Finding a Bash Script's Directory Reliably

Getting the directory path of a running script is a common requirement in bash, whether you’re sourcing files, building paths to config directories, or navigating relative to the script itself. The naive approach often fails with symbolic links or when the script is sourced rather than executed directly.

The Problem

Using $0 alone is unreliable. If your script is a symlink, dirname $0 returns the symlink’s directory, not the actual script location. If the script is sourced, $0 refers to the shell itself. These edge cases break path resolution in production scripts.

The Solution

Use readlink -f to resolve the absolute path, then pass it to dirname:

script_dir=$(dirname "$(readlink -f "$0")")

Breaking this down:

readlink -f "$0" — resolves all symlinks and returns the absolute path to the actual script file
dirname ... — extracts the directory portion of that path
script_dir=... — stores the result in a variable

Handling Sourced Scripts

When a script is sourced (rather than executed), $0 points to the parent shell, not the sourced file. Use ${BASH_SOURCE[0]} instead:

script_dir=$(dirname "$(readlink -f "${BASH_SOURCE[0]}")")

${BASH_SOURCE[0]} always references the currently executing file in the stack, so it works correctly whether the script is executed or sourced.

Practical Example

Here’s a complete script that sets up directories reliably:

#!/bin/bash

script_dir=$(dirname "$(readlink -f "${BASH_SOURCE[0]}")")
config_dir="$script_dir/config"
data_dir="$script_dir/data"

echo "Script location: $script_dir"
echo "Config directory: $config_dir"
echo "Data directory: $data_dir"

if [[ ! -d "$config_dir" ]]; then
    mkdir -p "$config_dir"
fi

Edge Cases to Consider

Symlinked script: If /usr/bin/myscript is a symlink to /opt/app/myscript.sh, readlink -f returns /opt/app/myscript.sh and script_dir becomes /opt/app/. This is the correct behavior for finding supporting files.

Script in PATH: If you run a script from PATH (e.g., just typing myscript), the absolute path resolution still works correctly.

Relative symlinks: Even complex chains of relative symlinks are resolved completely by readlink -f, unlike older -e flag implementations.

No readlink available: On some minimal systems, readlink -f may not be available. As a fallback:

script_dir=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)

This uses cd and pwd to normalize the path, though it doesn’t resolve symlinks. For most scripts, this is acceptable.

Best Practice

Use this as your standard:

script_dir=$(dirname "$(readlink -f "${BASH_SOURCE[0]}")")

It handles symlinks, sourcing, relative paths, and works across different invocation methods. Store it near the top of your script before any file operations that depend on the script’s location.

Finding a Bash Script’s Directory Reliably