Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?


Efficient duplicate file finder and remover

CircleCI Documentation License: MIT

fclones is a command line utility that identifies groups of identical files and gets rid of the file copies you no longer need. It comes with plenty of configuration options for controlling the search scope and offers many ways of removing duplicates. For maximum flexibility, it integrates well with other Unix utilities like find and it speaks JSON, so you have a lot of control over the search and cleanup process.

fclones treats your data seriously. You can inspect and modify the list of duplicate files before removing them. There is also a --dry-run option that can tell you exactly what changes on the file system would be made.

fclones has been implemented in Rust with a strong focus on high performance on modern hardware. It employs several optimization techniques not present in many other programs. It adapts to the type of the hard drive, orders file operations by physical data placement on HDDs, scans directory tree in parallel and uses prefix compression of paths to reduce memory consumption when working with millions of files. It is also friendly to page-cache and does not push out your data out of cache. As a result, fclones easily outperforms many other popular duplicate finders by a wide margin on either SSD or HDD storage.

fclones is available on a wide variety of operating systems, but it works best on Linux.


  • Identifying groups of identical files
    • finding duplicate files
    • finding files with more than N replicas
    • finding unique files
    • finding files with fewer than N replicas
  • Advanced file selection for reducing the amount of data to process
    • scanning multiple directory roots
    • can work with a list of files piped directly from standard input
    • recursive/non-recursive file selection
    • recursion depth limit
    • filtering names and paths by extended UNIX globs
    • filtering names and paths by regular expressions
    • filtering by min/max file size
    • proper handling of symlinks and hardlinks
  • Removing redundant data
    • removing, moving or replacing files with soft or hard links
    • removing redundant file data using native copy-on-write (reflink) support on some file systems
    • selecting files for removal by path or name patterns
    • prioritizing files to remove by creation, modification, last access time or nesting level
  • High performance
    • parallel processing capability in all I/O and CPU heavy stages
    • automatic tuning of parallelism and access strategy based on device type (SSD vs HDD)
    • low memory footprint thanks to heavily optimized path representation
    • variety of fast non-cryptographic and cryptographic hash functions up to 512 bits wide
    • doesn't push data out of the page-cache (Linux-only)
    • optional persistent caching of file hashes
    • accurate progress reporting
  • Variety of output formats for easy further processing of results
    • standard text format
      • groups separated by group headers with file size and hash
      • one path per line in a group
    • optional fdupes compatibility (no headers, no indent, groups separated by blank lines)
    • machine-readable formats: CSV, JSON


Copy-on-write file data deduplication (reflink) is not supported on Windows.

Some optimisations are not available on platforms other than Linux:

  • ordering of file accesses by physical placement
  • page-cache drop-behind


Let's first create some files:

$ mkdir test
$ cd test
$ echo foo >foo1.txt
$ echo foo >foo2.txt
$ echo foo >foo3.txt
$ echo bar >bar1.txt
$ echo bar >bar2.txt

Now let's identify the duplicates:

$ fclones group . >dupes.txt
[2021-06-05 18:21:33.358] fclones:  info: Started grouping
[2021-06-05 18:21:33.738] fclones:  info: Scanned 7 file entries
[2021-06-05 18:21:33.738] fclones:  info: Found 5 (20 B) files matching selection criteria
[2021-06-05 18:21:33.738] fclones:  info: Found 4 (16 B) candidates after grouping by size
[2021-06-05 18:21:33.738] fclones:  info: Found 4 (16 B) candidates after grouping by paths and file identifiers
[2021-06-05 18:21:33.739] fclones:  info: Found 3 (12 B) candidates after grouping by prefix
[2021-06-05 18:21:33.740] fclones:  info: Found 3 (12 B) candidates after grouping by suffix
[2021-06-05 18:21:33.741] fclones:  info: Found 3 (12 B) redundant files

$ cat dupes.txt
# Report by fclones 0.12.0
# Timestamp: 2021-06-05 18:21:33.741 +0200
# Command: fclones group .
# Found 2 file groups
# 12 B (12 B) in 3 redundant files can be removed
7d6ebf613bf94dfd976d169ff6ae02c3, 4 B (4 B) * 2:
6109f093b3fd5eb1060989c990d1226f, 4 B (4 B) * 3:

Finally we can replace the duplicates by soft links:

$ fclones link --soft <dupes.txt 
[2021-06-05 18:25:42.488] fclones:  info: Started deduplicating
[2021-06-05 18:25:42.493] fclones:  info: Processed 3 files and reclaimed 12 B space

$ ls -l
total 12
-rw-rw-r-- 1 pkolaczk pkolaczk   4 cze  5 18:19 bar1.txt
lrwxrwxrwx 1 pkolaczk pkolaczk  18 cze  5 18:25 bar2.txt -> /tmp/test/bar1.txt
-rw-rw-r-- 1 pkolaczk pkolaczk 382 cze  5 18:21 dupes.txt
-rw-rw-r-- 1 pkolaczk pkolaczk   4 cze  5 18:19 foo1.txt
lrwxrwxrwx 1 pkolaczk pkolaczk  18 cze  5 18:25 foo2.txt -> /tmp/test/foo1.txt
lrwxrwxrwx 1 pkolaczk pkolaczk  18 cze  5 18:25 foo3.txt -> /tmp/test/foo1.txt


The code has been thoroughly tested on Ubuntu Linux 21.10. Other systems like Windows or Mac OS X and other architectures may work. Help test and/or port to other platforms is welcome. Please report successes as well as failures.

Official Packages

Snap store (Linux):

snap install fclones

Homebrew (macOS and Linux)

brew install fclones

Installation packages and binaries for some platforms are also attached directly to Releases.

Third-party Packages

Building from Source

Install Rust Toolchain and then run:

cargo install fclones

The build will write the binary to .cargo/bin/fclones.


fclones offers separate commands for finding and removing files. This way, you can inspect the list of found files before applying any modifications to the file system.

  • group – identifies groups of identical files and prints them to the standard output
  • remove – removes redundant files earlier identified by group
  • link – replaces redundant files with links (default: hard links)
  • dedupe – does not remove any files, but deduplicates file data by using native copy-on-write capabilities of the file system (reflink)

Finding Files

Find duplicate, unique, under-replicated or over-replicated files in the current directory, including subdirectories:

fclones group .
fclones group . --unique 
fclones group . --rf-under 3
fclones group . --rf-over 3

You can search in multiple directories:

fclones group dir1 dir2 dir3

By default, hidden files and files matching patterns listed in .gitignore and .fdignore are ignored. To search all files, use:

fclones group --no-ignore --hidden dir

Limit the recursion depth:

fclones group . --depth 1   # scan only files in the current dir, skip subdirs
fclones group * --depth 0   # similar as above in shells that expand `*` 

Caution: Versions up to 0.10 did not descend into directories by default. In those old versions, add -R flag to enable recursive directory walking.

Finding files that match across two directory trees, without matching identical files within each tree:

fclones group --isolate dir1 dir2

Finding duplicate files of size at least 100 MB:

fclones group . -s 100M

Filter by file name or path pattern:

fclones group . --name '*.jpg' '*.png' 

Run fclones on files selected by find (note: this is likely slower than built-in filtering):

find . -name '*.c' | fclones group --stdin --depth 0

Follow symbolic links, but don't escape out of the home folder:

fclones group . -L --path '/home/**'

Exclude a part of the directory tree from the scan:

fclones group / --exclude '/dev/**' '/proc/**'

Removing Files

To remove duplicate files, move them to a different place or replace them by links, you need to send the report produced by fclones group to the standard input of fclones remove, fclones move or fclones link command. The report format is detected automatically. Currently, default and json report formats are supported.

Assuming the list of duplicates has been saved in file dupes.txt, the following commands would remove the redundant files:

fclones link <dupes.txt             # replace with hard links
fclones link -s <dupes.txt          # replace with symbolic links
fclones move target_dir <dupes.txt  # move to target_dir  
fclones remove <dupes.txt           # remove totally

If you prefer to do everything at once without storing the list of groups in a file, you can pipe:

fclones group . | fclones link

To select the number of files to preserve, use the -n/--rf-over option. By default, it is set to the value used when running group (which is 1 if it wasn't set explicitly). To leave 2 replicas in each group, run:

fclones remove -n 2 <dupes.txt

By default, fclones follows the order of files specified in the input file. It keeps the files given at the beginning of each list, and removes / replaces the files given at the end of each list. It is possible to change that order by --priority option, for example:

fclones remove --priority newest <dupes.txt        # remove the newest replicas
fclones remove --priority oldest <dupes.txt        # remove the oldest replicas

For more priority options, see fclones remove --help.

It is also possible to restrict removing files to only files with names or paths matching a pattern:

fclones remove --name '*.jpg' <dupes.txt       # remove only jpg files
fclones remove --path '/trash/**' <dupes.txt   # remove only files in the /trash folder

If it is easier to specify a pattern for files which you do not want to remove, then use one of keep options:

fclones remove --keep-name '*.mov' <dupes.txt           # never remove mov files
fclones remove --keep-path '/important/**' <dupes.txt   # never remove files in the /important folder

To make sure you're not going to remove wrong files accidentally, use --dry-run option. This option prints all the commands that would be executed, but it doesn't actually execute them:

fclones link --soft <dupes.txt --dry-run 2>/dev/null

mv /tmp/test/bar2.txt /tmp/test/bar2.txt.jkXswbsDxhqItPeOfCXsWN4d
ln -s /tmp/test/bar1.txt /tmp/test/bar2.txt
rm /tmp/test/bar2.txt.jkXswbsDxhqItPeOfCXsWN4d
mv /tmp/test/foo2.txt /tmp/test/foo2.txt.ze1hvhNjfre618TkRGUxJNzx
ln -s /tmp/test/foo1.txt /tmp/test/foo2.txt
rm /tmp/test/foo2.txt.ze1hvhNjfre618TkRGUxJNzx
mv /tmp/test/foo3.txt /tmp/test/foo3.txt.ttLAWO6YckczL1LXEsHfcEau
ln -s /tmp/test/foo1.txt /tmp/test/foo3.txt
rm /tmp/test/foo3.txt.ttLAWO6YckczL1LXEsHfcEau

Handling links

Files linked by symbolic links or hard links are not treated as duplicates. You can change this behavior by setting the following flags:

  • When --isolate is set:
    • links residing in different directory trees are treated as duplicates,
    • links residing in the same directory tree are counted as a single replica.
  • When --match-links is set, fclones treats all linked files as duplicates.

Consider the following directory structure, where all files are hard links sharing the same content:

  - file1
  - file2
  - file3
  - file4

Because all files are essentially the same data, they will end up in the same file group, but the actual number of replicas present in that file group will differ depending on the flags given:

Command Number of replicas Group reported Files to remove
fclones group dir1 dir2 1 No
fclones group dir1 dir2 --isolate 2 Yes file3, file4
fclones group dir1 dir2 --match-links 4 Yes file2, file3, file4

Symbolic links

The group command ignores symbolic links to files unless at least --follow-links or --symbolic-links flag is set. If only --follow-links is set, symbolic links to files are followed and resolved to their targets. If --symbolic-links is set, symbolic links to files are not followed, but treated as hard links and potentially reported in the output report. When both --symbolic-links and --follow-links are set, symbolic links to directories are followed, but symbolic links to files are treated as hard links.

Caution: Using --match-links together with --symbolic-links is very dangerous. It is easy to end up deleting the only regular file you have, and to be left with a bunch of orphan symbolic links.

Preprocessing Files

Use --transform option to safely transform files by an external command. By default, the transformation happens on a copy of file data, to avoid accidental data loss. Note that this option may significantly slow down processing of a huge number of files, because it invokes the external program for each file.

The following command will strip exif before matching duplicate jpg images:

fclones group . --name '*.jpg' -i --transform 'exiv2 -d a $IN' --in-place     


List more options:

fclones [command] -h      # short help
fclones [command] --help  # detailed help

Path Globbing

fclones understands a subset of Bash Extended Globbing. The following wildcards can be used:

  • ? matches any character except the directory separator
  • [a-z] matches one of the characters or character ranges given in the square brackets
  • [!a-z] matches any character that is not given in the square brackets
  • * matches any sequence of characters except the directory separator
  • ** matches any sequence of characters including the directory separator
  • {a,b} matches exactly one pattern from the comma-separated patterns given inside the curly brackets
  • @(a|b) same as {a,b}
  • ?(a|b) matches at most one occurrence of the pattern inside the brackets
  • +(a|b) matches at least occurrence of the patterns given inside the brackets
  • *(a|b) matches any number of occurrences of the patterns given inside the brackets
  • \ escapes wildcards on Unix-like systems, e.g. \? would match ? literally
  • ^ escapes wildcards on Windows, e.g. ^? would match ? literally


  • On Unix-like systems, when using globs, one must be very careful to avoid accidental expansion of globs by the shell. In many cases having globs expanded by the shell instead of by fclones is not what you want. In such cases, you need to quote the globs:

    fclones group . --name '*.jpg'       
  • On Windows, the default shell doesn't remove quotes before passing the arguments to the program, therefore you need to pass globs unquoted:

    fclones group . --name *.jpg
  • On Windows, the default shell doesn't support path globbing, therefore wildcard characters such as * and ? used in paths will be passed literally, and they are likely to create invalid paths. For example, the following command that searches for duplicate files in the current directory in Bash, will likely fail in the default Windows shell:

    fclones group *

    If you need path globbing, and your shell does not support it, use the builtin path globbing provided by --name or --path.

The Algorithm

Files are processed in several stages. Each stage except the last one is parallel, but the previous stage must complete fully before the next one is started.

  1. Scan input files and filter files matching the selection criteria. Walk directories recursively if requested. Follow symbolic links if requested. For files that match the selection criteria, read their size.
  2. Group collected files by size by storing them in a hash-map. Remove groups smaller than the desired lower-bound (default 2).
  3. In each group, remove duplicate files with the same inode id. The same file could be reached through different paths when hardlinks are present. This step can be optionally skipped.
  4. For each remaining file, compute a hash of a tiny block of initial data. Put files with different hashes into separate groups. Prune result groups if needed.
  5. For each remaining file, compute a hash of a tiny block of data at the end of the file. Put files with different hashes into separate groups. Prune small groups if needed.
  6. For each remaining file, compute a hash of the whole contents of the file. Note that for small files we might have already computed a full contents hash in step 4, therefore these files can be safely omitted. Same as in steps 4 and 5, split groups and remove the ones that are too small.
  7. Write report to the stdout.

Note that there is no byte-by-byte comparison of files anywhere. All available hash functions are at least 128-bit wide, and you don't need to worry about hash collisions. At 1015 files, the probability of collision is 0.000000001 when using a 128-bit hash, without taking into account the requirement for the files to also match by size.


You can select the hash function with --hash-fn (default: metro). Non-cryptographic hashes are much more efficient than cryptographic, however you probably won't see much difference unless you're reading from a fast SSD or if file data is cached.

Hash function Hash width Cryptographic
metro 128-bit No
xxhash3 128-bit No
blake3 256-bit Yes
sha256 256-bit Yes
sha512 512-bit Yes
sha3-256 256-bit Yes
sha3-512 512-bit Yes


This section provides hints on getting the best performance from fclones.

Incremental Mode

If you expect to run fclones group more than once on the same set of files, you might benefit from turning on the hash cache by adding the --cache flag:

fclones group --cache <dir>

Caching can dramatically improve grouping speed on subsequent runs of fclones at the expense of some additional storage space needed for the cache. Caching also allows for resuming work quickly after interruption, so it is recommended if you plan to run fclones on huge data sets.

The cache works as follows:

  • Each newly computed file hash is persisted in the cache together with some metadata of the file such as its modification timestamp and length.
  • Whenever a file hash needs to be computed, it is first looked up in the cache. The cached hash is used if the current metadata of the file strictly matches the metadata stored in the cache.

Cached hashes are not invalidated by file moves because files are identified by their internal identifiers (inode identifiers on Unix), not by path names, and moves/renames typically preserve those.

Beware that caching relies on file metadata to detect changes in file contents. This might introduce some inaccuracies to the grouping process if a file modification timestamp and file length is not updated immediately whenever a file gets modified. Most file systems update the timestamps automatically on closing the file. Therefore, changed files that are held open for a long time (e.g. by database systems) might be not noticed by fclones group and might use stale cached values.

The cache database is located in the standard cache directory of the user account. Typically, those are:

  • Linux: $HOME/.cache/fclones
  • macOS: $HOME/Library/Caches/fclones
  • Windows: $HOME/AppData/Local/fclones

Configuring Parallelism

The --threads parameter controls the sizes of the internal thread-pool(s). This can be used to reduce parallelism level when you don't want fclones to impact performance of your system too much, e.g. when you need to do some other work at the same time. We recommended reducing the parallelism level if you need to reduce memory usage.

When using fclones up to version 0.6.x to deduplicate files of sizes of at least a few MBs each
on spinning drives (HDD), it is recommended to set --threads 1, because accessing big files from multiple threads on HDD can be much slower than single-threaded access (YMMV, this is heavily OS-dependent, 2x-10x performance differences have been reported).

Since version 0.7.0, fclones uses separate per-device thread-pools for final hashing and it will automatically tune the level of parallelism, memory buffer sizes and partial hashing sizes based on the device type. These automatic settings can be overridden with -threads as well.

The following options can be passed to --threads. The more specific options override the less specific ones.

  • main:<n> – sets the size of the main thread-pool used for random I/O: directory tree scanning, file metadata fetching and in-memory sorting/hashing. These operations typically benefit from high parallelism level, even on spinning drives. Unset by default, which means the pool will be configured to use all available CPU cores.
  • dev:<device>:<r>,<s> – sets the size of the thread-pool r used for random I/O and s used for sequential I/O on the block device with the given name. The name of the device is OS-dependent. Note this is not the same as the partition name or mount point.
  • ssd:<r>,<s> – sets the sizes of the thread-pools used for I/O on solid-state drives. Unset by default.
  • hdd:<r>,<s> – sets the sizes of the thread-pools used for I/O on spinning drives. Defaults to 8,1
  • removable:<r>,<s> – sets the size of the thread-pools used for I/O on removable devices (e.g. USB sticks). Defaults to 4,1
  • unknown:<r>,<s> – sets the size of the thread-pools used for I/O on devices of unknown type. Sometimes the device type can't be determined e.g. if it is mounted as NAS. Defaults to 4,1
  • default:<r>,<s> – sets the pool sizes to be used by all unset options
  • <r>,<s> - same as default:<r>,<s>
  • <n> - same as default:<n>,<n>


To limit the parallelism level for the main thread pool to 1:

fclones group <paths> --threads main:1  

To limit the parallelism level for all I/O access for all SSD devices:

fclones group <paths> --threads ssd:1 

To set the parallelism level to the number of cores for random I/O access and to 2 for sequential I/O access for /dev/sda block device:

fclones group <paths> --threads dev:/dev/sda:0,2 

Multiple --threads options can be given, separated by spaces:

fclones group <paths> --threads main:16 ssd:4 hdd:1,1     


Different duplicate finders were given a task to find duplicates in a large set of files. Before each run, the system page cache was evicted with echo 3 > /proc/sys/vm/drop_caches.

SSD Benchmark

  • Model: Dell Precision 5520
  • CPU: Intel(R) Xeon(R) CPU E3-1505M v6 @ 3.00GHz
  • RAM: 32 GB
  • Storage: local NVMe SSD 512 GB
  • System: Ubuntu Linux 20.10, kernel 5.8.0-53-generic
  • Task: 1,460,720 paths, 316 GB of data
Program Version Language Time Peak Memory
fclones 0.12.1 Rust 0:34.59 266 MB
yadf 0.15.2 Rust 0:59.32 329 MB
czkawka 3.1.0 Rust 2:09.00 1.4 GB
rmlint 2.9.0 C, Python 2:28.43 942 MB
jdupes 1.18.2 C 5:01.91 332 MB
dupe-krill 1.4.5 Rust 5:09.52 706 MB
fdupes 2.1.1 C 5:46.19 342 MB
rdfind 1.4.1 C++ 5:53.07 496 MB
dupeguru 4.1.1 Python 7:49.89 1.4 GB
fdupes-java 1.3.1 Java > 20 minutes 4.2 GB

fdupes-java did not finish the test. I interrupted it after 20 minutes while it was still computing MD5 in stage 2/3. Unfortunately fdupes-java doesn't display a useful progress bar, so it is not possible to estimate how long it would take.

HDD Benchmark

  • Model: Dell Precision M4600
  • CPU: Intel(R) Core(TM) i7-2760QM CPU @ 2.40GHz
  • RAM: 24 GB
  • System: Mint Linux 19.3, kernel 5.4.0-70-generic
  • Storage: Seagate Momentus 7200 RPM SATA drive, EXT4 filesystem
  • Task: 51370 paths, 2 GB data, 6811 (471 MB) duplicate files

Commands used:

  /usr/bin/time -v fclones -R <file set root> 
  /usr/bin/time -v jdupes -R -Q <file set root>
  /usr/bin/time -v fdupes -R <file set root>
  /usr/bin/time -v rdfind <file set root>

In this benchmark, the page cache was dropped before each run.

Program Version Language Threads Time Peak Memory
fclones 0.9.1 Rust 1 0:19.45 18.1 MB
rdfind 1.3.5 C++ 1 0:33.70 18.5 MB
yadf 0.14.1 Rust 1:11.69 22.9 MB
jdupes 1.9 C 1 1:18.47 15.7 MB
fdupes 1.6.1 C 1 1:33.71 15.9 MB