Estimate "true" APFS disk usage of folder without counting cloned files multiple times

Question

I would like to compute how much "true" APFS disk space is used by a folder. This folder has many APFS "cloned" versions of the same files. A "true" count of the disk space used would only count each of these clones once.

One thought I've had is to do this:

1. Go through each file in the folder, recursively.
2. Somehow determine, for each file, which physical location on the drive that file maps to.
3. Build an array of these locations for all files; prune duplicates when they appear (which would be clones).
4. Get the total computed size.

I am curious if one could use this kind of method to get the true size of the folder. I'm also curious if there are any existing utilities to do this, or if not, how I could implement this on my own. Perhaps using fcntl and F_LOG2PHYS, from this answer?

Answer 1

Yes, that method would be a starting point in getting the "true size" of the folder.

In practice, you would need to start with defining what you think is the "true size" of a folder, as that could very much vary from person to person.

For example: If files are compressed in the file system (i.e. not using an user space compression tool such as gzip or xz), would you regard the compressed or uncompressed size as the "true size"?

For example: If files contains "holes" (so called 'sparse files'), would you consider the size of the holes as part of the "true size" or not?

For example: If the file data (i.e. payload) does not match completely with the block size, would you consider the payload size the "true size" or would you round it up to a multiple of the block size?

And the list can probably go on with personal preferences of "true sizing".