Duplicate files on a Mac: why APFS clones change the maths

On APFS, the format of every current Mac’s startup disk, a “duplicate” is often not a duplicate at all. When you choose File › Duplicate (Command-D) in Finder, or copy a file to another folder on the same volume, macOS usually makes a clone: a second file that points at the same data blocks as the first. It uses almost no extra space, so deleting it gives almost nothing back.

How APFS clones work

A clone is a new file whose contents are not copied. Both files point to the same blocks on disk. Only when one of them is changed does APFS write new blocks for the changed parts, a technique called copy-on-write. The unchanged parts stay shared.

macOS makes clones when:

• You use Duplicate or copy (Option-drag) within the same APFS volume in Finder.
• An app copies files with the system’s standard file-copy calls, which many do.
• You use cp -c in Terminal. Plain cp writes a full copy.

cp -c "Big Video.mov" "Big Video copy.mov"   # clone: shares data, almost no new space
cp "Big Video.mov" "Big Video copy.mov"      # full copy: writes all the data again

Copies to another volume, an external drive or a network share are always full copies. As Howard Oakley explains, clones also lose their sharing when they are restored from a Time Machine backup: each comes back as a separate file.

Why deleting a clone frees almost nothing

Space comes back only when the last file that uses a block is deleted, and no local snapshot still holds it. With a pair of clones, deleting one leaves every shared block in use by the other, so you recover a few kilobytes of file-system metadata. We tested it: cloning a 200 MB file reduced free space by about 4 KB, and a plain copy reduced it by 200 MB.

Partially edited clones sit in between. If you cloned a 10 GB video and then trimmed one copy, the blocks that differ belong to one file only; deleting that file frees those blocks and nothing else.

Why duplicate finders can overstate savings

Duplicate finders compare file contents, usually with a checksum. A clone has exactly the same contents as its original, so it is reported as a duplicate, and its full size is added to the “you could save” total. On a Mac with many clones, that total can be gigabytes too high.

The same blind spot explains why some tools disagree with each other. Finder’s Get Info and du add up each file’s size separately, so a folder with three clones of a 1 GB file looks like 3 GB. Your volume’s free space, shown in Disk Utility, is what really changes.

If you have many genuine copies, the better move is often to convert them into clones rather than delete them, so every path still works. That is what tools such as Hyperspace, by John Siracusa, do.

Logical size and allocated size

Every file has two sizes:

• Logical size: how many bytes it contains. ls -l and Finder’s Size show this.
• Allocated size: how much disk space it takes. Finder’s “on disk” figure and du show this.

They differ for clones (shared blocks), for sparse files such as disk images and virtual machines (unwritten regions take no space), for files compressed by APFS, and for cloud files whose download you removed. See logical vs physical size for more.

ls -lh "Big Video.mov"   # logical size
du -h "Big Video.mov"    # allocated size, counted as if nothing were shared

How to find real duplicates safely

Start with large files. Small duplicates rarely matter. In Finder, choose File › New Smart Folder, click +, set File Size › is greater than › 500 MB (choose Other… in the first menu if File Size is not listed), and switch to List view sorted by Name. Matching names and sizes are your candidates.

Confirm with a checksum. shasum ships with macOS and only reads files:

shasum -a 256 ~/Movies/"Trip.mov" ~/Desktop/"Trip copy.mov"

If the long hexadecimal strings match, the contents are identical. To check a whole folder of large files for matches:

find ~/Movies -type f -size +100M -print0 | xargs -0 shasum -a 256 \
  | awk '{n[$1]++; l[$1]=l[$1] $0 "\n"} END {for (h in n) if (n[h]>1) print l[h]}'

This prints groups of files with the same checksum. It deletes nothing. A matching checksum does not tell you whether the files are clones, so treat the space as “possibly recoverable”.

Use a reputable tool if you prefer. Open-source dupeGuru and well-known commercial apps work fine, provided you review each group, exclude libraries and system folders, and treat their savings figures as an upper limit.

The risk of deleting the “wrong” copy

Identical contents do not mean the copies are interchangeable.

• Photos libraries. Never delete files inside a .photoslibrary package in Finder; the library’s database still expects them and can be damaged. Since macOS Ventura, Photos has a Duplicates album under Utilities that merges duplicates properly. Deleted items wait in Recently Deleted.
• Music, Final Cut Pro, Lightroom and similar libraries reference files by path. Removing the “duplicate” that a library points to leaves a missing file.
• App bundles. Many apps ship identical frameworks. Deleting anything inside an .app breaks its code signature and can stop it opening. Uninstall the whole app instead; see how to uninstall Mac apps completely.
• Projects and code. A file in a project folder is referenced by builds and scripts; the copy on your Desktop is usually the one to remove.
• Backups. A deliberate second copy on another drive is a backup, not a duplicate.

When in doubt, keep the copy that lives where an app expects it, and move the other to the Trash so you can put it back.

How MoreRoom helps

MoreRoom’s Free Without Deleting finds identical files of 1 MB or more, for free. With Pro, it turns the extra copies into APFS clones, like Finder’s Duplicate: each copy stays where it is, with its own name, dates and tags, but the data is stored once, and nothing is deleted. Every file is checked again right before, and editing a copy later simply gives it its own space again. MoreRoom also measures your disk honestly: when it scans, cloned data is counted once, so its totals and treemap are not inflated by clones the way per-file sizes are. Its rules also flag common copies that are safe to review, such as old disk images left in Downloads. Anything you remove goes to the Trash first, with Undo. For Photos duplicates, use Photos’ own Duplicates album.