Passing some rows based on a threshold - awk

Question

I am wondering how can I adapt the code (or add extra calls) below to pass some rows based on a threshold (here, a percentage value). My xyz file:
21.4 7
21.5 141
21.6 4
21.7 43
21.8 26
21.9 133
22 305
22.1 216
22.2 93
22.3 33
22.4 13
22.7 23
22.8 2
22.9 10
23 39
23.1 22
23.2 33
23.3 8
23.4 9
23.5 2
23.6 270
23.7 724
23.8 2349
23.9 2
24 1
24.1 11
24.2 376
24.3 1452
24.4 92
with the following awk call I obtain the corresponding percentage of the value in the 3rd column:
awk 'FNR==NR { s+=$2; next; } { printf "%s\t%s\t%s%%\n", $1, $2, 100*$2/s }' xyz xyz | sort -k3 -g
which gives:
24 1 0.0155304%
22.8 2 0.0310607%
23.5 2 0.0310607%
23.9 2 0.0310607%
21.6 4 0.0621214%
21.4 7 0.108713%
23.3 8 0.124243%
23.4 9 0.139773%
22.9 10 0.155304%
24.1 11 0.170834%
22.4 13 0.201895%
23.1 22 0.341668%
22.7 23 0.357198%
21.8 26 0.403789%
22.3 33 0.512502%
23.2 33 0.512502%
23 39 0.605684%
21.7 43 0.667806%
24.4 92 1.42879%
22.2 93 1.44432%
21.9 133 2.06554%
21.5 141 2.18978%
22.1 216 3.35456%
23.6 270 4.1932%
22 305 4.73676%
24.2 376 5.83942%
23.7 724 11.244%
24.3 1452 22.5501%
23.8 2349 36.4808%
So, I want to automagically filter the last Nth rows if the sum of the last values in the 3rd column is just greater than 60%, in the case above it will be 36.4808% + 22.5501% + 11.244% = 70.2749%:
23.7 724 11.244%
24.3 1452 22.5501%
23.8 2349 36.4808%
Any hints are appreciated,

Answer 1

That could be done in a single awk command, but I think this version is shorter:
awk -v OFS="\t" 'FNR==NR { s+=$2; next; } { $3=100*$2/s "%" }1' xyz xyz |
sort -k3 -g |
awk '(t+=$3)>40'
prints out:
23.7 724 11.244%
24.3 1452 22.5501%
23.8 2349 36.4808%

Answer 2

Assumptions:
for rows with duplicate values in the 2nd column, precendence goes to rows with a higher row number (via awk/FNR)
One GNU awk idea:
awk '
BEGIN {OFS="\t" }
{ a[$2][FNR]=$1 # [FNR] needed to distinguish between rows with duplicate values in 2nd column
s+=$2
}
END { PROCINFO["sorted_in"]="#ind_num_desc" # sort array by numeric index (descending order)
for (i in a) { # loop through array (sorted by $2 descending)
for (j in a[i]) { # loop through potential duplicate $2 values (sorted by FNR descending)
pct=100*i/s
out[--c]=a[i][j] OFS i OFS pct "%" # build output line, store in out[] array, index= -1, -2, ...
sum+=pct
if (sum > 60) break # if the sum of percentages > 60 then break from loop
}
if (sum > 60) break # if the sum of percentages > 60 then break from loop
}
for (i=c;i<0;i++) # print contents of out[] array starting with lowest index and increasing to -1
print out[i]
}
' xyz
NOTE: requires GNU awk for:
multidimensional arrays (aka array of arrays)
PROCINFO["sorted_in"] support
This generates:
23.7 724 11.244%
24.3 1452 22.5501%
23.8 2349 36.4808%