string check in bash with -z and -n
I made a mistake in test syntax in bash but now I want to understand what really happens in string check with -n and -z. I wrote the following lines to get in LINENUM variable the line number from a grep. When the string is not found (there is only one in the file, for sure), the LINENUM variable is empty.
$ LINENUM=$(grep -w -n mystring myfile | cut -d: -f1) $ echo --$LINENUM-- ---- $ if [ -n $LINENUM ] ; then echo "Checked -n"; fi Checked -n $ if [ -z $LINENUM ] ; then echo "Checked -z"; fi Checked -z
Then I realized I forgot the double quotes and then the following check gave to me:
$ if [ -n "$LINENUM" ] ; then echo "Checked -n"; fi $ if [ -z "$LINENUM" ] ; then echo "Checked -z"; fi Checked -z
So, in the former tests, where I forgot the double quotes, versus what did the if test check , really, since I got two positive checks from both -n and -z ?
Without the quotes, your test statement (with either operator, represented with -X here), reduces to
if [ -X ]; then echo "Checked -X"; fi
According to the POSIX standard, the one-argument form of test (which you now have here) is true if the argument is non-null. Since the literal string -X is non-null (it's not an operator anymore), it evaluates to true.
With the quotes, you get
if [ -X "" ]; then echo "Checked -X"; fi
Since the quotes force an empty 2nd argument, you have a 2-argument form of test, and -X (whether -n or -z) is properly recognized as a primary operator acting on the 2nd, null, argument.