URL path with "/../"

[kkaiser]

yarl==1.2.6 converts a URL like https://www.example.com/../example.jpg to https://www.example.com/example.jpg just fine.

In yarl==1.3.0 that was changed and now throws a ValueError instead.

>>> from yarl import URL
>>> URL('https://www.example.com/../example.jpg')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/kkaiser/.pyenv/versions/ocean_ai/lib/python3.7/site-packages/yarl/__init__.py", line 174, in __new__
    cls._validate_authority_uri_abs_path(host=host, path=path)
  File "/Users/kkaiser/.pyenv/versions/ocean_ai/lib/python3.7/site-packages/yarl/__init__.py", line 623, in _validate_authority_uri_abs_path
    "Path in a URL with authority " "should start with a slash ('/') if set"
ValueError: Path in a URL with authority should start with a slash ('/') if set

Should that be the expected behaviour? It was already closed once: #84

[aio-libs-bot]

GitMate.io thinks possibly related issues are #138 (Question: how could I join to URL list path parts?), #185 (Allow joining URL and pathlib.Path), #84 (Incorrect handling of '..' in url path), #33 (Parse URL parameter), and #242 (Handle path argument of URL.build, which doesn't start from /).

[Soniyanayak51]

Hi, is this issue still open?

[asvetlov]

I think https://www.example.com/../example.jpg is an invalid path; there is no parent folder in /../example.jpg path context.

[kkaiser]

It is invalid but unfortunately appears on some websites that is why all browsers seem to convert it. Shouldn't yarl do the same?

[asvetlov]

yarl shouldn't be as forgiving as a human-oriented web browser.
Consider it as dividing by zero.