sed -> awk buffers ...

[A. Daniel King_1]

Alright shell gurus, here is an easy 10 points.

sed "s/^\([^:]*\):\([^:]*\)/\2:\1/" /tmp/somefile.txt

This sed snippet takes a colon delimited file and switches the first and second field. How do you do the same thing in awk?

[Pete Randall]

cat /tmp/somefile.txt | awk '{ print $2, $1 }'


Pete

[A. Daniel King_1]

So close ... "colon delimited" ...

How about sub-field matches? I'm really interested in the buffering \( \) in sed ... how do you do the same thing in awk?

[Pete Randall]

cat /tmp/somefile.txt | awk -F: '{ print $2, $1 }'

[Hein van den Heuvel · Accepted solution]

Do you know how many fields?
Then you could use:

awk -F: 'BEGIN {OFS=":"}{print $2,$1,$3,$4,$5}' /tmp/somefile.txt

for a variable field count, one suggestion is:

awk -F: '{print $2 ":" $1 substr ($0,match($0,":" $3))}' /tmp/somefile.txt


Hein.

[Thierry Poels_1]

awk -F':' .....

regards,
Thierry

[TwoProc]

Assuming there's only two fields...
cat /tmp/somefile.txt | cut -f 1,2 -d : | awk '{print $2," ",$1 }'

[Hein van den Heuvel]

Ah... the obligatory perl alternatives....

like SED:

perl -pe 's/(.*):(.*):/$2:$1:/' x


Split words in an array, swap, join and have printed:

perl -pe '@words=split(":"); $x=$words[0];$words[0]=$words[1];$words[1]=$x; $_=join(":",@words)' x


Split, join in new order, and print.
The final @words is in scalar context and is the number of elements of words.

perl -pe '@words=split(":"); $_=join(":",@words[1,0,2..@words])' x


grins,
Hein.

[Elmar P. Kolkman]

If you don't know the number of columns on each line, do it this way:

awk -F: '{printf "%s:%s",$2,$1;
for (i=3;i<=NF;i++)
printf ":%s",$i
printf "\n";
}' < /tmp/somefile.txt


You could also use split on $0 with a ":" seperator and swap the first two array entries, for instance. But the resulting could would be bigger as the above one, since I don't know of a join function in awk, like in perl.

Good luck,

Elmar

[Hein van den Heuvel]

Argh, I hate to reply to myself, but I hate it even more to leave something that is wrong.

My perl example for the SED look alike is wrong.

I forgot about the greedy-ness of the .* match. So you can not say 'anything' because that may include colons. You have to say 'any number of non-colon chars'

perl -pe 's/([^:]*):([^:]*):/$2:$1:/' x

Hein.

[A. Daniel King_1]

I needed the delimiter to be a colon when done, too! The substr item is exactly what I was looking for, though I was hoping for something as elegant as the sed/perl buffer.

Thanks, all.