Python os.path.join on Windows

to join a windows path, try

mypath=os.path.join('c:\', 'sourcedir')

basically, you will need to escape the slash

The reason os.path.join('C:', 'src') is not working as you expect is because of something in the documentation that you linked to:

Note that on Windows, since there is a
current directory for each drive,
os.path.join(“c:”, “foo”) represents a
path relative to the current directory
on drive C: (c:foo), not c:foo.

As ghostdog said, you probably want mypath=os.path.join('c:\', 'sourcedir')

To be pedantic, it’s probably not good to hardcode either / or as the path separator. Maybe this would be best?

mypath = os.path.join('c:%s' % os.sep, 'sourcedir')

or

mypath = os.path.join('c:' + os.sep, 'sourcedir')

To be even more pedantic, the most python doc consistent answer would be:

mypath = os.path.join('c:', os.sep, 'sourcedir')

Since you also need os.sep for the posix root path:

mypath = os.path.join(os.sep, 'usr', 'lib')

I’d say this is a (windows)python bug.

Why bug?

I think this statement should be True

os.path.join(*os.path.dirname(os.path.abspath(__file__)).split(os.path.sep))==os.path.dirname(os.path.abspath(__file__))

But it is False on windows machines.

Consent with @georg-

I would say then why we need lame os.path.join– better to use str.join or unicode.join e.g.

sys.path.append('{0}'.join(os.path.dirname(__file__).split(os.path.sep)[0:-1]).format(os.path.sep))

You have a few possible approaches to treat path on Windows, from the most hardcoded ones (as using raw string literals or escaping backslashes) to the least ones. Here follows a few examples that will work as expected. Use what better fits your needs.

In[1]: from os.path import join, isdir

In[2]: from os import sep

In[3]: isdir(join("c:", "\", "Users"))
Out[3]: True

In[4]: isdir(join("c:", "/", "Users"))
Out[4]: True

In[5]: isdir(join("c:", sep, "Users"))
Out[5]: True

For a system-agnostic solution that works on both Windows and Linux, no matter what the input path, one could use

def joinpath(rootdir, targetdir):
    return os.path.join(os.sep, rootdir + os.sep, targetdir)

On Windows:

>>> joinpath("C:", "Windows")
'C:\Windows'
>>> joinpath("C:\Windows", "src")
'C:\Windows\src'

On Linux:

>>> joinpath("usr", "lib")
'/usr/lib'
>>> joinpath("/usr/lib", "src")
'/usr/lib/src'

answering to your comment : “the others ‘//’ ‘c:’, ‘c:\’ did not work (C:\ created two backslashes, C: didn’t work at all)”

On windows using
os.path.join('c:', 'sourcedir')
will automatically add two backslashes \ in front of sourcedir.

To resolve the path, as python works on windows also with forward slashes -> ‘/’, simply add .replace('\','/') with os.path.join as below:-

os.path.join('c:\', 'sourcedir').replace('\','/')

e.g: os.path.join('c:\', 'temp').replace('\','/')

output : ‘C:/temp’

The proposed solutions are interesting and offer a good reference, however they are only partially satisfying. It is ok to manually add the separator when you have a single specific case or you know the format of the input string, but there can be cases where you want to do it programmatically on generic inputs.

With a bit of experimenting, I believe the criteria is that the path delimiter is not added if the first segment is a drive letter, meaning a single letter followed by a colon, no matter if it corresponds to a real unit.

For example:

import os
testval = ['c:','c:\','d:','j:','jr:','data:']

for t in testval:
    print ('test value: ',t,', join to "folder"',os.path.join(t,'folder'))

test value:  c: , join to "folder" c:folder
test value:  c: , join to "folder" c:folder
test value:  d: , join to "folder" d:folder
test value:  j: , join to "folder" j:folder
test value:  jr: , join to "folder" jr:folder
test value:  data: , join to "folder" data:folder

A convenient way to test for the criteria and apply a path correction can be to use os.path.splitdrive comparing the first returned element to the test value, like t+os.path.sep if os.path.splitdrive(t)[0]==t else t.

Test:

for t in testval:
    corrected = t+os.path.sep if os.path.splitdrive(t)[0]==t else t
    print ('original: %stcorrected: %s'%(t,corrected),' join corrected->',os.path.join(corrected,'folder'))

original: c:    corrected: c:  join corrected-> c:folder
original: c:   corrected: c:  join corrected-> c:folder
original: d:    corrected: d:  join corrected-> d:folder
original: j:    corrected: j:  join corrected-> j:folder
original: jr:   corrected: jr:  join corrected-> jr:folder
original: data: corrected: data:  join corrected-> data:folder

it can be probably be improved to be more robust for trailing spaces, and I have tested it only on windows, but I hope it gives an idea.
See also Os.path : can you explain this behavior? for interesting details on systems other then windows.

I got around this by using:

os.sep.join(list('C:', 'sourcedir'))

join here is not from os.path.join(), but from ''.join()

It can be useful in the following situation:

import os

some_path = r'C:some_foldersome_file.txt'
path_items = some_path.split(os.sep)
same_path = os.sep.join(path_items)

For simplicity’s sake, a workaround:

my_path = r'D:testtest2file.txt' # original path string
drive_letter = my_path.split(':')[0] # avoid os.path.join relative path bs
my_path_parts = os.path.normpath(my_path.split(':')[1]).split(os.sep)

# do some path cmponent modifications here if you want

if drive_letter: # if drive letter exists
    drive_letter += ':\'

my_new_path = drive_letter + os.path.join(*my_path_parts)
my_new_path

In windows using os.paht.join("/", "Temp") will result in /Temp by default but as strange as it sounds, there is no problem in using that path as a full path equivalent to "C:/Temp" and it works for both saving files and opening files.