[SOLVED] make a sum in the awk command

Drosera_capensis · 04-26-2020, 09:05 AM

Hello everyone,

I would like to make the ratio of all the 50 numbers that are in a column of a file. So, the ratio version of a column of numbers.
The ratio would be equal to row value / sum of all the column.

Code:

for f in {1..50};
  do 
   cat file | awk ' {sum += $1} {a = $1; b = sum; print (a/b)}' ; 
 done

But the result is zero at every loop. WOuld anyone know what is wrong with this command?

Thanks for the help!

Turbocapitalist · 04-26-2020, 09:25 AM

You'd have to read the file to know the sum of the column. So that means you'd have to make two passes through the file. You can compare the NR and FNR to see if you are on the "first" file or the "second" file. Since these are the same file, it is just whether you are on the first pass through it or not.

Code:

awk    'NR==FNR { sum=sum+$1; next }
        NR!=FNR { print $1,$1/sum }' \
        OFS="\t" \
        file file

So in this case cat must really be avoided.

Drosera_capensis · 04-26-2020, 09:50 AM

It is working, thanks Turbocapitalist. I have to spend more time studying the awk syntax.

I am a bit confused by the "file file" at the end of the command, but I suppose it is for each part of the awk command?

Turbocapitalist · 04-26-2020, 09:59 AM

AWK can read an arbitrary number of files. Here the same file is fed in twice.

Both instructions are read for each file, but only the first file has NR and FNR matching. Only the second and subsequent files will have NR and FNR not match. Take a quick look in the manual page for AWK to see what those built-in variables stand for. There are some others which are often handy.

Drosera_capensis · 04-27-2020, 04:52 AM

Thanks for the recommandations. I have one more question about the output format given by awk.

The former command print output in scientifict format (9.50368e-05). I wish to output results in plain numbers.

I have used printf to do so, but it outputs a unique string of results, without spaces or jump to next row.

I have tried the following syntax, with %s. , but it does not work.

Quote:

awk 'NR==FNR { sum=sum+$1; next }
NR!=FNR { printf %s.$1/sum }' \
OFS="\t" \
file file

Would anyone know if the printf formating allow to jump to next row?
Or if any command can suppress the scientific format of the output given by awk?

Turbocapitalist · 04-27-2020, 04:55 AM

Quote:

Originally Posted by Drosera_capensis

I have tried the following syntax, with %s. , but it does not work.

Re-check the syntax for AWK's printf() in the reference manual. It is rather similar to printf() as found elsewhere and quite different from what you wrote.

Code:

man awk

Drosera_capensis · 04-27-2020, 05:43 AM

I have read the BSD Library Functions Manual on printf(3) (link given by man awk). I have to admit I don't understand most of it. I have made some experiments with it.

Quote:

{print output, "\n"}

Is outputing results jumping a line, as I wished.

Quote:

{printf output, "\n"}

Is not changing the output of printf. A long string of output.

Could anyone give me some clues about this curious result?

shruggy · 04-27-2020, 05:55 AM

Quote:

Originally Posted by Drosera_capensis

{printf output, "\n"}

For printf, it should be in reverse order:

Code:

{printf "%s\n", output}

shruggy · 04-29-2020, 03:42 PM

With miller:

Code:

seq 50 |
 mlr --csv -N put -q '
   @sum+=$1;@num[NR]=$1;
   end{
     for (k,v in @num){print v."\t".v/@sum}
   }'

shruggy · 01-31-2022, 12:34 PM

Revisiting the old thread to give a simpler Miller solution (the more so as the solution outlined above won't work unchanged in Miller 6, see issue #909)

Code:

mlr --n2t -N fraction -f 1

mshare from the MCMD toolset is another possibility.

Code:

KG_VerboseLevel=1 mshare -nfn -x -q f=0