rmlint

Find duplicates in directory

$ rmlint [/path/to/directory]

$ rmlint [/path/to/directory]

$ rmlint -T df [/path/to/directory]

$ rmlint -k [/path/to/directory]

$ rmlint [/path/to/originals] // [/path/to/duplicates]

$ rmlint -T ef,ed [/path/to/directory]

$ rmlint --dry-run [/path/to/directory]

$ rmlint -o json [/path/to/directory]

rmlint [-T types] [-k] [-o output] [options] paths [// paths]

rmlint finds duplicate files, empty files, broken symlinks, and other lint. It generates scripts to remove or manage found items.
Duplicate detection uses progressive matching: size first, then partial hashes, finally full hashes or paranoid byte comparison. This minimizes I/O for large collections.
The double-slash (//) separator defines original vs duplicate paths. Files in paths before // are preferred originals; those after are marked as duplicates. This enables controlled cleanup of backup or mirror directories.
Output includes a shell script (rmlint.sh) with removal commands. The script is cautious by default, requiring confirmation and keeping originals. JSON and CSV outputs enable custom processing.
Sorting criteria (-S) determine which duplicate is kept: by age, path depth, basename length, or alphabetically. Multiple criteria combine for fine-grained control.
Additional lint types include: empty directories, broken symlinks, files with bad user/group, and non-stripped binaries.

-T, --types TYPES

Find types: df (duplicates), ef (empty files), ed (empty dirs).

Keep files in tagged (first) paths.

Require match in tagged path.

Output format: sh, csv, json, py.

Configure output handler.

Filter by file size.

Maximum directory depth.

Don't write output files.

Show progress bar.

Byte-by-byte comparison.

Sorting for original selection.

Only files newer than file.

Include hidden files.

Follow symbolic links.

Hash-based detection has theoretical collision risk. Large filesystems need significant memory for tracking. Follow-symlink mode can expand search dramatically. Removal scripts should be reviewed before execution. Some filesystems don't track modification time accurately.

rmlint was created by Christopher Pahl (SeeSpotRun) around 2012 as a fast, modern duplicate finder. Written in C, it replaced slower Python predecessors. The project emphasizes safety (generating review-able scripts) and performance (parallel hashing, incremental matching).

fdupes(1), jdupes(1), duperemove(1), rdfind(1)