Question

The example code is like this, it does the statistic of number of times the first column appears and sort the result.

    { dist[$1]+=1; }
END { for (i in dist) {
        print i,dist[i] | "sort"
      }
    }

In my opinion, the process is like this:

(WORKFLOW A)

1) print all element in the dist, save them all to a buffer

2) get all elements in the buffer, and pipe them to sort function

But in the example above, the process looks like this:

(WORKFLOW B)

1) print one element in the dist, and then pipe it to sort function

2) process the next element in the dist, until no new element in dist

I was wondering why I shouldn’t place the sort like this:

    { dist[$1]+=1; }
END { for (i in dist) {
        print i,dist[i] 
      } 
      | "sort"
    }

Anyone has any idea about the reason? And how can I write the pipe if I want to do the work like WORKFLOW B?

Thanks!

Answer 1

The reason you can’t do it the second way is because | "command" is part of the syntax of awk’s print command, it can’t be used with arbitrary statements or statement groups. The same thing goes for > filename.

The way it works is that the first time it encounters a redirection to a file or pipe, it opens that file/pipe, and keeps that descriptor open. Then every time you redirect to the same file/pipe, it sends the output to the corresponding descriptor.