[SOLVED] Convert lines to columns

sharky · 06-28-2014, 10:09 PM

I am betting there is a simple awk one liner that will do what I need. However, I haven't found it yet.

What I need is very simple. I want to take a mult-line input and convert to columns for each nth line.

Example: say n is 3 and the input is;

Quote:

some - text
more - text
even - more
yet - again
for - ever
stop - now

What I want returned is;

Quote:

some - text yet - again
more - text for - ever
even - more stop - now

Nothing more complicated than that. No special cases. Just convert n number of lines into columns with each column seperated by a space.

I guess there is one additional complication; In the example above if the number of lines on the input were 9 instead of 6 I would get 3 columns returned instead of two.

It is OK if the n is hard coded. As a matter of fact it's preferable.

I'm still digging so if I find a solution I will be sure to post it.

Cheers,

BenCollver · 06-28-2014, 11:05 PM

It may not be elegant, but below is a working script. example: ./test.sh input.txt 3

Code:

#!/bin/sh
fn="$1"
n="$2"
if [ ! -f "$fn" -o -z "$n" -o "$n" -lt 2 ]
then
    echo "Usage: test.sh filename lines"
fi
for i in $(seq 1 $n)
do
    tail -n +$i "$fn" | awk "NR == 1 || NR % $n == 0" | tr '\n' ' '
    echo ""
done
exit 0

sharky · 06-29-2014, 12:33 AM

That was close but not quite.

I created a test file that was 27 lines. The test file simply list 1 - 9 three times.

Quote:

1
2
3
4
5
6
7
8
9
1
2
3
4
5
6
7
8
9
1
2
3
4
5
6
7
8
9

The output from your script was this;

Quote:

1 9 9 9
2 1 1
3 2 2
4 3 3
5 4 4
6 5 5
7 6 6
8 7 7
9 8 8

Not sure how to fix it but it's definitely close.

Thanks for the help.

grail · 06-29-2014, 01:54 AM

Well this looks a little clumsy but you get the idea:

Code:

awk '{a=NR%3?NR%3:3;b[a]=b[a](b[a]?" ":"")$0}END{for(i=1;i<=3;i++)print b[i]}' file

syg00 · 06-29-2014, 02:02 AM

Quote:

Originally Posted by grail

... but you get the idea:

Luv it - how to make awk look like perl ....

ntubski · 06-29-2014, 06:56 AM

Using pr:

Code:

pr -t -s' ' --columns=$(($(wc -l < input) / 3)) input

Unfortunately, it needs to get the input twice.

@grail: you can simplify a bit by using 0 based indexing:

Code:

awk '{a=(NR-1)%3;b[a]=b[a](b[a]?" ":"")$0}END{for(i=0;i<3;i++)print b[i]}' file

grail · 06-29-2014, 07:14 AM

Cheers ntubski

I have a cold and I knew there was a way but the head was not playing the game

BenCollver · 06-29-2014, 07:35 AM

Ah, the script was off by one.

OLD: NR % $n
NEW: (NR-1) % $n

Code:

#!/bin/sh
fn="$1"
n="$2"
if [ ! -f "$fn" -o -z "$n" -o "$n" -lt 2 ]
then
    echo "Usage: test.sh filename lines"
fi
for i in $(seq 1 $n)
do
    tail -n +$i "$fn" | awk "NR == 1 || (NR-1) % $n == 0" | tr '\n' ' '
    echo ""
done
exit 0

sharky · 06-29-2014, 12:44 PM

Several solutions that work!

Thanks to everyone.

danielbmartin · 06-29-2014, 07:36 PM

For your consideration: a different approach.

Advantage: simple code
Disadvantage: uses work files

With this InFile ...

Code:

some - text
more - text
even - more
yet - again
for - ever
stop - now

... this code ...

Code:

rm $Work*; split -dl3 $InFile $Work; paste $Work* >$OutFile

... produced this OutFile ...

Code:

some - text	yet - again
more - text	for - ever
even - more	stop - now

With this InFile ...

Code:

some - text
more - text
even - more
yet - again
for - ever
stop - notyet
apple - fruit
honda - car
purple - color

... the same code produced this OutFile ...

Code:

some - text	yet - again	apple - fruit
more - text	for - ever	honda - car
even - more	stop - notyet	purple - color

Daniel B. Martin

shruggy · 08-03-2021, 11:06 AM

Just for fun, building upon Daniel's idea

Code:

#/bin/bash
input=$(seq -w 12)

# split in three, then paste each part sequentially
echo == 1 ==
<<<"$input" split -nr/3 --filter='paste -sd\ '
echo == 2 ==
getrow(){<<<"$input" split -nr/$1/3;}
paste -sd' ' <(getrow 1) <(getrow 2) <(getrow 3)

# paste three in parallel, then transpose
xpos=$(<<<"$input" paste - - -)
echo == 3 ==
getcol(){<<<"$xpos" cut -f$1;}
paste -sd' ' <(getcol 1) <(getcol 2) <(getcol 3)
echo == 4 ==
for i in 1 2 3;do <<<"$xpos" cut -f$i|paste -sd' ';done
echo == 5 ==
csvtool -t TAB -u ' ' transpose - <<<"$xpos"
echo == 6 ==
csvtk -tD' ' transpose <<<"$xpos"

The tools used in the last two solutions are csvtool and csvtk.

syg00 · 08-03-2021, 04:44 PM

I wonder if anyone (else) still cares all these years later ...

rnturn · 08-03-2021, 11:35 PM

Quote:

Originally Posted by syg00

Luv it - how to make awk look like perl ....

IMHO, only bonkers Perl programmers write code like that.

(I'm working on a project that uses a large number of Perl modules -- surely totaling well over 10,000 lines of code -- and none of what I've seen of the code base looks like that.)