David Golden > Path-Class-Rule-0.015 > Path::Class::Rule



Annotate this POD



Open  0
View/Report Bugs
Module Version: 0.015   Source   Latest Release: Path-Class-Rule-0.018


Path::Class::Rule - File finder using Path::Class


version 0.015


  use Path::Class::Rule;

  my $rule = Path::Class::Rule->new; # match anything
  $rule->file->size(">10k");         # add/chain rules

  # iterator interface
  my $next = $rule->iter( @dirs );
  while ( my $file = $next->() ) {

  # list interface
  for my $file ( $rule->all( @dirs ) ) {


This module iterates over files and directories to identify ones matching a user-defined set of rules. The API is based heavily on File::Find::Rule, but with more explicit distinction between matching rules and options that influence how directories are searched. A Path::Class::Rule object is a collection of rules (match criteria) with methods to add additional criteria. Options that control directory traversal are given as arguments to the method that generates an iterator.

Here is a summary of features for comparison to other file finding modules:




  my $rule = Path::Class::Rule->new;

Creates a new rule object that matches any file or directory. It takes no arguments. For convenience, it may also be called on an object, in which case it still returns a new object that matches any file or directory.


  my $common      = Path::Class::Rule->new->file->not_empty;
  my $big_files   = $common->clone->size(">1MB");
  my $small_files = $common->clone->size("<10K");

Creates a copy of a rule object. Useful for customizing different rule objects against a common base.

Matching and iteration


  my $next = $rule->iter( @dirs, \%options);
  while ( my $file = $next->() ) {

Creates a subroutine reference iterator that returns a single Path::Class object when dereferenced. This iterator is "lazy" -- results are not pre-computed.

It takes as arguments a list of directories to search and an optional hash reference of control options. If no search directories are provided, the current directory is used ("."). Valid options include:

Filesystem loops might exist from either hard or soft links. The loop_safe option prevents infinite loops, but adds some overhead by making stat calls. Because directories are visited only once when loop_safe is true, matches could come from a symlinked directory before the real directory depending on the search order. To get only the real files, turn off follow_symlinks. Turning loop_safe off and leaving follow_symlinks on avoids stat calls and will be fastest, but with the risk of an infinite loop and repeated files. The default is slow, but safe.

The error_handler parameter must be a subroutine reference. It will be called when a rule test throws an exception. The first argument will be the Path::Class object being inspected and the second argument will be the exception.

The Path::Class objects inspected and returned will be relative to the search directories provided. If these are absolute, then the objects returned will have absolute paths. If these are relative, then the objects returned will have relative paths.


  my @matches = $rule->all( @dir, \%options );

Returns a list of Path::Class objects that match the rule. It takes the same arguments and has the same behaviors as the iter method. The all method uses iter internally to fetch all results.


  if ( $rule->test( $path ) ) { ... }

Test a file path against a rule. Used internally, but provided should someone want to create their own, custom iteration algorithm.

Logic operations

Path::Class::Rule provides three logic operations for adding rules to the object. Rules may be either a subroutine reference with specific semantics (described below in "EXTENDING") or another Path::Class::Rule object.


  $rule->and( sub { -r -w -x $_ } ); # stacked filetest example
  $rule->and( @more_rules );

Adds one or more constraints to the current rule. E.g. "old rule AND new1 AND new2 AND ...". Returns the object to allow method chaining.


    sub { -r -w -x $_ },

Takes one or more alternatives and adds them as a constraint to the current rule. E.g. "old rule AND ( new1 OR new2 OR ... )". Returns the object to allow method chaining.


  $rule->not( sub { -r -w -x $_ } );

Takes one or more alternatives and adds them as a negative constraint to the current rule. E.g. "old rule AND NOT ( new1 AND new2 AND ...)". Returns the object to allow method chaining.



Takes one or more alternatives and will prune a directory if any of the criteria match. For files, it is equivalent to $rule->not($rule->or(@rules)). Returns the object to allow method chaining.

This method should be called as early as possible in the rule chain. See "skip_dirs" below for further explanation and an example.


Rule methods are helpers that add constraints. Internally, they generate a closure to accomplish the desired logic and add it to the rule object with the and method. Rule methods return the object to allow for method chaining.

File name rules


  $rule->name( "foo.txt" );
  $rule->name( qr/foo/, "bar.*");

The name method takes one or more patterns and creates a rule that is true if any of the patterns match the basename of the file or directory path. Patterns may be regular expressions or glob expressions (or literal names).


  $rule->iname( "foo.txt" );
  $rule->iname( qr/foo/, "bar.*");

The iname method is just like the name method, but matches case-insensitively.


  $rule->skip_dirs( @patterns );

The skip_dirs method skips directories that match one or more patterns. Patterns may be regular expressions or globs (just like name). Directories that match will not be returned from the iterator and will be excluded from further search.

Note: this rule should be specified early so that it has a chance to operate before a logical shortcut. E.g.

  $rule->skip_dirs(".git")->file; # OK
  $rule->file->skip_dirs(".git"); # Won't work

In the latter case, when a ".git" directory is seen, the file rule shortcuts the rule before the skip_dirs rule has a chance to act.

File test rules

Most of the -X style filetest are available as boolean rules. The table below maps the filetest to its corresponding method name.

   Test | Method               Test |  Method
  ------|-------------        ------|----------------
    -r  |  readable             -R  |  r_readable
    -w  |  writeable            -W  |  r_writeable
    -w  |  writable             -W  |  r_writable
    -x  |  executable           -X  |  r_executable
    -o  |  owned                -O  |  r_owned
        |                           |
    -e  |  exists               -f  |  file
    -z  |  empty                -d  |  directory, dir
    -s  |  nonempty             -l  |  symlink
        |                       -p  |  fifo
    -u  |  setuid               -S  |  socket
    -g  |  setgid               -b  |  block
    -k  |  sticky               -c  |  character
        |                       -t  |  tty
    -T  |  ascii
    -B  |  binary

For example:

  $rule->file->nonempty; # -f -s $file

The -X operators for timestamps take a single argument in a form that Number::Compare can interpret.

   Test | Method
    -A  |  accessed
    -M  |  modified
    -C  |  changed

For example:

  $rule->modified(">1"); # -M $file > 1

Stat test rules

All of the stat elements have a method that takes a single argument in a form understood by Number::Compare.

  stat()  |  Method
       0  |  dev
       1  |  ino
       2  |  mode
       3  |  nlink
       4  |  uid
       5  |  gid
       6  |  rdev
       7  |  size
       8  |  atime
       9  |  mtime
      10  |  ctime
      11  |  blksize
      12  |  blocks

For example:


Depth rules


The min_depth and max_depth rule methods take a single argument and limit the paths returned to a minimum or maximum depth (respectively) from the starting search directory.

Perl file rules

  # All perl rules

  # Individual perl file rules
  $rule->perl_module;     # .pm files
  $rule->perl_pod;        # .pod files 
  $rule->perl_test;       # .t files
  $rule->perl_installer;  # Makefile.PL or Build.PL
  $rule->perl_script;     # .pl or 'perl' in the shebang

These rule methods match file names (or a shebang line) that are typical of Perl distribution files.

Version control file rules

  # Skip all known VCS files

  # Skip individual VCS files

Skips files and/or prunes directories related to a version control system. Just like skip_dirs, these rules should be specified early to get the correct behavior.

Other rules



The dangling rule method matches dangling symlinks. Use it or its inverse to control how dangling symlinks should be treated. Note that a dangling symlink will be returned by the iterator as a Path::Class::File object.



The shebang rule takes a list of regular expressions or glob patterns and checks them against the first line of a file.

Negated rules

Most rule methods have a negated form preceded by "not_".


Because this happens automatically, it includes somewhat silly ones like not_nonempty (which is thus a less efficient way of saying empty).

Rules that skip directories or version control files do not have a negated version.


Custom rule subroutines

Rules are implemented as (usually anonymous) subroutines callbacks that return a value indicating whether or not the rule matches. These callbacks are called with two arguments. The first argument is a Path::Class object, which is also locally aliased as the $_ global variable for convenience in simple tests.

  $rule->and( sub { -r -w -x $_ } ); # tests $_

The second argument is a hash reference that can be used to maintain state. Keys beginning with an underscore are reserved for Path::Class::Rule to provide additional data about the search in progress. For example, the _depth key is used to support minimum and maximum depth checks.

The custom rule subroutine must return one of three values:

The 0 but true value will shortcut logic (it is treated as "true" for an "or" rule and "false" for an "and" rule). For a directory, it ensures that the directory will not be returned from the iterator and that its children will not be evaluated either. It has no effect on files -- it is equivalent to returning a false value.

For example, this is equivalent to the "max_depth" rule method with a depth of 3:

    sub {
      my ($path, $stash) = @_;
      return $stash->{_depth} <= 3 ? 1 : "0 but true";

Files of depth 4 will not be returned by the iterator; directories of depth 4 will not be returned and will not be searched.

Generally, if you want to do directory pruning, you are encouraged to use the "skip" method instead of writing your own logic using 0 but true.

Extension modules and custom rule methods

One of the strengths of File::Find::Rule is the many CPAN modules that extend it. Path::Class::Rule provides the add_helper method to provide a similar mechanism for extensions.

The add_helper class method takes three arguments, a name for the rule method, a closure-generating callback, and a flag for not generating a negated form of the rule. Unless the flag is true, an inverted "not_*" method is generated automatically. Extension classes should call this as a class method to install new rule methods. For example, this adds a "foo" method that checks if the filename is "foo":

  package Path::Class::Rule::Foo;

  use Path::Class::Rule;

    foo => sub {
      my @args = @_; # do this to customize closure with arguments
      return sub {
        my ($item) = shift;
        return if $item->is_dir;
        return $item->basename =~ /^foo$/;


This allows the following rule methods:


The add_helper method will warn and ignore a helper with the same name as an existing method.


If you run with lexical warnings enabled, Path::Class::Rule will issue warnings in certain circumstances (such as a read-only directory that must be skipped). To disable these categories, put the following statement at the correct scope:

  no warnings 'Path::Class::Rule';


This is an early release for community feedback and contribution. Some features are still unimplemented:

Filetest operators and stat rules are subject to the usual portability considerations. See perlport for details.

Performance suffers somewhat from all of the abstraction layers of Path::Class and File::Spec. Hopefully, convenience makes up for that.


There are many other file finding modules out there. They all have various features/deficiencies, depending on your preferences and needs. Here is an (incomplete) list of alternatives, with some comparison commentary.

File::Find is part of the Perl core. It requires the user to write a callback function to process each node of the search. Callbacks must use global variables to determine the current node. It only supports depth-first search (both pre- and post-order). It supports pre- and post-processing callbacks; the former is required for sorting files to process in a directory. File::Find::Closures can be used to help create a callback for File::Find.

File::Find::Rule is an object-oriented wrapper around File::Find. It provides a number of helper functions and there are many more File::Find::Rule::* modules on CPAN with additional helpers. It provides an iterator interface, but precomputes all the results.

File::Next provides iterators for file, directories or "everything". It takes two callbacks, one to match files and one to decide which directories to descend. It does not allow control over breadth/depth order, though it does provide means to sort files for processing within a directory. Like File::Find, it requires callbacks to use global variables.

Path::Class::Iterator walks a directory structure with an iterator. It is implemented as Path::Class subclasses, which adds a degree of extra complexity. It takes a single callback to define "interesting" paths to return. The callback gets a Path::Class::Iterator::File or Path::Class::Iterator::Dir object for evaluation.

File::Find::Object and companion File::Find::Object::Rule are like File::Find and File::Find::Rule, but without File::Find inside. They use an iterator that does not precompute results. They can return File::Find::Object::Result objects, which give a subset of the utility of Path::Class objects. File::Find::Object::Rule appears to be a literal translation of File::Find::Rule, including oddities like making -M into a boolean.

File::chdir::WalkDir recursively descends a tree, calling a callback on each file. No iterator. Supports exclusion patterns. Depth-first post-order by default, but offers pre-order option. Does not process symlinks.

File::Find::Iterator is based on iterator patterns in Higher Order Perl. It allows a filtering callback. Symlinks are followed automatically without infinite loop protection. No control over order. It offers a "state file" option for resuming interrupted work.

File::Find::Declare has declarative helper rules, no iterator, is Moose-based and offers no control over ordering or following symlinks.

File::Find::Node has no iterator, does matching via callback and offers no control over ordering.

File::Set builds up a set of files to operate on from a list of directories to include or exclude, with control over recursion. A callback is applied to each file (or directory) in the set. There is no iterator. There is no control over ordering. Symlinks are not followed. It has several extra features for checksumming the set and creating tarballs with /bin/tar.


Bugs / Feature Requests

Please report any bugs or feature requests through the issue tracker at https://rt.cpan.org/Public/Dist/Display.html?Name=Path-Class-Rule. You will be notified automatically of any progress on your issue.

Source Code

This is open source software. The code repository is available for public review and contribution under the terms of the license.


  git clone git://github.com/dagolden/path-class-rule.git


David Golden <dagolden@cpan.org>


This software is Copyright (c) 2011 by David Golden.

This is free software, licensed under:

  The Apache License, Version 2.0, January 2004
syntax highlighting: