Behaviour of os.path.join

[Mats Wichmann]

On 5/26/20 10:57 AM, BlindAnagram wrote:
> On 26/05/2020 16:59, Mats Wichmann wrote:
>> On 5/26/20 8:56 AM, BlindAnagram wrote:
>>> I came across an issue that I am wondering whether I should report as an
>>> issue.  If I have a directory, say:
>>>
>>>   base='C:\\Documents'
>>>
>>> and I use os.path.join() as follows:
>>>
>>>   join(base, '..\\..\\', 'build', '')
>>>
>>> I obtain as expected from the documentation:
>>>
>>> 'C:\\Documents\\..\\..\\build\\'
>>>
>>> But if I try to make the directory myself (as I tried first):
>>>
>>>   join(base, '..\\..\\', 'build', '\\')
>>>
>>> I obtain:
>>>
>>> 'C:\\'
>>>
>>> The documentation says that an absolute path in the parameter list for
>>> join will discard all previous parameters but '\\' is not an absoute path!
>>
>> But it is - an absolute path is one that starts with the pathname separator.
> 
> On Windows, when part of a string representing a file or directory path,
> '\\' acts as a directory separator.  It may imply, as you say, that it
> is a path on the current drive but it also behaves as a move to a
> sub-directory of what has appeared to its left.  And this is how it is
> treated in other os.path  functions. For example:
> 
> base = "C:\\Documents\\Finance\\"
> abspath(base)
> 'C:\\Documents\\Finance'
> 
> where it can be seen that '\\' is being treated as a directory
> separator.  In my view it is not sensible to have some functions
> treating '\\' in one way and others differently.  I would hence expect
> join() to accept '\\'  as an addition to a file path on Windows and not
> a signal of an absolute path.

Unlike the string join method, which would behave as you seem to want,
the os.path.join method "knows" it is working on paths, so it's going to
assign some meaning to the content of the pieces.   If it was up to me I
might not have chosen the approach Python did of "if we see a piece that
looks like an absolute path, discard what came before", but that's the
one that was chosen, presumably for good reasons, and it is so
documented. And the pathlib library module chose to retain that
behavior: "When several absolute paths are given, the last is taken as
an anchor (mimicking os.path.join()’s behaviour)". So it's just
something to get used to?