Getting the total count of IDs from a comma delimited list of IDs that can contain ranges with awk

I'm trying to get the total count of CPUs allocated to a job from SLURM's scontrol --details --oneliner show job output. The format is a comma delimited list of CPU IDs that can contain hyphen-delimited ranges.

Here's an extract of my code, which doesn't work (I get 40 instead of 22):

echo 0,2-11,48-58 |

awk '
    {
        cpus = 0;
        m = split($0, arr, ",");
        for (j = 1; j <= m; j++)
            cpus += ( idx = index(arr[j], "-") \
                 ? 1 + substr(arr[j], idx+1) - substr(arr[j], 1, idx-1) \
                 : 1 \
            );
        print cpus
    }
'

40

What's happening ? I'm using GNU Awk 4.2.1, API: 2.0 (GNU MPFR 3.1.6-p2, GNU MP 6.1.2)

Instead of the intended:

( idx = index(arr[j], "-")) ? … : 1; # compute idx, then if it is > 0 do …

your parenthesis are:

( idx = index(arr[j], "-") ? … : 1);

which is equivalent to:

idx = (index(arr[j], "-") ? … : 1); # set idx to "if index() > 0 …"

Thus when using idx in the … it is not defined yet (thus 0),
or, on the later iterations, equal to the integer extracted on the previous iteration.

Your full script with parenthesis at the right place:

echo 0,2-11,48-58 |

awk '
    {
        cpus = 0;
        m = split($0, arr, ",");
        for (j = 1; j <= m; j++)
            cpus += ( idx = index(arr[j], "-")) \
                 ? 1 + substr(arr[j], idx+1) - substr(arr[j], 1, idx-1) \
                 : 1 \
            ;
        print cpus
    }
'

(… I understood the problem while adding an idx = …; print …; to diagnose before your ? :, and was surprised to get the intended 22)