When do I need `expr ......` and when not? $(($aaa - $bbb)) simplyfiable?
Wolfgang Meister
calculate mathematical expressions. When I do for example the following
firstsec=345
secondsec=723
diffsec=$((secondsec - firstsec))
echo diff=$diffsec
then it works although I did not used "expr" for calculate the difference.
So when do I have to use expr and when not?
By the way: When can I write just
diffsec=$(secondsec - firstsec)
and when
diffsec=$((secondsec - firstsec))
and when
diffsec=$(($secondsec - $firstsec))
Wolfgang
Stephane CHAZELAS
> As I learned it is not always necessary to use `expr ....` (with backticks) to
> calculate mathematical expressions. When I do for example the following
>
> firstsec=345
> secondsec=723
> diffsec=$((secondsec - firstsec))
> echo diff=$diffsec
>
> then it works although I did not used "expr" for calculate the difference.
>
> So when do I have to use expr and when not?
You need expr in that case when writing scripts meant to be
portable to systems that don't have a POSIX shell. Mostly more
than 15 year old systems I'd say, or embedded systems that come
with a torn down version of a shell.
The Bourne shell and early versions of the Almquist shell didn't
have $((...)). In modern Unices, sh is no longer a Bourne shell
but an interpreter or the other of a standard sh language
specified by POSIX. And that language has $((...)).
The ":" operator of "expr" can still be uselful nowadays.
You may still want to use expr when dealing with 0 padded
decimal numbers. For $((...)), 010 is 8 (octal 010) while for
expr, it's 10.
--
Stéphane
Wayne
>
> The ":" operator of "expr" can still be uselful nowadays.
Indeed, expr can be used with if to test if some value is
a number, something that test alone can't do. And while
sed can be used to return a testable value when matching
some string against a regular expression, it is much
easier with expr.
>
> You may still want to use expr when dealing with 0 padded
> decimal numbers. For $((...)), 010 is 8 (octal 010) while for
> expr, it's 10.
Of course you can also easily strip off leading zeros from
numbers in shell variables in POSIX shells, using ${NUM##0}:
$(( ${aaa##0} - ${bbb##0} ))
Stephane, I still say you should write a book. Your
knowledge and experience of shells is invaluable.
-Wayne
Stephane CHAZELAS
[...]
>> The ":" operator of "expr" can still be uselful nowadays.
>
> Indeed, expr can be used with if to test if some value is
> a number, something that test alone can't do. And while
> sed can be used to return a testable value when matching
> some string against a regular expression, it is much
> easier with expr.
Yes, and sed operates on lines by default, while expr operates
on a string that may contain several lines.
expr can also be used for string comparison.
expr "x$string1" "<" "x$string2"
which the POSIX "[" can't do.
>> You may still want to use expr when dealing with 0 padded
>> decimal numbers. For $((...)), 010 is 8 (octal 010) while for
>> expr, it's 10.
>
> Of course you can also easily strip off leading zeros from
> numbers in shell variables in POSIX shells, using ${NUM##0}:
> $(( ${aaa##0} - ${bbb##0} ))
[...]
The problem with that is that it strips only one 0, and for the
number "0" itself, it converts it to an empty string.
${aaa#"${aaa%%[!0]*}"}
removes every leading 0, but doesn't solve the other problem and
becomes hard to read. You'd probably want to use a function to
do the 0 stripping.
strip_zeros() {
eval "$1=\${$1#\"\${$1%%[!0]*}\"}; $1=\${$1:-0}"
}
strip_zeros aaa
for instance.
--
Stéphane
Jignesh
> Stephane CHAZELAS wrote:
>
> > The ":" operator of "expr" can still be uselful nowadays.
>
> Indeed, expr can be used with if to test if some value is
> Stephane, I still say you should write a book. Your
> knowledge and experience of shells is invaluable.
>
> -Wayne
Yes, Stephane, You should :). I have read your tips on 'Advanced Bash
scripting Guide' at TLDP.
Jignesh