expr command help

[piyush mathiya]

Gurus,
I am very new in shell scripting, i am going to create a script which is calculating the average of io utilization. in that script i want to get the average of floating values like.
5.62
2.75
3.41
1.08
2.41
2.69
----

that means total of all valuses / no. of values. i had used SUM=`expr $SUM + $VALUE`, it shows the following error.

expr: Syntax error
expr: An integer value was expected.

what should I do to get the average of floationg value.

Thanks
Piyush Mathiya

[Yogeeraj_1]

hi Piyush Mathiya,

A much easier method is using awk:

$ cat sum.awk
#sum.awk
{ sum+=$COL}
END { print sum }

$ cat values.txt
5.62
2.75
3.41
1.08
2.41
2.69

$ awk -f sum.awk COL=1 < values.txt
17.96
$

hope this helps!
kind regards
yogeeraj

[piyush mathiya]

Thanks Yogeeraj,
I got your concept but i am not geeting how can i write this statment in a script as i had cut that filed from another file. it is do .. while loop and the floating value is stored in "T" variable.

Thanks
Piyush Mathiya

[Yogeeraj_1]

hi Piyush,

please clarify further. Maybe you can post an extract of the script in question.

thanks

kind regards
yogeeraj

[piyush mathiya]

Yogeeraj,
What I am doing is.. I have four fields in a file called p1.txt, i have cut 3rd field from the file and i want to calculate the average of that column in a variable "sum".
sum=0
cat p1.txt | grep -v "%" | while read line1
do
T=`echo $line1 | tr -s " " | cut -f 3 -d " "`
sum=`expr $sum + $T`
echo $sum
done

where the "T" is floating value.

[Dennis Handly]

>i had cut that field from another file. it is do .. while loop and the floating value is stored in "T" variable.

You need to change it to have awk directly read your file and extract that field and average it. Or take your while loop and create a new file or pipe that you pass to awk (which is probably simpler):

while read stuff; do
# some cut stuff
echo $T
done < first-file | awk -v COL=1 '
{ sum+=$COL }
END { print sum / NR } '

[Dennis Handly · Accepted solution]

> I have four fields in a file called p1.txt, i have cut 3rd field from the file and i want to calculate the average of that column in a variable "sum".

Do it all in awk.
awk -v COL=3 '
BEGIN { getline } # skip first line
{ sum+=$COL }
END { print sum / NR } ' p1.txt

[Yogeeraj_1]

hi again,

I think Dennis "Do it all in awk." would be simpler to implement in your case.

Thank you Dennis for the detailed input.

Good luck!
kind regards
yogeeraj

[piyush mathiya]

Denis & Yogeeraj.
Thanks for your prompt and realy very helpful answers. Still I want a help. anyone have a good document (Examples) of awk & sed commands. so i can read and learn about this two commands.
Problem is resolved. again thanks.!
Regards,
Piyush Mathiya

[Yogeeraj_1]

hi Piyush,

Start with "man awk" and "man sed"

Also have a look at the following URLs:

http://www.grymoire.com/Unix/Sed.html
http://www.grymoire.com/Unix/Awk.html

hope this helps!

kind regards
yogeeraj

[Dennis Handly]

>anyone have a good document (Examples) of awk & sed commands. so i can read and learn about this two commands.

I typically just use the man page. But a quick google search finds for awk:
http://www.ibm.com/developerworks/library/l-awk1.html
http://www.gnulamp.com/awkexamples.html
http://www.linuxfocus.org/English/September1999/article103.html
http://www.softpanorama.org/Tools/awk.shtml
http://www.vectorsite.net/tsawk_3.html
http://sparky.rice.edu/~hartigan/awk.html
http://www.cs.hmc.edu/qref/awk.html
http://www.gentoo.org/doc/en/articles/l-awk2.xml
http://www.freeos.com/guides/lsst/ch07sec09.html

And the same for sed:
http://student.northpark.edu/pemente/sed/sed1line.txt
http://www.ibm.com/developerworks/linux/library/l-sed1.html
http://www.ibm.com/developerworks/linux/library/l-sed2.html
http://www.grymoire.com/Unix/Sed.html
http://www.gnulamp.com/sed.html
http://www.cs.utk.edu/~vose/c-stuff/sed_tutorial.html
http://people.debian.org/~caiqian/programming/sed/l-sed1.html
http://www.gentoo.org/doc/en/articles/l-sed1.xml
http://www.freeos.com/guides/lsst/ch07sec14.html
http://www.emunix.emich.edu/~khailany/files/sed-examples.htm

There is probably some duplication in the links.

[piyush mathiya]

Thanks,

[James R. Ferguson]

Hi:

Most shells only support integer arithmetic as you have found. While I would probably also use 'awk' (or Perl), the Korn93 shell supports floating point arithmetic.

In HP-UX the Korn shell in '/usr/bin/ksh' is the Korn88 variant and this lacks the support you seek.

However, the Korn93 shell lives in '/usr/dt/bin/dtksh'. For example:

# cat ./float
#/usr/dt/bin/dtksh
typeset -F R1
typeset -F3 R2
typeset -E R3
let R1=1/8
echo $R1
let R2=1/8
echo $R2
let R3=1/8
echo $R3

# ./float
.1250000000
.125
0.125

A google search for 'Korn shell' will yield more information.

Regards!

...JRF...