MultiplexedPath only traverses first directory

[jooste]

I ran into the following issue while trying to update my MultiplexedPath example in #264 to the current structure. Let's say you have the following path structure:

data01
  |_x

data02
  |_x
    |_y
      |_z.txt

If you combine data01 and data02 in aMultiplexedPath, data01/x will make it impossible to find data02/x/y, see the following example:

from importlib_resources.readers import MultiplexedPath

p = MultiplexedPath('data01', 'data02')

In this case if you do:

print(p / 'x/y')

this will give data01/x/y instead of data02/x/y, and

for f in (p / 'x/y').iterdir():
    print(f)

will raise a FileNotFoundError.

The reason for this is that in Traversable.joinpath(), only the first path segment is used to search for a match:

names = itertools.chain.from_iterable(
    path.parts for path in map(pathlib.PurePosixPath, descendants)
)
target = next(names)

and then, only the first match is returned:

matches = (
    traversable for traversable in self.iterdir() if traversable.name == target
)
try:
    match = next(matches)

The solution to this problem could be related to the suggestion in #264, unless returning MultiplexedPaths isn't the desired behaviour of MultiplexedPath.joinpath().

[jaraco]

Agreed, MultiplexedPath is meant to honor all paths, recursively. Your suggestion looks approximately correct.

Do you have any interest in developing the fix (as a pull request), including tests to capture the missed expectation?

[jooste]

Sure! My only worry was that my implementation is based on a previous structure of MultiplexedPath. But if it looks ok I'll work on making it into a PR.