Passing arguments by reference

It's 2018, and this question deserves an update. At least in Bash, as of Bash 4.3-alpha, you can use namerefs to pass function arguments by reference:

function boo() 
{
    local -n ref=$1
    ref='new' 
}

SOME_VAR='old'
echo $SOME_VAR # -> old
boo SOME_VAR
echo $SOME_VAR # -> new

The critical pieces here are:

• Passing the variable's name to boo, not its value: boo SOME_VAR, not boo $SOME_VAR.

• Inside the function, using local -n ref=$1 to declare a nameref to the variable named by $1, meaning it's not a reference to $1 itself, but rather to a variable whose name $1 holds, i.e. SOME_VAR in our case. The value on the right-hand side should just be a string naming an existing variable: it doesn't matter how you get the string, so things like local -n ref="my_var" or local -n ref=$(get_var_name) would work too. declare can also replace local in contexts that allow/require that. See chapter on Shell Parameters in Bash Reference Manual for more information.

The advantage of this approach is (arguably) better readability and, most importantly, avoiding eval, whose security pitfalls are many and well-documented.

Use a helper function upvar:

# Assign variable one scope above the caller.
# Usage: local "$1" && upvar $1 value [value ...]
# Param: $1  Variable name to assign value to
# Param: $*  Value(s) to assign.  If multiple values, an array is
#            assigned, otherwise a single value is assigned.
# NOTE: For assigning multiple variables, use 'upvars'.  Do NOT
#       use multiple 'upvar' calls, since one 'upvar' call might
#       reassign a variable to be used by another 'upvar' call.
# See: http://fvue.nl/wiki/Bash:_Passing_variables_by_reference
upvar() {
    if unset -v "$1"; then           # Unset & validate varname
        if (( $# == 2 )); then
            eval $1=\"\$2\"          # Return single value
        else
            eval $1=\(\"\${@:2}\"\)  # Return array
         fi
    fi
}

And use it like this from within Newfun():

local "$1" && upvar $1 new

For returning multiple variables, use another helper function upvars. This allows passing multiple variables within one call, thus avoiding possible conflicts if one upvar call changes a variable used in another subsequent upvar call.

See: http://www.fvue.nl/wiki/Bash:_Passing_variables_by_reference for helper function upvars and more information.

The problem with:

eval $1=new

is that it's not safe if $1 happens to contain a command:

set -- 'ls /;true'
eval $1=new  # Oops

It would be better to use printf -v:

printf -v "$1" %s new

But printf -v cannot assign arrays.

Moreover, both eval and printf won't work if the variable happens to be declared local:

g() { local b; eval $1=bar; }  # WRONG
g b                            # Conflicts with `local b'
echo $b                        # b is empty unexpected

The conflict stays there even if local b is unset:

g() { local b; unset b; eval $1=bar; }  # WRONG
g b                                     # Still conflicts with `local b'
echo $b                                 # b is empty unexpected

From the Bash man-page (Parameter Expansion):

    If the first  character of parameter is an exclamation  point (!), a
    level of variable indirection is  introduced. Bash uses the value of
    the variable  formed from the rest  of parameter as the  name of the
    variable; this variable  is then expanded and that value  is used in
    the rest  of the  substitution, rather than  the value  of parameter
    itself. This is known as indirect expansion.

Therefore a reference is the variable's name. Here is a swap function using variable indirection that does not require a temporary variable:

function swap()
{   # 
    # @param VARNAME1 VARNAME2
    #
    eval "$1=${!2} $2=${!1}"
}

$ a=1 b=2
$ swap a b
$ echo $a $b
2 1