need to get a list of files that fit a set of filename globs input by the user. The problem is, the files may be in a different directory. For example, if the user enters:

  $area = "public/test/";
  @files = ("*.txt", "_*");

The function should return something like:

  readme.txt, files.txt, _index, _sortopts, ... etc.

I'm trying to use the glob function, but if you specify another directory, that directory name is prepended to every filename. Here's my current function:

@file_list =
    map { /^$area(.*)/; $1 }
    glob(join(' ',map { $area.$_ } @files));

Works as expected, but that seems like a lot more work than necessary. Is there an easier / clearer / more efficient way to do this?

[merlyn]

If you don't mind a chdir, then add it first:

chdir $area or die "chdir $area: $!";
@file_list = glob("@files");

-- Randal L. Schwartz, Perl hacker

[Zoogie]

I considered this, but I'd have to return the original working directory at the sub, since other subs assume that the working dir hasn't changed.

I suppose I could save the original working dir with
$prevdir = `pwd`;
But executing the shell command doesn't seem like it would be much more efficient than mapping the directory name on to each glob and then chopping it off...

[chromatic]

Use Cwd. It's a standard module.

use Cwd;

my $dir = cwd();
chdir $otherdir;
# do something
chdir $dir;

[Corion]

As most of my scripts have to run under Win32, I don't use glob :) (Note: glob() is also available under Win32, but another process must be started that reads the directory in, not really elegant IMO). I use the

  DIR = opendir( "$area" ) or die "Can't read '$area' : $!";
  @files = readdir( DIR ) or die "Error reading from '$area' : $!";
  closedir( DIR ); # Don't care about errors here

combo. If you simply want to process every file in a directory, instead of   @files = readdir(DIR) you could use

  while ($file = readdir(DIR)) {
    ... do your stuff to $file ...
  }


The drawback of readdir() against glob() is of course, that you have to descend through the directories yourself if you want recursive processing, but I like programs more that give the user some feedback (like painting a dot for each directory traversed as appropriate) ...

[merlyn]

In 5.6, glob() was rewritten to use readdir (even on Win32), so your complaint disappears.

-- Randal L. Schwartz, Perl hacker

[t0mas]

I often use readdir in a recursive mode, like :

&makeTree(0,'/where/to/start');

sub makeTree {
  my ($level,$Dir)=@_;

  # Get all directories
  opendir(DIR,$Dir) || die "$!";
  my @Dirs = grep { /^[^.].*/ && -d "$Dir/$_" }
    readdir(DIR);
  closedir(DIR);

  # Read files and do stuff or whatever...

  # Call self
  foreach my $currDir (@Dirs) {
    &makeTree($level+1,join('/',$Dir,$currDir));
  }
};



/brother t0mas

[Corion]

In your code, you don´t need to re-read the directory to get at the files in the directory - although it would be difficult to "compute" the difference of the array elegantly (well, difficult to me at least). I mostly use the following way, intermixing files and directories :

  # Untested code - use at your own risk
  sub handledirectory {
    my ($directory) = @_;
    my ($entry, @direntries);
    opendir( DIR, $directory ) or die "Can´t read $directory : $!\n";
    @direntries = readdir( DIR ) or die "Error reading $directory : $!
+\n";
    closedir DIR;

    foreach $entry, @direntries {
      # File::Spec gives us cross-platform path utilities
      # and comes with every Perl standard distribution
      require File::Spec;
      my $fullpath;
      
      # skip current and parent directory entries
      next if $entry =~ /^\.\.?$/;

      $fullpath = File::Spec->catfile( $directory, $entry );
      if (-d $fullpath ) {
        &handledirectory($fullpath);
      } elsif ( -f $fullpath ) {
        # This second call to stat() (implicit in the "-f")
        # could be done away by using some other short
        # variable that does caching, but that would maybe
        # confuse the readers ...
        ... do stuff ...
      } else {
        # something strange ...
      };
  }

[Zoogie]

Hurm... is there a way readdir() will work with specifications like "*.txt" or "dir??.log"?

[Corion]

readdir() will always read in a complete directory, there is no way around that - and for what I know, there is no function in Perl to hint to the operating system that you are only interested in a certain subset of the files either, as Perl comes from a UNIX background and Unix does not have the notion of wildcards in the file system. Wildcard expansion is always done by the shell under UNIX.

To accomplish what you want done, you do (untested!) the following :

  opendir DIR, $directory or die "Couldn't open $directory : $!\n";
  @files = readdir( DIR ) or die "Couldn't read from $directory : $!\n
+";
  closedir( DIR );

  @files = grep @files, { /\.txt$/ && -f $_ };

(I hope that this test thing in grep() works the way I want it. The RE part checks if the name ends with ".txt", and the -f part checks if the name corresponds to a file (and not a directory)). Another solution for you could be to let the user specify all files (in UNIX-style) on the command line, there even is a module called GlobArgv, which does wildcard expansion for you automagically (under Win32).

[BBQ]

Does this mean that glob will now return . and .. too? (I haven't upgraded yet)

#!/home/bbq/bin/perl
# Trust no1!