Shell Engineering

Comprehensive guidelines for writing production-quality shell scripts based on Google’s Shell Style Guide.

When to Use Shell

• Small utilities and simple wrapper scripts
• Scripts calling other tools with straightforward logic
• Rewrite in a structured language (Go, Python) when exceeding ~100 lines or using complex control flow

Shell Choice

• Bash is the only permitted shell for executables
• Start scripts with #!/bin/bash with minimal flags
• Libraries must have .sh extension and not be executable
• SUID/SGID are forbidden on shell scripts

File Structure

#!/bin/bash
#
# Brief description of the script's purpose.

set -euo pipefail

# Constants and environment variables (UPPERCASE)
readonly SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
readonly LOG_FILE="/tmp/script.log"

# Source libraries
source "${SCRIPT_DIR}/lib/utils.sh"

# Function definitions (lowercase_with_underscores)
my_function() {
  local arg1="$1"
  # ...
}

# Main function
main() {
  # Script logic here
}

main "$@"

Formatting Rules

Indentation and Length

• 2 spaces for indentation (no tabs)
• 80 characters maximum line length
• Split long pipelines with pipe at line start:

command1 \
  | command2 \
  | command3

Control Structures

• ; then and ; do on same line as if/while/for:

if [[ -n "${var}" ]]; then
  # ...
fi

for file in "${files[@]}"; do
  # ...
done

Quoting

• Always quote strings with variables, command substitutions, or spaces
• Use "${var}" format with braces for clarity
• Use "$@" not $* for argument lists
• Use arrays for lists with spaces in elements

Naming Conventions

Preferred Syntax

Use These

# Command substitution
result=$(command)

# Test conditions
if [[ -n "${var}" ]]; then

# Arithmetic
if (( count > 10 )); then
total=$(( a + b ))

# Local variables in functions
my_func() {
  local name="$1"
}

# Arrays for lists
files=("file1.txt" "file2.txt" "file with spaces.txt")
for f in "${files[@]}"; do

Avoid These

# Backticks (use $() instead)
result=`command`

# Single brackets (use [[ ]] instead)
if [ -n "$var" ]; then

# let, expr, $[ ] (use $(( )) instead)
let count=count+1

# eval (security risk)
eval "$cmd"

# Piping to while (loses variable scope)
cat file | while read line; do

# alias in scripts (use functions)
alias ll='ls -la'

# Unquoted wildcards
for f in *; do  # Use ./* instead

Error Handling

STDERR for Errors

err() {
  echo "[$(date +'%Y-%m-%d %H:%M:%S')] ERROR: $*" >&2
}

if ! process_file "${file}"; then
  err "Failed to process ${file}"
  exit 1
fi

Check Return Values

# Direct if check
if ! mv "${file}" "${dest}"; then
  err "Failed to move file"
fi

# Pipeline status
tar -cf - . | gzip > archive.tar.gz
if (( PIPESTATUS[0] != 0 || PIPESTATUS[1] != 0 )); then
  err "Archive creation failed"
fi

Comments and Documentation

File Header (Required)

#!/bin/bash
#
# Script description explaining purpose and usage.
#
# Usage: script.sh [options] <input_file>

Function Documentation

#######################################
# Process a data file and output results.
# Globals:
#   OUTPUT_DIR
# Arguments:
#   $1 - Input file path
#   $2 - Output format (csv|json)
# Outputs:
#   Writes processed data to OUTPUT_DIR
# Returns:
#   0 on success, non-zero on error
#######################################
process_data() {
  local input_file="$1"
  local format="${2:-csv}"
  # ...
}

TODO Comments

# TODO(username): Handle edge case for empty input

Testing and Validation

• Use ShellCheck to identify bugs
• Test string emptiness explicitly:

# Good
if [[ -z "${var}" ]]; then  # empty
if [[ -n "${var}" ]]; then  # non-empty

# Avoid
if [[ "${var}" ]]; then

Built-in Preference

Prefer bash builtins over external commands:

# Good: parameter expansion
filename="${path##*/}"
extension="${filename##*.}"
basename="${filename%.*}"

# Avoid: external commands
filename=$(basename "$path")
extension=$(echo "$filename" | sed 's/.*\.//')

Quick Reference