Shell Programming and Scripting

HI,
This one.

code:

Code:

awk 'BEGIN{FS="|"}
{
for (i=1;i<=NF;i++)
if($i!=" ")
printf("%s|",substr($i,index($i,":")+1))
printf("\n")
}' filename

You can use the same regex in awk:

Code:

> echo 'P1|10,9:12/123456789|,10:1|11,10:60|15,9:5/12345|,10:1|'|awk '{gsub(/\|[^|]*:/,"|",$0);print}'
P1|12/123456789|1|60|5/12345|1|

And another awk version:

Code:

awk '$1=$1' OFS= FS="[^|]*:"

Use nawk or /usr/xpg4/bin/awk on Solaris.

Or a bit modified version of the previous sed solution:

Code:

sed 's/[^|]*://g'

Very nice i didn't know OFS/FS could be a regex in awk.

Thx radoulov.

wow! so many answers, which one to choose?

just another question though...when I have the choice of solving something like this using awk or sed, which one should i choose and why? is on better than the other in terms of performance?

Yes,
and GNU Awk (and may be tawk?) supports regular expressions
even for RS.

ou might be interested in this post by me, which was written somewhat with tongue-in-cheek.

To cut a long story short: When you need to work context-oriented *) use awk, as doing this in sed is a nightmare. As long as speed is a consideration try to use sed as often as possible as it is much faster than awk. Something like

print - "$string" | awk '{print $4}'

is wasting a mighty tool on trivial problems - use sed and/or cut for this.

Still "wasting" is not always the same. Compare the above line to:

Code:

(( i = 0 ))
while [ $i -le 100000 ] ; do
     print - "$string" | awk '{print $4}'
     (( i += 1 ))
done

Same line, quite different effect as the line is not executed once but a 100000 times. The effect of changing from awk to sed/cut will probably be negligable in the first example but will probably be very big in the second one.

Or, other aspect:

Code:

cat file | while read string ; do
     print - "$string" | awk '{print $4}'
done

In this case it will depend on how big "file" is: the loop will get executed once for every line of the input file. If this is 5 lines long it won't matter, if it is a database dump with millions of lines it will make a big difference.

As a general consideration: try to learn both tools equally well. Both have their merits and avoid getting in a situation where "the only tool you have is a hammer and therefore everything starts looking like nail".

I hope this helps.

bakunin

*) "Context-oriented" means: problems like "get all numbers in a file totalling them" or "get the content of a certain line and apply changes to the other lines following it depending on this content" or something like that.