diffstat

diffstat [options] [file]

diffstat reads output from `diff` (or any unified/context diff, including `git diff`, `svn diff`, `hg diff`, or a `patch` file) and produces a histogram showing insertions, deletions, and modifications per file. It is commonly used to summarize patch size and churn.By default `diffstat` reads from stdin; a file argument reads from that file. Filenames in the header lines drive grouping; the `-p` option strips leading path components the same way as `patch -p`.

-C

Strip ANSI color escape sequences before parsing. Useful with colordiff output.

Show concise output with a single histogram code per file: `+` (insert), `-` (delete), `!` (modify).

Directory of result files to compute unchanged line counts.

Filename format: 0=path, 1=name, 2=auto, 3=dir, 4=count.

List filenames only.

Merge insert/delete counts from each chunk to approximate modified lines.

Strip num leading path components (like `patch -p`).

Directory containing original files (before applying diff).

Output as a comma-separated table.

Assume patch was created with old and new files swapped.

Set output width (histogram is never shorter than 10 columns).

Histograms are scaled — the width (controlled by `-w`) only sets the maximum bar length, not the absolute number of changes. When running on a diff with both added and removed files, use `-c` for a compact summary. Binary files and empty diffs produce no output.

diffstat was written by Thomas E. Dickey in 1992 as a companion tool to GNU diff. It ships in most Linux distributions in the `diffstat` package.

diff(1), patch(1), git-diff(1), colordiff(1)