Steps for file::finder
## See File::Finder for normal use of steps
## subclassing example:
BEGIN {
package My::File::Finder;
use base File::Finder;
sub _steps_class { "My::File::Finder::Steps" }
}
BEGIN {
package My::File::Finder::Steps;
use base File::Finder::Steps;
sub bigger_than { # true if bigger than N bytes
my $self = shift;
my $bytes = shift;
return sub {
-s > $bytes;
}
}
}
my $over_1k = My::File::Finder->bigger_than(1024);
print "Temp files over 1k:\n";
$over_1k->ls->in("/tmp");
\*(C`File::Finder::Steps\*(C' provide the predicates being tested for \*(C`File::Finder\*(C'.
These methods are called on a class or instance to add a \*(L"step\*(R". Each step adds itself to a list of steps, returning the new object. This allows you to chain steps together to form a formula.
As in find, the default operator is \*(L"and\*(R", and short-circuiting is performed.
Like find's \*(C`or\*(C'.
Like a left parenthesis. Used in nesting pairs with \*(C`right\*(C'.
Like a right parenthesis. Used in nesting pairs with \*(C`left\*(C'. For example: my $big_or_old = File::Finder ->type('f') ->left ->size("+100")->or->mtime("+90") ->right; find($big_or_old->ls, "/tmp"); You need parens because the \*(L"or\*(R" operator is lower precedence than the implied \*(L"and\*(R", for the same reason you need them here: find /tmp -type f '(' -size +100 -o -mtime +90 ')' -print Without the parens, the -type would bind to -size, and not to the choice of -size or -mtime. Mismatched parens will not be found until the formula is used, causing a fatal error.
Alias for \*(C`left\*(C'.
Alias for \*(C`right\*(C'.
Like find's \*(C`!\*(C'. Prefix operator, can be placed in front of individual terms or open parens. Can be nested, but what's the point? # list all non-files in /tmp File::Finder->not->type('f')->ls->in("/tmp");
Always returns true. Useful when a subexpression might fail, but you don't want the overall code to fail: ... ->left-> ...[might return false]... ->or->true->right-> ... Of course, this is the find command's idiom of: find .... '(' .... -o -true ')' ...
Always returns false.
Like \s-1GNU\s0 find's \*(L",\*(R". The result of the expression (or subexpression if in parens) up to this point is discarded, and execution continues afresh. Useful when a part of the expression is needed for its side effects, but shouldn't affect the rest of the \*(L"and\*(R"-ed chain. # list all files and dirs, but don't descend into CVS dir contents: File::Finder->type('d')->name('CVS')->prune->comma->ls->in('.');
Enables symlink following, and returns true.
True if basename matches \s-1NAME\s0, which can be given as a glob pattern or a regular expression object: my $pm_files = File::Finder->name('*.pm')->in('.'); my $pm_files_too = File::Finder->name(qr/pm$/)->in('.');
Like find's \*(C`-perm\*(C'. Leading \*(L"-\*(R" means \*(L"all of these bits\*(R". Leading \*(L"+\*(R" means \*(L"any of these bits\*(R". Value is de-octalized if a leading 0 is present, which is likely only if it's being passed as a string. my $files = File::Finder->type('f'); # find files that are exactly mode 644 my $files_644 = $files->perm(0644); # find files that are at least world executable: my $files_world_exec = $files->perm("-1"); # find files that have some executable bit set: my $files_exec = $files->perm("+0111");
Like find's \*(C`-type\*(C'. All native Perl types are supported. Note that \*(C`s\*(C' is a socket, mapping to Perl's \*(C`-S\*(C', to be consistent with find. Returns true or false, as appropriate.
Prints the fullname to \*(C`STDOUT\*(C', followed by a newline. Returns true.
Prints the fullname to \*(C`STDOUT\*(C', followed by a \s-1NUL\s0. Returns true.
Not implemented yet.
True if the owner is \s-1USERNAME\s0 or \s-1UID\s0.
True if the group is \s-1GROUPNAME\s0 or \s-1GID\s0.
True if the entry doesn't belong to any known user.
True if the entry doesn't belong to any known group.
Like find's \*(C`-links N\*(C'. Leading plus means \*(L"more than\*(R", minus means \*(L"less than\*(R".
True if the inode number meets the qualification.
True if the file size meets the qualification. By default, N is in half-K blocks. Append a trailing \*(L"k\*(R" to the number to indicate 1K blocks, or \*(L"c\*(R" to indicate characters (bytes).
True if access time (in days) meets the qualification.
True if modification time (in days) meets the qualification.
True if inode change time (in days) meets the qualification.
Forks the child process via \*(C`system()\*(C'. Any appearance of \*(C`{}\*(C' in any argument is replaced by the current filename. Returns true if the child exit status is 0. The list is passed directly to \*(C`system\*(C', so if it's a single arg, it can contain \*(C`/bin/sh\*(C' syntax. Otherwise, it's a pre-parsed command that must be found on the \s-1PATH\s0. Note that I couldn't figure out how to horse around with the current directory very well, so I'm using $_ here instead of the more traditional \*(C`File::Find::name\*(C'. It still works, because we're still chdir'ed down into the directory, but it looks weird on a trace. Trigger \*(C`no_chdir\*(C' in \*(C`find\*(C' if you want a traditional find full path. my $f = File::Finder->exec('ls', '-ldg', '{}'); find({ no_chdir => 1, wanted => $f }, @starting_dirs); Yeah, it'd be trivial for me to add a no_chdir method. Soon.
Like \*(C`exec\*(C', but displays the command line first, and waits for a response. If the response begins with \*(C`y\*(C' or \*(C`Y\*(C', runs the command. If the command fails, or the response wasn't yes, returns false, otherwise true.
Sets $File::Find::prune, and returns true.
Not yet implemented.
Not yet implemented.
Ah yes, the master escape, with extra benefits. Give it a coderef, and it evaluates that code at the proper time. The return value is noted for true/false and used accordingly. my $blaster = File::Finder->atime("+30")->eval(sub { unlink }); But wait, there's more. If the parameter is an object that responds to \*(C`as_wanted\*(C', that method is automatically called, hoping for a coderef return. This neat feature allows subroutines to be created and nested: my $old = File::Finder->atime("+30"); my $big = File::Finder->size("+100"); my $old_or_big = File::Finder->eval($old)->or->eval($big); my $killer = File::Finder->eval(sub { unlink }); my $kill_old_or_big = File::Finder->eval($old_or_big)->ls->eval($killer); $kill_old_or_big->in('/tmp'); Almost too cool for words.
Like find's \*(C`-depth\*(C'. Sets a flag for \*(C`as_options\*(C', and returns true.
Like find's \*(C`-ls\*(C'. Performs a \*(C`ls -dils\*(C' on the entry to \*(C`STDOUT\*(C' (without forking), and returns true.
Not yet implemented.
Not yet implemented.
Incorporate a \*(C`File::Find::Rule\*(C' object as a step. Note that this must be a rule object, and not a result, so don't call or pass \*(C`in\*(C'. For example, using \*(C`File::Find::Rule::ImageSize\*(C' to define a predicate for image files that are bigger than a megapixel in my friends folder, I get: require File::Finder; require File::Find::Rule; require File::Find::Rule::ImageSize; my $ffr = File::Find::Rule->file->image_x('>1000')->image_y('>1000'); my @big_friends = File::Finder->ffr($ffr) ->in("/Users/merlyn/Pictures/Sorted/Friends");
True if the file contains \*(C`pattern\*(C' (either a literal string treated as a regex, or a true regex object). my $plugh_files = File::Finder->type('f')->contains(qr/plugh/); Searching is performed on a line-by-line basis, respecting the current value of $/.
A step consists of a compile-time and a run-time component.
During the creation of a \*(C`File::Finder\*(C' object, step methods are called as if they were methods against the slowly-growing \*(C`File::Finder\*(C' instance, including any additional parameters as in a normal method call. The step is expected to return a coderef (possibly a closure) to be executed at run-time.
When a \*(C`File::Finder\*(C' object is being evaluated as the \*(C`File::Find\*(C' \*(C`wanted\*(C' routine, the collected coderefs are evaluated in sequence, again as method calls against the \*(C`File::Finder\*(C' object. No additional parameters are passed. However, the normal \*(C`wanted\*(C' values are available, such as $_, $File::Find::name, and so on. The \*(C`_\*(C' pseudo-handle has been set properly, so you can safely use \*(C`-X\*(C' filetests and \*(C`stat\*(C' against the pseudo-handle. The routine is expected to return a true/false value, which becomes the value of the step.
Although a \*(C`File::Finder\*(C' object is passed both to the compile-time invocation and the resulting run-time invocation, only the \*(C`options\*(C' self-hash element is properly duplicated through the cloning process. Do not be tempted to add additional self-hash elements without overriding \*(C`File::Finder\*(C''s \*(C`_clone\*(C'. Instead, pass values from the compile-time phase to the run-time phase using closure variables, as shown in the synopsis.
For simplicity, you can also just mix-in your methods to the existing \*(C`File::Finder::Steps\*(C' class, rather than subclassing both classes as shown above. However, this may result in conflicting implementations of a given step name, so beware.
File::Finder
None known yet.
Randal L. Schwartz, <[email protected]>
Copyright (C) 2003,2004 by Randal L. Schwartz, Stonehenge Consulting Services, Inc.
This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself, either Perl version 5.8.2 or, at your option, any later version of Perl 5 you may have available.