From: Rob Biedenharn <Rob@...>
Date: 2009-02-23T13:36:36+09:00
Subject: Re: FileUtils.mv behaving not as documented

On Feb 22, 2009, at 9:59 PM, Alex DeCaria wrote:
> Rob Biedenharn wrote:
>> On Feb 22, 2009, at 8:06 PM, Alex DeCaria wrote:
>>
>>> The documentation for FileUtils.mv(src, dest) says that if src and
>>> dest
>>> are on different disk partitions then it is supposed to act as a  
>>> copy
>>> instead.  But when I use it to copy a file from the "C:" drive to a
>>> flash drive or mp3 player using another drive letter such as "E:" it
>>> moves it and doesn't copy it as advertised.  Is this a bug, or is  
>>> the
>>> documentation incorrect?  I am using Ruby 186-26 on Windows.
>>>
>>> -Alex
>>
>> I think that the difference the documentation is attempting to point
>> out is that a mv on a single partition just changes names and  
>> pointers
>> in directory entries so it makes no difference the size of the file.
>> On different partitions, the size of the file will matter because the
>> effect is a copy followed by an unlink of the original file.
>>
>> If you want to copy, call FileUtils.cp
>> If you want to move, continue to call FileUtils.mv
>>
>> -Rob
>>
>> Rob Biedenharn    http://agileconsultingllc.com
>> Rob@AgileConsultingLLC.com
>
> Thanks.  Can you explain another curious bit of behavior I've noticed?
> I found that if I use the FileUtil.cp(src, dest) to copy a large file
> from say the C: to the E: drive, it takes much longer than if I first
> use FileUtil.cp(src, tmp) to copy the file to a temporary directory on
> the C: drive and then use FileUtil.mv(tmp, dest) to move the file from
> the temporary directory to the E: drive.  This has the same result as
> just doing FileUtil.cp(src, dest), but takes much less time.  If I'm
> doing this on a lot of files the time saving is significant, but I  
> don't
> understand why.
>
> In other words, if I do
>
> FileUtil.cp("C:/somefile", "C:/Temp/somefile")
> FileUtil.mv("C:/Temp/somefile", "E:/")
>
> it is faster than if I do
>
> FileUtil.cp("C:/somefile", "E:/")
>
> -Alex


I'm making an educated guess (but it is nonetheless speculation) that  
the underlying operating-system-level code that implements mv is using  
a better buffer or at least tighter code which the implementation of  
cp uses some higher-level Ruby code (or a different/smaller buffer).   
It's also possible that Windows does something quite different between  
cp and mv.

If you're truly interested, dig into the code and find out the  
differences yourself.  (I don't work on the Windows platform so I have  
no direct access [or interest].)

-Rob

Rob Biedenharn		http://agileconsultingllc.com
Rob@AgileConsultingLLC.com