in reply to Re^2: File::Find stat question
in thread File::Find stat question

I'm still getting a 0 for directory size

Does your script run on a Windows Perl (Strawberry, ActiveState or the like)?

Windows does not have a stat() system call. Perl emulates it, but not perfectly. I assume - without looking at the perl sources - that the struct stat is simply zeroed out and then filled depending on what the Windows API functions return. For directories, size is simply not touched and stays 0.

Note that FAT and NTFS are very different from filesystems found on Unix systems, so having a "size" field for a directory may not make any sense at all.

Also note that the "size" field is not at all related to the size of the files in the directory. On Unix systems, there is -- depending on the filesystem type -- a relation to the highest number of files in the directory over the lifetime of the directory, and maybe the length of the filenames. If you want the sum of the file sizes, calculate it by calling stat for each file in the directory.

POSIX leaves the meaning of the st_size field undefined for directories. It is defined for regular files, symlinks, shared memory objects, and typed memory objects, but not for directories. ("For other file types, the use of this field is unspecified.") So, depending on your operating system, stat() may legally return 0 or just garbage for the size of a directory.

My result for stat on a directory on Win7 NTFS and FAT, Strawberry Perl 5.14.2 is 0, too.

On Linux 3.10.17, perl 5.18.1, ext3, ext2, procfs, sysfs, tmpfs, the results vary from 0 (procfs, sysfs) to 12288 (heavily used directory on ext3), matching the results from ls. The directory size of 0 on procfs is documented in stat(2), sysfs behaves in the same way.

Alexander

--
Today I will gladly share my knowledge and experience, for there are no sweeter words than "I told you so". ;-)

Replies are listed 'Best First'.
Re^4: File::Find stat question
by Anonymous Monk on Nov 27, 2014 at 22:16 UTC

    Thanks Alexander for your reply, I'm using the following version of Activestate Perl. This is perl 5, version 16, subversion 2 (v5.16.2) Unfortunately I'm limited to what the I.T Dept installs and limited to what modules they have installed. I'll give your idea a go and add up all the file sizes in the directories, I assume this will be slower than a stat on a dir?

[reply]
      I'm using ... Activestate Perl

      So you are very likely using Windows. There are ActiveState Perls for some Unix systems, but as most Unix systems have some Perl out of the box, ActiveState's Perl is rarely found on Unix systems.

      I'll ... add up all the file sizes in the directories, I assume this will be slower than a stat on a dir?

      Did you actually read my first answer? Let me repeat the two important points:

      • The size value returned by stat() for a directory is IN NO WAY RELATED to the size of the files the directory contains.
      • The size value returned by stat() for a directory may legally be COMPLETE NONSENSE (according to POSIX).

      What is your intention?

      • If you want to know the total number of bytes that the files in a directory use, THERE IS NO OTHER WAY than summing up their individual sizes returned by stat(). The size value returned by stat returns nonsense.
      • If you want to know how much disk space is "wasted" by the directory itself, not by the files it contains, the size value returned by stat() for the directory itself MAY BE useful - IF your operating system cooperates (Windows DOES NOT) AND IF the actual filesysem has this information (remember that sysfs and procfs on Linux usually return 0).

      Alexander

      --
      Today I will gladly share my knowledge and experience, for there are no sweeter words than "I told you so". ;-)
[reply]
[d/l]
[select]

        Thanks Alexander

        I'll do this.

        If you want to know the total number of bytes that the files in a directory use, THERE IS NO OTHER WAY than summing up their individual sizes returned by stat(). The size value returned by stat returns nonsense

[reply]

In Section Seekers of Perl Wisdom