Git: Run a command on all files in a repository

A deconstructed example

For example, to get the line counts of all Python files:

$ git ls-files -z -- '*.py' | xargs -0r wc -l
...
     255 example/urls.py
     138 example/utils.py
    3160 example/views.py
   78123 total

Let’s take that apart.

You can take the line count of a file with wc and its -l flag:

$ wc -l example/models.py
    7142 example/models.py

git ls-files -- '*.py' lists all files known to Git that end in .py, in the current folder:

$ git ls-files -- '*.py'
...
example/urls.py
example/utils.py
example/views.py

Adding -z uses null byte separator, which isn’t exactly human readable:

$ git ls-files -z -- '*.py'
...example/urls.pyexample/utils.pyexample/views.py

Terminals don’t typically render the null byte, so all the filenames are jammed together. But the null byte separators are vital to correctly pass filenames with spaces, newlines, and weirder characters to xargs.

xargs takes a list of files from its input, and runs the command you give it on them, in groups as required. Its -0 option makes it split on the null byte separator, rather than newlines. The -r option avoids running the command if there are no files to process.

So in the combination command, it receives the list of files from git ls-files, and runs wc -l on them in batches.

Running Fixer Tools

You can also use this technique to run “fixer” tools. For example to run my tool django-upgrade on all Python files in your repository:

$ git ls-files -z -- '*.py' | xargs -0r django-upgrade --target-version 4.0
...
Rewriting example/models.py
Rewriting example/urls.py

Fin

May your mass edits not require mass reverts,

—Adam