unexpected behavior of pathlib with empty string

[sarnold]

Bug report

Bug description:

While os.path.exists("") correctly returns False, the same input with Path("").exists() returns True.

>>> from pathlib import Path
>>> import os
>>> Path("").exists()
True
>>> os.path.exists("")
False
>>>

CPython versions tested on:

3.13

Operating systems tested on:

Linux

Linked PRs

• gh-140971: Doc: Clarify pathlib.Path("") behavior #140993

[inventshah]

This happens because __fspath__ defaults to the current directory when the constructed path is the empty string (to support Path() being the current directory). There is a test for this behavior:

cpython/Lib/test/test_pathlib/test_pathlib.py

Lines 266 to 269 in 947bb46

	def test_empty_path(self):
	# The empty path points to '.'
	p = self.cls('')
	self.assertEqual(str(p), '.')

The docs say

When pathsegments is empty, the current directory is assumed

I wonder if Path("") counts as empty, and this should be treated as a docs issue? A simple patch also breaks this test

cpython/Lib/test/test_dbm_sqlite3.py

Lines 256 to 259 in 947bb46

	def test_misuse_empty_filename(self):
	for flag in "r", "w", "c", "n":
	with self.assertRaises(dbm_sqlite3.error):
	db = dbm_sqlite3.open("", flag="c")

So maybe others depend on this behavior?

[LiuQhahah]

Hi
Please assign the task to me
Thanks

[ZeroIntensity]

So maybe others depend on this behavior?

This is almost certainly the case, especially if we're explicitly testing for this. I think we should document it instead.

[sarnold]

👀 ok, that seems a teensy bit weird, but thanks for the doc clarity; I tend to use that repeatedly