Perl Cookbook

Globbing, or Getting a List of Filenames Matching a Pattern

Problem

You want to get a list of filenames similar to MS-DOS’s *.* and Unix’s *.h (this is called globbing).

Solution

Perl provides globbing with the semantics of the Unix C shell through the glob keyword and < >:

@list = <*.c>;
@list = glob("*.c");

You can also use readdir to extract the filenames manually:

opendir(DIR, $path);
@files = grep { /\.c$/ } readdir(DIR);
closedir(DIR);

The CPAN module File::KGlob does globbing without length limits:

use File::KGlob;

@files = glob("*.c");

Discussion

Perl’s built-in glob and <WILDCARD> notation (not to be confused with <FILEHANDLE>) currently use an external program to get the list of filenames on most platforms. This program is csh on Unix,^[16] and a program called dosglob.exe on Windows. On VMS and the Macintosh, file globs are done internally without an external program. Globs are supposed to give C shell semantics on non-Unix systems to encourage portability. The use of the shell on Unix also makes this inappropriate for setuid scripts.

To get around this, you can either roll your own selection mechanism using the built-in opendir or CPAN’s File::KGlob, neither of which uses external programs. File::KGlob provides Unix shell-like globbing semantics, whereas opendir lets you select files with Perl’s regular expressions.

At its simplest, an opendir solution uses grep to filter the list returned by readdir:

@files = grep { /\.[ch]$/i } readdir(DH);

You could also do this with the DirHandle ...