Usage of -n and -z in test built in - Bash

Question

Consider the following in bash:

root@debian-lap:/tmp I=$(echo)
root@debian-lap:/tmp echo "$I"

root@debian-lap:/tmp [ -z "$I" ] && echo "TRUE" || echo "FALSE"
TRUE

This means that variable $I is zero. The same I could achieve with negation test to see if variable is non zero, and ! makes test reverse so it checks is variable zero

root@debian-lap:/tmp ! [ -n "$I" ] && echo "TRUE" || echo "FALSE"
TRUE
root@debian-lap:/tmp

So, my question is, are there any special cases when to use -z and ! -n , or vice versa ! -z and -n as they are basically doing the same test?

Thanks

Answer 1

1. [ -z "$var" ]
2. [ ! -n "$var" ]
3. ! [ -n "$var" ]
4. [ ! "$var" ]
5. [ "$var" = '' ]
6. [ ! "$var" != '' ]
7. [ '' = "$var" ]...²

Are all functionally equivalent in POSIX shells¹. The Bourne shell¹ didn't have the ! keyword, so wouldn't support 3 above and had all sorts of bugs with special values of $var (like =, -n, !...) in all but 7 (many other shells also had similar bugs).

In a similar vein, you can (and should) use [ "$a" = "$b" ] && [ "$x" = "$y" ] in place of [ "$a" = "$b" -a "$x" = "$y" ], the latter being deprecated and non-reliable.

The ! of 3 is a POSIX shell feature. It still makes sense of the test/[ utility to support its own ! operator (as in 2/6) for other shells or non-shell applications that may want to use it and negate conditions.

The test utility was introduced by Unix v7. Before that, there was a if utility that had !, =, !=, -a/-o, but no -n/-z.

I suppose -n/-z were added for convenience (test -n "$var" being a short form for test "$var" != '' and -z being added for consistency)

---

¹ Technically, while test/[ is built in virtually all Bourne-like shells, that's not mandated by POSIX and early versions of the Bourne shell didn't have it built-in. While the Almquist shell always had a test builtin, some BSDs did not enable it so some BSDs did not have a [/test built in sh until the 2000s. In any case, regardless of whether a particular shell has a [/test command built-in, all POSIX systems will also have a [/test command on the file system (which may or may not behave the same as the [/test builtin command of any shell on the system).

² That's not to say that's the only commands that can be used to test if a variable is the empty string.

case $var in
  "") ...
esac

is one way that doesn't involve running any command.

expr "x$var" : 'x$' > /dev/null
${var:+"false"}
awk 'BEGIN{exit(ARGV[1] != "")}' "$var"
[ "${#var}" -eq 0 ]

Are more examples of convoluted ways to do that.

Some Korn-like shells like bash also have some built-in conditional and arithmetic constructs that aim at replacing the [ command ([[ -z $var ]], ((${#var} == 0))...).

Answer 2

You are given -n and -z for the same reason that other test suites give you both == and !=, or AND and NOT. Some test cases can be made a lot clearer to future maintainers by eschewing double-negatives. Also, as mentioned in an above comment, ancient incarnations of sh (i. e. the Bourne and Thompson shells), as opposed to modern POSIX sh did not have a ! keyword to negate the truthiness of test expressions.