As Ciro Santilli notes using command substitutions will drop trailing newlines. Their workaround adding trailing characters is great, but after using it for quite some time I decided I needed a solution that didn't use command substitution at all.
My approachMy approach now uses read along with the printf builtin's -v flag in order to read the contents of stdin directly into a variable.
# Reads stdin into a variable, accounting for trailing newlines. Avoids needing a subshell or
# command substitution.
read_input() {
# Use unusual variable names to avoid colliding with a variable name
# the user might pass in (notably "contents")
: "${1:?Must provide a variable to read into}"
if [[ "$1" == '_line' || "$1" == '_contents' ]]; then
echo "Cannot store contents to $1, use a different name." >&2
return 1
fi
local _line _contents
while read -r _line; do
_contents="${_contents}${_line}"$'\n'
done
_contents="${_contents}${_line}" # capture any content after the last newline
printf -v "$1" '%s' "$_contents"
}
# Reads stdin into a variable, accounting for trailing newlines. Avoids
# needing a subshell or command substitution.
# Note that NUL bytes are still unsupported, as Bash variables don't allow NULs.
# See https://stackoverflow.com/a/22607352/113632
read_input() {
# Use unusual variable names to avoid colliding with a variable name
# the user might pass in (notably "contents")
: "${1:?Must provide a variable to read into}"
if [[ "$1" == '_line' || "$1" == '_contents' ]]; then
echo "Cannot store contents to $1, use a different name." >&2
return 1
fi
local _line _contents=()
while IFS='' read -r _line; do
_contents+=("$_line"$'\n')
done
# include $_line once more to capture any content after the last newline
printf -v "$1" '%s' "${_contents[@]}" "$_line"
}
This supports inputs with or without trailing newlines.
Example usage:
$ read_input file_contents < /tmp/file
# $file_contents now contains the contents of /tmp/file