Reason for ksh obsoleting -eq
The latest version of ksh obsoletes using -eq within [[ ]] blocks, prefering (( )) instead. Why is this? I can’t find any documentation on the advantages of (( )) over [[ ]] anywhere, and I find that syntax considerably more obvious.
For example:
#!/bin/ksh
var=1
if [[ $var -eq 1 ]]
then
echo ALPHA
fi
gives
[adam@hendrix test]$ ksh -n test.sh
test.sh: warning: line 3: -eq within [[...]] obsolete, use ((...))
If you do
$ ksh -n 'if [[ 1 -eq 1 ]]; then echo hi; fi'
you get the message
ksh: warning: line 1: -eq within [[...]] obsolete, use ((...))
as you’ve seen. Now try this:
$ ksh -n 'if (( 1 -eq 1 )); then echo hi; fi'
ksh: 1 -eq 1 : arithmetic syntax error
This works:
$ ksh -n 'if (( 1 == 1 )); then echo hi; fi'
Remember that the first message is only a warning. You can continue to use that form. I doubt that it will be removed since it would break too many existing scripts. By the way, this is accepted without a warning:
$ ksh -n 'if [ 1 -eq 1 ]; then echo hi; fi'
One of the main reasons that double parentheses is preferred is that the comparison operators are the more familiar <, <=, ==, etc., (at least when compared to other languages).
Double parentheses also work in Bash and zsh. A related form, arithmetic substition, works in all of them, plus it’s specified by POSIX.
$ a=$((3 * 4))
Korn, Bash and Z can also do it this way:
$ (( a = 3 * 4 ))
Even though Dash, as an example of a POSIX shell, doesn’t support double parentheses comparisons in the form if (( ... )), you can still do them using arithmetic substitution, but the result is the opposite from what you’d expect (this is also true for the others).
$ echo $(( 1 == 1 )); (( 1 == 1 )); echo $? # in Dash, use true instead of the second statement
1
0
$ echo $(( 1 == 2 )); (( 1 == 2 )); echo $? # in Dash, use false instead of the second statement
0
1