[C++] Handle URI paths for filesystems?

[lidavidm]

Describe the enhancement requested

Given a URI like s3://bucket/foo/bar.parquet, right now, you can turn that URI into a FileSystem instance + a non-URI path. That is useful, but then the FileSystem APIs don't work with URIs. For instance, it's explicitly rejected here:

arrow/cpp/src/arrow/filesystem/s3fs.cc

Lines 416 to 419 in 0844924

	if (internal::IsLikelyUri(s)) {
	return Status::Invalid(
	"Expected an S3 object path of the form 'bucket/key...', got a URI: '", s, "'");
	}

So if you have a set of URIs and want to work with them with filesystems, you have to parse the URIs yourself. Given we already do the URI munging internally, it might be nice for FileSystem instances to expose a method to parse URIs into a path as well (raising an error if the URI is incompatible).

Component(s)

C++

[westonpace]

Ah, so arrow::fs::FileSystemFromUri is sufficient for the single file case but not the multi-file case?

Maybe it would be simpler to add arrow::fs:FileSystemFromUris?

Otherwise I worry that there may be many entry points into filesystem that accept a path (and there are many filesystem implementations). For example, this change would touch the S3 filesystem and the GCS filesystem and so on. I think there is some advantage in keeping the filesystem implementations as simple as possible.

[lidavidm]

That would also work. In my use case, I would like to get all the paths back to pass on to another system (which does also use Arrow filesystems), but there is one step that would at least like to parse and validate that all the URIs are consistent for a filesystem.