Upload client-side assets to S3 via AWS CLI

upload-to-s3.sh

#!/bin/bash

SOURCE_FILES=($(find public/*.{js,css,gz,br,map} 2>/dev/null))

DESTINATION_BUCKET="hashed-assets-eu"

DESTINATION_FOLDER="page-kit"

for FILE in ${SOURCE_FILES[@]}; do
	if [[ $FILE == *".gz" ]]; then
		ENCODING="gzip"
	elif [[ $FILE == *".br" ]]; then
		ENCODING="br"
	fi

	# the AWS CLI can guess content types but not the original type of compressed files
	# <https://github.com/aws/aws-cli/issues/3817>
	case $FILE in
		*".js"|*".js.gz"|*".js.br")
			TYPE="application/javascript"
			;;
		*".css"|*".css.gz"|*".css.br")
			TYPE="text/css"
			;;
		*".map")
			TYPE="application/octet-stream"
			;;
	esac

	aws s3 cp $FILE "s3://$DESTINATION_BUCKET/$DESTINATION_FOLDER/$FILE" \
		--cache-control=31536000 \
		--content-type="$TYPE; charset=utf-8" \
		--content-encoding="$ENCODING" \
		--acl="public-read"
done

Question - with variables like $FILE - is it a thing that's it's safer for them to be wrapped in ""? I vaguely remember being told this was 'safer' and Bats throwing an error if not "$VAR" but can't find a convincing explanation as to why in this particular case. https://stackoverflow.com/questions/10067266/when-to-wrap-quotes-around-a-shell-variable

if it's not quoted, and the filename contains spaces, then bash will expand it to multiple arguments to the command you're passing the variable to, which we probably don't want

i've forked the gist to add appropriate quoting and wrapped things in functions for proper local variable scoping https://gist.github.com/quarterto/4e8bbb7391455d15dbe40be74150dcd3