The format of the report should exactly match the format of the sample
report. Since this script is to be standalone, it might be easier as a
shell script that can pipe the results of "last" into awk, rather than
a pure awk script.
It is important to note that the output of ttylast does not contain
multiple uses of a tty by a user -- only the most recent. This makes
your life much easier.
In order to keep the assignment fun, you are allowed to use only the
awk (awk, nawk or gawk) and last commands -- well, and, of course, a
shell.
First, begin by looking at the output from my program. Then, take a look
at the output of last. Make sure you understand the name of the game.
To help you out, I included a copy of last's output captured from
within my script, as well as my script's output.
There are many ways of solving this problem. As a hint, I'll tell you
how I did it. As I worked my way through last's output one line at a time,
I kept three arrays:
- one of users (by user)
- one of ttys (by tty)
- one of the "whole lines" indexed by their "userid,tty" pair
The first two arrays amount to nothing more than a list of users and a list
of ttys, respectively. In reality, I really only cared about the keys, not
the values.
The next array gave me the mapping. By building the "userid,tty" key", I
gave myself an easy way to solve the problem. Essentially, I could find
any pair. Invalid pairs would be the empty string "". So,
if I'd look up "gkesden, pts/8" and gkesden didn't use pts/8, the result
would be "".
So, to produce the output, I used two nested loops. The outer loop walked
through the list of ttys. The inner loop walked through the lists of users.
The code inside printed out the line of last mapped to "user,tty", well,
at least unless that lookup yielded the empty string ("").
The last detail was the annoying extraneous line in last's output:
wtmp begins Tue Sep 12 04:06:14 2006
To avoid this, I filtered by "action" so that it wouldn't happen
if the pattern contained a "wtmp" in the first field.
One detail that I haven't given you: How to keep the most recent,
rather than the least recent, use. Think about this one a bit. You'll
likely want to reverse the output of last. This can be done with
sed. Or, as an alternative, you can use nl to number
the lines and then sort in reverse, and then whack the line numbers
using, for example, awk or cut.