Weird result with as_uri() on s3 files about universal_pathlib HOT 4 CLOSED

Hi @mberlinger3,

thank you for opening an issue! It seems as_uri is broken in general for UPath.
An easy way would be to implement it as:

class UPath(...):
    def as_uri(self):
        return str(self)

would you be interested in adding tests and a patch in a PR?

Cheers,
Andreas 😃

from universal_pathlib.

Hi @mberlinger3,

thank you for opening an issue! It seems as_uri is broken in general for UPath. An easy way would be to implement it as:

class UPath(...):
    def as_uri(self):
        return str(self)

would you be interested in adding tests and a patch in a PR?

Cheers, Andreas smiley

I think this would make more sense on the non-local (cloud?) implementations because it looks like the as_uri() method is currently inherited from the standard pathlib implementation which is correct for local files. Perhaps a specific wrapper in the UPath class to explicitly call the pathlib implementation and separate methods for the differing schemes.

from universal_pathlib.

Local paths that aren't provided via a file uri are currently returned as their corresponding pathlib.PosixPath and pathlib.WindowsPath instances. (which is going to change in the near future, see: #90)

As it stands right now, as_uri seems to be broken in general for all UPath subclasses:

>>> import upath
>>> upath.UPath("file://a/b/c.dat").as_uri()
'file://file%3A//a/b/c.dat'

So a general fix in upath.UPath.as_uri seems appropriate. If we add a test to upath/tests/cases.py we can ensure that we do the right thing for the implemented subclasses and address potential issues there with custom implementations.

from universal_pathlib.

I see what is issue is here. If you instantiate UPath using a uri as above it breaks; however, if you use a file path then as_uri works as expected.

>>> from upath import UPath
>>> print( UPath("/a/b/c.dat").as_uri() )
file:///a/b/c.dat

This could likely be solved by a simple conditional to check if a scheme exists and pass it to pathlib if it doesn't. We could probably implement the scheme check as a simple property.

An alternative solution would be something like what s3path does. They have a method S3Path().from_uri(uri) that takes in a uri while the base init assumes a path syntax.

from universal_pathlib.