Git clean

The example creates a new Git repository in the git_clean_test directory. It then proceeds to create a tracked_file which is added to the Git index, additionally, an untracked_file is created, and an untracked_dir. The example then invokes git status which displays output indicating Git's internal state of tracked and untracked changes. With the repository in this state, we can execute the git clean command to demonstrate its intended purpose.

At this point, executing the default git clean command may produce a fatal error. The example above demonstrates what this may look like. By default, Git is globally configured to require that git clean be passed a "force" option to initiate. This is an important safety mechanism. When finally executed git clean is not undo-able. When fully executed, git clean will make a hard filesystem deletion, similar to executing the command line rm utility. Make sure you really want to delete the untracked files before you run it.

Given the previous explanation of the default git clean behaviors and caveats, the following content demonstrates various git clean use cases and the accompanying command line options required for their operation.

The -n option will perform a “dry run” of git clean. This will show you which files are going to be removed without actually removing them. It is a best practice to always first perform a dry run of git clean. We can demonstrate this option in the demo repo we created earlier.

The output tells us that untracked_file will be removed when the git clean command is executed. Notice that the untracked_dir is not reported in the output here. By default git clean will not operate recursively on directories. This is another safety mechanism to prevent accidental permanent deletion.

The force option initiates the actual deletion of untracked files from the current directory. Force is required unless the clean.requireForce configuration option is set to false. This will not remove untracked folders or files specified by .gitignore. Let us now execute a live git clean in our example repo.

The command will output the files that are removed. You can see here that untracked_file has been removed. Executing git status at this point or doing a ls will show that untracked_file has been deleted and is nowhere to be found. By default git clean -f will operate on all the current directory untracked files. Additionally, a < path > value can be passed with the -f option that will remove a specific file.

The -d option tells git clean that you also want to remove any untracked directories, by default it will ignore directories. We can add the -d option to our previous examples:

Here we have executed a 'dry run' using the -dn combination which outputs untracked_dir is up for removal. Then we execute a forced clean, and receive output that untracked_dir is removed.

A common software release pattern is to have a build or distribution directory that is not committed to the repositories tracking index. The build directory will contain ephemeral build artifacts that are generated from the committed source code. This build directory is usually added to the repositories .gitignore file. It can be convenient to also clean this directory with other untracked files. The -x option tells git clean to also include any ignored files. As with previous git clean invocations, it is a best practice to execute a 'dry run' first, before the final deletion. The -x option will act on all ignored files, not just project build specific ones. This could be unintended things like ./.idea IDE configuration files.

Like the -d option -x can be passed and composed with other options. This example demonstrates a combination with -f that will remove untracked files from the current directory as well as any files that Git usually ignores.

In addition to the ad-hoc command line execution we have demonstrated so far, git clean has an "interactive" mode that you can initiate by passing the -i option. Let us revisit the example repo from the introduction of this document. In that initial state, we will start an interactive clean session.

We have initiated the interactive session with the -d option so it will also act upon our untracked_dir. The interactive mode will display a What now> prompt that requests a command to apply to the untracked files. The commands themselves are fairly self explanatory. We'll take a brief look at each in a random order starting with command 6: help. Selecting command 6 will further explain the other commands:

Is straight forward and will exit the interactive session.

Will delete the indicated items. If we were to execute 1: clean at this point untracked_dir/ untracked_file would be removed

will iterate over each untracked file and display a Y/N prompt for a deletion. It looks like the following:

Will display an additional prompt that takes input used to filter the list of untracked files.

Here we input the *_file wildcard pattern which then restricts the untracked file list to just untracked_dir.

Similar to command 2, command 3 works to refine the list of untracked file names. The interactive session will prompt for numbers that correspond to an untracked file name.

To recap, git clean is a convenience method for deleting untracked files in a repo's working directory. Untracked files are those that are in the repo's directory but have not yet been added to the repo's index with git add. Overall the effect of git clean can be accomplished using git status and the operating systems native deletion tools. Git clean can be used alongside git reset to fully undo any additions and commits in a repository.