A git hook that forces you to respect Commit Convention

commit-msg.sh

#!/bin/bash
#
# 	Commit Convention Hook | alemi <[email protected]> May 2022
#
# A simple git hook to enforce (sort of) Commit Convention.
# 		www.conventionalcommits.org
# 
# This script uses bash builtin regex matching, so it's not sh compatible.
# 
# Commit message will be stripped of comments (lines starting with #) and matched 
#  against a regular expression enforcing Commit Convention.
# If no match is found, this script will try to point out commit message issues.
# 
# A brief explaination of the Commit Convention can be requested with COMMIT_CONVENTION_HELP=1
# A custom regex can be provided with COMMIT_CONVENTION_REGEX
# By default, accepted commit types are
# 	build|chore|ci|docs|feat|fix|perf|refactor|revert|style|test
# but these can be changed with COMMIT_CONVENTION_TYPES="your|types"
# The default separator is a colon ':', but can be customized with COMMIT_CONVENTION_SEPARATOR="|"
# Disable issues hints with COMMIT_CONVENTION_HINTS_DISABLED=1
# 
# This hook (and commit checking) can be bypassed by adding flag --no-verify to git command
# 
# To install this hook, simply put it into the .git/hooks folder of your project with name 'commit-msg',
#  and make sure that it's executable. Bash needs to be installed.
# 

# Check environment for options and set defaults
if [[ -z $COMMIT_CONVENTION_TYPES ]]; then
	COMMIT_CONVENTION_TYPES="build|chore|ci|docs|feat|fix|perf|refactor|revert|style|test"
fi
if [[ -z $COMMIT_CONVENTION_SEPARATOR ]]; then
	COMMIT_CONVENTION_SEPARATOR=":"
fi
if [[ -z $COMMIT_CONVENTION_REGEX ]]; then
	COMMIT_CONVENTION_REGEX="^($COMMIT_CONVENTION_TYPES)([[(]"$'(\w+)'"[])]|)(!|)$COMMIT_CONVENTION_SEPARATOR (.{1,45})"$'(\n\n.*$|\n$|$)'
fi

# strip commit message of comments
INPUT_COMMIT_MESSAGE=$(sed -e "s/^#.*$//g" "$1")

# test against commit convention regex
if [[ $INPUT_COMMIT_MESSAGE =~ $COMMIT_CONVENTION_REGEX ]]; then
	# we also check length and empty line thanks to regex above. length is not perfect: we have
	#  to assume that commit type is short (5 chars), but it may be longer.
	#  this allows for some leeway in commit title length, but precise error reporting 
	#  if commit doesn't comply with convention.
	#  using very long scopes can bypass length check, but if we're using scopes that long,
	#  I think we can go with an exception to the rule
	exit 0 # Everything good
fi

echo "Failed creating commit: message does not comply with Commit Convention"

# Print help message giving some info about Commit Convention if requested
if [[ $COMMIT_CONVENTION_HELP ]]; then
	# sed works line by line, so cannot be used. tr replaces 1 char with 1 char, so cannot add indent
	TABBED_COMMIT_MESSAGE=$(echo "$INPUT_COMMIT_MESSAGE" | awk 1 ORS='\n    ')
	cat <<-ENDSTRING
	A correctly formatted commit message should look like
	    <type>[(<optional scope>)][!]: <description>
	
	    [<optional body>]
	
	    [<optional footer>]
	
	Valid types are:
	    $COMMIT_CONVENTION_TYPES

	Scope can be anything, could reference a branch or tag
	Add an optional ! bang before colon to signal breaking changes
	Type, scope and description should all be in one line and less than 80 characters long
	Your body has no length or format restrictions, just leave an empty line
	It's good practice to reference contributors in a footer
	
	Examples:
	    fix: memory leak
	    feat!: added ability to define buffer sizes at launch time
	    refactor(config): reworked config storage
	
	Input commit message:
	    $TABBED_COMMIT_MESSAGE
	
	Issues:
	ENDSTRING
fi

# Attempt to parse commit message more leniently and suggest issues if not disabled
if [[ -z $COMMIT_CONVENTION_HINTS_DISABLED ]]; then
	IFS=$'\n' readarray INPUT_COMMIT_LINES <<< "$INPUT_COMMIT_MESSAGE" # split commit message at newlines

	if [[ ${#INPUT_COMMIT_LINES[0]} -ge 50 ]]; then
		echo "* commit title too long (${#INPUT_COMMIT_LINES[0]})"
	fi

	if [[ ${#INPUT_COMMIT_LINES[*]} -gt 1 ]]; then
		if [[ ${#INPUT_COMMIT_LINES[1]} -gt 1 ]]; then # newline is included
			echo "* no blank line after commit title"
		fi
	fi

	LENIENT_COMMIT_REGEX="(\w*)((\!|)[[(](.*)[])]|)(!|)($COMMIT_CONVENTION_SEPARATOR|)( |)(.*?)"$'(\n.*|$)'
	if [[ "$INPUT_COMMIT_MESSAGE" =~ $LENIENT_COMMIT_REGEX ]]; then
		if [ -n "${BASH_REMATCH[3]}" ]; then
			echo "* breaking mark before scope (should be after)"
		fi
		if [[ -z "${BASH_REMATCH[6]}" ]]; then
			echo "* missing separator ($COMMIT_CONVENTION_SEPARATOR) after type and scope"
		fi
		if [[ -z "${BASH_REMATCH[7]}" ]]; then
			echo "* missing whitespace after separator"
		fi
		if [[ -z "${BASH_REMATCH[8]}" ]]; then
			echo "* missing description"
		fi
		# This must happen last (even if I'd want it first) because it overrides regex matches
		TYPE="${BASH_REMATCH[1]}"
		TEST="^($COMMIT_CONVENTION_TYPES)$"
		if [[ $TYPE =~ $TEST ]]; then :; else # TODO how to invert regex?
			echo "* unknown type '$TYPE', expected ($COMMIT_CONVENTION_TYPES)"
		fi
	else
		echo "* empty or unparsable commit"
	fi
fi

# Fail, making git abort this commit
exit 1