Question: how could I join to URL list path parts?

[kxepal]

Assume I have an URL URL('http://localhost:5984/base') and some list of path segments like ['foo', '/bar', 'baz%']. What's the proper way to join them to url to haveURL(http://localhost:5984/base/foo/%2fbar/baz%25) as result?

So far I came to:

url = URL('http://localhost:5984/base')
segments = ['foo', '/bar', 'baz%']
for seg in sements:
  url /= seg.replace('/', '%2F')

But that doesn't looks like good solution. Or it does?

[asvetlov]

In [1]: from yarl import URL

In [2]: url = URL('http://localhost:5984/base')

In [3]: segments = ['foo', '/bar', 'baz%']

In [4]: url2 = url / '/'.join(segments)

In [5]: url2
Out[5]: URL('http://localhost:5984/base/foo//bar/baz%25')

[samuelcolvin]

foo//bar?

[kxepal]

Out[5]: URL('http://localhost:5984/base/foo//bar/baz%25')

Problem is here: //bar vs /%2Fbar. I didn't found a way so far to avoid quoting on my side /:

[kxepal]

Also url2 = url / '/'.join(segments) will fail if first segment starts with /, which should be quoted in final path.

[asvetlov]

yarl has not with_path_segments method because https://docs.python.org/3/library/pathlib.html has no such method.
I not sure if we need to add it -- just for sake of keeping public API simple.
On other side I have no strong objection.

Opinions?

[samuelcolvin]

I think yarl should match pathlib with respect to double slash //, currently it doesn't:

In [11]: Path('foobar') / 'x//y'
Out[11]: PosixPath('foobar/x/y')

In [12]: URL('http://localhost:5984/base') / 'x//y'
Out[12]: URL('http://localhost:5984/base/x//y')

[kxepal]

Well, a few. First one is a with_path_segments as you already mentioned. Second is about to allow / work with list / tuple on the right side:

>>> url = URL('http://localhost/base')
>>> url / ['foo', '/bar']
URL('http://localhost/base/foo/%2Fbar)

I believe pathlib doesn't have such issue because it doesn't have quote anything at all, so path segment couldn't have / character in any form. For URL's it's possible.

[kxepal]

@samuelcolvin
It's not quite the same: for paths, you can't have a empty name object (dir, file, etc.). For URLs, path is an resource identifier. It have no obligations to follow FS rules, so base/x//y - is quite valid path and it's not equals to base/x/y - these are two different resources.

[samuelcolvin]

makes sense.

[asvetlov]

@kxepal returning to the issue.
Accepting list/tuple as right param for / operator looks very native.
If we decide to implement this -- no need for with_path_segments, with_path can accept just the same list/tuple.

Unfortunately I don't find a time for the issue at least up to New Year.
Do you want to make a PR?

[kxepal]

Sounds good for me. Will do.