Warning: The awk that I use does not provide the asort() function; so all of my comments are just based on reading the man pages and visually inspecting your code; I was not able to test your code to verify my hypothesis.

I believe there are two problems with your program:

First, in your END clause, when you call split(a[i],b,"@") with the array a[] containing the elements:

Code:

a[line1] = line1|error_type_b@5; error_type_c@10; error_type_a@15
a[line2] = line2|error_type_a@1; error_type_f@3

the array b[] will either contain the elements:

Code:

line1|error_type_b
5; error_type_c
10; error_type_a
15

or the elements:

Code:

line2|error_type_a
1; error_type_f
3

I'm guessing that this is not what you wanted.

And, second, after you run asort(), the elements in b[] will be changed to a list of numbers like (1, 3, 4, 2) or (1, 2, 3), respectively. Then the for loop in the END clause is going to print:

Code:

line1 FS contents of a[b[1]]
line1 FS contents of a[b[2]]
line1 FS contents of a[b[3]]
line1 FS contents of a[b[4]]
line2 FS contents of a[b[1]]
line2 FS contents of a[b[2]]
line2 FS contents of a[b[3]]

Obviously, the FS in the above lines will be replaced by | and the standards leave it is unspecified whether the "line1" lines will come before or after the "line2" lines. But, since the only defined subscripts in the a[] array are "line1" and "line2", the a[b[j]] (AKA a[1], a[2], a[3], and sometimes a[4]) will all expand to empty strings and, therefore, the contents of a[digit] in all of the output lines will be empty.

I think this explains what went wrong, but your description didn't clearly explain what output you were trying to get, so I can't suggest a way to fix it. As a (maybe not too wild) guess, I would think you might want to replace:

Code:

    a[$1] = a[$1] (a[$1]?"; ":"")$2

with something like:

Code:

    c[a[$1]] = $2

and then sort c[a[i]] or some manipulation of it in the END clause instead of trying to split elements of a[] and sort the split elements.

Hope this helps.