«

env: python 2.7.3

6 test files' name in a directory as below:
12ab Abc Eab a1bc acd bc

the following is test code:
for root, dirs, files in os.walk(os.getcwd()):
print files

the output in win32 platform is:
['12ab', 'a1bc', 'Abc', 'acd', 'bc', 'Eab']

but in linux is：
['Eab', 'acd', 'a1bc', '12ab', 'bc', 'Abc' ]

they are so different. a bug?

On Fri, Mar 1, 2013 at 12:43 AM, Honghe Wu <> wrote:
> env: python 2.7.3
>
> 6 test files' name in a directory as below:
> 12ab Abc Eab a1bc acd bc
>
> the following is test code:
> for root, dirs, files in os.walk(os.getcwd()):
> print files
>
> the output in win32 platform is:
> ['12ab', 'a1bc', 'Abc', 'acd', 'bc', 'Eab']
>
> but in linux is：
> ['Eab', 'acd', 'a1bc', '12ab', 'bc', 'Abc' ]
>
> they are so different. a bug?

Nope. When os.walk() fetches a listing of the contents of a directory,
it internally uses os.listdir() (or a moral equivalent thereof). The
docs for os.listdir() state that "The [returned] list is in arbitrary
order.". The order is dependent on the OS and filesystem, and likely
also more obscure factors (e.g. the order in which the files were
created). The lack of any required ordering allows for improved I/O
performance in many/most cases.

Cheers,
Chris

On Fri, Mar 1, 2013 at 12:43 AM, Honghe Wu <> wrote:
> env: python 2.7.3
>
> 6 test files' name in a directory as below:
> 12ab Abc Eab a1bc acd bc
>
> the following is test code:
> for root, dirs, files in os.walk(os.getcwd()):
> print files
>
> the output in win32 platform is:
> ['12ab', 'a1bc', 'Abc', 'acd', 'bc', 'Eab']
>
> but in linux is：
> ['Eab', 'acd', 'a1bc', '12ab', 'bc', 'Abc' ]
>
> they are so different. a bug?
> --

The function doesn't specify a particular order, just that it will
hand you a list of files. It grabs those from the underlying file
system. It looks like Windows sorts it alphabetically and Linux just
does whatever (maybe sorted by creation time?). I don't think it's a
bug. If the order matters to you, sort it yourself.

Thanks! Cause I need sorted returnd list, and the arbitrary list makes the
other procedure go wrong. Maybe the I/O speed is more important in other
cases.
On Mar 1, 2013 4:55 PM, "Chris Rebert" <> wrote:

> On Fri, Mar 1, 2013 at 12:43 AM, Honghe Wu <> wrote:
> > env: python 2.7.3
> >
> > 6 test files' name in a directory as below:
> > 12ab Abc Eab a1bc acd bc
> >
> > the following is test code:
> > for root, dirs, files in os.walk(os.getcwd()):
> > print files
> >
> > the output in win32 platform is:
> > ['12ab', 'a1bc', 'Abc', 'acd', 'bc', 'Eab']
> >
> > but in linux is：
> > ['Eab', 'acd', 'a1bc', '12ab', 'bc', 'Abc' ]
> >
> > they are so different. a bug?
>
> Nope. When os.walk() fetches a listing of the contents of a directory,
> it internally uses os.listdir() (or a moral equivalent thereof). The
> docs for os.listdir() state that "The [returned] list is in arbitrary
> order.". The order is dependent on the OS and filesystem, and likely
> also more obscure factors (e.g. the order in which the files were
> created). The lack of any required ordering allows for improved I/O
> performance in many/most cases.
>
> Cheers,
> Chris
>

On Fri, 01 Mar 2013 17:24:05 +0800, Honghe Wu wrote:

> Thanks! Cause I need sorted returnd list, and the arbitrary list makes the
> other procedure go wrong. Maybe the I/O speed is more important in other
> cases.

You can sort the lists of files and subdirectories with e.g.:

for root, dirs, files in os.walk(os.getcwd()):
dirs[:] = sorted(dirs)
files = sorted(files)
...

Note that modifying the directory list in-place will affect which
subdirectories are traversed and in what order.