CODEWITHSUNDEEP
awk
Joanne
Joanne

Reputation: 127

awk: filter for number of non-zeros not filtering

I have previously used awk to reduce an enormous data table which has mostly zeros, to a smaller table with just the interesting rows (those with not too many zeros), with something like this:

awk -F '\t' '{count=0} {for(i=2; i<30; i++) if($i==0) count++} {if(count<5) print $0}' BigTable > SmallerTable

Now I would like to filter a similar table, to find rows with non-zero values in most of the "female" columns and zeros in most of the "male" columns. I tried to use the same awk logic, but my code returns all lines of the input file. 

#! /usr/bin/awk -f 
FS="\t"
{countF=0} {for(i=2; i<7; i++) if($i==0) countF++} 
# count zeros in female columns 2-6
{countM=0} {for(i=7; i<12; i++) if($i==0) countM++} 
# count zeros in male columns 7-12
{if (countF<2 && countM>3) {print $0}}
# if fewer than 2/5 females AND more than 3/5 males are zero, print line

My input file starts like this: 

NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN F_CR1   F_CR2   F_CR3   F_CR4   F_CR6   M_CR10  M_CR5   M_CR7   M_CR8   M_CR9
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA 14727   13526   13318   13862   11040   18975   21411   20079   16285   15611
CCGGTGTGACAACTGTAGTGAACTCAGCTCA 23  32  26  15  28  28  42  29  8   22
AACCAAATCTACAAACAGGAGATGTTGTTCT 107 110 118 106 95  100 121 132 92  90
GAAATAGAACAGGCCTGGAAGCCATGTCAAA 15  15  16  12  11  31  23  19  9   28

Have I messed up the syntax in the print line? Any advice much appreciated!

Upvotes: 0

Views: 126

Answers (1)

Ed Morton
Ed Morton

Reputation: 203324

Change FS="\t" to BEGIN{FS="\t"}. Right now the result of that assignment is a true condition which invokes the default action of printing every line.

Then change your shell script to:

/usr/bin/awk '
BEGIN { FS="\t" }
{
    # count zeros in female columns
    countF=0
    for(i=2; i<=6; i++) {
        if ($i==0) {
            countF++
        }
    }

    # count zeros in male columns
    countM=0
    for(; i<=NF; i++) {
        if ($i==0) {
            countM++
        }
    }
}
# if fewer than 2/5 females AND more than 3/5 males are zero, print line
countF<2 && countM>3
' "$@"

so it's more awk-ish and easier to enhance later if/when you need to separate shell args into awk args and awk variable assignments (shebangs are not useful for this).

Also consider abbreviating it and removing the hard-coded Male/Female limits but get them from the header line instead:

/usr/bin/awk '
BEGIN { FS="\t" }
FNR==1 {
    for (i=2; i<=NF; i++) {
        sub(/_.*/,"",$i)
        gender[i] = $i
    }
    next
}
{
    for (i=2; i<=NF; i++) {
        count[gender[i]] += ($i==0)
    }
}
count["F"]<2 && count["M"]>3
' "$@"

The above is untested since you didn't provide the expected output for us to test with.

Upvotes: 3

Related Questions