Re: Re: filesize module on win32 systems

There are caveats, of course:

NTFS supports transparent compression of files. There is the "actual" file size, and the "compressed" size. Which does -s return?
NTFS supports sparse files. If a large section of a file contains NULL, the filesystem can save space by not allocating room for the NULL data. Here, I would assume -s would return the full size, not the "on disk" size.
NTFS supports "alternate streams". A file can have many alternate streams, taking up data that I assume would never be reported by -s.
Depending on the file size & cluster size, most files probably take up more space "on disk" than the contents of the file would suggest. Which does -s report here?

This is probably not an exhaustive list, I'm sure I'm forggeting something.

Edited by Chady -- closed ul tag.

Comment on Re: Re: filesize module on win32 systems

Replies are listed 'Best First'.
Re^3: filesize module on win32 systems by Aristotle (Chancellor) on Mar 02, 2004 at 10:15 UTC
I'll make an educated guess about compressed files later. Most every modern filesystem supports sparse files, and the call underlying `-s` reports the virtual, not on-disk, size, on most of them. I assume NTFS is no different. Programs have to specifically be "sparse-aware" to deal with such files properly. The size of alternative streams is not reported by the call underlying `-s`. Again, programs need to be "alternate-stream-aware" to deal with this situation correctly. It reports the byte size of the file, not the cluster size. Unless you are writing a lowlevel filesystem manipulation/report tool, you shouldn't concern yourself with this distinction in the first place. You'll notice that the defaults are such that a program naively copying the contents of a file to another with taking the existence of advanced filesystem features into account will still work (and that's on OS level, not Perl `-s` level). In light of that trend I'd suggest that `-s` reports the uncompressed size of NTFS compressed files. And that's probably all the OP needed, too. Makeshifts last the longest.	[reply]
Re: Re: Re: filesize module on win32 systems by BrowserUk (Patriarch) on Mar 02, 2004 at 11:40 UTC
Perl calls the C runtime fstat() to get the information. I the case of MSVC the actual call is _fstati64() which in turn calls GetFileInformationByHandle(). That returns a structure called BY_HANDLE_FILE_INFORMATION, which contains to DWORDS fields that contain the filesize. The filesize as it would be if you read the whole think into memory. Ie. decompressed, de-sparsed etc. To get the actual on-disk size you would need to call GetCompressedFileSize(). Examine what is said, not who speaks. "Efficiency is intelligent laziness." -David Dunham "Think for yourself!" - Abigail	[reply] [d/l]


P is for Practical
	PerlMonks