How to run a program and parse the output using KornShell - shell

Question

I am pretty new to KornShell (ksh). I have a project that needs to be done with ksh. The question is:
Please write a ksh script which will run the ‘bonnie’ benchmark
utility and parse the output to grab the values for block write, block
read and random seeks/s. Also consider how you might use these values
to compare to the results from previous tests. For the purpose of
this test, please limit yourself to standard GNU utilities (sed, awk,
grep, cut, etc.).
Here is the output from the ‘bonnie’ utility:
# bonnie -s 50M -d /tmp
File '/tmp/Bonnie.2001096837', size: 52428800
Writing with putc()...done
Rewriting...done
Writing intelligently...done
Reading with getc()...done
Reading intelligently...done
Seeker 1.S.e.eker 2.S.e.eker 3...start 'em...done...done...done...
-------Sequential Output-------- ---Sequential Input-- --Random--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
50.0 36112 34.1 138026 1.9 179048 7.0 51361 51.1 312242 4.3 15211.4 10.3
Any suggestion of how to write this script would be really appreciate.
Thanks for reading!

Answer 1

Here's a simple solution to experiment with, that assumes the last line will always contain the data you want:
# -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
# Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
# 50.0 36112 34.1 138026 1.9 179048 7.0 51361 51.1 312242 4.3 15211.4 10.3
# block write, block read and random seeks/s
bonnie++ \
| awk '
{line=$0}
END{
# print "#dbg:last_line=" $line
split(line, lineArr)
printf ("blkWrt=%s\tblkRd=%s\tRandSks=%s\n", lineArr[4], lineArr[8], lineArr[12])
}' # > bonnieOutput
# ------^^ remove # to write output to file
(Note that the \ char after bonnie++ must be the last character on the line, NO SPACES OR TABS allowed!!! (It will blow up otherwise!) ;-) )
Awk reads all lines of input passed thru a pipe. When you're in the END{} block of awk, you put the last line read into the lineArr[], and then print out just the elements you want from that line, using the index number of the field in your data, so lineArr[4] is going to return the 4th field in that last line of data, lineArr[12], the 12th, etc. You may have to adjust what index number you use to get the data you want to display. (You'll have to figure that out! ;-)
To save the data to a file, use the shell redirection by uncommenting (removing the # char between }' and > bonnieOutput. Leave the # char in place until you get the output you need, THEN you can redirect it to a file.
Needless to say the labels that I've used in the printf like blkWrt= are mostly for debugging. Once you are sure about what data you need to capture, and that it reliably appears in the same location each time, then you can remove those labels and then you'll have a nice clean datafile that can process with other programs.
Keep in mind that almost all Unix toolbox utilities are line oriented, that is they expect to process 1 line of data at a time and there are often tricks to see what is being processed. Note the #dbg line I've included at the top of the END{} block. You'll have to remove the '#' to uncomment it to see the debug output.
There's a lot more than can be done, but if you want to learn ksh/unix toolbox with awk, you'll have to spend the time understanding what the features are. If you've read the chapter that included the question you're working with and don't understand how to even start solving this problem, maybe you better read the chapter again, OK? ;-)
Edit
Note that in awk, the variable $0 contains all text in the current line (as defined by the RS variable value, usually the Unix line ending char, \n). Other numbered values, i.e. $1, $2, indicate the first, or second "field" on the current line ($0).
Based on my new understand from you comment below, you want to extract values from lines that contain the text "Latency". This is even easier to process. The basic pattern will be
bonnie++ \
| awk '
/Latency/{
# print "#dbg:latency_line=" $0
printf ("blkWrt=%s\tblkRd=%s\tRandSks=%s\n", $4, $8, $12)
}' # > bonnieOutput
So this code says, read all output from bonnie++ into awk, through the pipe, and when you find a line containing the text "Latency", print the values found in the 4th, 8th, and 12th fields, using the printf format string that contains self-describing tags like blkWrt, etc.
You'll have to change the $4, etc to correctly match the number in the current line for each element of data. I.E. maybe it $5, $9, $13, or $3, $9, $24? OK?
Note that /Latency/ is case sensitive, and if there are other places in the output where the word appears, then we'll have to revise the reg-exp "rule" used to filter the output.
As a learning exercise, and as a very basic tool that any Unix person uses every day, skip awk, and just see what bonnie++| grep 'Latency' gets you.
IHTH

Answer 2

Just got the answer by the help from Shellter!
bonnie++\
| awk '/Machine/ {f=1;next}f{
print "#dbg: line_needed=" $0
printf("blkWrt=%s\t blkRd=%s\t RandSks=%s\n", $4, $8, $12);exit
}'