[rfile] ListKeys: don't include the root directory itself in the listing

[silverweed]

This Pull request:

changes the behavior of RFile::ListKeys to exclude the root directory itself from the list of returned keyInfos.

Previously ListKeys("a", kListDirs) would return (e.g.) { "a", "a/b", "a/c", "a/d" }; now it only returns { "a/b", "a/c", "a/d" }.

Rationale: you generally don't care about the root dir in the listing, since you already know it's there.

[github-actions]

Test Results

    21 files      21 suites   3d 10h 18m 53s ⏱️
 3 863 tests  3 802 ✅   0 💤 61 ❌
72 810 runs  72 642 ✅ 107 💤 61 ❌

For more details on these failures, see this check.

Results for commit ef1c2d7.

[jblomer]

I'm not sure if it is true that we generally know that the directory to be listed is present. Can we distinguish between an empty directory and a non-existing directory?

[silverweed]

I'm not sure if it is true that we generally know that the directory to be listed is present. Can we distinguish between an empty directory and a non-existing directory?

That is a good point. You can get that information by calling file->GetKeyInfo("root_dir"), so the question is which choice is the most convenient for typical use cases: needing code to skip the first element of the array vs calling a separate function to know if the directory is there.

My impression is that the first choice is superior because when you call ListKeys you typically want to perform some action on each child of the folder and you probably don't care if the list is empty because there is an empty directory of because there is no directory.

In fact, the whole concept of "empty directory" is pretty second class in RFile: you cannot create an empty directory through RFile and you are not really supposed to distinguish the case of empty vs non-existing dir semantically.

[vepadulano]

I think it's fine as long as it is documented, thanks!