If Bash arrays can't contain a null string, how does this array do it?

Question

I've recently discovered that you can't pass a null string to a Bash array. This is tricky, as I'm trying to insert an empty string value into an array.

I've noticed that the Bash completion variable COMP_WORDS, however, can contain an empty value.

I've written a dummy completion script to illustrate this, with the debugging options on:

  1 #!/bin/bash
  2 _comp() {
  3     set -xv
  4     for i in ${!COMP_WORDS[@]}; do
  5         echo "<${COMP_WORDS[$i]}>"                                                                                         
  6     done
  7     set +xv
  8     local cur=${COMP_WORDS[$COMP_CWORD]}
  9 
 10     local arr=(apple banana mango pineapple)
 11     COMPREPLY=()
 12     COMPREPLY=($(compgen -W "${arr[*]}" -- $cur))
 13 
 14 }
 15 
 16 complete -F _comp dummy

If I source this file, and then enter the following:

dummy apple   banana

And then if I move the cursor back to the middle space between apple and banana (indicated below by an underscore) ...

dummy apple _ banana

And then if I hit tab, I get the following output from the resulting debug, showing four array elements, 'dummy' 'apple' '' 'banana'

for i in '${!COMP_WORDS[@]}'
+ echo '<dummy>'
<dummy>
+ for i in '${!COMP_WORDS[@]}'
+ echo '<apple>'
<apple>
+ for i in '${!COMP_WORDS[@]}'
+ echo '<>'
<>
+ for i in '${!COMP_WORDS[@]}'
+ echo '<banana>'
<banana>

So the COMP_WORDS array is clearly capable of storing a null string of some kind. Although interestingly, this does not happen if I use a regular for loop, accessing the array elements rather than the indices. I.e., if I change the for loop in the debug section to this ...

for i in ${COMP_WORDS[@]}; do
    echo "<$i>"
done 

Then I get the following debug output showing only three array elements: 'dummy' 'apple' 'banana'

for i in '${COMP_WORDS[@]}'
+ echo '<dummy>'
<dummy>
+ for i in '${COMP_WORDS[@]}'
+ echo '<apple>'
<apple>
+ for i in '${COMP_WORDS[@]}'
+ echo '<banana>'
<banana>

And I've tried echoing the length of the array with ${#COMP_WORDS[@]} - for this input, it shows a length of four. So it clearly does contain four elements, one being a null string.

This seems like really strange behaviour to me. My question is twofold:

1. What is the explanation for the COMP_WORDS array being able to store null strings when normal arrays apparently cannot?

2. Why am I unable to see the null string when I use a regular for loop, but I can see it when I access the elements of the array by their indices?

Answer 1

You seem to be confusing the empty string "" with the NULL character \0. They are not the same. You can use empty strings in arrays and variables:

$ a=(aa "" "b")
$ echo "${#a[@]}"
3
$ printf -- '- %s\n' "${a[@]}"
- aa
- 
- b
$ for i in "${a[@]}"; do
> printf '%s\n' "$i"
> done
aa

b

Some of the conditional expressions in Bash even allow testing whether a string is empty:

-z string

True if the length of string is zero.

-n string string

True if the length of string is non-zero.

See also When is double-quoting necessary?