How to assign a zero-length, non-null string

Question

Based on this, amongst many other things I've read, my understanding is that a='' makes a both null and zero-length. But, then, how does one create a zero-length, non-null string? Or, is there no such thing? To the extent this is shell-dependent, I'm working in bash.

@JeffSchaller It sounds like your answer to the second half of my question, "Or, is there no such thing?", is "yes". — TTT, Jun 12 '19 at 17:26
I just wasn't sure if you intended something else, like "set" vs "unset", because it sounded like a contradiction to me. — Jeff Schaller, Jun 12 '19 at 17:33
It seems to me that the answers to the question you linked to would answer this one; do you agree, or what part is left over? e.g. `a null string in that context is a string of length 0 containing no byte at all` — Jeff Schaller, Jun 12 '19 at 17:36
Frankly, just being able to be certain whether, in the context of a set variable, "zero-length", "empty", and "null" are the same thing is what I've been struggling with. — TTT, Jun 12 '19 at 17:50

Kusalananda · Accepted Answer · 2019-06-12T18:00:42.263

"A null string" means "a zero length (empty) string". See e.g. the POSIX definition of "null string". This means that there is no such thing as a non-null zero length string.

However, there is such a thing as an unset variable.

var=''
unset uvar

There's now a difference between var and uvar after running the above code. For example, ${var-hello} would expand to the empty string since var is set, while ${uvar-hello} would expand to hello since uvar is unset. Likewise, ${var+hello} would expand to hello since var is set, and ${uvar+hello} would expand to an empty string since uvar is unset (see standard parameter expansions)

In bash, you can also use the -v test to test whether a variable is set or not:

if [ -v variable ]; then
    echo variable is set
fi

Again, a variable being "set but empty" is different from a variable being "unset". A string (contents of a variable) can't be null and at the same have non-zero length.

In other languages, an array of characters may contain nul bytes (\0), which means that you may have an array starting off with a nul byte, and then containing some text after that (terminated with another nul byte). When that is interpreted as a string, that string would have zero length, but the contents of the array would not be empty.

Most shells (apart from zsh) does not allow nul bytes in variables though.

You switch from "null" to "nul", are they different? Most shells don't allow nul(l) bytes **anywhere** in variables, or just at the start? — TTT, Jun 12 '19 at 17:55
@TTT "nul byte" means a `\0` character. "Null" means "empty". They are distinct. Most shells don't allow nul bytes anywhere in variables (as it is what _terminates the internal representation of a string_). — Kusalananda, Jun 12 '19 at 17:57

How to assign a zero-length, non-null string

1 Answers1