Count files by extension
########################

:date:      2025-07-11 07:23:00 +0300
:category:  Linux
:tags:      find, sed, sort, uniq

TL;DR:

.. code-block:: bash

    find . -type f | sed 's/.*\///; s/^\.//; /\./! s/$/./; s/.*\.//' | sort | uniq -c | sort -nr

.. TEASER_END

The most popular command you can google appears to be this:

.. code-block:: bash

    find . -type f | sed 's/.*\.//' | sort | uniq -c

`Here <https://askubuntu.com/a/749005>`__ for example.

The problem is that it incorrectly counts files without extensions:

.. code-block:: text

    1 css
    1 csv
    1 git/COMMIT_EDITMSG
    1 git/config
    1 git/description
    1 git/HEAD
    4 gitignore
    ...

So it needs an improvement.
The first step is to cut the path and keep only the base name:

.. code-block:: bash

    find . -type f | sed 's/.*\///; s/.*\.//' | sort | uniq -c

Now we no longer have paths:

.. code-block:: text

    1 COMMIT_EDITMSG
    1 config
    1 css
    1 csv
    ...

The next step would be adding a dot at the end of extension-less files so empty extension gets counted separately:

.. code-block:: bash

    find . -type f | sed 's/.*\///; /\./! s/$/./; s/.*\.//' | sort | uniq -c

Now we have a useful output:

.. code-block:: text

    263
      1 bash
      7 coffee
      1 css
      1 csv
      1 exe
      1 fish
      4 gitignore
      1 iml
    ...

Next step is to add a fix for hidden files without extension (like ``.gitignore``):

.. code-block:: bash

    find . -type f | sed 's/.*\///; s/^\.//; /\./! s/$/./; s/.*\.//' | sort | uniq -c

Now we have a perfect list sorted by extension.
To sort by count, you can add ``sort -n`` for ascending order or ``sort -nr`` for descending:

.. code-block:: bash

    find . -type f | sed 's/.*\///; s/^\.//; /\./! s/$/./; s/.*\.//' | sort | uniq -c | sort -nr

.. code-block:: text

    1058 php
     268
      47 md
      36 json
    ...

The final explanation:

* ``find . -type f`` find every file in the current directory
* ``sed``:

  * ``'s/.*\///'`` cut the path, keeping only basename
  * ``'s/^\.//'`` remove leading dot because it does not indicate an extension
  * ``'/\./! s/$/./'`` add a dot if there is no dot at all
  * ``'s/.*\.//'`` remove the file name, keeping only an extension
* ``sort`` because uniq works only on sorted output
* ``uniq -c`` find duplicates and count them
* ``sort -nr`` optional part, sort extensions by a number of entries

It does not handle common double extensions like ``.tar.gz`` but it was always enough for me.