get_url() is slow when using with many files

[derdre]

Context

I am using DVC to track a remote data repository repo with revision rev. My goal is to directly access the files in the DVC cache. To get the relative paths within this repository I use

from dvc.api import DVCFileSystem

fs = DVCFileSystem(repo, rev)
files = fs.find('/', details=False, only_dvc=True)

I still need to resolve the relative paths to the actual paths in the cache, which can be done via

from dvc.api import get_url

paths = []
for f in files:
    paths.append(get_url(f, repo, rev))

Question

It seems to me that there is some overhead in get_url which makes it slow, if I want to resolve the cache paths for many files. Am I missing something here? Is there a way to get the cache paths directly from the DVCFileSystem (which has already be initialized)?

I have seen that I can get the md5 hashes also from DVCFileSystem. However, the checksum alone will not give me the path within the cache. For example, if I need to cater for different DVC versions (2.x vs 3.x) then I need to be able to distinguish the pattern

./dvc_cache/xx/xxxxxxxxxx

from

./dvc_cache/files/md5/xx/xxxxxxxxxx