Refactor nt._path_is* & nt._path_[l]exists

[nineteendo]

Feature or enhancement

Proposal:

Quoting @eryksun:

I agree that all of these _path_is* and _path_[l]exists helpers are a lot of code to maintain, taken together. It could be refactored into smaller inline helper functions that can be reused, which would also make the code more readable.

Has this already been discussed elsewhere?

This is a minor feature, which does not need previous discussion elsewhere

Links to previous discussion of this feature:

• gh-117841: Add C implementation of ntpath.lexists #117842

Linked PRs

• gh-118507 : Refactor nt._path_is* #118755
• [3.13] gh-118507 : Refactor nt._path_is* to improve applicability for other cases (GH-118755) #119360
• [3.12] gh-118507 : Fix ntpath.isfile for pipes. #119381
• [3.12] gh-118507 : Speed up ntpath.isjunction() & ntpath.lexists() with native helpers (GH-118755) #119420
• gh-118507: Amend news entry #120817
• [3.13] gh-118507: Amend news entry to mention ntpath.isfile bugfix (GH-120817) #121365

[nineteendo]

Eryk, do you have any ideas to clean them up (besides the simpler error handling)?

[nineteendo]

Good start, we probably need a follow_symlinks option for _testFileTypeByName to use LSTAT() though.

[nineteendo]

Did you speed something up, because it's only 3 lines less than the old code.

[eryksun]

My goal was to remove duplicated code and divide the work into two separate operations that can be understood and modified independently, written in a way that I think is easy to understand. This reduces the maintenance burden. I also added internal support for checking for mount points, in case we implement _path_isjunction().

I added a GetFileType() check in the by-handle code. When combined with the diskOnly parameter, the GetFileType() check makes it safer to check an open file descriptor since the check is implemented directly in the I/O manager without trying to acquire the file lock. If GetFileInformationByHandleEx() is called on a handle for a synchronous-mode pipe, it could block indefinitely. For example, with the current implementation in 3.12:

>>> pr, pw = os.pipe()
>>> threading.Thread(target=os.read, args=(pr, 1)).start()
>>> os.path.isfile(pr)
^C

Also, given the system's named-pipe filesystem (NPFS) support for basic file information (anonymous pipes are also in NPFS, but they're like unlinked open files in POSIX), the GetFileType() check avoids classifying pipes as regular files. Again, with the current implementation in 3.12:

>>> pr, pw = os.pipe()
>>> os.path.isfile(pr)
True

[nineteendo]

OK, that explains why the code isn't much shorter. I added a test for the second issue. The first one seems difficult to test.