Multi tool use

How to sort or rearrange numbers from multiple column into multiple row [fixed into 4 columns]?

I have 1 text file, which is test1.txt.

text1.txt contain as following:

Input:

##[A1] [B1] [T1] [V1] [T2] [V2] [T3] [V3] [T4] [V4]## --> headers
1 1000 0 100 10 200 20 300 30 400
40 500 50 600 60 700 70 800
1010 0 101 10 201 20 301 30 401
40 501 50 601
2 1000 0 110 15 210 25 310 35 410
45 510 55 610 65 710
1010 0 150 10 250 20 350 30 450
40 550


Condition:
A1 and B1 -> for each A1 + (B1 + [Tn + Vn])
A1 should be in 1 column.
B1 should be in 1 column.
T1,T2,T3 and T4 should be in 1 column.
V1,V2,V3 and V4 should be in 1 column.

How do I sort it become like below?

Desire Output:

## A1 B1 Tn Vn ## --> headers

1 1000 0 100
10 200
20 300
30 400
40 500
50 600
60 700
70 800
1010 0 101
10 201
20 301
30 401
40 501
50 601
2 1000 0 110
15 210
25 310
35 410
45 510
55 610
65 710
1010 0 150
10 250
20 350
30 450
40 550


Here is my current code:

First Attempt:

Input

cat test1.txt | awk ' { a=$1 b=$2 } { for(i=1; i<=5; i=i+1) { t=substr($0,11+i*10,5) v=substr($0,16+i*10,5) if( t ~ /^ +[0-9]+$/ || t ~ /^[0-9]+$/ || t ~ /^ +[0-9]+ +$/ ){ printf "%7s %7d %8d %8d n",a,b,t,v } }}' | less


Output:

1 1000 400 0
40 500 800 0
1010 0 401 0
2 1000 410 0
1010 0 450 0


I'm trying using simple awk command, but still can't get the result.
Can anyone help me on this?

Thanks,

Am

1 1000

4 Answers
4

This is a rather tricky problem that can be handled a number of ways. Whether bash, perl or awk, you will need to handle to number of fields in some semi-generic way instead of just hardcoding values for your example.

bash

perl

awk

Using bash, so long as you can rely on an even-number of fields in all lines (except for the lines with the sole initial value (e.g. 1010), you can accommodate the number of fields is a reasonably generic way. For the lines with 1, 2, etc.. you know your initial output will contain 4-fields. For lines with 1010, etc.. you know the output will contain an initial 3-fields. For the remaining values you are simply outputting pairs.

1010

1, 2

4-fields

1010

3-fields

The tricky part is handling the alignment. Here is where printf which allows you to set the field-width with a parameter using the form "%*s" where the conversion specifier expects the next parameter to be an integer value specifying the field-width followed by a parameter for the string conversion itself. It takes a little gymnastics, but you could do something like the following in bash itself:

printf

"%*s"

integer

(note: edit to match your output header format)

#!/bin/bash

declare -i nfields wd=6 ## total no. fields, printf field-width modifier

while read -r line; do ## read each line (preserve for header line)
arr=($line) ## separate into array
first=${arr[0]} ## check for '#' in first line for header
if [ "${first:0:1}" = '#' ]; then
nfields=$((${#arr[@]} - 2)) ## no. fields in header
printf "## A1 B1 Tn Vn ## --> headersn" ## new header
continue
fi
fields=${#arr[@]} ## fields in line
case "$fields" in
$nfields ) ## fields -eq nfiles?
cnt=4 ## handle 1st 4 values in line
printf " "
for ((i=0; i < cnt; i++)); do
if [ "$i" -eq '2' ]; then
printf "%*s" "5" "${arr[i]}"
else
printf "%*s" "$wd" "${arr[i]}"
fi
done
echo
for ((i = cnt; i < $fields; i += 2)); do ## handle rest
printf "%*s%*s%*sn" "$((2*wd))" " " "$wd" "${arr[i]}" "$wd" "${arr[$((i+1))]}"
done
;;
$((nfields - 1)) ) ## one less than nfields
cnt=3 ## handle 1st 3 values
printf " %*s%*s" "$wd" " "
for ((i=0; i < cnt; i++)); do
if [ "$i" -eq '1' ]; then
printf "%*s" "5" "${arr[i]}"
else
printf "%*s" "$wd" "${arr[i]}"
fi
done
echo
for ((i = cnt; i < $fields; i += 2)); do ## handle rest
if [ "$i" -eq '0' ]; then
printf "%*s%*s%*sn" "$((wd+1))" " " "$wd" "${arr[i]}" "$wd" "${arr[$((i+1))]}"
else
printf "%*s%*s%*sn" "$((2*wd))" " " "$wd" "${arr[i]}" "$wd" "${arr[$((i+1))]}"
fi
done
;;
* ) ## all other lines format as pairs
for ((i = 0; i < $fields; i += 2)); do
printf "%*s%*s%*sn" "$((2*wd))" " " "$wd" "${arr[i]}" "$wd" "${arr[$((i+1))]}"
done
;;
esac
done


Rather than reading from a file, just use redirection to redirect the input file to your script (if you want to just provide a filename, then redirect the file to feed the output while read... loop)

while read...

Example Use/Output

$ bash text1format.sh <dat/text1.txt
## A1 B1 Tn Vn ## --> headers
1 1000 0 100
10 200
20 300
30 400
40 500
50 600
60 700
70 800
1010 0 101
10 201
20 301
30 401
40 501
50 601
2 1000 0 110
15 210
25 310
35 410
45 510
55 610
65 710
1010 0 150
10 250
20 350
30 450
40 550


As between awk and bash, awk will generally be faster, but here with formatted output, it may be closer than usual. Look things over and let me know if you have questions.

awk

bash

awk

As between awk and bash, awk will generally be faster, but here with formatted output, it may be closer than usual

read

echo

Unlike what is stated elsewhere, there's nothing tricky about this at all, you're just using fixed width fields in your input instead of char/string separated fields.

With GNU awk for FIELDWIDTHS to handle fixed width fields it really couldn't be much simpler:

$ cat tst.awk
BEGIN {
# define the width of the input and output fields
FIELDWIDTHS = "2 4 5 5 6 5 6 5 6 5 6 99"
ofmt = "%2s%5s%6s%5s%6s%sn"
}
{
# strip leading/trailing blanks and square brackets from every field
for (i=1; i<=NF; i++) {
gsub(/^[[s]+|s]+$/,"",$i)
}
}
NR==1 {
# print the header line
printf ofmt, $1, $2, $3, "Tn", "Vn", " "$NF
next
}
{
# print every other line
for (i=4; i<NF; i+=2) {
printf ofmt, $1, $2, $3, $i, $(i+1), ""
$1 = $2 = $3 = ""
}
}


.

$ awk -f tst.awk file
## A1 B1 Tn Vn ## --> headers
1 1000 0 100
10 200
20 300
30 400
40 500
50 600
60 700
70 800
1010 0 101
10 201
20 301
30 401
40 501
50 601
2 1000 0 110
15 210
25 310
35 410
45 510
55 610
65 710
1010 0 150
10 250
20 350
30 450
40 550


With other awks you'd use a while() { substr() } loop instead of FIELDWIDTHS so it'd be a couple more lines of code but still trivial.

while() { substr() }

FIELDWIDTHS

The above will be orders of magnitude faster than an equivalent shell script. See https://unix.stackexchange.com/questions/169716/why-is-using-a-shell-loop-to-process-text-considered-bad-practice.

This isn't easy because it is hard to identify when you have the different styles of row — those with values in both column 1 and column 2, those with no value in column 1 and a value in column 2, and those no value in column 1 or 2. A first step is to make this easier — sed to the rescue:

sed

$ sed 's/[[:space:]]{1,}$//
s/^....../&|/
s/|....../&|/
:a
s/|( *[0-9][0-9]* )( *[^|])/|1|2/
t a' data
1 | 1000 | 0 | 100 | 10 | 200 | 20 | 300 | 30 | 400
| | 40 | 500 | 50 | 600 | 60 | 700 | 70 | 800
| 1010 | 0 | 101 | 10 | 201 | 20 | 301 | 30 | 401
| | 40 | 501 | 50 | 601
2 | 1000 | 0 | 110 | 15 | 210 | 25 | 310 | 35 | 410
| | 45 | 510 | 55 | 610 | 65 | 710
| 1010 | 0 | 150 | 10 | 250 | 20 | 350 | 30 | 450
| | 40 | 550
$


The first line removes any trailing white space, to avoid confusion. The next two expressions handle the fixed-width columns 1 and 2 (6 characters each). The next line creates a label a; the substitute finds a pipe |, some spaces, some digits, a space, and some trailing material which doesn't include a pipe; and inserts a pipe in the middle. The t a jumps back to the label if a substitution was done.

a

|

t a

With that in place, it becomes easy to manage awk with a field separator of |.
This is verbose, but seems to do the trick:

awk

|

awk -F '|' '
$1 > 0 { printf "%5d %4d %3d %3dn", $1, $2, $3, $4
for (i = 5; i <= NF; i += 2) { printf "%5s %4s %3d %3dn", "", "", $i, $(i+1) }
next
}
$2 > 0 { printf "%5s %4d %3d %3dn", "", $2, $3, $4
for (i = 5; i <= NF; i += 2) { printf "%5s %4s %3d %3dn", "", "", $i, $(i+1) }
next
}
{ for (i = 3; i <= NF; i += 2) { printf "%5s %4s %3d %3dn", "", "", $i, $(i+1) }
next
}'


Output:

1 1000 0 100
10 200
20 300
30 400
40 500
50 600
60 700
70 800
1010 0 101
10 201
20 301
30 401
40 501
50 601
2 1000 0 110
15 210
25 310
35 410
45 510
55 610
65 710
1010 0 150
10 250
20 350
30 450
40 550


If you need to remove the headings, add 1d; to the start of the sed script.

1d;

sed

sed

awk

0 0

0 0

diff -b

This might work for you (GNU sed):

sed -r '1d;s/^(.{11}).{11}/&n1/;s/^((.{5}).*n)2/1 /;s/^(.{5}(.{6}).*n.{5})2/1 /;/S/P;D' file


Delete the first line (if the header is needed see below). The key fields occupy the first 11 (the first key is 5 characters and the second 6) characters and the data fields occupy the next 11. Insert a newline and the key fields before each pair of data fields. Compare the keys on adjacent lines and replace by spaces if they are duplicated. Do not print any blank lines.

If the header is needed, use the following:

sed -r '1{s/[[^]]+]s*//5g;y// /;s/1/n/3g;s/B/ B/;G;b};s/^(.{11}).{11}/&n1/;s/^((.{5}).*n)2/1 /;s/^(.{5}(.{6}).*n.{5})2/1 /;/S/P;D' file


This does additional formatting on the first line to remove superfluous headings, 's, replace 1's by n, add an additional space for alignment and a following empty line.

1

n

Further more. By utilising the second line of the input file as a template for the data, a sed script can be created that does not have any hard coded values:

hard coded

sed -r '2!d;s/s*S*//3g;s/.>/&n/;h;s/[^n]/./g;G;s/[^n.]/ /g;s#(.*)n(.*)n(.*)n(.*)#1d;s/^(12)12/&n1/;s/^((1).*n)2/13/;s/^(1(2).*n1)2/14/;/S/P;D#' file |
sed -r -f - file


The script created from the template is piped into a second invocation of the sed as a file and run against the original file to produce the required output.

Likewise the headers may be formatted if need be as so:

sed -r '2!d;s/s*S*//3g;s/.>/&n/;h;s/[^n]/./g;G;s/[^n.]/ /g;s#(.*)n(.*)n(.*)n(.*)#s/^(12)12/&n1/;s/^((1).*n)2/13/;s/^(1(2).*n1)2/14/;/S/P;D#' file |
sed -r -e '1{s/[[^]]+]s*//5g;y// /;s/1/n/3g;s/B/ B/;G;b}' -f - file


By extracting the first four fields from the second line of the input file, Four variables can be made. Two regexp and two values. These variables can be used to build the sed script.

N.B. The sed script is created from strings extracted from the template and the variables produced are also strings so they can be concatenated to produce further new regexp's and new values etc etc

sed -r '1d;s/^(.{11}).{11}/&n1/;P;D' file

sed -r '2!d;s/s*S*//3g;s/.>/&n/;h;s/[^n]/./g;G;s/[^n.]/ /g;s#(.*)n(.*)n(.*)n(.*)#s/^(12)12/&n1/;s/^((1).*n)2/13/;s/^(1(2).*n1)2/14/;/S/P;D#' file | sed -r -e '1{s/[[^]]+]s*//5g;y// /;s/1/n/3g;s/B/ B/;G;b}' -f - file

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.